Air Tightness

TOPIC
Air-tightness
Page....................................7-117
Full papers - NSB 2014 page 6

Full paper no: 1
A quantitative evaluation of airtightness measurement

experiences
Thomas Olofsson, Ph.D and Ingrid Allard, M.Sc.
Applied Physics and Electronics, Umeå University, Sweden
KEYWORDS: Air tightness, Buildings, Field Measurements, Survey, ZEF

SUMMARY:
Evaluation of building air tightness based on field measurements is an important aspect in the process
to provide good indoor environment and energy efficient buildings. The measurements are generally
conducted by experts. To improve field methods for evaluation of building air tightness, experience of
these experts can be useful.
The scope of this paper is to problematize usefulness of methods for airtightness measurements, how
and when methods are used and potential for future development. Surveyed Swedish and Finnish
experts have valued the performance and usefulness of methods in the perspective of being used for
singe family, multifamily, new, inhabited, under construction, renovation, leaky and air-tight
buildings. They also valued the future potential, as well as their own level of experience and expertise,
of the methods.
Although the results of the survey were based on a small set of collected data, it indicated that the
experts favored methods with their highest assessed experience and expertise, i.e. Blower door and
Surface temperature measurements/Thermography. Potential of future development within
quantitative and component methods was assessed to be positive in general and in particular for the
favored methods.
1. Introduction
Uncontrolled ventilation from air leakage has an impact on the indoor environment and the building
energy efficiency. In that context, the air tightness is generally an important issue calling for
evaluation. However, that evaluation is not an easy task (Sherman 2004, Sikander 2008, Gränne 2001,
Bankvall 2013). Evaluation of building airtightness, based on field measurements for new buildings,
buildings under construction, as well as for inhabited buildings is often conducted by trained experts.
For their use there are several available methods. Based on the conditions in question and the demands
on the evaluation, each method demonstrates its strengths and drawbacks, described in the literature
(Allard 2012, Nirvan 2010, Lee 2011, Sherman 1992, Hassan 2013). The reported evaluations were
often based on experimental conditions. It can be assumed that the experts’ experience from field
measurements can differ from what is documented under controlled laboratory conditions. Field
experiences can supplement the experimental findings in general but also be useful for future
development of field measurements.
In this paper we have investigated how a number of Finnish and Swedish experts value their expertise,
the performance and usefulness, and expected future use of a selected number of identified methods
for measuring airtightness. The scope of the paper is to problematize usefulness of the methods, how
and when they are used and their potential for future development. The intension was for the findings
of this pre-study to be used for a future similar, but more qualitative, investigation.

2. Survey background
The survey was based on statements and answered using a parametric Likert scale, with the range 1 to
5, in accordance with an increasing level of agreement. The survey was conducted on a selected set of
methods and specified conditions of building types and life stages, introduced in this section.
2.1 Surveyed methods
The surveyed methods are arranged in three different groups: quantitative methods, component
methods and experimental methods, see table 1 below. The methods in the quantitative group are used
for verification of air tightness. The methods in the component group are used to enhance air tightness
of building components, e.g. a wall. The methods in the experimental group are not used
commercially today, but may have advantages and future potential.
Table 1. The methods analyzed in the study, arranged in categories.
Quantitative Component Experimental
1. Fan pressurization - blower 4. Ocular inspection/light 8. Acoustic measurements:
door: A common method that uses measurements: Ocular (visual) Pressurization operates by inducing
a mounted fan to maintain a inspection of air tightness on the a sinusoidal volume change pulse
defined pressure inside the test object. It requires experience of about 1 Hz. By measuring the
building, in order to estimate the and knowledge by the user. Ocular pressure pulse, information about
air permeability and is inspection is not standardized air leakages can be extracted and is
standardized in (ISO 9972/EN (Allard 2012). described by (Sherman 2004)
13829).
2. Fan pressurization - ventilation: 5. Surface temperature 9. Pulse pressurization, SECA:
A method that uses the fan in the measurements/thermography: Pulse pressurization or sudden
ventilation system to maintain a Surface temperature measurement expansion of compressed air
defined pressure inside the determines the temperature (SECA) can be obtained from e.g.
building, in order to estimate the distribution, performed with a a compressed air tank. A power
air permeability and is surface thermometer or an IR law equation predicts a finite
standardized in (ISO 9972/EN camera. It demands a temperature recovery time for the decay and
13829). difference > 5°C (Nordtest 1997), can be used with the measurements
standardized in (SS EN 13187). to calculate leakage and volume
(Lee 2011).
3. Tracer gas: SF6 or N2O, is 6. Air velocity measurements/hand
diffused in the building or a single method: Carried out with
zone. With time the tracer gas will suppressed pressure inside the
become diluted by inflowing air. building. The anemometer (or
Based on the gas decay and hand) is moved along the analyzed
knowing the building geometry, area and air velocity (leakage) can
the air permeability can be be detected and is described in
calculated, described in (NT VVS (ISO 7726, ISO 7730).
055).
7. Smoke methods: With an
overpressure, inside air leakages
can be detected using smoke
released with a smoke pen, bottle,
machine or pistol smoke and is
described in (ASTME 1186).
2.2 Surveyed building types for the quantitative methods

Five building types where investigated with the quantitative methods: Single-zone, Multi-zone, High
rise, Leaky building and Air tight buildings. Single-zone buildings are defined as single family
buildings. Multi-zone buildings are for example apartment buildings or other houses with single-zone
living compartments that border both to the outside and another single-zone. High rise buildings are
defined as buildings with four floors or more. This definition is based on building conditions in the
northern parts of Europe. Leaky buildings are buildings with high air permeability and air tight

buildings have low air permeability. The latter are often some kind of energy efficient building, such
as a passive house.
2.3 Surveyed life stages for the methods
The different life stages of buildings, considered for component and experimental methods, are listed
in table 2 below, with comments. Component methods can be used for all building types without
significant mutual disadvantages. Experimental methods are considered too untested to analyze with
respect to building type, and thus with concern of the questionnaire length, the questions for
experimental methods related to all building types. New buildings are buildings which are built, but
the inhabitants have not moved in. Forecast and background questions were asked considering all life
stages and building types simultaneously.
Table 2. The life stages considered and comments.
Life stage Comment
1. Under construction or renovation Not used for quantitative questions
2. New building Climate envelope is finished but inhabitants have not
moved in
3. Inhabited Building Inhabitants are living in the building
3. Questioner
The questionnaire used for the survey is developed by a Finnish marketing company (ZEF 2013) and
utilizes a graphical 2D-interface.The results are reported in Fig 1-3, below, where the smaller circle
with a number (1..n) is referring to the evaluated questions. The position of the point is based on the
average of all who have answered the question. Around each circle there is a darker area/ellipse,
calculated with a standard deviation formula. It shows, in a graphical way, how much answers from
the respondents (experts) vary from each other. The smaller the ellipse is, the more persons agree on
that topic. The model of the ellipse also tells you in which direction/criteria (horizontal/vertical)
agreement is smaller or bigger.
The questions were categorized into three levels; Background, Recommendations and Forecast, all
further introduced in this section.
3.1 Background: Expert awareness questions
The background question is supposed to give information about expertise and practical experience of
the methods among the participants. This shows which methods are most commonly used within the
expert group. The background question was stated as: “I assess my expertise and practical experience
to be good”. Therefore, both information on expertise (which may be only theoretical) and practical
experience will be gathered. This information may be used for qualitative assessments of the answers
in combination with the other questions.
3.2 Recommendations: Method recommendation questions
The recommendations are supposed to indicate which methods that is appropriate to use under certain
circumstances. The circumstances, or scenarios, depend on building type and life stage. The
recommendation question was stated as: “I would recommend this method by the terms of ease of use
and performance for this building type and life stage”. The question was modified to apply to the
different building types and life stages.
3.3 Forecast: Future method use questions
The questions were asked about the potential and future use of the investigated methods. The forecast
question was stated as: “I believe there is likelihood of use and potential of performance for this
method in the future”. The term ‘future’ is used since the exact date is not important.

4. Results
The survey was conducted on a number of identified Finnish and Swedish experts. In total 16 Finnish
and 4 Swedish experts were asked to participate. Finally, 8 Finnish and 2 Swedish experts completed
the survey.
In this section, the results are illustrated in graphs based on the method categories: quantitative,
qualitative and experimental. The results are also presented as quantitative scoring based on the Root
Means Square of the assessed value of x-axes (0-100) and of the y-axes (0-100), where the Likert
scoring of 1 to 4 was transformed to a scale 0-100. The calculated scoring is referred to as “high” if it
is larger than 67. The scoring is referred to as “low” if it is less than 33. “Medium” is used if the
scoring it is between 45 and 55. A scoring of 55 to 67 is referred to as “above medium”, and a scoring
of 33 to 45 is referred to as “below medium”.
4.1 Qualitative methods

For the three investigated quantitative methods, the experts assessed their expertise and practical
experience of fan pressurization with a blower door as high, marked with circle (1) in Fig 1a. The
expertise and practical experience of fan pressurization using ventilation (2) and the tracer gas method
(3), in Fig 1a, was assessed below medium.
When the experts assessed recommendations for the three methods, blower door (which the experts
were most familiar with) gained high and the best score, applied to all investigated types of new
buildings; i.e. single-zone, multi-zone, high rise, leaky, and air tight, see the circle with (1,4,7,10,13)
in Fig 1b, as well as for all investigated types of inhabited buildings, see (1,4,7,10,13) in Fig 1c. The
performance and ease of use of fan pressurization using ventilation was assessed above medium and
lower for all investigated types of new buildings, see circle with (2,5,8,11,14) in Fig 1b, as well as for
all investigated types of inhabited buildings, see (2,5,8,11,14) in Fig 1c. The tracer gas method was
rated below medium and lowest for all investigated types of new buildings, see circle with (3,6,9,
12,15) Fig 1c , as well as for all investigated types of inhabited buildings, see (3,6,9,12,15) Fig 1d.
The investigation of potential and future use indicated that the by experts favored blower door method
gained a high score. Fan pressurization using ventilation was rated lower, but high, and the tracer gas
methods was rated below medium, see Fig 1d.
An overall indication of the investigation of qualitative methods is that the assessment was rated
according to the experience of the experts, i.e. blower door method, which the experts seamed
accustomed to, were found superior in all investigated situations.
FIG 1a. Expertise and practical experience of FIG 1b. Performance and ease of use for
fan pressurization with a blower door, fan qualitative methods: Recommendations for
pressurization using ventilation and the tracer single-zone, multi-zone, high rise, leaky as well
gas methods as air tight and new buildings

FIG 1c. Performance and ease of use for FIG 1d. A forecast of potential and future use of
qualitative methods: Recommendations for fan pressurization with a blower door, fan
single-zone, multi-zone, high rise, leaky as well pressurization using ventilation and the tracer
as air tight and inhabited buildings gas methods
4.2 Component methods
The experts assessed their expertise and practical experience of ocular inspection/light measurements
and surface temperature measurements/thermography to be good and above medium, see circle with
(1,2) in Fig 2a. The experts assessed their expertise and practical experience of air velocity
measurements/hand method and smoke methods to be medium, see (3,4) in Fig 2a.
The experts assessed a high recommendation for surface temperature measurements/thermography for
buildings under construction or renovation, new buildings and inhabited buildings, see circle with
(2,5,6), as well as ocular inspection and light measurements for a building under construction or
renovation, see (1), in Fig 2b. The recommendation of air velocity measurements/hand method and
smoke methods, for a building under construction or renovation were assessed above medium, see
(3,4) in Fig 2b.
The experts made a forecast of potential and future use of component methods. The scoring of
potential and likelihood of use were ranked similar as the scoring of expertise and practical
experience. The forecast of ocular inspection/light measurements and surface temperature
measurements/thermography was scored high, see circle with (1,2) Fig 2c. Air velocity
measurements/hand method and smoke methods was scored above medium, see (3,4) Fig 2c.
The results indicate that the experts in general assessed high rating for methods they are accustomed
to, i.e. the favored ocular inspection/light measurements and surface temperature
measurements/thermography.

FIG 2a. Expertise and practical FIG 2b. Recommendations for FIG 2c. A forecast of potential and
experience of ocular inspection and buildings under likelihood for component methods
light measurements, surface construction/renovation, new
temperature measurements and buildings or inhabited buildings for
thermography, air velocity component methods
measurement/hand method, and
smoke methods
4.3 Experimental methods
The expertise and experience of the experts was assessed below medium for acoustic methods see (1)
in Fig 3a. The expertise and experience of pulse pressurization or sudden expansion of compressed air
(SECA) was assessed low, see (2) in Fig 3a.
The experts recommendations of acoustic measurements was rated above medium for new buildings,
see (2) in Fig 3b and medium for new buildings and inhabited buildings, see circle with (1,4) in Fig
3b. Sudden expansion of compressed air (SECA) was assessed below medium, lowest for buildings
under construction and new buildings, see circle with (3,5) in Fig 3b.
The experts assessed the likelihood of future use and potential of performance for acoustic
measurements above medium, which can be interpreted as a small optimism for the future, see (1) in
Fig 3c. The forecast for pulse pressurization or sudden expansion of compressed air (SECA) was
assessed lower, medium, see (2) in Fig 3c.
It can be assumed that the experts have limited practical experience of these methods and
consequently the interpretation of the results has less significance.
FIG 3a. Expertise and experience FIG 3b. Performance and ease of FIG 3c. Potential and likelihood of
of experimental methods use for experimental methods use for experimental methods
4.4 General results
The identification of experts for survey in this pre-study was based on mapping procedure supported
by the national research institutes. It was found difficult to identify Swedish and Finnish experts of the
investigated field measurement methods. For the study, 20 experts were identified and the

participation rate was 50%. From a statistical perspective that is small. For a larger study it would be
beneficial with a more comprehensive mapping of experts.
Although the statistical material was limited, the standard deviation in the rating of the experts was
relatively small, illustrated as the darker area around each circle in Figs 1-3. Additionally, the experts
claimed their own expertise to be rather high. For the three categories of methods: quantitative,
qualitative and experimental, the expert rating of recommendations were in accordance with their
rated expertise and practical experience, which in general is rated medium to high. This indicates that
the surveyed experts have sufficient expertise to answer the questionnaire. Thus it is assumed that the
results can at least be used to give some guidance and some general trends can be discussed. An
overall illustration of the results is presented in Tab 3.
Tabel 1. Summary of the results for the investigated methods. The calculated root mean square can be
found within the brackets.
Categories Methods Expertise & Recommendations Future
Experience forecast
Quantitative Fan pressurization - High (70) High for all applications (74) High (83)
blower door
Fan pressurization – Below medium Above medium for all High (71)
ventilation (43) applications (56)
Tracer gas Below medium Below medium for all Below
(42) applications (43) med. (44)
Component Ocular inspection/light Above medium High for buildings under High (69)
measurements (64) construction (74)
Surface temp measur./ Above medium Best component method, high High (75)
Thermography (66) for all stages (78)
Air velocity measure- Medium Above medium for buildings Above
ments/hand method (54) under construction (65) med. (64)
Smoke methods Medium (48) Above medium for buildings Above
under construction (56) med. (59)
Experimental Acoustic Below medium Above medium for new build., Above
measurements (41) medium under construction medium
and inhabited buildings (55) (66)
Pulse pressurization, Low(16) Below medium for new and Medium
SECA inhabited buildings(38) (54)
The results showed an obvious correlation between high assessed recommendations and future
expectations and high assessed experience and expertise of, e.g. Blower door and Surface temp
measurements/Thermography. According to the experts, the potential for future development of
quantitative and component methods reflects the recommendations of the favored methods. However,
for experimental methods, the experts express a positive expectation of development, although the
present recommendation was moderate. The rating of the future forecast in the study corresponds to
the rated expertise and practical experience. The experts also indicated a general difficulty to measure
air tightness in high rise and multi zone buildings, regardless of the method used.
Some overall suggestions of explanation can be proposed. The experts seems to prefer methods they
are accustomed to before selecting method from actual condition, such as building type and the life
stage of the building. Additionally, there is a general positive expectation from the experts in terms of
a general large and growing future potential of field measurements in general for all three method
categories.
5. Conclusions
The study was made on a rather small set of participating experts and should thus be interpreted as a
pre-study. The investigated methodology was found applicable and useful and some trends and

findings were found that can be subject for future studies, e.g. of how and when the methods are used
as well as future expectation.
For an interpretation in more detail, a future extended investigation is proposed, where more effort
should be taken to get a larger set of experts, perhaps including more countries. Further, the study in
this paper was conducted on a single survey. To collect more information and get more comprehensive
results, it could be based on a methodology such as the Delphi method. In a future study experience
based on findings in the literature could also be used to complement the findings from the surveyed
experts.
6. Acknowledgement
This project has been funded by the Intereg IV project IEEB and the Kolartic project SBHN. The
foundation of the survey presented in this paper was made by the Energy Engineering MSc-student
Ragnar Björkén, at Umeå University, who is gratefully acknowledged. The authors would also express
their gratitude to the experts who participated is the survey.
References
I. Allard, T. Olofsson, and O.B. Hassan, (2012) “Methods for air tightness analysis for residential
buildings in Nordic countries" WIT Transaction on Ecology & the Environment, pp. 311–322
ASTME 1186-03 (2009), Standard Practices for Air Leakage Site Detection in Building Envelopes
and Air Barrier Systems
C. Bankvall (013), "Luftboken : luftrörelser och täthet i byggnader", Studentlitteratur
F. Gränne (2001), “Air and Water Tightness in Building Envelopes - Evaluation of Methods for
Quality Assurance,” KTH, Stockholm
O Hassan (2013), An alternative method for evaluating the air tightness of building components,
Building and Environment, Vol. 67, pp. 82-86
ISO 7726 (2001), Ergonomics of the thermal environment, Instruments for measuring physical
quantities (British Standard)
ISO 7730 (2005), Ergonomics of the thermal environment -- Analytical determination and
interpretation of thermal comfort using calculation of the PMV and PPD indices and local
thermal comfort criteria
ISO 9972:2006 (2006), Thermal performance of buildings -- Determination of air permeability of
buildings -- Fan pressurization method
M. J. Lee, N. Il Kim, and H. S. Ryou (2011), “Air tightness measurement with transient methods using
sudden expansion from a compressed chamber,” Building and Environment, vol. 46, no. 10,
pp. 1937–1945
G. Nirvan, F. Haghighat, L. Wang, and H. Akbari (20012), “Contaminant transport through the garage
– House interface leakage,” Building and Environment, vol. 56, pp. 176–183
Nordtest (1987), “Buildings: Total Outdoor Air Inflow (NT VVS 055),” pp. 1–11
M. Sherman (1992), “A Power-Law Formulation of Laminar Flow in Short Pipes,” Berkeley, Journal
of Fluids Engineering, Vol.114, pp 601-605
M. H. Sherman and R. Chan (2004), “Building Airtightness: Research and Practice,” Lawrence
Berkeley National Laboratory Report No. LBNL-53356, Berkeley
E. Sikander and P. Wahlgren (2008), “Alternativa metoder för utvärdering av byggnadsskalets
lufttäthet,” SP-Borås
SS-EN 13187 (1999), Byggnaders termiska egenskaper - Kvalitativ metod för lokalisering av termiska
ofullkomligheter i klimatskärmen - Infraröd metod
ZEF Solutions Ltd.,” 2013, Available: http://www.zef.fi/en/. [Accessed:10 Nov 2013]

Full paper no: 2
Analysis and prediction of transient flow during night ventilation

in a full scale test room
Sarah Leenknegt, PhD1
Dirk Saelens, Professor1
1
Building Physics Section, Department of Civil Engineering, KU Leuven, Leuven, Belgium
KEYWORDS: measurement, mechanical ventilation, unsteady ventilation, tracer gas, night

ventilation, full scale test room
SUMMARY
A full scale test room was constructed for the analysis of the energy performance of a room with night
ventilation with air supply through a hopper window, supported by an extraction fan. Contrary to
typical climate rooms, the room was exposed to realistic climatic conditions. This paper illustrates by
means of measurements the air flow phenomena during night ventilation. The room includes a
concrete floor and ceiling and detailed monitoring is done of local air and mass temperatures, air
flow rate and local air velocities and surface heat flux. Based on a selection of summer nights, an
assessment is made of the flow pattern occurring during night ventilation. Furthermore, to allow
taking into account such flow changes in BES models, the timing of the flow changes are connected to
the Richardson number. Three air flow typologies were identified: forced flow with short-circuit
between supply opening and outlet (Coandă at ceiling), transition flow with limited Coandă effect, and
buoyancy dominated flow with supply air falling down upon entry. The implications for assumptions
used in BES models are discussed shortly, as the flow pattern will strongly influence the surface heat
transfer.
1. Introduction
Night ventilation is a promising technique, able to reduce the cooling load through use of cold outdoor
air at high air change rates to cool down the internal thermal mass of a building in moderate climates
with sufficiently cold nights. A previous study focussed on the determination of the transient surface
convection during night ventilation through numerical simulations using the commercial CFD package
Fluent (Leenknegt et al. 2013). A simplified 2D geometry was simulated with unsteady RANS, using
constant boundary conditions. Various supply air temperatures and velocities were studied, for a flow
time of up to eight hours. Results from this numerical study showed the prevalence of unstable flow
behaviour during constant high ventilation rates. Three flow regimes were observed, illustrated in FIG
1. Buoyant flow was defined as air falling down upon entry, with the main flow path over the floor
surface. At a critical point, the air detaches from the wall under the inlet, moves upwards and attaches
to the ceiling. The transition flow regime is then defined as a flow adhering to the ceiling with variable
attachment length. During the simulations, this flow regime was only an intermediate regime, typically
lasting only up to half an hour. Finally, the room air flow stabilizes into forced flow, with maximum
ceiling attachment length and supply air short-circuiting via the ceiling towards the outlet. These flow
regimes strongly influence the surface convection on the different surfaces during the intensive
ventilation period. This influence was particularly noticeable at the ceiling where the evolution from
buoyant to forced flow increased the surface averaged convective heat transfer coefficient (CHTC) by
a factor 10. This illustrates the importance of taking into account this unsteady flow behaviour in
rooms with night ventilation. However, this is currently not taken into account in Building Energy
Simulation (BES) models.

(a) Buoyant flow (b) Transition flow (c) Forced flow
FIG 1. Three flow regimes, observed in numerical simulations (Leenknegt et al. 2013)
Indeed, the most advanced but pragmatic convection model, used in BES models, is the adaptive
convection algorithm, developed by Beausoleil-Morrison (2000), which defines five categories,
assuming stable convection regimes during the (de)activation of HVAC. For instance, night ventilation
would fall in the fifth category of mixed convection, as there is a system that supplies air to the room,
combined with large temperature differences. Mixed convection correlations are then applied at the
internal surfaces (Beausoleil-Morrison 2000). However, the numerical study showed that, during
buoyant flow, the ceiling displayed very low heat transfer, corresponding to natural convection with
stratified flow. To allow implementation of these insights in a refined convection model in BES, a
criterion is required to predict the flow pattern. Therefore, the flow development was coupled to the
dimensionless Richardson number. A critical value of Ri = 1 was suggested by the simulations, to
indicate the transition from buoyant to forced flow.
This test room was constructed inside the Vliet test building of the Building Physics Section of KU
Leuven, Leuven, Belgium. This is a 25.2 m long and 7.2 m wide test building, designed for the study
of wall, window and roof components under real field conditions. The room differs from typical full
scale test rooms in similar research (Artmann et al. 2010; Goethals et al. 2012), as it is exposed to real
climatic conditions through a south-west oriented hopper window. High air change rates are provided
through an extractor fan. As ventilation is done with outdoor air, the supply air temperature is not
controlled and the thermal response of the room in case of realistic boundary conditions can be
investigated. This limitation in ventilation control is compensated by more than 200 sensors, allowing
a detailed monitoring of the air, surface and mass temperatures, as well as local air velocities, surface
heat flux and pressure difference. This paper focusses on the determination of the air flow pattern
during intensive night ventilation. It is investigated whether the same flow regimes are observed
during the measurements as during the simulations. Furthermore, the prediction of the flow regime is
studied.
2. Description of test cell

Although several researchers have performed detailed measurements regarding night ventilation
(Artmann et al. 2010; Goethals et al. 2012), these measurements were limited to well controlled
climate rooms under constant ventilation rate and supply air temperature. It was deemed interesting to
investigate the thermal response of a full scale room during more realistic boundary conditions.
2.1 Location
The main orientations of the building are north-east and south-west, with the room located on the
north-western side, façade facing south-west. The most prevalent wind direction in Belgium is south-
west, so the building was oriented to have one long façade exposed to normal wind forces. However,
as the room is located at the corner of the building, the wind climate at the window will be more
unstable. The wind climate during the measurements was summarized in FIG 2b, showing the wind
velocity as a function of the wind direction during the measurements. These were measured using an
ultrasonic anemometer, placed on a mast (M) at a height of 10 m at approximately 20 m south-west of
the building. Remaining outdoor climatic conditions were measured in the weather station (W),

located on the roof of the building. The outdoor air temperature varied between 12.8 °C and 20.4 °C
during 80 % of the time.
FIG 3 shows a vertical section of the test room. The room internal dimensions are 1.80 m wide, 3.45 m
long and 2.40 m high. A window with dimensions 1.25 m wide and 1.60 m high is placed centrally in
the façade, starting at a height of 0.68 m above room floor level. The top part of the window is a
hopper window, with a maximum opening angle of 30°. Opposite the window, an outlet opening is
placed, with dimensions of 1.00 m wide and 0.10 m high and located at a height of 2.10 m and at
0.20 m distance from the ceiling. In front of the outlet, a grid is placed with horizontal lamellae with
2 cm spacing. The outlet opens into a metal plenum box, which is 40 cm deep. The plenum is
connected with an extractor fan via a 1.50 m long flexible channel with 125 mm diameter, able to
generate up to 18 h-1. An exhaust opening is made in the NW-façade of the building and provided with
a wind shield. A flexible duct leads 3.20 m upwards, connecting the fan to the exhaust opening.
(a) Orientation of Vliet building and room (b) Wind velocity as a function of the wind direction
(black) (M = wind mast, W = weather
station)
FIG 2. Overview of Vliet test building orientation and environmental parameters
FIG 3. Vertical section over the length of the test room with sensor locations (round black:
thermocouples, black triangle: tracer gas injection (Kin) and sampling (K), black diamonds:
anemometers (V)

2.2 Measurement equipment
Thermal monitoring of the room air is provided by 26 thermocouples (TC), whose locations are shown
in FIG 3 with black dots. One extra TC is placed in the extraction plenum to measure exhaust
temperature. This figure further shows the measurement planes used to organise the TC locations. Five
measurement planes are defined over the length of the room, from window to outlet: F, FM, M, MB
and B, at a distance from the façade of respectively 25, 47, 165, 285 and 315 cm. Furthermore, five
measurement planes have been defined over the height, from floor to ceiling: 1, 2, m, 3 and 4, at
heights of respectively 10, 90, 120, 150 and 230 cm. Finally, three measurement planes have been
defined over the width, from left to right (when facing the window): a, b, c, with distances from the
right wall of respectively 120, 90 and 60 cm. For each TC in FIG 3, its location on the a/b/c planes is
indicated. The temperatures are logged by two Fluke 2625a data loggers. All recorded values are
instantaneous values, recorded every 30 s.
The air flow in the room is estimated from the measurement of the local air velocity in front of the
inlet and outlet opening, as well as the pressure difference over window and extraction duct orifice.
The sensor locations are indicated in FIG 3 with black diamonds (V). Three hot-wire anemometers
from Sensor Electronics (AirDistSys 5000) are located in a vertical gradient, in the symmetry plane at
50 cm from the façade, just adjacent to the FM-plane (V). Two additional hot-wire anemometers (TSI,
type 8475) are placed just in front of the outlet grid, in the middle of the opening height and 25 cm
from the left and right edge. They are used to verify the calculated air flow rate. A 125 mm iris valve
(Air Trade Center) was placed in the exhaust channel before the extractor fan. The pressure difference
over this orifice is measured with a Halstrup differential pressure meter with range of 0 to 250 Pa. The
orifice has seven manually adjustable opening positions. As the pressure-flow curves of the orifice are
known, the flow rate can be calculated. This value was corrected based on tracer gas measurements,
taking into account two-way flow at the window opening (Leenknegt & Saelens 2014).
2.3 Measurement campaign

A dataset of 60 nights is selected, consisting of six periods, spread from June to September 2012. The
used data are taken from summer nights with night ventilation, but without mixing fans for tracer gas
measurements. From June 29th on a 40 cm deep perforated box was placed in front of the window. The
purpose of the box is to limit the influence of wind fluctuations and stabilize the flow rate. Five
different ventilation settings were used during the selected nights, resulting in roughly four ranges of
ACH: 10 h-1, 14 h-1, 16 h-1 and 17.5 h-1. These values are based on the pressure difference in the
exhaust channel, with the above mentioned correction based on the tracer gas measurements. These
ventilation rates are high, compared to values expected in realistic implementations. The reasons are
twofold: the uncertainty on the measured ACH is lower in case of higher ACH and the high flow rate
made it possible to study more dynamic flow behaviour. The room is free-floating, so no pre-
conditioning was made before each night ventilation cycle.
3. Results
3.1 Classification according to flow type
The same three flow typologies that were defined in the numerical study, namely buoyant flow (BF),
transition flow (TF) and forced flow (FF), can be used as well to classify the flow pattern observed
during the measurement campaign. However, the following differences with regard to timing and
occurrence were noted. Firstly, the forced flow pattern was only observed at near isothermal cases or
when the supply temperature is warmer than the room air temperature, placing the room in heating
modus. Secondly, the transition flow, defined by a ceiling attachment length xcl larger than zero but
smaller than room depth D, will be shown to be a near stationary condition, rather than a short

intermezzo between buoyant and forced flow. Finally, the buoyant flow pattern was clearly observed,
with supply air falling down immediately upon entry.
In order to classify the measurement data according to these three flow patterns, formal definitions are
required. These are based on FIG 4, visually illustrating the different typologies. The forced flow
pattern is active whenever the outdoor temperature is warmer than the indoor temperature at the
window opening. For the locations and coding of the TCs, the reader is referred to FIG 3, with TAFb4
located in the F-plane just adjacent to the window opening, in the symmetry plane (plane b) and at a
height of 230 cm (plane 4). In case of forced flow, the incoming warm air will initially attach to the
ceiling and short-circuit to the outlet, also known as the Coandă effect. The ceiling is then actually in
heating mode during this period. Depending on the outdoor temperature, this may continue throughout
the night or cease due to a decrease of the outdoor temperature. In this dataset, the former does not
occur, but the latter is seen during 18 nights.
The buoyant flow pattern, indicated by a higher temperature at the window opening edge (AF-b4) than
outside, is defined for nights where the incoming air flows downwards on either side of the hopper
window. In this case, the outdoor air temperature is often more than 1 K lower than the temperature in
AF-b4. The transition flow is defined for nights during which the incoming air flows over the
horizontal edge of the hopper window and penetrates at least shortly into the room, which causes a
limited Coandă effect at the ceiling. In this case, the air temperature in AF-b4 closely follows the
outdoor air temperature. Many of these cases display a variable penetration depth throughout the
night. Of course, the flow pattern is not stable, so some nights display a shift from forced flow to
transition or buoyant flow. On FIG 4a, three typical air temperature profiles are shown, comparing the
temperature in AF-b4 with the outdoor air temperature (Text). In black, one night is shown which
displays all three flow types. In dark and light grey, a full night with transition and buoyant flow are
shown respectively. To sort the data into flow patterns, a number of criteria are defined, based on the
characteristics of the supply air, and summarized in TABLE 1, with Te the outdoor air temperature.
(a) Air temperatures for different flow types (b) Schematic visualisation of flow types
FIG 4. Illustration of different flow types
TABLE 1. Criteria to sort measurement data per night according to flow development
Flow type Criteria Number of nights
forced (FF) Te > TAFb4 0
transition (TF) Te < TAFb4 & Te ≈ TAFb4 18
buoyant (BF) Te < TAFb4 & Te << TAFb4 20
FF → TF Te > TAFb4 → Te ≈ TAFb4 5
FF (→ TF) → BF Te > TAFb4 →Te << TAFb4 3
TF → BF Te ≈ TAFb4 → Te << TAFb4 14

The forced flow type only displays Coandă effect when the room is not in cooling mode, and night
ventilation should be avoided. Note that a realistic case with night ventilation should have a control
system to limit ventilation, in case the indoor-outdoor temperature difference is too low or even
negative. In combination with the low supply air velocity, typical for this configuration, it is expected
that the buoyant flow pattern will be prevalent most of the time, with transition flow occurring only
rarely. Therefore, the transition and buoyant flow regimes are most relevant. However, the evolution of
forced to transition or buoyant flow can provide useful information regarding the timing of the flow
changes, which is required to improve the convection modelling in BES. Literature study showed that
attempts have been made to use the Richardson or Archimedes number, both of which are defined as
Gr/Re² (Leenknegt & Saelens 2013), with Gr the Grashof number and Re the Reynolds number.
Therefore, the Richardson number will be used to predict flow pattern changes, as it was already used
for the numerical study. The definition of RiH is given by equation (1).
, −
= (1)
⁄ ℎ
Where g gravitational acceleration (m/s²)
β thermal expansion coefficient of air (K-1)
H room height (m)
Tr,c central room air temperature (°C)
Te outdoor air temperature (°C)
volumetric air flow rate (m³/h)
b, h opening width and height of supply window in façade (m)
3.2 Global evaluation of flow type as a function of Ri

A total of 39 nights were found with fluctuations. Out of this dataset, 12 nights showed a drastic
change from forced to buoyant flow. They start with the outdoor air warmer than or nearly equal to the
indoor air, which results in a forced flow pattern at the start of the night ventilation with Coandă effect
over the full length of the ceiling. As the outdoor air cools down, the attachment length was seen to
decrease rapidly. One such night is illustrated in FIG 5. On the left hand side, the local air
temperatures are shown and the changes are indicated. The right hand side graph gives the
corresponding Richardson number as a function of the supply air velocity. Two reference temperatures
are compared: either the central air temperature (Tcentral) or the temperature at the window opening
(AF-b4). It is clear that the central room air temperature is a more relevant reference value, even
though Ri also indicates local flow variations. As soon as the Richardson number reaches zero, i.e. the
outdoor temperature is no longer higher than the central room temperature, the air starts to detach from
the ceiling at the outlet, indicated by a diverging air temperature in point AB-b4. About two hours
later, the temperature profile at the centre of the ceiling (AM-b4) also shifts towards stratified flow,
and one more hour later, the air detaches next to the window (AFM-4) as well. This full change in
flow pattern takes about three hours.
A similar analysis is performed by hand for these 39 nights with flow pattern changes for which the
Richardson number at the time of the change is logged. Five flow pattern changes are defined here,
presented schematically in FIG 6. They are categorized by comparing the local air temperatures in
following four locations: AF-b4, AFM-4, AM-b4 and AB-b4. Flow pattern changes A, B and C
represent forced flow or transition flow, with air flowing over the horizontal edge of the window and
at least partially continuing along the ceiling. A indicates that the Coandă effect at the ceiling
decreases to at least the middle of the ceiling (AM-b4). B and C indicate a further decrease,
respectively past AFM-4 and AF-b4. D and E represent buoyant flow, with air predominantly falling
down on either side of the hopper window. D then indicates the change from transition flow to
buoyant flow, but with turbulences at the window opening, resulting in very similar temperatures in
AF-b4 and AFM-4. Finally, E marks that AFM-4 lies within the stratified zone at the ceiling. This

analysis results in a range of values for Ri corresponding to each of the flow pattern changes, which is
shown in FIG 7. Note that the A/B/C ranges show some overlap, indicating for example that on some
nights, the full transition from forced flow to transition flow with minimal ceiling attachment length
occurs between Ri-values of 3-5 and on other nights between 0 and 2. More measurement data are
needed to deduce a more definite critical value.
(a) Change in flow pattern (b) Richardson number

FIG 5. Indication of Richardson value during flow changes
FIG 6. Schematic presentation of flow pattern changes
FIG 7. Range of Richardson values for which flow pattern changes occur (dataset: 39 nights)

4. Conclusions
A full scale test room was constructed for the study of night ventilation under realistic conditions. Air
was supplied through a south-west oriented hopper window, supported by an extractor fan, ensuring
high flow rates. Previous work on the transient surface convection during night ventilation through
numerical simulations showed the importance of the unsteady flow development. Therefore, the main
topic of this paper is to analyse the flow patterns in a dataset of 60 summer nights with ventilation
rates ranging from 10 to 17.5 h-1. Three flow regimes were identified, which were also observed
during the numerical simulations, i.e. buoyant, transition and forced flow. Forced flow displayed a
short-circuit flow between supply opening and outlet (Coandă at ceiling), transition flow was seen in
case of limited Coandă effect, and buoyancy dominated flow was classified when the supply air falls
down upon entry. The CFD simulations illustrated that the flow pattern strongly influences the surface
convection. Therefore, the aim of this research is to develop criteria to predict the dynamic flow
behaviour.
The level of buoyancy of the supply air was expressed with the dimensionless Richardson number.
This number was defined using the room height, the supply air velocity and the indoor-outdoor
temperature difference. Based on four local air temperatures, five flow pattern changes were defined.
The first one, corresponding to the shift from forced to transition flow, was situated for a Richardson
number between 0 and 2. Various stages of transition flow could be seen for Ri between 0 and 9.
Buoyant flow can be assured for Ri higher than 19. Further measurements are required to refine the
determination of the flow changes.
5. Acknowledgements
This research was funded by the Research Foundation Flanders (FWO Vlaanderen). Their financial
contribution is greatly appreciated. Furthermore, Patricia Elsen was a great help in conducting the
measurements.
References
Artmann, N. et al., 2010. Experimental investigation of heat transfer during night-time ventilation.
Energy and Buildings, 42(3), pp.366–374.
Beausoleil-Morrison, I., 2000. The adaptive coupling of heat and air flow modelling within dynamic
whole-building simulation. University of Strathclyde, Glasgow, UK.
Goethals, K. et al., 2012. Experimental investigation of the impact of room/system design on mixed
convection heat transfer. Energy and Buildings, 49, pp.542–551.
Leenknegt, S. et al., 2013. Numerical study of convection during night cooling and the implications
for convection modeling in Building Energy Simulation models. Energy and Buildings, 64,
pp.41–52.
Leenknegt, S. & Saelens, D., 2014. Determination of ventilation rate in a full scale night ventilation
experiment through tracer gas measurements. In submitted to Nordic 2014.
Leenknegt, S. & Saelens, D., 2013. Use of dimensionless numbers to classify room air flow and
convection regimes: part 1 - literature review. Building and Environment, submitted,.

Full paper no: 3
Determination of the ventilation rate in a full scale night

ventilation experiment through tracer gas measurements
Sarah Leenknegt, PhD1
Dirk Saelens, Professor1
1
Building Physics Section, Department of Civil Engineering, KU Leuven, Leuven, Belgium
KEYWORDS: measurement, mechanical ventilation, unsteady ventilation, tracer gas, night

ventilation, full scale test room
SUMMARY
A full scale test room was constructed for the analysis of the energy performance of a room with night
ventilation with air supply through a hopper window, supported by an extraction fan.. Contrary to
typical climate rooms, the room was exposed to realistic climatic conditions. The determination of the
air change rate is paramount in a night ventilation system. This paper compares two methods for the
determination of the air change rate: (1) the pressure difference over the orifice in the exhaust
channel and (2) tracer gas measurements using the averaging method on the tracer gas mass balance.
The latter method allowed accounting for the back flow through the hopper window, which was not
captured via the orifice method. A large underestimation was seen in case of the pressure based ACH,
as this method does not take into account two-way flow via the large supply opening. Therefore, a
correlation was developed to correct the pressure based flow rate, to be able to determine the flow
rate in future measurement campaigns, which cannot be combined with tracer gas experiments.
1. Introduction
Night ventilation is a promising technique, able to reduce the cooling load through use of cold outdoor
air at high air change rates to cool down the internal thermal mass of a building in moderate climates
with sufficiently cold nights. To investigate the energy performance of night ventilation, a full scale
test room was constructed inside the Vliet test building of the Building Physics Section of KU Leuven,
Leuven, Belgium. The room differs from typical full scale test rooms in similar research (Artmann et
al. 2010; Goethals et al. 2012), as it is exposed to real climatic conditions through a south-west
oriented hopper window. High air change rates are established through an extractor fan. As ventilation
is done with outdoor air, the supply air temperature is not controlled and the thermal response of the
room in case of realistic boundary conditions can be investigated. This limitation in ventilation control
is compensated by more than 200 sensors, allowing a detailed monitoring of the air, surface and mass
temperatures, as well as local air velocities, surface heat flux and pressure difference. One of the key
parameters in the analysis of the energy performance is the ventilation rate. Therefore, it is crucial to
have a reasonable estimate of the air change rate during the night ventilation experiments. This is
complicated by the influence of wind on the large supply opening. Therefore, the objective of this
paper is to determine the air change rate in this mechanically ventilated case.
One of the commonly used methods to asses ventilation rate is the use of tracer gas, which is also the
main method discussed here. Alternatives found in literature are hot wire anemometry or pressure
difference measurements either by wind on the façade or over an orifice in a extraction/supply duct
(Van Buggenhout et al. 2009; Shen et al. 2012). Furthermore, an orifice was placed in the exhaust
channel to allow determination of the flow rate through the orifice using the measured pressure
difference. This paper will discuss the determination of the ventilation rate through tracer gas

measurements, and compare these results with the calculated ACH based on the pressure difference
over the orifice in the exhaust channel. As the window is exposed to the influence of wind, unsteady
air change rates are observed, combined with two-directional flow, even at high fan power. This latter
behaviour will not be captured by the orifice method. In the following, first the test room is described.
This is followed by a discussion of the results, starting with a short theoretical introduction on tracer
gas measurements and continuing with a comparison of both methods.
2. Description of test cell and measurement methodology

Although several researchers have performed detailed measurements regarding night ventilation
(Artmann et al. 2010; Goethals et al. 2012), these measurements were limited to well controlled
climate rooms under constant ventilation rate and supply air temperature. It was deemed interesting to
investigate the thermal response of a full scale room during more realistic boundary conditions.
2.1 Location
The Vliet test building is a 25.2 m long and 7.2 m wide building, designed for the study of wall,
window and roof components under real field conditions. The main orientations of the building are
north-east and south-west, with the room located on the north-western side, façade facing south-west.
The most prevalent wind direction in Belgium is south-west, so the building was oriented to have one
long façade exposed to normal wind forces. However, as the room is located at the corner of the
building, the wind climate at the window will be more unstable. The wind climate during the
measurements was summarized in FIG 1b and c, showing a box plot of respectively the wind direction
and wind velocity. These were measured using an ultrasonic anemometer, placed on a mast (M) at a
height of 10 m at approximately 20 m south-west of the building. The box plot is based on the 10th
and 90th percentile, with conditions occurring only 2 % of the time not considered in the calculation
of the spread. Remaining outdoor climatic conditions were measured in the weather station (W),
located on the roof of the building. FIG 2 shows a vertical section of the test room. The room internal
dimensions are 1.80 m wide, 3.45 m long and 2.40 m high. A window with dimensions 1.25 m wide
and 1.60 m high is placed centrally in the façade, starting at a height of 0.68 m above room floor level.
The top part of the window is a motorized hopper window, with a maximum opening angle of 30°.
Opposite the window, an outlet opening is placed, with dimensions of 1.00 m wide and 0.10 m high
and located at a height of 2.10 m and at 0.20 m distance from the ceiling. In front of the outlet, a grid
is placed with horizontal lamellae with 2 cm spacing. The outlet opens into a metal plenum box, which
is 40 cm deep. The plenum is connected with an extractor fan via a 1.50 m long flexible channel with
125 mm diameter. A maximum air change rate of up to 18 h-1 could be realized. An exhaust opening is
made in the NW-façade of the building, provided with wind shield. A flexible duct leads 3.20 m
upwards, connecting fan to exhaust.
(a) Orientation of Vliet building (b) Spread on wind direction (c) Spread on wind velocity
and room (black) (M = wind mast, during analysed nights at 10 m high during
W = weather station) analysed nights
FIG 1. Overview of Vliet test building orientation and environmental parameters

FIG 2. Vertical section over the length of the test room with sensor locations (black circles:
thermocouples, black triangles: tracer gas dosing (Kin) and sampling (K) points, black diamonds:
anemometers (V))
2.2 Measurement equipment

The air flow in the room is monitored through measurement of the local air velocity in front of the
inlet and outlet opening. Furthermore, the air flow rate is monitored through the pressure difference
over window and exhaust channel orifice. The sensor locations are indicated in FIG 2.Three hot-wire
anemometers from Sensor Electronics (AirDistSys 5000) are located in a vertical gradient, in the
symmetry plane at 50 cm from the façade, just adjacent to the FM-plane (V). Two additional hot-wire
anemometers (TSI, type 8475) are placed just in front of the outlet grid, in the middle of the opening
height and 25 cm from the left and right edge. They are used to verify the calculated air flow rate. The
pressure difference over the window is monitored through a Halstrup differential pressure meter with
range 0 to 25 Pa. Additionally, a 125 mm iris orifice (Air Trade Center) was placed in the exhaust
channel before the extractor fan. The pressure difference over this orifice is measured with a Halstrup
differential pressure meter with range of 0 to 250 Pa. The orifice has seven manually adjustable
opening positions, going from maximum opening at setting 1 to minimum opening at 7. As the
pressure-flow characteristics of the orifice are known, the flow rate can be calculated.
The tracer gas equipment consists of a dosing- and sampling unit (type 1303) from Bruel & Kjær
Innova, containing six dosing and six sampling channels. The air samples are analysed in a multi-gas
monitor (type 1302), which measures the gas concentration using photo-acoustic gas detection. The
gas used for the tracer gas experiments is sulphur hexafluoride or SF6. The health related threshold for
SF6 is 1000 ppm and it is an artificial gas. However, it has the downside that it is about five times
heavier than air, with a density of 6.2 kg/m³. This results in a risk of stratification, where the gas sinks
to the floor, especially for low air velocity situations. Furthermore, it is a strong greenhouse gas,
whose use should be limited. The location of the six sampling points (K) and two injection points
(Kin) is illustrated in FIG 2. Two small mixing fans are located on the floor adjacent to the injection
points, directed away from window and outlet opening.

2.3 Tracer gas techniques
The standard tracer gas analysis techniques are well known and will not be further discussed here,
instead the interested reader is referred to Sherman (1990). All methods are based on the tracer gas
mass balance, also called the continuity equation, given in equation (1).
V = − (1)
Where V zone volume (m³)
C tracer gas concentration in the zone (10-6 mol/mol)
t time (h)
QT injected tracer gas volumetric flow rate (m³/h)
Q volumetric flow rate (m³/h)
Sometimes however, the standard techniques are not usable, for example if the air change rate is
unstable and the controller is too slow to follow the changes. When no assumptions are made
regarding constant injections and concentrations apart from the continuity equation itself, Sherman
describes three different approaches for determination of the ventilation directly from the
concentration data: (1) regression, (2) integral and (3) averaging techniques, of which the latter
technique is used here (Sherman 1990). The continuity equation cannot be used to determine
instantaneous ventilation, but it can be averaged over a measurement period T to get the average
ventilation, as given in equation (2).
= − ln (2)
Where average value over period T

T averaging period (h)
final/initial: final and initial tracer gas concentration during averaging period
3. Results
The following results will start with a description of the ventilation rate based on the averaging
method applied to the measured tracer gas concentrations. The influence of wind climate is shortly
discussed, resulting in unstable flow rates. These results can then be compared to the air change rate
based on the pressure difference over the orifice in the exhaust channel. The problems concerning this
latter method are identified and addressed.
3.1 Tracer gas based determination of ACH

Over the course of three months, starting in May 2012, 45 tracer gas experiments have been
conducted. Each measurement had a duration of minimum two and up to eight hours and was
conducted during night. Two different tracer dosing control methods were used: constant concentration
(target 20 or 50 ppm) and constant dosing (18 mg/s). FIG 3 shows the range of the unsteady air change
rate nTr30m for each measurement, based on the averaging technique by equation (2), using an
averaging period of 30 minutes. Each box plot represents one experiment. The results have been sorted
according to increasing fan power and orifice opening.
At fan power 1, only small orifice openings were tested, as this setting would mainly be used to
achieve low ACH. The minimum ACH achieved here however is still at least 5 h-1, though most ranges
are higher than 7 h-1. Remarkable is the large range of calculated ACH for each measurement, despite
the constant fan power and orifice opening. During most measurements, there is a significant
fluctuation of the wind direction (not shown here). Even though the wind velocity is relatively low
(most of the time < 3 m/s), this causes an unstable flow rate. Indeed, analysis of the air temperatures in
the room when the mixing fans were not active, showed a highly unsymmetrical flow, illustrating the

strong influence of the wind. Therefore, a perforated box was installed in front of the window to
equalize the wind pressure and reduce the effect of wind direction fluctuations. The box was placed
halfway through the measurement period at the end of June 2012. It is 40 cm deep, constructed with 4
mm thin plywood, and about 13 % of the area is perforated with 8 cm diameter holes. In FIG 3, the
measurements conducted with the box in front of the window are coloured grey-black, instead of grey.
FIG 3. ACH based on averaging technique from equation (2) over 30 minute periods as a function of
orifice opening and fan setting (wind box present: black bars, no wind box: grey bars)
For the measurements with fan power 1, the effect of the wind box is clearly visible, with much more
stable flow rates. At fan power 2, all measurements are conducted with the wind box in place. Most of
the calculated ACH display only a very limited range, indicated by the box plot. Nevertheless, the
minima and maxima, indicated by the vertical lines, are still quite far apart. The measurements with
the widest ACH-range correspond again to the moments with the most unstable wind direction.
At fan power 3, measurements were conducted with and without wind box. At this high fan power, the
correlation between wind direction and ACH range is less strong. Some measurements show a very
stable ACH, despite a large variation in wind direction. Nevertheless, the wind box again results in
much more stable air flow rates. One case at fan power 3 and orifice opening 1 is an exception, with
the air change rate ranging from 17 to nearly 23 h-1 in one night, illustrated in more detail in FIG 4a,
displaying the ACH, along with current wind direction and velocity. The tracer gas based air change
rate nTr is calculated for a 30 minute averaging period (nTr30m), which is indicated on the graphs by
horizontal bars. The wind velocities (WindVel) are rather low, i.e. equal to or lower than 2 m/s.
However, the wind direction (WindDir) is continuously nearly perpendicular to the window. This
could explain why the moderate wind still has such a high influence on the air flow rate. However, the
increase in ACH over the presented period seems not correlated with the wind climate and could not
be explained.
Another interesting case is shown in FIG 3 at fan power 1, orifice opening 5 and with wind box, but
nevertheless with a large spread. The main dataset is very stable at 8.5 to 9 h-1, though a peak up to 10
h-1 is seen. This day is shown in more detail in FIG 4b. Cross-referencing with wind climate explains
this variation. During the stable phase, the wind velocity is rather low, i.e. equal to or lower than 2
m/s. However, the local peak in ACH coincides with a sudden increase in wind velocity up to 5.5 m/s.
3.2 Pressure difference based determination of ACH

The ACH was also calculated based on the pressure difference over the orifice in the extraction duct.
The pressure differential deducer in the exhaust channel has an accuracy of ± 0.2 % on the reading
with a minimum of ± 0.15 Pa. The ventilation rate of the room is calculated based on the airflow rate

determined through the pressure-flow characteristics, supplied by the product information of the
orifice. The product information stipulates an uncertainty on these curves of ± 7 % if the minimum
straight pipe length after a bend and before the orifice is higher than the orifice diameter, which is the
case in this experimental set-up.
(a) Tracer gas based ACH on day 62 (F3V1) (b) Tracer gas based ACH on day 48 (F1V5)
FIG 4. Detail of tracer gas based ACH, correlation to climatic data
FIG 5. Comparison of pressure based ACH (nΔP) and tracer gas based ACH (nTr30m), see equation (3)
for Δnrel (filled markers = with wind box, not filled markers = without wind box)
During the analysed periods of intensive ventilation, the pressure difference is always 10 Pa or higher,
resulting in an uncertainty on the calculated ACH of approximately ± 7 %, as the uncertainty on the
measured pressure difference is very small. FIG 4 shows for both days also the pressure based ACH,
indicated by nΔP. The uncertainty is indicated by the vertical error bars, based on the uncertainty on the
pressure difference measurement, combined with the pressure-flow characteristics. Also the outdoor
temperature is indicated by Text. Two conclusions can be drawn: firstly, the pressure based air change
rates nΔP are significantly lower than the tracer gas based air change rates nTr, and secondly, they are
independent of the wind climate. They may fluctuate somewhat by the indoor-outdoor temperature
difference, although no clear correlation was found. Furthermore, the ventilation rate is driven by the
extraction fan in all measurements. The pressure based ACH for the full measurement period are
shown in box plots in FIG 5. The tracer gas based median values nTr30m of each measurement are
shown as well. To illustrate the difference between both methods, the relative difference Δnrel
according to equation (3) has been plotted on the right axis, with nTr30m based on the 30 minutes

averaging value and nΔP based on the pressure difference over the orifice. Filled markers refer to
measurements with wind box. The time-average of the full measurement was used. Often a large
difference between both methods to determine the ACH is observed. Particularly at low fan power nTr
is often 30 to 50 % higher than nΔP.
− ∆
∆ = (3)
3.3 Corrected ACH

The difference between the two calculated air change rates, based either on the pressure difference or
on the tracer gas measurements, is illustrated in FIG 6a, showing the relative difference as a function
of nΔP. The overestimation of the tracer gas based results can be explained by air leaking back through
the window to outside, which means that the incoming air could be contaminated by tracer gas,
causing an underestimation of the ACH in the tracer gas analysis so far. To check this,19 further
measurements were performed in winter showing that 1 to 9 % of the tracer gas concentration in the
room is found outside the window, depending on the strength of the extractor fan. At air changes rates
as high as 10 h-1, a tracer gas concentration of 5 % of the inside value was measured outside, which
went up to 9 % at lower ACH. Based on these measurements, an exponential function was found,
correlating the calculated air flow rates with the tracer gas percentage outside. With this correlation,
the previously calculated ventilation rates were corrected. The average difference with the initially
calculated values was only 3.4 % with a total range of 1 to 6 % of underestimation.
(a) Relative difference between tracer gas based (b) Correlation between pressure-based on tracer
and pressure based ACH gas based ACH
FIG 6. Correlation to predict actual ACH based on pressure-based ACH
Future research is aimed at predicting the air flow pattern in the room during intensive night
ventilation, as this strongly influences the local surface convection. This was shown in previous
research through numerical simulations (Leenknegt et al. 2013) . The presence of mixing fans would
disrupt the flow pattern, which is a crucial part of the research. Therefore, it is impossible to perform
tracer gas measurements during the night ventilation experiments and the ACH must be determined
based on the pressure difference over the orifice. It was shown before that this value only takes into
account the flow rate leaving through the extraction duct and ignores the additional air leaving through
the window opening. As such, the pressure based ACH is an underestimation of the actual value and a
correlation was sought and found between nTr and nΔP to obtain a corrected value nΔP,corr. This
correlations is illustrated in FIG 6b and given by equation (4). Despite the good correlation, there is
still some scatter, and the corrected value will have an uncertainty which is estimated at ± 15 %, based
on the typical difference between the measured values and the correlation.
.'(
∆ ,!" = 2.35 ∆ (4)

4. Conclusions
A full scale test room was constructed for the study of night ventilation under realistic climatic
conditions. Air was supplied through a south-west oriented hopper window, supported by an extractor
fan, ensuring high flow rates. For the analysis of the night ventilation performance, knowledge of the
ventilation rate is paramount. This paper analyses the results of two methods to determine the
ventilation rate. As a first method, an orifice was placed in the exhaust channel, calibrated to allow
calculation of the volumetric flow rate based on the opening size and pressure difference. However,
this method does not provide an evaluation of the unsteady air flow rate. Indeed, due to the influence
of wind on the large hopper window opening, unstable air ventilation rates were observed, even at
high fan power. This could however be evaluated with tracer gas measurements, the second method.
The averaging technique was applied to the continuity equation to determine a profile of the unsteady
air flow rate. Furthermore, the results of these two methods could be compared. It was shown that the
pressure based air change rate was consistently lower than that based on the tracer gas method. This
was attributed to the back flow at the large supply opening of the hopper window, caused by the wind
and resulting in two-way flow. The influence of the indoor-outdoor temperature difference was low, as
the ventilation rate was driven by the extraction fan throughout all measurements. The results were
corrected, taking into account the tracer gas contamination of the supply air. As the pressure based
ACH only takes into account the air leaving the room via the exhaust channel, this value
underestimates the actual air flow rate. However, future research is focussed on the energy
performance of this night ventilated room. A crucial part of the research is dedicated to the prediction
of the air flow pattern, as this is closely related to the local surface convection regimes. Therefore, the
activation of mixing fans is not possible and tracer gas experiments cannot be used for the
determination of the air change rate. To solve this, a correlation was developed, providing a correction
on the pressure based ACH.
5. Acknowledgements
This research was funded by the Research Foundation Flanders (FWO Vlaanderen). Their financial
contribution is greatly appreciated. Furthermore, Patricia Elsen was a great help in conducting the
tracer gas measurements.
References
Artmann, N. et al., 2010. Experimental investigation of heat transfer during night-time ventilation.
Energy and Buildings, 42(3), pp.366–374. Available at:
http://linkinghub.elsevier.com/retrieve/pii/S0378778809002412.
Van Buggenhout, S. et al., 2009. Influence of sampling positions on accuracy of tracer gas
measurements in ventilated spaces. Biosystems Engineering, 104(2), pp.216–223.
Goethals, K. et al., 2012. Experimental investigation of the impact of room/system design on mixed
convection heat transfer. Energy and Buildings, 49, pp.542–551.
Leenknegt, S. et al., 2013. Numerical study of convection during night cooling and the implications
for convection modeling in Building Energy Simulation models. Energy and Buildings.
Shen, X., Zhang, G. & Bjerg, B., 2012. Comparison of different methods for estimating ventilation
rates through wind driven ventilated buildings. Energy and Buildings, 54, pp.297–306.
Sherman, M.H., 1990. Tracer-gas techniques for measuring ventilation in a single zone. Building and
Environment, 25(4), pp.365–374.

Full paper no: 4
The impact of leakage properties onto the flow-through of single
leakages
Jens Schmidt, 1
Oliver Kornadt, Professor 2
1, 2
Technical University of Kaiserslautern, Germany
KEYWORDS: convection, leakage flow, discharge coefficient, air tightness, simulation
SUMMARY
Leakages in the building envelope may lead to increase of primary energy demand and carbon foot-
print of buildings. Additionally, particle and bacteria infiltration or exfiltration are possible. Fur-
thermore, the building shell may be damaged by hygric enveloping surface infiltration. For investiga-
tion of the impact of leakages in the building envelope, different measurements and simulations are
used. The disadvantage of the both is that the knowledge on discharge coefficient of leakages in real
buildings is limited. Therefore, the discharge coefficient is estimated or based on sharp edged, circu-
lar standard orifice. But real leakages are different, e.g. the shape is not comparable to a standard
orifice. Due to these, the flow through different leakages in perforated air tightness layers of light-
weight timber frame constructions was analysed. The applied pressure differences were real meas-
ured values of long term investigations on building shells. At these specific pressure differences, the
flow profile through leakages is different to studies of standard orifice. The volumetric flow through
leakages is significantly influenced by this difference. The results show the impact of the different
leakage parameters onto the discharge coefficient. Thus, simulations should not be based on standard
orifice investigations.
Introduction
In walls of new or refurnished buildings, it is common to integrate different insulation. Thus, a
reduction of energy losses through the building envelope is possible. However this climate separation
of interior and environment represents also a potential danger. If the water vapour diffusion resistance
of the room side materials is too small, condensation may arise in the wall construction. For
prevention, room-side diffusion inhibiting or blocking materials e.g. vapour barriers are arranged.
According to (Schulze H. 2011), awareness regarding diffusive moisture transport processes has
increased. Therefore, moisture problems caused by diffusion in constructions of new buildings are
infrequent at present time. A far bigger problem is given by the deficient air tightness of the building
envelope. According to (Biskop R. 2008), these leakages are caused by mistakes in planning and
execution. Especially during the heating season, bacteria, particle and moisture may be transported by
convection from the interior to the environment. This transport of matter may lead to deposits in the
wall structure. This might result in growth of mould and mildew or the formation of heat bridges e.g.
in case of condensation in non-capillary active components. Additionally, the convective heat
exchange leads to increased energy losses. Light-weight timber constructions are particularly at risk
concerning leakages. Damages in the air tight and protection layer are easily possible and the used
insulation materials are frequently air permeable. In addition, the convective leakage flow is promoted
by the non-obligatory air tightness of the weather protection layer. These findings led to the
development of simulation software with the aim of analysing the disadvantages of leakage flows on
buildings. The best-known algorithms are implemented in the software of Delphin and WUFI. These
algorithms are based on leakage flows caused by density differences (Häupl P. et al. 1997, Künzel
H.M. 2010). Only a few calculation models are developed for leakage flow analyses under real

pressure difference conditions of wind and buoyancy. They are frequently based on the existing
calculation approaches of Delphin and WUFI (Langmans J. 2013, de With G. et al. 2009, Zirkelbach
D. 2009, Kurnitski J. et al. 2000). In these new approaches, the contraction influence on the leakage
flow is either estimated or refers to unpublished references. Currently, standards and knowledge about
the discharge coefficient are exclusively based on standard orifice and materials which are used in
technical building equipment (ASHRAE 2009, DIN EN ISO 5167-2). However, leakages in buildings
may differ e.g. in shape, compared to standard orifice. Nevertheless, for investigation of the
equivalent leakage area with the pressure difference method, also a generalized discharge coefficient
is used (DIN EN 13829, ASTM E779-10, ASTM E1186-03). Furthermore, the friction coefficient of
pipeline materials for air and water transport is different, compared to the short channels which arise
by perforation of the room-side layers e.g. in materials of light-weight timber frame constructions.
Therefore, this work aims to analyse the impact of the leakage shape and size in air barriers onto the
flow-through. Also, the influence of deformation of the air barriers onto the discharge coefficient was
measured. Finally, a comparison of the volumetric flow through a single leak and several leakages of
the same total leakage area and texture was carried out.
Basics
The pressure difference between interior and the surrounding area is the potential of flows through
leakages in the building envelope. In naturally ventilated buildings, these are influenced by wind and
buoyancy. The change of pressure differences occurs in intervals of seconds (Schmidt J. et al. 2010b).
According to the Bernoulli equation, the leakage flow is from the higher to the lower pressure level.
In order to analyse and simulate the exfiltration flow through a room-side layer with a single leak and
adjacent permeable insulation in a light-weight timber structure, knowledge on the pressure loss
caused by the leak is required. Commonly, the cross-section of a leak in the building shell is much
smaller than the cross-section area of the room, which forces a constriction of the exfiltration flow.
This results in contraction and velocity losses. The contraction factor Ψ is affected by the ratio of
stream and leakage cross-section. In addition, the shape of leakage edge is crucial (fig. 1, 2). Ψ of e.g.
sharp edged standard orifice is 0.61 … 0.64 (Bohl W. et al. 2005). The velocity coefficient Φ of the
leakage flow is influenced by the macroscopic and microscopic geometry of the opening, the viscosity
and the surface tension (if not air) of the fluid. Φ is in case of sharp edged standard orifice 0.97 (Bohl
W. et al. 2005). The product of Ψ and Φ gives the discharge coefficient ζ. “Caused by the difficulty to
investigate Ψ and Φ experimentally separated from each other, values of these coefficients are only
scarcely documented in the literature.” (Bohl W. et al. 2005) Theoretical considerations were already
carried out by Torricelli and Borda in 17th and 18th century (Borda J. C. 1769).
FIG 1. Circular opening FIG 2. Sharpe edged opening

By comparison of pressure and impulse force, they determined a theoretical contraction coefficient of
0.5 (Bohl W. et al. 2005). Later, investigations of the discharge coefficient of standard orifice by E.
Buckingham (1931) and J. Unger (1979) have resulted in the values given in (DIN EN ISO 5167-2).
The discharge coefficient ζ of the air flow through a horizontal leak may be derived by the Bernoulli
and continuity equation according eq. (1):
2 ⋅ ∆p
AL ⋅ ζ ⋅ = Vreal (1)
r ⋅ (1 − N ² )

In eq. (1) AL represents the leakage area, ∆p the pressure difference and r the density of air. The ratio
between room and leakage cross-section is given by N. Corresponding to eq. (2), it is possible to
determine the discharge coefficient by experimental investigations. In this context represents Vreal the
measured and Vtheo the undisturbed, theoretical volumetric flow.
Vreal
⋅ (1 − N ² ) = ζ (2)
V
theo
According to (DIN EN ISO 5167-2), the parameter N is powered by 4. An explanation of the

exponent is missing in DIN EN ISO 5167. Due to the fact that N approaches to 0 the root expression
is negligible. In contrary to the Darcy-Weisbach formulation in eq. (3), according to the (ASHRAE
2009) it is possible to investigate ζ without knowledge about channel friction, profile and velocity of
the leakage flow through perforations of layers. Furthermore, the influences of inlet and outlet losses
for integration in eq. (3) are scarcely documented (Langmans J. 2013). The impacts of these losses are
integrated in ζ.
 f ⋅L   ρ ⋅u ² 
 + ΣC  ⋅   = ∆p (3)
 dh   2 
In eq. (3) ΣC is representing the sum of local losses. f is the Darcy-Weisbach friction factor. In case of
fully developed laminar flow it is equal 96/Re (Langmans J. 2013). The other parameters in eq. (3) are
L the length, dh the hydraulic diameter and the mean square velocity u ² in the channel. Re denotes the
Reynolds number, which defines the stream profile (laminar, Re ≤ 2300). DIN EN ISO 5167 provides
additional opportunities for estimation of ζ under conditions of 0.05 ≤ N ≤ 0.64 and 5 · 104 < Re <
107. A tabulation of some ζ-values of standard orifice is implemented in this standard. Furthermore
the Reader-Harris/Gallagher equation may be used for estimation under knowledge of Re at N ≥ 0.1
(Böswirth L. 2007). It is only applicable to pipes with circular standard orifice. Caused by the
approach of N  0 and the laminarity of flows through common leakages in buildings, (Langmans J.
2013) values that these models are unsuitable for forecasts of pressure losses. Hence, the discharge
coefficient was analysed using eq. (2).
Measurements
The investigations were carried out on vapour barrier foils of about 1 x 1 m in the “Measurement
Setup for research on convective moisture transfer (MCMT)” (Schmidt J. et al. 2010a). Every
investigated leakage was sharp edged by use of special tools. The climate conditions between the over
and under pressure chamber were equal at all measurements. The perforated vapour barrier was the
separation of these two connected chambers. For the early investigations, the foil was stiffened
without influencing the flow. The volumetric flows through the leaks were measured parallel by use
of Tracer-gas-system “TGS” and calorimetric flow meter “MSD”. According to (Böhle M. et al. 2002,
Leick Ph. 2008), the flow-through was under open outflow wherein the air flow after overcoming the
obstacle could open conically. First, the pressure difference between the inflow and outflow-side were
calculated by weather data of a German test reference year (Christoffer J. et al. 2004). Later, these
values were arranged based on long term measurements at real building facades. According to the
design of the MCMT (Schmidt J. et al. 2012), the maximum total leak size was 10 cm ², at a
maximum pressure difference of 350 Pa. Under real measured pressure differences (Schmidt J. 2013)
the leakage flow is usually laminar. Due to poor execution of real buildings, cavities between the
insulation and airtightness layer may occur. In order to analyse their influence on the leakage flow of
perforated vapour barrier foils, double chamber measurements were carried out under free arching of

foils like before. Additionally, the deformation of 4-sided clamped, perforated vapour barrier foils was
investigated in single chamber tests with onesided over pressure. For this, a 2D- traversal with
rangefinder was installed behind the leakage at the under-pressure side.
Results
The first part of the study related to the analysis of the flow of differently shaped leaks. Therefore,
leakages of the same cross-section and shape of edge were investigated in stiffened PE-foil.
FIG 3. Volumetric flow through different shaped FIG 4. Comparison of volumetric flow through
leaks, AL = 0.2 cm² different shaped leaks to the calculated
leakage flow, AL = 0.2 cm²
Figure (3) shows the leakage flow through 0.2 cm² leakages of different shape at different pressure
differences. In this context, the flow may be approximated by a leak according to (DIN EN 13829,
ASTM E779-10, ASTM E1186-03) as a function of the applied pressure difference by an exponential
function. As leakages in figure (3, 4) were used: an edged rectangular leak (0.1 x 2 cm), a squared
leak (0.45 x 0.45 cm) and a circular leak (d = 0.5 cm). The measurement error between the carried out
parallel measurements was about 3%. The association for air tightness in the building trade “Flib”
(2008) suggests using a ζ-value of 0.61 to analyse the equivalent leakage area of a building.
Calculated flows, using this ζ-value, were compared with the measurement results, see figure (3, 4).
According to figure (4), the calculated flows had an average difference of about 20 % compared to the
measured flows. When using the hydraulic diameter to calculate the flow rate, the error would
increase further. For e.g.the rectangular leak, it would yield dh = 0.19 cm. Thus, a much smaller
calculation cross-section would be given.
The discharge coefficient is calculable by use of eq. (2) and neglecting N. According to figure (5), a
linear relationship exists between theoretical and measured volumetric flow. The inclination of the
linear fit defines the correlation between the two flow rates. Thus, the inclination resembles a
resistance coefficient respectively the discharge coefficient. The properties of the ζ-value were
investigated on sharp edged, circular leaks of different cross-section. As shown in figure (6), the
discharge coefficient ζ is hyperbolic, approaching a limit value. This relationship can be
approximated by an exponential function. The result for circular holes was similar (figure 7). Both,
figure 6 and 7 show that the discharge coefficient approaches the theoretical, calculated contraction
coefficient according to Torricelli and Borda (1769) with increasing leakage size. This suggests that
the influence of the velocity on Φ decreases when the leakage area increases.

FIG 5. Linear relationship of theoretical and FIG 6. Change of the discharge coefficient
measured volumetric flow AL = 0.5 cm², in comparison of the change of the
sharp edged, circular leak leakage area of circular, sharp edged
leaks
FIG 7. Change of the discharge coefficient in comparison of the change of the leakage area of
annular, circular, sharp edged leaks
In the next step of this investigation, the deformation behaviour of a 1 m² large, 0.2 mm thick,
perforated PE-foil was examined. In figure (8) is shown that the PE-foil is arching like a nozzle due to
the one-sided pressure load. The maximum of arching was 7.4 ± 0.1 cm at a leak size of 0.2 cm² and
350 Pa of pressure difference (figure 9).
FIG 8. Arching of perforated PE-foil at ∆p = 2Pa, FIG 9. Maximum of arching of perforated PE-
AL = 0.2 cm², circular, sharp edged leak foil at different ∆p, AL = 0.2 cm²,
circular, sharp edged leak

The values for the volumetric flow through the holes in an arched and stiffened PE-foil were
compared with the calculated flows for ζ = 0.61 according to (Flib 2008). As can be seen in figure
(10), the formation of a nozzle in the air tightness layer may lead to a change of the leakage flow at
different pressure differences.
FIG 10. Impact of foil arching onto the leakage FIG 11. Comparison of leakage flows through a
flow compared to calculated flow with single leak with the total cross-section
ζ =0.61, AL = 0.2 cm², circular, sharp of 5 small leaks of the same properties
edged hole in deformable PE-foil
In figure (10) is shown that the smallest variation between the flows through arched and stiffened PE-
foil occurs for circular leaks. The difference of leakage flow between deformed and undeformed foil
increases with increasing pressure difference (figure 10) for leakages of non-circular shape. The ratio
between calculated and measured volumetric flows was never 1 for all measurements (figure 10).
These results show that only one value for ζ is not generally valid for every single leakage.
Further investigations should clarify if the ζ-value of the volumetric flow of one leak is equal to the ζ-
value of several leakages of the same total cross-section and properties. For the experiment, a leakage
area of 1 cm² was chosen. The reference was the flow through a single, sharp edged, centred leakage
in deformable PE-foil. The total flow at different pressure differences for the single leak was
compared to 5 leakages of AL = 0.2 cm² (total 1 cm²) with different positions.
In figure (11), circle means a uniform distribution of the small holes with a 30 cm radius around one
central leak. The definition star describes a hole distribution with one centre leak and 4 leaks in a
distance of 5cm on the ordinate and abscissa. Coincidence “coinc.” denotes a freely chosen, non-
definable distribution of the small leakages. Figure (11) shows, that the total leakage flows are similar
for all 5-leak-tests. The relationship of the flow through the single leak and the 5 holes of the same
total cross-section is linear. However, the inclination of the flow fit between the 5 small to the one
large leak shows a deviation of 11 to 13 %. In order to be able to define only one discharge coefficient
for existing leakages in the building envelope, e.g. in Zone models or to determine the equivalent
leakage rate, accurate knowledge about the existing leaks is essential. Even the size of the leakage can
be critical.
Conclusions
The investigations of leakage flows through PE-foil show that the volumetric flow through a leak in
lightweight timber structures and the adjacent layers is significantly influenced by the discharge
coefficient ζ of the room-side leak (exfiltration). It is shown, that the assumption of a discharge
coefficient of standard orifices to describe a leakage flow under real pressure differences (wind +
buoyancy) may contain errors. The ζ-values of building leakages deviate to ζ-values of standard

orifice in pipes due to the ratio of usual leakage area and room cross-section that approaches zero.
Furthermore, leakage flows through building structures under real pressure differences are mainly
laminar, in deviation to studies of standard orifice. The investigations show that the discharge
coefficient of building leaks is significantly influenced by size, shape and edge properties of a leak.
Additionally, leakage edge deformations may influence the ζ-value for PE-foil. Furthermore, caused
by poor workmanship, cavities between airtightness and insulation layer may affect the flow through
the leakages. In case of PE-foil as air tightness layer, arching of the foil is possible. Depending on the
shape of leakage and the applied pressure difference, this may affect the leakage flow in various ways.
Only for circular holes were similar volumetric flows for stiffed and deformable foils detected. The
theoretical value of ζ = 0.61 for the discharge coefficient (based on shape-independent, stable
standard orifices at N ≥ 0.1) could not be measured for stiffened or deformable foils with single leaks.
Furthermore, the discharge coefficient of a single leakage and several leakages with the same
properties and the same total cross-section did not match. The obtained results demonstrate that the
assumption of only one ζ-value for the whole variety of leakages in the building envelope, e.g. in zone
models and for the determination of the equivalent leakage area in differential pressure tests, may be
incorrect. The findings of this study can help to optimize simulation models of convective transports
through leaks in the building envelope of lightweight timber constructions. However, the study refers
only to the leakage flow through perforated air tightness foils. Therefore, the presented measurements
should be continued for multi-layered, room boundary layers. It can be assumed that non-negligible
factors will be the channel roughness, the material thickness and the propagation behaviour of the air
stream in the leak channel.
References
ASTM E1186-03, 2009, Standard Practices for Air Leakage Site Detection in Building Envelopes and
Air Barrier Systems.
ASTM E779-10, 2010, Standard Test Method for Determining Air Leakage Rate by Fan
Pressurization.
ASHRAE Standards Committee, American National Standards Institute, 2009, ASHRAE Handbook -
Fundamentals. Atlanta: ASHRAE Customer Service.
Biskop R., 2008, Bauphysikalische Probleme - Fallbeispiele aus der Praxis des mo-dernen Bauens, 1.
Workshop InnoProfile / nuBau, CD, Bauhaus-Universität Weimar (editor), Germany.
Bohl W. & Elmendorf W., 2005, Technische Strömungslehre, Stoffeigenschaften von Flüssigkeiten
und Gasen, Hydrostatik, Aerostatik, inkompressible Strömungen, kompressible Strömungen, Strö-
mungsmesstechnik, 13.th ed., Vogel, Würzburg, Germany.
Böhle M., Etling D. et al., 2002, Prandtl - Führer durch die Strömungslehre, Grundlagen und Phäno-
mene, 11th ed., Vieweg, Braunschweig, Germany.
Borda J. C., 1769, Mémoire sur l'Ecoulement des. Fluides par les Orifices des Vases, Mémoires de
l'Académie, Royale des Sciences, Paris, France.
Böswirth L., 2007, Technische Strömungslehre: Lehr- und Übungsbuch, 7th ed., Vieweg, Braun-
schweig, Germany.
Buckingham E., 1931, Beitrag zur Berechnung der Kontraktionszahl, Mitteilung des Bureau of
Standards, Washington, Rundschau, Forschungsband, vol. 2., 5th ed., Fritz W. (editor), Berlin,
Germany, pp. 185–192.
Christoffer J., Deutschländer Th. et al., 2004, Testreferenzjahre von Deutschland für mittlere und
extreme Witterungsverhältnisse TRY, Selbstverl. des Deutschen Wetter-dienstes, Offenbach a.
Main, Germany.

de With G., Cherry N. et al., 2009, Journal of Building Physics, Thermal Benefits of tiled roofs with
above-sheathing ventilation, vol. 33, 2nd ed., pp. 171–194.
DIN EN 13829, 02/2001, Bestimmung der Luftdurchlässigkeit von Gebäuden, Differenzdruck-
verfahren.
DIN EN ISO 5167-2, 01/ 2004, Durchflussmessungen von Fluiden mit Drosselgeräten in voll
durchströmten Leitungen mit Kreisquerschnitt-Teil 2: Blenden.
„FLIB“ Fachverband Luftdichtheit im Bauwesen e.V. (editor), 2008, Gebäude-Luftdichtheit, vol. 1,
Kassel, Germany.
Häupl P., Grunewald J. et al., 1997, International Journal of Heat and Mass Transfer, Coupled heat air
and moisture transfer in building structures, vol. 40, 7th ed., pp. 1633–1642.
Künzel H.M., 2010, Holzbau die neue quadriga, Trocknungsreserven schaffen, Ein-fluss des
Feuchteeintrags aus Dampfkonvektion, 1st ed., pp. 28–32.
Kurnitski J. & Vuolle M., 2000, Simultaneous calculation of heat, moisture, and air transport in a
modular simulation environment, Proceedings of the Estonia Academy of sciences / Eesti Teaduste
Akadeemia Toimetised, Engineering / Tehnikateadused, vol. 1, Aben H. & Kurnitski V. (editors),
The Academy, Tallinn, Estonia, pp. 25–47.
Langmans J., 2013, Feasibility of exterior air barriers in timber frame constructions, PhD Thesis, KU,
Building Physics Section, Leuven, Belgium.
Leick Ph., 2008, Quantitative Untersuchungen zum Einfuss von Düsengeometrie und Gasdichte auf
den Primärzerfallsbereich von Dieselsprays, Dissertation, Technischen Universität Darmstadt,
Fachbereich Maschinenbau, Germany.
Schmidt J., 2013, Konvektiver Feuchtetransport, Schlussbericht zum InnoProfile
Forschungsvorhaben: Methoden und Baustoffe zur nutzerorientierten Bausanierung, Vogel A.,
Völker C. et al. (editors), Universitätsverlag, Weimar, Germany, pp. 39–41.
Schmidt J. & Kornadt O., 2010a, Verfahren und Untersuchungsvorrichtung zur Untersuchung eines
konvektiven Feuchtetransports in einem Bauwerkelement, Bauhaus-Universität Weimar
(applicant), 09.07.2010, publication number: DE102010031141.3.
Schmidt J. & Kornadt O., 2010b, Konvektiver Feuchtetransport durch Bauteilleckagen in
Bestandsgebäuden, Nutzerorientierte Bausanierung, vol. 1., 1st ed., Völker C. & Kornadt O.
(editors), Universitätsverlag, Weimar, Germany, pp. 109–116.
Schmidt J. & Kornadt O., 2012, Der Bausachverständige, Konvektiver Feuchtetrans-port durch die
Gebäudehülle – neuartiges Messsystem zur Klärung relevanter Fra-gen, 4th ed., pp. 29–33.
Schulze H., 2011, Feuchtebedingte Schäden an Wänden, Decken und Dächern in Holzbauart,
Schadenfreies Bauen, vol. 5, 2nd ed., Fraunhofer IRB Verlag, Stuttgart, Germany.
Unger J., 1979, Forschung im Ingenieurwesen, Strömungen in zylindrischen Kanälen mit Versperrung
bei hoher Reynolds-Zahl, vol. 45, 3rd ed., pp. 69–80.
Zirkelbach D., 2009, Dampfkonvektion wird berechenbar-Instationäres Modell zur Berücksichtigung
von konvektivem Feuchteeintrag bei der Simulation von Leichtbaukonstruktionen, 4th
International Symposium on Building and Ductwork Air tightness, Berlin, Germany, pp. 1–8.
www.hoki.ibp.fhg.de, last call 19.01.2012.

Full paper no: 5
Effect of different façade systems on the cooling capacity of a
cooled ceiling – Tracer gas measurement of the air flow
Katharina Eder 1
Hannes Konder 1
Thomas Bednar 1
1
Vienna University of Technology, Institute for Building Construction and Technology, Department
for Building Physics and Building Acoustic, Austria
KEYWORDS: Air flow, tracer gas measurement, comfort, façade systems, cooling ceiling, cooling
capacity
SUMMARY:
The aim of the work is to clarify how the façade system and the cooling ceiling impacts the air
movement within a room. Therefore tracer gas measurements with different inject locations and
different room set-ups were done.
One part of the results show the impact of different façade systems – a single skin façade and a single
storey double skin façade – on the air movement. The measurements show an increase of
concentration within the ceiling zone which is attributed to a higher air flow rate over the ceiling.
Furthermore different set ups of the cooling ceilings were measured – one with natural ventilation
over the suspended cooling ceiling and one without natural ventilation. The measurement results
show that there is a decrease of concentration due to missing natural ventilation.
The measurement results show a clear difference in the rate of concentration on different room points
due to solar radiation, the location of impact and the room set up.
1. Introduction
This paper presents the impact of different room boundary conditions (façade system and ventilation
rate over cooling ceiling) on the air movement inside an office room. The motivation for this work is
that the calculation of comfort in the common areas of buildings plays a key role for the acceptance of
new concepts. In the completeness the calculation of complex flow processes in closed and open
façades, the dynamic of building elements' performance in combination with the control system is at
the moment, if possible, very time consuming. The aim is to develop a simplified room model for the
evaluation of building concepts in terms of energy efficiency and comfort.
Measurements for the model validation are one part of the development. A series of measurements
according to the operative temperature and the cooling capacity have been carried out and the results
show that there is an impact on the cooling capacity with respect to different façade systems. There is
an increase of the cooling capacity of about 20 W/m²cooling area in dependency of the temperature
difference between the operative temperature in the middle of the room and the mean medium
temperature due to a second internal glazing (Eder, 2012).
For verification of the assumption, that the capacity increase is due to a higher air flow rate over the
ceiling and therefore to a higher convection load, a mathematical model is built. The validation of this
model is done by a series of tracer gas measurements and the results of these measurements are
presented in this paper. Chen (2010) describes an overview of ventilation assessment methods. The
methods are divided into analytical, empirical, small-scale experimental, full-scale experimental,

multizone network, zonal and CFD models for ventilation in buildings. The advantage of full-scale
measurements is, that the real situation is studied, although they are mostly limited by the numbers of
measurement points in the tested area.
2. In-situ measurement –Building Information

The measurement was done in an existing office building. On the 34th floor a test room of about
10.8 m² has been built, this is typical for a single person office in Austria. As the room is situated on
the 34th floor, there is no shading due to other buildings or geographical surroundings. The façade is a
west-orientated totally glazed façade.
Figure 1 shows the floor plan (left side) and the section view (right side) of the measurement room.
The floor is raised which is caused to its air leading properties (the supply air is realized over the
raised floor) open throughout the whole storey. It is not possible to close the floor, because then the
test room would be without fresh air. The suspended ceiling of the measurement rooms is separated by
a foreclosure (mineral wool) around the room. The suspended ceiling in the corridor area is used as
exhaust air duct.
FIG 1. Floor plan (left side) / section (right side); location of tracer gas injection is highlighted; floor
plan: direct injection to the room (1); concrete ceiling (2) cooling ceiling (3); supply air area (4);
façade area (5)
The U-Values of the surrounding areas of the room are summarized in the following list:
 glazing: U = 1.10 W/m²K
 frame (percentage 8%): U = 2.95 W/m²K
 ceiling / floor: U = 1.47 W/m²K
 Interior wall (gypsum cardboard): U = 0.65 W/m²K
 Interior door: U = 1.80 W/m²K
There is the possibility to change the façade system from a single skin façade to a single storey double
skin façade by adding a second internal glazing element. These two façade systems were tested, both
with an internal shading element. The difference between these two façade variations are shown in
figure 3, there the difference is highlighted between picture 1 and 3, and between 2 and 4 respectively.
2.1 Cooling
The cooling of the rooms is done by a suspended cooling ceiling with capillary pipe mats, which are
placed in the ceiling panels. The cooling area is splitted into a façade area and a room area, these are
overlapping so that there is a higher capacity in the façade area. For the room area of the measurement
room the active area of the cooling ceiling is 5.56 m² and for the façade area it is 3.19 m². Overall

there is an active cooling area of 8.75 m², based on the room area (10.8 m²) there is a rate of cooling
area of about 80 %.
Through the measurement the boundary conditions of the suspended cooling ceiling was changed, one
time with and one time without natural ventilation over the cooling ceiling. The difference on room
conditions is shown in figure 3, for the cooling ceiling picture 1 and 3 are without natural ventilation
and the 2nd and the 4th one are with natural ventilation.
2.2 Ventilation
There is a mechanical ventilation system. The design of the ventilation system is as full air
conditioning – heating, cooling / dehumidification and humidification.
The supply air for the room is realized via the raised floor, the whole floor is air leading. The air
outlet is over “slot diffusors” next to the façade. There is a flow rate of about 50 m³/h for the room,
which means an air change rate of about 2.0 1/h with a supply air temperature of about 21 °C.
The exhaust air is realized by exhaust air lights (light fixture combined with extract air terminals).
There are flexible ducts between the lights and the suspended ceiling in the corridor. The idea of
doing the return air in this way is to remove the heat were it is generated. The suspended ceiling in the
corridor is air leading as well.
3. Measurement Set up
The goal of the measurement is to paint a picture of the air flow movement in the room and to see if it
is changing with different façade situations. Therefore a series of tracer gas measurements was done.
The ASHREA Handbook of fundamentals (2009) describes three different tracer gas methods to
determine the air change rate:
 concentration decay or growth method
 constant concentration method
 constant injection method
The concentration decay method, measuring the decay of tracer gas is the simplest and most often
used method. The following three conditions should strictly be satisfied using single point
measurements. For single point measurements the tracer gas should be uniformly mixed in the
enclosure, there should be no unknown tracer gas sources and the ventilation flow should be the
dominant means of the tracer gas removal from space to get reliable results (according ASHREA
Handbook of fundamentals (2009)).
According to Charlesworth (1988) it is very difficult to achieve a uniformly mixture of the tracer gas
in the entire enclosure of in-situ measurements. To solve this problem tracer gas concentrations
should be measured at several points of the considered areas and it is assumed that the mean value is
the average concentration for the entire enclosure.
As tracer gas carbon dioxid (CO2) was used, as described in Laussmann (2011) it fulfils a number of
the specification of a good tracer gas.
1.1 Sensor Location

To measure the room a grid of 27 CO2 sensors was set up. In figure 2 the position of the CO2 sensors
is shown, where the left side is the floor plan and the right side shows the section of the room. To
measure the air flow in the room, a grid of CO2 sensors was built.
The measurement itself was done by CO2 injections and the concentration at the measurement points
was logged. A defined amount of CO2 was injected on five different positions of the room (Figure 1
shows the different impact locations of tracer gas):

 Inj 1: directly into the room, floor area (by a frame with small openings)
 Inj 2: ceiling – height of the concrete ceiling
 Inj 3: ceiling – height of the suspended cooling ceiling
 Inj 4: Supply air area
 Inj 5: Façade area
FIG 2. Measurement set up – floor plan (left side) and section (right side) including the sensor
numbers, where the first 4 numbers are the maximum concentration rate and the last 2 are the
numbers referring to the evaluation diagrams
1.2 Measurement Variations

The measurement was done with a single skin façade and a single storey double skin façade and with
an open suspended ceiling and a closed suspended ceiling (figure 3).
The measurement was done with the following 4 different room types:
 Room type 1: single skin façade, closed suspended cooling ceiling (sskin+closedc)
 Room type 2 single skin façade, open suspended cooling ceiling (sskin+openc)
 Room type 3: single storey double skin façade, closed suspended cooling ceiling
(dskin+closedc)
 Room type 4: single storey double skin façade, open suspended cooling ceiling
(dskin+openc)
FIG 3. Schematic overview of the four different room types; from left to right: single skin
façade+closed suspended cooling ceiling; single skin façade+open suspended cooling ceiling, single

storey double skin façade+closed suspended cooling ceiling; single storey double skin façade+open
suspended cooling ceiling
In total there were measured 40 different variants with different boundary conditions. Each variant
was measured twice, one time with a clear sunny sky and the other time with a cloudy sky. The
measurement period for each variant is 1 hour due to the air change rate of the room. The
measurement interval is 10 seconds.
2. Results
2.1 Decay curves – CO2 concentration
As a result of the measurement there are curves of concentration for each variant and each sensor. The
measurement period for one variant is 1 hour, but the main air movement is within the first 10
minutes, then the concentration is equal within the room (see figure 4 and 5).
Figure 4 represents the measurement results for room type 4 (single storey double skin façade+open
ceiling) for the different locations in the room. The results are for a clear sunny day and the CO2
impulse was injected directly into the room. In the diagram there are results for different room
heights, where the black line represents the measurement points at 1m height, the black dotted line the
measurement results at 2m room height and the grey line the measurement results for the ceiling. The
location of the measurement is marked in the diagram at the right corner. The figure shows the change
of concentration for different locations within the room and the measurement results indicate that
there is no big difference of concentration with respect to the room location, the existing difference
can be explained by the supply air location next to the façade.
FIG 4. Measurement result of CO2 concentration a different location in the room with the impact in
the room for room type 4, the numbers shown in the legend are referring to the numbers of the sensor
(last two numbers in fig 2). The measurement results of the air temperatures (t) for the different room
zones are: troom= 24.5 °C, tfacade = 31.2 °C, tceiling = 23.5 °C and tfloor= 21.8 °C.
2.2 Impact Location of Injection

The summary of the results for the difference of the CO2 concentration in the different room areas by
different location of injection are shown in figure 5 for room type 04 (single storey double skin
façade+open suspended ceiling).

The grey line represents the concentration in the raised floor and is constant within the whole
measurement period. The marked black lines show the different room areas (the room itself, the
façade area and the ceiling area). The diagram shows that there is an “air roll”, the highest
concentration rate with the injection direct to the room is in the façade area, the second highest
concentration rate is in the ceiling area and the room itself has the smallest concentration rate. With
the injection of tracer on the cooled layer (third part of the diagram) the results show a rapid
declension of a cold airstream, first the concentration raises within the room and then in the façade
and the ceiling area.
FIG 5. Measurement result of CO2 concentration with respect to different impact location for room
type 4 (dskin+openc); with solar radiation; measurement period 1 hour for each variant; different
concentrations in the specified areas within the first 10 minutes, after the first 10 minutes the
concentration is equal in each area independent of the impact location
2.3 Summary of results for a day with solar radiation

In figure 6 there is the summary of the measurement results for the four different room types for a day
with solar radiation visualized.
Figure 6 shows the first 10 minutes after the CO2 injection directly into the room. The values in the
diagram are the mean values of concentration for the room area, the façade area, the suspended ceiling
and the raised floor. The concentration of the raised floor is not changing during the whole
measurement, so there is no mass flow from the room to the floor.
The different rate of concentration in the suspended cooling ceiling show the difference in the air flow
over the cooling ceiling. The measurement results show an increase of air flow due to the single
storey double skin façade (comparing room type sskin+openc and dskin+openc) and a decrease of air
flow due to the closed ceiling (comparing room type sskin+openc and sskin+closedc to room types
dskin+openc and dskin+closedc).
2.4 Summary of results for a day without solar radiation

In figure 7 the results for the four different room types for a day without solar radiation are visualized.
There is almost no difference of concentration between the room types. Due to the lack of buoyancy
there is no air roll and the air movement is very small.

In figure 7 the first 10 minutes after the CO2 injection direct into the room are shown. The values in
the diagram are the mean values of concentration for the room area, the façade area, the suspended
ceiling and the raised floor.
The different rate of concentration in the suspended cooling ceiling show the difference in the air flow
over the cooling ceiling due to the tested boundary conditions of the room. The measurement results
for the case without solar radiation show almost no difference of concentration.
FIG 6. Measurement results of CO2 concentration for the different room types with solar radiation,
the tracer gas injection directly into the room (floor area). The diagram shows an increase of
concentration in the ceiling area due to the single stores double skin façade and a decrease due to the
lack of natural ventilation for the cases with the closed suspended ceiling.
FIG 7. Measurement results of CO2 concentration for the different room types without solar
radiation, the tracer gas injection directly into the room (floor area). The diagram shows that there is
almost no difference of concentration between the room types

3. Conclusion
Through an intensive set of measurements the impact of air flow due to different room conditions
(façade system, cooling system) were evaluated. The tracer gas measurement results show the impact
by different concentration rates. The location of injections has no impact on the decay curves after the
first 10 minutes.
The single storey double skin façade leads to a significant increase of tracer gas concentration within
the ceiling zone. The maximum concentration within the suspended ceiling variate between 720 ppm
(room type 1) and 1491 ppm (room type 4). Due to the second glazing there is an increase of
350 ppm. The difference between the open and the closed ceiling is about 420 ppm.
For further evaluation of the measurement results an ideal multi zone model was assumed and mass
balances were calculated. Furthermore a mathematical model (CFD) is built and the measurements are
used to validate the air flow model.
The measurement results show an impact on the air flow due to different boundary conditions. The
next step for the model development is the detailed mathematical modeling of the air flow network
within the room, which is done by CFD. To reach the aim of the research, which is the development of
a computational model that can determine the impact on comfort with reasonable accuracy, the set of
tracer gas measurement is one part of the model validation.
Based on the validated room model, recommendations for future building designs in net energy
producing buildings can be formulated.
4. Acknowledgements
The doctoral program “Energy Systems 2030 (ENSYS 2030)” is an internal research project funded
by the Vienna University of Technology, which has the main goal to provide doctoral students within
the research area "Energy and Environment," a remarkably structured and interdisciplinary education.
10 PhD positions are advertised and supported as part of the College of the Vienna University of
Technology.
Energy and Environment is part of the development plan of the Vienna University of Technology
2010+ as one of the five research priorities. The doctoral program ENSYS 2030 has the following key
themes: solar energy, energy storage, electro mobility and energy active building.
References
ASHRAE. Handbook of fundamentals.2009. Atlanta, USA: American Society of Heating,
Refrigerating and Air-Conditioning Engineers, Inc.
Charlesworth P.S.1988. Air exchange rate and airtightness measurement techniques - an applications
guide. Coventry, UK: Air Infiltration and Ventilation Centre
Chen, Q., Lee, K., Mazumdar, S., Poussou, S., Wang, L., Wang, M., and Zhang, Z. 2010. “Ventilation
performance prediction for buildings: Model Assessment,” Building and Environment, 45(2), 295-
303.
Eder K., Steininger C., Bednar T. 2012 “Enhancing the performance of a cooling ceiling by an
innovative façade system”; Proceedings of the 5th International Building Physics Conference, IBPC,
2012, Kyoto
Laussmann D., Helm D. 2011. Air Change Measurements Using Tracer Gases: Methods and Results.
Significance of air change for indoor air quality. Chemistry, Emission Control, Radioactive Pollution
and Indoor Air Quality, Dr. Nicolas Mazzeo (Ed.), ISBN: 978-953-307-316-3, pp. 365-399

Full paper no: 6
Effect of different façade systems on the cooling capacity of a
cooled ceiling – CFD modeling
Katharina Eder 1
Maximilian Neusser 1
Thomas Bednar 1
1
Vienna University of Technology, Institute for Building Construction and Technology, Department
for Building Physics and Building Acoustic, Austria
KEYWORDS: façade systems, natural ventilated double skin façade, computational fluid dynamics,
operative temperature, cooling capacity, comparison measurement and CFD calculation
SUMMARY:
One of the key issues for the calculation of a room’s condition and the resulting thermal comfort is
the knowledge of the air flow distribution in the room. A simplified but still accurate enough CFD
model should be developed to determine the effect of different façade types on the cooling capacity of
a cooling ceiling. A parametric study using the program COMSOL Multiphysics 4.3a was done. The
output of the calculations was compared with the results of a series of in-situ measurements and is
presented in this paper.
The measurement and the calculation results show an impact on the cooling capacity by changing air
flow conditions within the room. The design of the façade system can increase the capacity of a
cooling ceiling and therefore decrease the operative temperature in the room. Neglecting the flow
characteristics in the room can lead to incorrect results, especially for complex façade systems in
combination with component activation (e.g. cooling ceiling).
Therefore an enhanced building model for the design of future buildings, taking into account the
interaction of building envelope and building service systems should be developed using the results of
the CFD calculation. Recommendations for energy active buildings related to room conditions, room
comfort and the energy demand can be formulated by simulations over yearly periods.
1 Introduction
This study presents results of the cooling capacity of a cooling ceiling in dependence of two different
façade systems – a single skin façade and a single storey double skin façade. The influence on the
cooling capacity due to the façade systems was measured by an intensive set of measurement and the
results are presented in Eder et al. (2012) and are used for the validation of the mathematical model.
Due to multiple coupled physical phenomena (air flow, convection, conduction and radiation) the
behaviour of double skin façades is very complex. Both, experimental and numerical models for the
optimization of the performance of double skin façades have been developed e.g. network models
(Tanimoto et al., 1997), zonal models (Jiru et al., 2008), energy simulation with coupled air flow
models (Stec et al. 2005) and detailed computational fluid dynamics studies (Manz, 2004).
Most of the developed models are for external double skin facades. This paper studies the effect of a
(internal) single storey double skin façade on the HVAC system. The diving force is natural buoyancy,
therefore the air flow is not easy to control and the dependence on weather conditions (mainly solar
radiation) makes it discontinuous. The validated results of the air flow calculation should be

integrated in an enhanced room model which takes flow phenomena that influences the cooling
capacity, the operative temperature - and therefore the comfort within the room - into account.
2 Theoretical background
The software program COMSOL Multiphysics version 4.3a was used, with the aim to calculate the
impact of air movement with different façade systems. As physical model non isothermal flow was
used to describe the problem. The main equations are the Navier Stokes (momentum balance and mass
balance), heat balance and energy transport (convection and conduction).
One of the main topics is the modeling of the façade. Because the simulation program COMSOL
Multiphysics version 4.3a consider all radiation as the same, it is not possible to make a distinction
between large infrared radiation and the solar spectrum (UV, light & short IR). Within this study the
interaction between the façade system and the cooling ceiling is the main topic, therefore the surface
to surface radiation has to be considered. For this work the incident solar radiation was realized by
heat impacts to the individual layers of the façade according to its absorption and solar transmission
coefficient.
3 2-D Model
3.1 Geometry and mesh properties
The geometry of the simulation model is according to the architectural plans, a section through the
room. A simplification of the model, from three dimensions to two was done. The symmetry is along
the depth of the room (see figure 1).
The main geometric data of the room is summarized in table 1.
FIG 1. Simplification of the model geometry from three dimensions to two dimensions
TABLE 1. Room / model dimensions

Outside pane height in m width in m
Room height 3.40 -
Room depth 4.10 -
Façade – exterior glass 3.07 0.01
Argon 3.07 0.01
Façade – interior glass 3.07 0.01
Gap between glass and screen 3.07 0.03
Screen 3.07 0.01

Gap between screen and 2nd glazing 3.07 0.11
2nd glazing extern 2.74 0.01
2nd glazing argon 2.74 0.01
2nd glazing intern 2.74 0.01
3.2 Façade systems

With the CDF Model the impact of two façade systems – Façade system 1: single skin façade with
internal shading element; Façade system 2: single storey double skin façade with a shading element
between the 2 glazing elements should be tested. Figure 2 shows the schematic overview and table 2
summarize the technical data for the façade systems. The total solar energy transmittance g-value of
the glazing is g = 0.36 according to the technical data sheet. The air permeability of the shading
systems was not considered within this study.
shading shading
2nd glazing
façade type 1 façade type 2
FIG 2. Façade type 1 (left side): single skin façade with internal shading element / façade type 2
(right side): single storey double skin façade with a shading element between the glazing elements.
TABLE 2. Technical data façade system

Parameter description Outside pane Shading system 2nd skin
Total solar energy
g = 0.36 --- ---
transmittance:
Solar transmittance:  = 0.33  = 0.11  = 0.55
Reflectance:  = 0.41  = 0.39  = 0.24
Absorptance outside pane: ac = 0.25 a = 0.50 ac = 0.25
Absorptance inside pane: ai = 0.01 --- ai = 0.01
Light transmittance: --- L = 0.10 ---
3.3 Boundary Conditions

The heat flux of the external wall is defined by an inward heat flux, the heat transfer coefficient for
the external wall is 20 W/m²K and the external temperature is 32 °C. The surface temperature for the
cooling ceiling panels was measured by thermography with a reference temperature measurement.
The temperatures of the enclosing surfaces are summarized in table 3.
TABLE 3. Temperatures surrounding surfaces
Surface Temperature in °C
Floor 23.0
Partition wall 25.0
Ceiling – room 23.0

Ceiling façade 32.0
Cooled ceiling (bottom) 18.5
Cooled ceiling (top) 18.5
3.4 Material Properties and Initial conditions

The material properties of the surrounding surfaces and used materials are shown in table 4 and the
the surface emissivity is shown in table 5. The initial temperature condition t0 is 294.15 K, the initial
pressure pref is 1 atm. The dynamic viscosity, the heat capacity and the thermal conductivity of the
layer ”air” are temperature dependent and the density is calculated in dependency of the air
temperature and the air pressure.
TABLE 4. Material data
heat
ration of
dynamic capacity at thermal
specific density
viscosity a constant conductivity
heat
pressure
Pa s - J/kg K kg/m³ W/mK
rho(pA[1/Pa
Air (room) eta(T[1/K]) 1.4 cp(T[1/K]) k(T[1/K])
],T[1/K])
Façade external glazing 480 2200 1.1
Argon 520 1.7837 0.0177
Façade internal glazing 480 2200 1.1
Screen 1700 1150 0.26
2nd glazing – external 480 2200 1.1
2nd glazing – internal 480 2200 1.1
Aluminium 900 2700 160
TABLE 5. Surface emissivity

Surface Surface emissivity
Floor 0.97
Aluminium – grey paint 0.95
Glass 0.80
Screen 0.88
Concrete – ceiling 0.96
Plaster – partition wall 0.94
3.5 Mesh and Solver setting

One output of the CFD calculation is the cooling capacity of the cooling ceiling, due to this issue, the
calculation was done with the Low Reynolds number k- turbulence model. The mesh of the complete
model consists of 100298 (façade type 1) and 120508 (façade type 2) elements.
The summary of the solver settings is shown in table 6
TABLE 6. Solver settings
description type
Type of analysis: non isothermal flow Non-isothermal flow
Linear system solver: MUMPS

Relative tolerance 0.001
Pivot threshold 0.1
Memory allocation factor 1.2
4 Results
4.1 Operative Temperature
In figure 3 the results of the operative temperature in the middle of the room are shown for the two
façade systems. The results for the single skin façade show an increase of the operative temperature
from 24.7 to 28.5 °C with an increasing solar impact. The mean capacity of the cooling ceiling variate
between 40 and 80 W/m²cooling area. The effect of the single storey double skin façade is a smaller
increase of the operative temperature in the middle of the room due to a higher efficiency of the
cooling ceiling. The diagram shows, that the operative temperature stays around 26 °C and the cooling
capacity increases with respect to the solar impact.
FIG 3. Comparison of operative temperature and the cooling capacity for façade type 1 (left side) and
façade type 2 (right side); the temperature is below 27°C with façade type 2, the cooling capacity is
increasing with increasing solar radiation; with façade type 1 the operative temperature is getting
above 27°C and the cooling capacity is lower.
4.2 Cooling capacity

The definition for the test bed measurements for the cooling capacity of cooling ceilings is done in the
standard EN 14240. According to the standard the cooling capacity is based on the average
temperature difference between the operative temperature in the middle of the room and the mean
medium (flow and return) temperature and is referring to the active cooling area.
The cooling capacity was calculated for both façade types with different incident radiation (from 100
W/m² to 800 W/m²): In figure 4 the results on cooling capacity in relation to the room depth for
façade type 2 for a temperature difference between the mean medium temperature and the operative
temperature in the middle of the room of 9 K are visualized.

FIG 4. Cooling capacity in dependence of the room depth for façade system 2; temperature difference
between mean medium temperature and operative temperature in the room is 9 K
The results for the cooling capacity of the cooling ceiling as a function of the depth of the room for
the 2 façade systems is shown in figure 5. The calculation results show that there is an increase of the
cooling capacity by comparing the single skin façade (grey line) and the single storey double skin
façade (black line). Especially next to the façade but also throughout the room, the difference is
minimal where the cooling ceiling is arranged overlappingly.
FIG 5. Cooling capacity in dependence of the depth of the room with an incident solar radiation of
800W/m² and a temperature difference of 9 K between the mean medium temperature and the
operative temperature in the middle of the room.
Figure 6 shows the resulting mean cooling capacity with different impacts due to the incident solar
radiation for the two façade systems. The temperature difference between the mean medium
temperature and the operative temperature in the middle of the room is 9 K.

FIG 6. Resulting cooling capacity in dependence of the incident solar radiation; the temperature
difference between the mean medium temperature and the operative temperature in the middle of the
room is 9 K
The comparison of the measurement results and the results of the CFD calculation is shown in figure
7. There is a good agreement between both for the operative temperature and the cooling capacity.
The measurement results are taken by a series of measurement in an existing building, the description
is found in Eder et al. (2012).
In figure 7 on the left side the operative temperatures for both façade types are shown, the left part in
this diagram represents the result of the CFD calculation and the right part the measurement results.
The difference of the operative temperature in the middle of the room is about 0.2 K. The right side
diagram in figure 7 show the results of the cooling capacity in dependence of the temperature
difference between the mean medium temperature and the operative temperature in the centre of the
room. The lines represent the measurement results. The points are the results out of the CFD
calculation for the different tested boundary conditions. The diagram shows a good agreement
between the measurement results and the results of the CFD calculation.
FIG 7. Measurement results on operative temperature (left side) and cooling capacity (right side;
(M)..measurement; (C)..CFD calculation) – there is a good accordance between the measurement
results and the CFD calculation; the difference of operative temperature is about 0.2 K;

5 Conclusion
The comparison of in-situ measurements of operative temperature and the cooling capacity of a
cooling ceiling and the calculation results of the CFD calculation show good agreement. The CFD
model can therefore be used to characterize different façade systems regarding their impact on
operative temperature and the capacity of the cooling ceiling.
The comparison of the two façade systems – the single skin façade (type 1) and the single storey
double skin façade (type 2) – shows that there is a decrease in operative temperature and an increase
of the cooling capacity.
The CFD calculation results show the cooling capacity in dependence of the solar radiation the
temperature difference between the mean medium temperature and the operative temperature in the
middle of the room and the depth of the room. The results show, that there is an impact due to the
solar radiation and the characteristics of the cooling capacity within the room geometry. The increase
of the cooling capacity due to a higher air flow rate is limited on the first 3 m from the façade, there is
no longer an impact due to solar radiation further in the back.
The results of the CFD calculation can be used for an enhanced building model for future building
design. Recommendations for net energy producing building according to room comfort and energy
demand can be formulated by simulations over yearly periods.
6 Acknowledgements
The doctoral program "Energy Systems 2030 (ENSYS 2030)" is an internal research project funded
by the Vienna University of Technology, which has the main goal to provide doctoral students within
the research area "Energy and Environment," a remarkably structured and interdisciplinary education.
10 PhD positions are advertised and supported as part of the College of the Vienna University of
Technology
Energy and Environment is part of the development plan of the Vienna University of Technology
2010+ as one of the five research priorities. The doctoral program ENSYS2030 has the following key
themes: solar energy, energy storage, electro mobility and energy active building.
References
Eder K., Steininger C., Bednar T. 2012 “Enhancing the performance of a cooling ceiling by an
innovative façade system”; Proceedings of the 5th International Building Physics Conference, IBPC,
2012, Kyoto
EN 14240. 2004. Ventilation for buildings - Chilled ceilings - Testing and rating
Jiru T.E., Haghighat F., 2008 “Modeling ventilated double skin façades – a zonal approach”, Energy
and Buildings 40, 1567–1576.
Manz H. 2004 “Total solar energy transmittance of glass double façades with free convection”,
Energy and Buildings 36, 127–136.
Safer, N., Woloszyn, M., Roux, J.J., 2005 "Three-dimensional simulation with a CFD tool of the
airflow phenomena in single floor double-skin façade equipped with a venetian blind", Solar Energy
79, 193–203.
Stec W.J., Van Paassen A.H.C. 2005, “Symbiosis of the double-skin façade with the HVAC system”,
Energy and Buildings 37, 461–469.
Tanimoto J., Kimura K., 1997 “Simulation study on an airflow window system with an integrated roll
screen”, Energy and Buildings 26, 317–325.

Full paper no: 7
Air leakage and hygrothermal performance of an internally
insulated log house
Üllar Alev, M.Sc 1
Andres Uus, M.Sc 2,3
Marko Teder, M.Sc 2
Martti-Jaan Miljan, M.Sc 2
Targo Kalamees, Professor 1
1
Tallinn University of Technology, Estonia
2
Estonian University of Life Sciences, Estonia
3
Hobbiton OÜ, Estonia
KEYWORDS: air leakage, internal insulation, log house, test-house, corner notch
SUMMARY:
In this study the air leakage of four different log corners and hygrothermal performance of three
different interior insulation materials are studied in a log test house. Field measurements of the air
leakages were analysed in a test house built with different types of log junctions: corner post, dovetail
notch, Scandinavian saddle notch, and double notch with wind lock. One wall made of logs (average
thickness of 270 mm) in a test house was internally insulated with three different materials: cellulose
fibre, mineral wool and reed mat. The air leakage rate of the overall house was also measured using
the standardized building pressurization technique. Hygrothermal performance of walls was measured
with t&RH sensors and heat flow plates.
The mean air leakage rate of the overall house at the pressure difference of 50 Pa was q50=2.8
m3/(h·m2) and mean air change rate was n50=5.0 h-1. The air leakage rate lowered slowly over time.
The lowest air leakage was through the Scandinavian saddle notch - 6.7 m3/(h⋅m).
Simulation models on the hygrothermal performance of the studied test walls were validated based on
the measurement results. Temperature, relative humidity and heat flux showed good agreement
between the measured and the calculated results.
1. Introduction
Log houses have a long history and they represent a variety of building techniques employed in
Estonia and in other Nordic countries. Results of measurements of old log houses have shown that
this building type is typical of the highest air leakage (Alev & Kalamees 2013). In a new log house,
the quality of the envelope has to be higher than a century ago, because the requirements for comfort,
function, and energy-efficiency of today’s residents are different.
Uncontrolled air movement through a building envelope leads to problems related to the hygrothermal
performance, health, energy consumption, performance of the ventilation systems, thermal comfort,
noise and fire resistance. Air leakage through the building envelope depends on the results of the air-
pressure differences across the envelope, the distribution of air leakage places and the airtightness of
the building envelope.
Many studies have analyzed the possibilities to use the internal thermal insulation for improving the
thermal resistance of external walls, including those focused on stone walls (Stopp et al. 2001; Häupl
et al. 2004; Toman et al. 2009) and only few have studied the internal insulation of log walls (Ojanen
2007; Alev et al. 2012; Arumägi & Kalamees 2012; Arumägi et al. 2011). Ojanen (2007) studied the

internal insulation only numerically and assumed that the log wall was completely airtight, but vertical
air channels between the log and the insulation layer were suggested to reduce the moisture level.
Arumägi & Kalamees (2012) and Alev et al. (2012) studied old walls with high leakage rate and found
that the log wall is not completely airtight and to validate the simulation model the air change through
the log wall had to be added. Other differences between mentioned studies include the type and usage
of the house, used insulation materials, moisture excess etc. Therefore the measurements of new log
wall were needed.
This study had two main objectives:
• to study the air leakages of different corners and the whole building of a new log house;
• to validate the simulation model of three different internal insulation constructions (materials).
2. Methods
2.1 Tested house
The field measurements were carried out in a small test house (one room with a net area of 18 m2)
specially designed and built for current study (FIG 1). The house was made of square logs with an
average thickness of 200 mm (except the back wall with half round logs and an average thickness of
270 mm, insulated from the internal side). Every corner had a different type of log junction (FIG 2).
A B C
Studied wall with

internal insulation
FIG 1. View of the test house from the back (left) and internally insulated wall (right).
1 2 3 4
FIG 2. Different corner notches of the log wall: double notch with wind lock (1), Scandinavian saddle
notch (also known as Norwegian notch) (2), dovetail notch (3) and corner post (4).

The types of corner notches were selected based on popularity among log house construction
companies in Estonia. In the corner notches, a special self-expanding sealing tape for log houses
(Classic Log Home Tape LHC 20-20-06) was used.
One wall was insulated internally with three different materials (FIG 1, left): mineral wool, cellulose
fibre (both covered with vapour retarder and gypsum board) and reed mat with clay plaster. The
internally insulated wall faced west and it was shaded with wooden boards (not present in FIG 1 (left),
between the boards and the log wall there was a ventilated air gap of 250 mm) to protect the wall and
sensors from direct sunlight. The indoor climate conditions created in the house were based on the
results of a study in Estonian wooden apartment buildings (Kalamees, Arumägi, et al. 2011). The
room was heated with an air-air heat pump with the heating setpoint of 21oC. The room was also
humidified, the target of the automatic humidification system was to hold the moisture excess of
5.5g/m3 in the winter period and 2.5g/m3 in the summer period constant.
2.2 Measurement methods

The air tightness of the entire test building was measured with the standardized (EVS-EN13829 2001)
fan pressurization method, using ‘‘Minneapolis Blower Door Model 4’’ equipment with an automated
performance testing system.
A special timber framed structure with the width of 1414 mm made of 20x80 mm planks was set
inside the pre-milled 50mm deep gaps in the log walls at each corner of the test house. The timber
frame, ceiling and floor were sealed with a 0.15mm thick PVC-membrane with an opening of
ø110mm to measure air leakage of every corner separately using an anemometer. A special self-
expanding sealing tape was used to ensure the air tightness between the log wall and timber frame. In
the calculations the air leakage of the plain wall was subtracted from the total air leakage rate of the
corner based on the wall area and the air leakage rate of the plain wall.
To determine the air tightness of the building envelope, depressurizing and pressurizing tests were
conducted. Measurements were made at 10 Pa pressure difference step from 0 to 60 Pa. An
exponential trend line was calculated according to the measurement points and an exact 50 Pa reading
was taken from the trend line. Measurements of every corner were repeated for 11 times during one
year. The pressure difference was generated with “Minneapolis Blower Door Model 4” equipment, the
air flow was measured with an anemometer (Ahlborn FVA 915 MA1; measurement range 0.2…20
m/s ±0.5%) and the pressure difference was measured with a pressure sensor (FD8612DPS;
measurement range was 1 mbar (100 Pa) ±1%), both were continuously recorded with a data logger
(Almemo® 2890-9) over a period of 30 s with a 1 s interval.
The values of the temperature, the RH both inside and outside the wall and the heat flux were
measured over a one-year period at one-hour intervals. The following sensors were used: temperature
sensors (TMC6-HD; measurement range: -40 °…+100 °C, accuracy: ±0.25 °) with HOBO U12-013
data loggers; temperature and RH sensors (Rotronic HygroClip SC05 ∅5mm×51mm, measurement
range: -40 °…+100 °C and 0…100%, accuracy: ±0.3 °C and ±1.5%); and heat flux plates (Hukseflux
HFP-01-05, measurement range ±2000 W/m2, accuracy: ±5%). Measurement results were saved with
a Grant Squirrel SQ2020 data logger. Together with airtightness measurements the moisture level in
the logs at different heights was measured with Gann Hydromette HT 85 (with puncture probe M40,
accuracy ±2%) to observe the drying out process of the logs.
2.3 Simulations
The measurement results were compared with a complex hygrothermal simulation model, WUFI 5.1
Pro. The comparison was made to validate the simulation model for future simulations with different
initial and climatic conditions as well as with different dimensions of the building envelope layers.

3. Results
3.1 Air tightness of the corners and the envelope
The airtightness of the entire house was measured for seven times during a one-year period. The
average air leakage rate of all measurements of the entire envelope at the pressure difference of 50 Pa
was q50=2.8 m3/(h⋅m2) (FIG 3 left) and the mean air change rate was n50=5.0 h-1. The air leakage rate
lowered slowly over time until the two last measurements in spring; these measurements show more
than 1 m3/(h⋅m2) higher value than before. The air-tightness of the log house is influenced by several
factors: volume changes of the logs due to water vapour (de-)sorption processes, the efficiency of
insulation and sealing in notches and grooves (joints of logs), accuracy of notches and grooves,
tightening of walls caused by the weight of the roof, and the overall building quality. The change of
airtightness in time can be explained in several ways. The reduction of air leakages was probably
caused by the weight of the roof that tightened the joints between the logs (grooves and notches). By
the time of the last two measurements, the weight of the roof was significantly decreased, because the
thick layer of snow had melted. Another reason for increased air leakages during spring was the drying
process of the logs caused by intensive solar radiation (FIG 3 left). The drying process is more
intensive near the ends of the logs and therefore the shrinkage of the logs near the corners is higher
and the cracks (grooves) will widen. The wider cracks near the corners increased the air leakage both
of the whole house and also in every corner (FIG 3 right).
The air leakages of the corners were measured for 11 times during the study. The average air leakage
rate of all measurements in every corner was 10 m3/(h⋅m). The air leakage of the corners were (FIG 3
right):
• Scandinavian saddle notch 6.7 m3/(h⋅m);
• dovetail notch 6.8 m3/(h⋅m);
• corner post 10.4 m3/(h⋅m);
• double notch with wind lock 17.3 m3/(h⋅m).
When comparing the change of the air leakage rate in time, all the corners except the Scandinavian
saddle notch started to improve after the second measurement. In the case of the Scandinavian saddle
notch, it can be explained as the combination of drying of the logs (shrinkage of volume),
effectiveness of sealing and tightening of the wall due to weight. During winter the snow layer added
extra weight to the roof and to the wall, thus the tightening process of the corners was the greatest.
FIG 3. Air leakage rate of the total house (left) and the corners (right).
3.2 Hygrothermal performance and validation of the simulation model

The daily average outdoor microclimate near the test wall varied between -27 °C and +28 °C. The
average outdoor temperature during summer (June…August) was +17.9 oC and during winter
(December…February) was -4.5 oC and the RH was accordingly 72% and 89%.

Modified material properties were used from the WUFI database. Table 1 shows the properties of the
materials used in comparison of the simulated measured and results when the best correlation was
obtained. Although the wall was made as airtight as possible, the air change through the log wall was
necessary to match the calculated results with the measured results. The air change rate of 0.3 h-1 was
added during the cold period (November to the February). The RH level of kiln dried logs at the time
of building was 21.6% at the depth of 3 cm.
Table 1. Main hygrothermal properties of the materials used in the simulation
Wooden Mineral Cellulose Reed “Intello” Clay Gypsum
log wool fiber mat membrane mortar board
3
Bulk density ρ, kg/m 390 60 60 136 425 1568 850
Porosity f, m3/m3 0.75 0.95 0.95 0.9 0.001 0.41 0.65
Specific heat capacity c, J/(kg·K) 1600 850 2000 2000 2300 488 850
Thermal conductivity λ, W/(m·K) 0.12 0.04 0.037 0.075 0.17 0.48 0.2
Vapour diffusion resistance factor µ, - 108 1.3 1.5 2.0 37500 20 8.3
Built-in moisture w, kg/m3 65 4.5 4.5 6.0 0 100 2.0
Two methods may be used to compare the thermal performance of the test walls: to compare
temperatures on the inner surface of the log, or to compare the measurement results of the heat flux
plate. The simulation model and this experiment were validated in both ways. The figures below show
the difference between the measured and the calculated temperatures (FIG 4), RH (FIG 5) and water
vapour pressure (FIG 6) between the log wall and the insulation layer for three materials. Thin lines on
both figures represent the measured values and thick lines the calculated values. There is a good
correlation between the measured and the calculated results in addition to the temperature and the RH,
also with the heat flux. Differences in the temperatures reflect the different thermal resistances of the
insulation materials. Reed mat has higher thermal conductivity and therefore the temperatures between
the insulation layer and the log wall were higher during the winter period.
FIG 4. Measured and calculated temperatures on the inner surface of the log wall.
The log wall with an average thickness of 27 cm had a thermal transmittance of U=0.49 W/m2K. After
adding internal insulation, the thermal transmittance decreased to U=0.31 W/m2K in the case of
cellulose fibre, U=0.28 W/m2K in the case of mineral wool, and U=0.36 W/m2K in the case of reed
mat.
The RH level of different materials was different due to different construction methods. Insulation
with cellulose fibre was most critical, because it was installed by a wet spray method. While the

cellulose was covered with water vapour barrier, the drying out of moisture was also slow (about six
months). The reed mat itself was dry, but the added plaster layer added a significant amount of water
into the insulation layer, therefore the RH was high during the first month. The drying process was
faster than in the cellulose fibre part, because the water vapour resistance of the clay layer is much
lower than the resistance of the PE-membrane used for water vapour barrier on the cellulose fibre and
on the mineral wool part. The mineral wool part had no additional moisture during installation and
therefore the RH level was low at the beginning of the measurements. The RH level increased because
of the moisture dried out from the logs. The RH in the reed mat in summer was about 10% lower than
in the mineral wool and cellulose fibre because the moisture can dry out to the room side easily. In the
mineral wool and cellulose fibre part the moisture mainly dried out to the external side due to air
convection through the log wall, which is much more intensive during the cold period (beginning from
November).
FIG 5. Measured and calculated RH on the inner surface of the log wall.
FIG 6. Water vapour pressure on the inner surface of the log wall.
The required moisture excess was guaranteed over the first half year without the use of air humidifier,
caused by the drying out of the logs and the absence of ventilation. The vertical temperature gradient
was small due to air movement in the room caused by the air-air heat pump.
4. Discussion
Comparison of airtightness measurements of this specially built house (2.8 m3/(h·m2)) and previous
measurements in the log houses in Estonia reveals substantial improvements using the new sealing
method and the quality of work: the average airtightness of 12 log dwellings was 9.2 m3/(h·m2)
(Kalamees 2008); the average of 35 measurements in wooden apartment buildings (made of logs) 10

m3/(h·m2) (Kalamees, Arumägi, et al. 2011) and the average airtightness of 24 old rural log houses
was 15 m3/(h·m2) (Kalamees, Alev, et al. 2011).
When sealing different corners, especially the double notch with the wind lock, the placement and the
choice of seal has an important role. A wide seal (15 cm) expanded less than a narrow one (2 cm).
Also, the logs should be as dry as possible before erecting the house to prevent the cracks caused by
the shrinking of the logs.
In this study the process and influence of the drying out of the logs on the internal insulation layer is
visible (FIG 5). On the one hand, it is reasonable to let the logs dry out more easily as in reed mat part.
On the other hand, the moisture level in reed mat raised to the same level as other insulation materials
during the winter period when the humidifier produces extra moisture to the room to hold the moisture
excess of 5.5g/m3. In the cellulose fibre part the RH stayed above the critical level for mould growth
(80%) more than half a year. The mould growth index (calculated according to Hukka & Viitanen
(1999)) on the surface of the log behind the cellulose fibre insulation layer had the highest value of 1.3
during April, which means that there could be some growth detected by microscopy. After removal of
insulation there was no mould growth detected visually and it was not viewed with a microscope.
5. Conclusions
The average air leakage rate of all measurements of the entire envelope at the pressure difference of 50
Pa was q50=2.8 m3/(h⋅m2). The average air leakage rate of all measurements in every corner was 10
m3/(h⋅m). The lowest air leakage was through the Scandinavian saddle notch - 6.7 m3/(h⋅m), almost
the same was through the dovetail notch - 6.8 m3/(h⋅m)), the corner post had much higher leakage rate
of 10.4 m3/(h⋅m) and the double notch with the wind lock was the leakiest - with 17.3 m3/(h⋅m).
The internal insulation parts were constructed differently from the hygrothermal point of view. The
mineral wool part was constructed as a dry wall, cellulose fibre was installed using a wet method and
the reed mat was covered with clay plaster that added also moisture to the reed mat. Cellulose fibre
and mineral wool were covered with water vapour barrier, which prevented these wall parts to dry to
the room side. The reed mat with clay plaster dried out within less than a month. The RH in the
cellulose fibre part was over 80% for over 6 months and the RH level in the mineral wool part had
increased to the same level as in the cellulose fibre part after 6 months and both started to dry out to
the external side during the winter period. During the warm period the RH was about 10% lower than
in other parts, during the cold period the RH level increased at the same rate as in other parts.
The simulation models of three internally insulated log walls were validated using long-term field
measurements. The WUFI was selected for the hygrothermal performance simulations. A good
correlation between the calculated results and the measured values was achieved after the modification
of the material properties and adding a factor as the air change rate in the material layers inside the
wall. The thermal transmittance decreased by 27..43% in different internally insulated wall parts as
compared to an uninsulated wall. Drying and wetting are determined more accurately if the convective
air flow is included in the hygrothermal simulation model. The drying out moisture from fresh, but a
kiln-dried log with an average RH level of 21.6% caused significant moisture excess to the room and
increased the RH in the internal insulation layer. The validated model will be used in our further
studies to focus on the performance of the internally insulated log wall in cold climates.
6. Acknowledgements
The research has been conducted as part of the projects IUT1-15 “Nearly-zero energy solutions and
their implementation on deep renovation of buildings” financed by Estonian Research Council.

References
Alev, Ü. et al., 2012. Comparison of thermal performance of mineral wool and reflective insulation on
internally insulated log wall. In Proceedings of Healthy Buildings 2012. Brisbane: Queensland
University of Technology.
Alev, Ü. & Kalamees, T., 2013. Field Study of Airtightness of Traditional Rural Houses in Estonia. In
M. L. Karel Kabele, Miroslav Urban, Karel Suchý, ed. Proceedings of CLIMA 2013. p. 6882.
Arumägi, E. et al., 2011. Field study of hygrothermal performance of log wall with internal thermal
insulation. In International Conference on Durability of Building Materials and Components.
Proceedings. pp. 811 – 819.
Arumägi, E. & Kalamees, T., 2012. Validation of a Simulation Model for Hygrothermal Performance
of Log Wall with Internal Thermal Insulation in Cold Climate. In Proceedings of 5th IBPC.
Kyoto, pp. 345–352.
EVS-EN13829, 2001. Thermal performance of buildings—determination of air permeability of

buildings—fan pressurization method., Estonian Centre for Standardisation.
Häupl, P., Fechner, H. & Petzold, H., 2004. Interior retrofit of masonry wall to reduce energy and
eliminate moisture damage: Comparison of modelling and field performance. In Thermal
Performance of the Exterior Envelopes of Buildings IX. Florida.
Hukka, A. & Viitanen, H.A., 1999. A mathematical model of mould growth on wooden material.
Wood Science and Technology, 33, pp.475–485.
Kalamees, T., Arumägi, E., et al., 2011. Eesti eluasemefondi puitkorterelamute ehitustehniline
seisukord ja prognoositav eluiga, Tallinn.
Kalamees, T., 2008. Elamupiirete õhupidavus: Uurimistöö „Elamute õhulekkearvu baasväärtuse

väljaselgitamine ja õhulekkearvu muul viisil tõendamise metoodika väljatöötamine“ raport.,
Kalamees, T., Alev, Ü., et al., 2011. Maaelamute sisekliima, ehitusfüüsika ja energiasääst I, Tallinn:
Tallinna Tehnikaülikool.
Ojanen, T., 2007. Low Energy Log Walls Under Cold Climate Conditions. In Proc. Thermal
Performance of the Exterior Envelopes of Whole Buildings X International Conference.
Clearwater Beach, Florida, p. 9.
Stopp, H. et al., 2001. The hygrothermal performance of external walls with inside insulation. In
Thermal Performance of the Exterior Envelopes of Buildings VIII. Clearwater Beach, Florida.
Toman, J., Vimmrová, A. & Černý, R., 2009. Long-term on-site assessment of hygrothermal
performance of interior thermal insulation system without water vapour barrier. Energy and
Buildings, 41(1), pp.51–55.

Full paper no: 8
Apartment's air-tightness in an apartment block area from the
1960s
Dan Jönsson, PhD 1

Dennis Johansson, PhD 2
Hans Bagge, PhD 3
1
Building Services, Lund University, Sweden
2
Building Services, Lund University, Sweden
3
Building Physics, Lund University, Sweden
Keywords: Air Leakage, Pressure Testing, Thermography.
Summary
It is important that the building envelope is airtight, to prevent outside air to leak into the house and
cause draught, excessive energy use and moisture convection which affects people's health and
satisfaction. Increased energy use is caused by cold outside air that blows into the building which
must be heated to the indoor temperature. Moisture convection can cause damage to the exterior walls
and roof when warm moist air condenses inside the wall and ceiling of the exterior parts that are
colder. The result for the ten apartments showed values between 0.59 to 2.29 l / (s ∙ m2exterior). In
today's Swedish building regulations, there is no particular requirements but a certain annual energy
use is required. In other regulations there are requirements in the form of a specific value. This value
should then be less than 0.3 l / (s ∙ m2exterior) according the passive house regulation in Sweden and less
than 0.8 l / (s ∙ m2exterior) in the former building regulation. This study shows that the studied multi-
family houses that were built of concrete with infill walls of lightweight concrete have a lower air
leakage than a multi-family house that was built of concrete with infill walls with wood frame.
Thermal imaging showed that a large leakage flow occurred at the junction between the outer wall
and the concrete floor slab. This is probably due to poor sealing. In an ongoing project these
apartments block will be investigated with destructive methods to help renovations and remodeling of
the large existing housing stock with the construction techniques from that time.
1 Introduction
It is important that the building envelope is airtight, to prevent outside air to leak into the house and
cause draught, excessive energy use and moisture convection which affects people's health and
satisfaction. Increased energy use is caused by cold outside air that blows into the building which must
be heated to the indoor temperature. Moisture convection can cause damage to the exterior walls and
roof when warm moist air condenses inside the wall and ceiling of the exterior parts that are colder.
Air leakage as the result of poor air tightness can be measured by a pressurization test where a fan is
controlled to keep a 50 Pa pressure difference between indoors and outdoors. The airflow through the
fan is measured and presented with a functional unit that can be for example envelope area or floor
area. The EN-13829 standard states that the resulting value should be weighted from one test with
under pressure and one with over pressure. Simultaneously with the pressurization test, it is common
practice to perform thermal imaging of the interior building envelope to find locations of thermal
bridging and air leakage.

Swedish building regulations have subsequently introduced requirements on air leakage after the 60’s.
The instruction to the building regulations of 1960 does not mention any requirement for air leakage,
but there is a text that a vapor barrier should be provided in the construction where required to avoid
buildup of moisture in the materials contained in the building structure. In the building regulation BBR
10 2002:19 there was a requirement that the leakage flow through the building envelope must not
exceed 0.8 l / (s∙m2) with the area referring to envelope area. This was replaced 2006 with a text in the
current regulation BBR 19 that the building envelope shall be designed with sufficient air tightness
regarding both energy use and moisture convection (The Swedish National Board of Housing,
Building and Planning 1).
A lot of studies have been presented investigating the air tightness of buildings, both in Sweden and
internationally. Internationally, usually, the functional area unit is the air exchange per hour at
measurements pressure. An example of a Swedish survey is Stein (2008) who examined the
uncertainty depending on leakage between apartments which is not described in the standard EN
13829, in a building built in 2001. In Stein's measurements of the air leakage at 50Pa pressure
difference, the apartments ranged between 0.74 and 2.70 l / (s ∙ m2) with an average of 1.53 l / (s ∙ m2)
referring to the exterior envelope area, and between 0.20 to 0.69 l / (s ∙ m2) with an average of 0.43 l /
(s ∙ m2) referring to the interior apartment envelope area. The leakage flow per apartment volume was
between 0.21 and 0.74 l / (s ∙ m3) which are equivalent to an air exchange of between 0.77 to 2.67 /h
with an average of 1.66 ACH. Leakage flow per apartment floor area was between 0.59 and 1.97 l / (s ∙
m2) with an average of 1.23 l / (s ∙ m2).
An example of an international investigation is Kalamees (2006) who examined 32 new houses built
in the years 2000-2001 in Estonia, of which only 41% passed the Estonian criterion of 3 m3 / ( h ∙ m2)
with the function denominator area as envelope area. The leakage depended largely on leaks at holes
of the air barrier and at junctions between ceilings and walls, and at junctions between exterior walls
and interior walls. For the nine one floor houses the average resulting value was 1.9 m3 / (h ∙ m2). The
23 houses that had professional supervision during the construction got a average value of 3.0 m3 / ( h ∙
m2) and the 15 houses with prefabricated wall or volume elements got an average value of 2.9 m3 / (h ∙
m2). Kalamees had an average value of 4.2 m3 / (h ∙ m2) for all his measurements.
The existing measurements of air tightness in buildings were generally performed in new buildings.
To meet the future energy requirements of the society, there is a need for extensive energy measures
and renovation in existing buildings. There is a lack of measurements of air tightness in old, existing
buildings. In this study ten apartments were measured to add to the knowledge of deviations between
apartments and the value and location of the leakage for houses built in the 1960’s that are due to be
renovated soon, and prior to such renovation, knowledge of shortcomings is important.
2 Method
For the pressurization tests, according to the standard EN 138 229, all air inlets and devices and other
intentional openings through the building envelope must be sealed. Before the measurements were
carried out, all supply and exhaust devices were taped and, sewage traps were filled with water and all
doors and windows were closed. A Minneapolis Blowerdoor with fan model 4 was used. The fan was
mounted in an outgoing balcony door. The fan was mounted in a Blowerdoor custom canvas that was
stretched on adjustable aluminum rails with a rubber strip on the side facing the door frame to get it
completely tight in the opening.
A pressure meter of model DG-700 measured the pressure difference between indoor and outdoor. A
pc application, Tectite Express 3.1, controled the fan to obtain a set point pressure difference and
another pressure meter gives the airflow through the fan. The standard specifies a number of set point
pressure differences that will be tested up to 50 Pa.

The pressurization tests were carried out simultaneously with the thermal imaging with a Flir
ThermoCAM 45 both before and during pressurization testing. By thermal imaging before
pressurization and during under pressure indoors, the combination will make it possible to specify
whether a cold spot is due to thermal bridging or cold air leaking in. Notes on the walls were used to
compare cold spots before and after pressurization. When positive and negative pressure tests were
carried out, the fan was set to 50 Pa for about 15 minutes so that cold outdoor air could cool down
surfaces at locations with air leakage.
The apartments were located in Kiruna, northern Sweden, 500 meters above sea level and about 200
kilometers north of the Arctic Circle. Normal winter is seldom colder than -20 °C and the period of
snow extends from late October to late May (Kiruna a, 2013). Predominant wind direction is between
south and west, while the winds that give large amounts of snow and rain along with strong winds
often come from the west or northwest (Kiruna b, 2013). Average annual temperature is -1.6 °C,
average annual vapor content is 3.7 g/m3 and the mean annual relative humidity is 76% (Elmarsson
and Nevander , 2006). The residential area where the test was performed was built in the 1960s. It had
natural ventilation hydronic heating with a boiler in a particular building. The residential area was
built with four groups of houses surrounding a recreation area in form of a lawn with integrated
playground. Each group of houses consists of three three-floor buildings and one two-floor house. The
two-floor houses had lightweight concrete walls on bottom floor, gables and infills. At the second
floor the end walls and the apartment separation walls were of lightweight concrete, while the
remaining walls were built with wooden frame. All walls in the three-story buildings were of concrete,
except for non-supporting infill walls were of lightweight concrete while the infill walls above and
below the window section in the living room were built with a wooden frame. All concrete floor slab
was built of concrete. On top of the attic floor a low-sloped roof was built.
3 Results
During the testing day the weather was sunny, the wind light and the temperature -6 ° C. The
pressurization test showed that apartment number seven which is an apartment on the second floor of a
two-floor house, has an airflow at 60 Pa of 499 l / s, compared to the nine other apartments that have a
maximum airflow of between 82 and 159 l / s. The high value of apartment seven was due to a
construction defect close to the shaft penetration to the attic space. Because of this large leakage in
apartment seven, its result was excluded from further analysis. However, it is included in Table 2 to
show the difference in airflow. Table 2 shows the leakage flows normalized to quantities depending on
the different surfaces and volumes in the apartments. This is to compare the apartments with each
other but also with the results of other studies. Table 1 gives the quantity description used in Table 2.
Table 1 Quantities of Table 2
Quantity Unit Description

V50 l/s Leakage airflow
Q50 l/(s∙m2) Leakage airflow per apartment envelope area
Q50,ext l/(s∙m2) Leakage airflow per apartment exterior envelope area
Q50,V l/(s∙m3) Leakage airflow per apartment volume
W50 l/(s∙m2) Leakage airflow per floor area
n50 m3/(h∙m3) Leakage air Change rate
Aext/AO m2/m2 Ratio between apartment exterior envelope and envelope area
B50 l/(s∙m) Leakage airflow per concrete floor slab edge length
The study did not measure airflow to adjacent apartments which makes it impossible to compare the
survey results with the requirement from the former building regulation, which specified airflow per
envelope area for the entire building. Leakage airflow in apartments can also enter adjacent
apartments. The flow at 50 Pa per exterior envelope area (Q50, ext) varies between 0.59 (APT 8) to 2.29

(APT 3) l/s. APT 7 had a leakage of 1.43 l / (s ∙ m2) for Q50 while the other apartments in the test have
values ranging from 0.28 to 0.45. This gives a ratio of 3 between APT 7 and APT 5 which have the
highest value of the other. However, the difference is not as extensive for the leakages through the
building envelope as APT 7 is located on the top floor of two-floor houses and has an exterior surface
of 60 percent of the enclosing surface. It gives a ratio of 1.05 (2.4 / 2.29) between APT 7 and APT 3
which have the highest value of the other.
Table 2 Air leakage when tested with a differential pressure of 50 Pa. In the column to the right of
each quantity its placement in the ranking is given. The largest value is ranked number 1.
APT V50/(l/s) Q50/(l/(s∙m2)) Q50,ext/(l/(s∙m2)) Q50,V/(l/(s∙m3))

1 117,50 4 0,42 3 0,86 8 0,51 3
2 119,17 3 0,40 4 1,76 5 0,49 4
3 72,22 9 0,35 6 2,29 1 0,48 5
4 85,00 7 0,30 8 1,79 4 0,39 8
5 132,78 2 0,45 1 1,86 3 0,54 2
6 133,33 1 0,44 2 1,60 6 0,55 1
8 78,06 8 0,28 9 0,59 9 0,35 9
9 116,11 5 0,39 5 1,98 2 0,48 6
10 93,33 6 0,32 7 1,27 7 0,42 7
Max/Min 1,85 1,61 3,88 1,57
7 438,89 1,43 2,40 1,77
APT W50/(l/(s∙m2)) n50/(m3/m3) (Aext/AO)/(m2/m2) B50/(l/(s∙m))

1 1,28 3 1,85 3 48% 1 6,48 1
2 1,22 4 1,76 4 23% 6 4,40 7
3 1,21 5 1,75 5 15% 9 5,09 2
4 0,98 8 1,41 8 17% 8 4,48 6
5 1,35 2 1,95 2 24% 5 4,66 4
6 1,36 1 1,96 1 27% 3 4,01 8
8 0,88 9 1,27 9 46% 2 4,56 5
9 1,19 6 1,72 6 20% 7 4,95 3
10 1,06 7 1,52 7 26% 4 3,18 9
Max/Min 1,54 1,54 3,2 2,04
7 4,42 6,37 60% 13,07
B50 / (l/(s·m))
APT 1
7
6 APT 3
5 APT 4 APT 8
APT 10
4
APT 2 APT 5 APT 9
3 APT 6
2
1
0
0 20 40 60 80 100
Ratio of apartments below / %
Figure 1. Leakage airflow B50 per concrete floor slab edge length.
Figure 1 shows apartments’ leakage airflow per concrete floor slab edge length to the outdoor sorted in
order. The apartments belonging to two-floor buildings is marked blue, italic, and the label is

positioned under the curve while the apartments from the three-floor houses are marked red, bold and
the tag is located above the curve.
Figure 2 gives the leakage airflow parameters Q50, Q50, ext and n50. The values are sorted. APT 8 had
the lowest leakage by 0.59 for Q50, ext and 0.28 for Q50 and minimum air change rate of 1.27 for n50
while it is different for the other apartments. APT 3 had the highest leakage of 2.29 l/ (s∙m2) based on
on exterior envelope area while APT 5 had the highest leakage based on the envelope area with an
airflow of 0.45 l / (s ∙ m2). Based on the envelope area, the apartments have a flow of 0.28 to 0.45
which means that there is a ratio of 1.6 between them, while based on the exterior envelope area, the
airflow was 0.59 to 2.29 l / (s ∙ m2) with a ratio of 3.9. APT 8 had the lowest air change rate of 1.27 /h
and APT 6 had the highest of 1.96 /h, a ratio of 1.54 regarding n50. APT 2, 5, 6 and 9 in the two-floor
houses generally had higher leakage airflow than the apartments 1, 3, 4, 8 and 10 in the three-floor
houses. This applies to both Q50 and Q50, ext.. The ratio between the highest and lowest values was 3.9
for Q50, ext, 1.5 for n50 and 1.6 for Q50. The ratio of external wall area was lower for the apartments that
had adjacent apartments on both sides and above and below compared to a gable apartment at the
highest floor, which leads to a rather high difference in dynamics between Q50, ext and Q50.
Q50 / (l/(s·m2); Q50,ext / (l/(s·m2) n50 / (m3/m3)

2,5 APT6 2,5
APT9
2 APT2 APT4 APT5 2,0
APT9
APT10 APT5 APT6
1,5 APT8
APT6 APT3 APT2 APT1 1,5
APT6
Q50
APT10
1 1,0 Q50, ext
APT1
N50
0,5 APT8 0,5
APT2 APT1 APT6 APT5
APT8 APT4 APT10 APT3 APT9
0 0,0
0 10 20 30 40 50 60 70 80 90 100
Ratio of apartments / %
Figure 2. Leakage airflow parameters for the apartments.

APT 1 had the maximum leakage airflow of 6.48 l / (s∙m) regarding concrete floor slab edge length.
APT 10 had the lowest of 3.18 l / (s ∙ m). This gives a ratio of 2.04 regarding B50, max/B50, min. Much of
the leak seemed to occur in the junctions between building elements in the apartments which Figure 3
indicates, since the leakage airflow increased with the concrete floor slab edge length in the
apartments. APT 3 had 14 m concrete floor slab edge length and a leakage airflow of 72 l / s, which is
the smallest in the range. APT 6 had the longest concrete floor slab edge length and the maximum
leakage flow. Two points do not follow the indicated correlation line of the curve. APT 1 had a
concrete floor slab edge length of 14 m and a leakage flow of 117 l / (s∙m). APT 10 had a concrete
floor slab edge of 29 m and a leakage flow of 93 l / (s∙m), both apartments in three-floor houses.
Figure 4 shows that the air change rate in the three-floor houses are overall larger than in the two-floor
houses.

v50 / (l/s)
140,00 APT5 APT6
APT9 APT2
APT1
120,00
100,00 APT10
APT4
APT8
80,00 APT3
60,00
40,00
20,00
0,00
0 5 10 15 20 25 30 35
Concrete floor slab edge/ (m)
Figure 3. Leakage airflow as a function of concrete floor slab edge length.
n50 / (m3/(h·m3) )
2,50
APT5 APT6
2,00 APT9 APT2
APT3 APT1
1,50
APT4 APT10
APT8 Two storey houses
1,00
Three storey houses
0,50
0,00
0 20 40 60 80 100
Ratio of apartments %
Figure 4. Air change rates.

The results of thermal imaging indicated leakage in the junctions between building elements. Figure 5
and 6 shows the outer wall of the bedroom in APT 1 where it is suspected that the seal between the
wall elements and the concrete slab are not in order.
41.7°C

14.3°C
Figure 5. Left side of the bedroom in APT 1 towards outdoors.
44.2°C
14.7°C
Figure 6. Right side of the bedroom in APT 1 towards outdoors.
4 Discussions and Conclusion

For the purpose of increasing the available reference material and showing variations in apartments in
old, existing buildings and analyze where such leakages occur, this study measured leakage from ten
apartments with the help of thermal imaging to distinguish between thermal bridging and leaks. The
result from the measurements of leakage in ten apartments with one excluded due to extremely high
leakage, are between 3.2 and 4.9 m3 / (h∙ m2exterior) with an average of 4.2 m3 / (h∙ m2exterior) to be
compared with Kalamees (2007) who found an average of 4.2 for all their measurements. Kalamees
however, had 41 percent below the 3 m3 / (h ∙ m2 exterior). The rather high variation between different
apartments indicates a need to measure in many apartments when buildings are verified.
The comparison with Stein (2008) is interesting because the investigated building constructed in the
south of Sweden in 2001 have approximately the same intensive leakage airflows as the buildings in
the present study which were built about 40 years before. In comparison, Q50 and W50 were even
slightly lower for the 40 years old apartments. For Q50, ext and W50, the results are close.
According to the thermal imaging results, leakage occurred especially along the joints between the
wall and floor and roof. Here, it appeared that thermal bridging and leakage points were the same in
the apartments for the entire apartment stock in the studied residential area. This suggests that either
the structure had been degenerated over time or that the mounting of the building elements were
poorly done as it was not now sealed in a proper way. It was clear that the seal between the floor slabs
and wall light weight concrete elements was not satisfactory and more or less leaked along the entire

element length. The present study as well as the study by Kalamees (2008) indicates that much of the
leakage occurs along the concrete floor slab edges. The results from the studied apartments built in the
1960s when the air tightness was not regulated by building regulations give values close to the much
newer building studied by Stein (2008). There was no special air sealing in the buildings. Gable walls
of concrete and lightweight concrete are supposed to be airtight if there are no cracks, but infill walls
built by wooden frame lack air barriers. The presently studied buildings were located in an area where
cracks can be suspected due to mining below the buildings, which can have influenced the leakage but
reasonably only towards higher values.
The measurements were performed in apartments without pressurization of adjacent apartments. There
are several reasons for using this method that was also used in the referred literature. It is more
economical to perform and there are reasons such as smell, fire, heat transfer and rebuilding to avoid
leakage through the internal walls and therefore there are reasons for having requirements on
apartment level. The different presented parameters show different results. N50 does not normalize for
the size of the apartment, which Q50 and Q50, ext do. The shown difference between Q50 and Q50, ext
indicates both that the ratio between them varies and that there is a leakage through internal walls that
are in the same magnitude as the external walls. Even if the result from this study is a help for
determining where to put efforts in a renovation, it would be interesting to extend the measurements
and pressurize adjacent apartments to determine the flow patterns though different internal walls,
which will be a matter of future research.
5 Acknowledgements
This study is funded by the Research Council Formas.
6 References
Stein J. (2008). Air tightness in multifamily - measurements and analysis. THID-08/5098 (In Swedish)
Kalamees T. (2007). Airtightness and air leakages of new lightweight single-family detached houses
in Estonia. Building and Environment, 42 (6), pp2369 – 2377
Nevander, L.E. and Elmarsson, B. (2006). Moisture Manual - practice and theory. ISBN 9173331562
Stockholm: Swedish Building Centre Ltd
Kiruna a 2013, Climate. [www]. Retrieved from <http://www.kiruna.se/kommun/Kommun-
politik/Kommunfakta/Geografi/Klimat/>
Kiruna b 2013, Event Handling new city hall. [www] . Retrieved from
<http://www.kiruna.se/PageFiles/6614/134368%20t%C3%A4vlingsprogram_uppslag_low.pdf?epslan
guage=sv>
The Swedish National Board of Housing, Building and Planning 1 2013,
<http://www.boverket.se/Bygga--forvalta/Regler-om-byggande/Boverkets-byggregler-BBR/>

Full paper no: 9
On the coupling of a zonal model with a multizone building
energy simulation model
Lien De Backer, M.Sc 1
Jelle Laverge, Ph.D 1
Arnold Janssens, Professor 1
Michel De Paepe, Professor 2
1
Ghent University, Department of Architecture and Urban Planning, Belgium
2
Ghent University, Department of Flow, Heat and Combustion Mechanics, Belgium
KEYWORDS: Building simulation, vertical temperature distribution, coupling, multizone model,

verification
SUMMARY:
The conditions in one zone are often treated as ‘fully mixed’ in Building Energy Simulation programs
(BES) e.g. TRNSYS. By contrast, Computational Fluid Dynamics (CFD) are too complex and time-
consuming to predict temperature and relative humidity in a building for a longer time period, such as
one year. Nevertheless, several building applications require a prediction of the vertical distribution
of the indoor relative humidity and temperature under transient boundary conditions. This paper
presents the development and possibilities of a coupled TRNSYS-zonal model, which allows
accounting the stratification in a room. The zonal model is based on the so-called block model
proposed by Togari[1] and describes a one-dimensional heat and mass transfer between horizontal
layers. The goal of this preliminary study was to examine the effect of parameters such as the number
of layers, the value of the heat convection coefficient and the value for the time step. To validate the
model, the case of natural convection in a test room (3m x 3m x 2.5m ) described by Arai and Togari
has been used .
1. Introduction
‘Energy use’ is a concept which we could not ignore nowadays: every day we are flooded with
information and publicity about energy saving measures and cutting down on power consumption.
However, when dealing with historical buildings, architectural and deontological criteria e.g., the rules
of the Flemish monumental Guard complicate the idea of sustainability. Due to increased thermal
comfort expectations from churchgoers, in recent decades heating systems have been installed in the
historical church buildings. Even today the demand for new or additional heating is still large. So
despite the idea of energy saving and despite the fact that a church building represents a large volume
to be heated, one can ask whether heating a church is ecologically and financially wise.
There also is the additional problem that a lot of churches contain valuable and historic pieces of art.
Hence, the heating systems, which often only operate during service, significantly alter the
microclimate in the church and these fast-changing indoor conditions may result in a faster
deterioration of the artworks. To avoid the damage of the artworks, guidelines can be found in
literature for the acceptable indoor temperature and relative humidity variations (Anon 2011). To
reconcile these preservation needs for the artworks with the heating demand, computer simulations
provide an alternative for experiments and measurements. Because heating often leads to stratification
in this type of buildings, the simulation study must correctly estimate the stratification. Therefore, the
calculation of the airflow in the space is also necessary, next to the calculation of the energy exchange.
To predict this temperature and humidity distribution in building, different modelling approaches have
been used; namely the CFD-method and a zonal airflow model. The CFD-method is a widespread
approach to simulate the airflow in a building. The models based on the CFD-method predict

accurately the temperature, velocity and other flow parameters by the Navier-Stokes equations.
However, in order to assess the damage risk of art works such as wooden panel paintings, the
fluctuations in temperature and humidity over a longer time period are of interest. To simulate this, the
mentioned CFD models are less suitable for their need for powerful computers with a large amount of
memory, even for a short-time simulation. In reality, these requirements are often not available. So to
be able to predict the airflow in a building in a fast way and for a longer time period, the use of
macroscopic airflow models offers a solution. In literature, several macroscopic airflow models can be
found. An overview of the most common models is given by Megri and Haghighat (Megri and
Haghighat 2007). The different macroscopic airflow models differentiate themselves by the
simplifications they make in the conservation equations. A suitable zonal model to predict the
temperature distributions in a large space is the so-called temperature zonal model proposed by Togari
(Togari et al. 1993), which is based on temperature differences (Heiselberg et al. 1998).
The goal of this paper is to develop a suitable dynamic simulation tool to predict the temperature and
humidity distribution in a church building. To this end, the authors expanded the Togari block-model
with mass conservation equations for water vapour and integrated this zonal model into the BES-
software TRNSYS v17.
2. Simplified Modelling of the vertical temperature and relative humidity

distribution
The model selected for this work was first proposed by Togari et al.(Togari et al. 1993) and has been
used by others (Wang et al. 2009, Arai et al. 1994, Takemasa et al. 1996). This model is a simplified
model for calculating the vertical temperature distribution in a large space building. To calculate the
temperature distribution in the room the model starts from a given inside wall temperature. The
authors assumed the supply airstreams and the airflows along the vertical wall surfaces to be the main
components of the air movement in the large space. Further, it was also assumed that the horizontal
temperature was uniform, except for the regions affected by supply air jet ventilation. This zonal
model would be more useful when it could be used in an energy simulation, starting from other
boundary conditions than the wall temperature, e.g. given outside conditions. Therefore the zonal
model was implemented in an existing BES-software. To do so, the software TRNSYS was used.
2.1 Governing equations of the zonal model

In the zonal model, the space is divided into a finite number of horizontal layers or blocks. Each layer
consists of a core cell and wall cells, as displayed in
FIG 1. The core cell represents a horizontal layer and when the layer is bounded by a wall, a wall cell
is defined which accounts for the mass flow along the wall. This wall cell is a fictive cell and has no
dimensions. Flows are defined as positive in the upward direction and from the wall cells to the layers.
There is an exchange of mass and energy between a “wall cell” and a “zone cell”, but, the
conservation of momentum in the zone is not considered (Schlichting and Gersten 1979) because the
velocity in the “zone cell “ is assumed to be very low.
2.1.1 Conservation of mass

The mass balance for a layer i is expressed as:
0 = ms + me + mcur + mlay (1)
Where ms net air mass flow from a source or to a sink, kg/s

me net air mass flow from a jet, kg/s
mcur net air flow between the wall current and the zone, =min(i,K) − mout(i,K) , kg/s
mlay net air flow to the layer underneath (mi−1) and above( mi ), kg/s
mout(I,K) mass of flow from layer i to wall current of wall K, kg/s

min(I,K) mass of flow from wall current of wall K to layer i, kg/s
The mass transfer along the wall is modelled using a wall current model based on the boundary layer
theory for natural convection for a vertical wall (Schlichting and Gersten 1979). The model assumes
that heat convection drives mass flow mout(I,K) with an average temperature TD(i,K) from layer i to its
related boundary layer (
FIG 1). To calculate mout(I,K) with temperature TD(i,K), the following assumptions were made:
TD(i,K) = 0.75Ti + 0.25Tw(i,K) (2)
αC(i,K) .Aw(I,K)
mout(i,K) = 4 Ci
(3)
Where Tw(i,K) the average temperature of the wall K adjacent to layer i, °C

Ti the average temperature of the layer i, °C
TD(i,K) the average temperature of the mass mout(I,K) , °C
αC(i,K) heat convection coefficient, W/m²K
Aw(I,K) area of wall K adjacent to layer i, m²
Ci specific heat of the air in layer i, J/kgK
When the currents are descendent, as shown on

FIG 1, the currents flow from wall cellK,i+1 to wall cellK,i and will combine with the current flow from
layer i to form a total flow with mass 𝑚𝑚(𝑖,𝐾) and an average temperature 𝑇𝑚(𝑖,𝐾) , yielding:
mm(i,K) = mout(i,K) + mmd(i+1,K) (4)
mout(i,K) TD(i,K) +mmd(i+1,K) Tm(i+1,K)

Tm(i,K) = mm(i,K)
(5)
Where mm(i,K) the mass of the wall current of wall K adjacent to layer i, kg/s
mmd(i+1,K) the mass of the wall current of wall K adjacent to layer i+1 or i-1 to layer i, kg/s
Tm(i,K) the temperature of the wall current of wall K adjacent to layer i, °C
Some of the air returns to the air layer i (𝑚𝑖𝑛(𝑖,𝐾) ) and some continues to the cell down/up(𝑚𝑚𝑑(𝑖,𝐾) ).
The splitting of the mass 𝑚𝑚(𝑖,𝐾) into 𝑚𝑖𝑛(𝑖,𝐾) and 𝑚𝑚𝑑(𝑖,𝐾) is calculated by the ratio P(i,K).
2.1.2 Conservation of vapour

The moisture balance equation for a layer i can be expressed as:
(Yi t+∆t,m − Yi t ) (6)
ρa Vi = Glay + Gwalls + Gcurrents + Gs
∆t
Where Glay the mass flow rate of water vapour from layer i-1 and layer i, kg/s
= max(mc,i−1 , 0) Yc,i−1 + min(mc,i−1 , 0)Yc,i − max(mc,i , 0)Yc,i − min(mc,i , 0)Yc,i+1
Gwalls the mass flow rate of water vapour at the wall surface K, kg/s
Gcurrents the mass flow rate of water vapour of the wall current of wall K adjacent to layer
i, kg/s, = Yi,K min(i,K) − Yi mout(i,K)
Gs vapour flow produced by people, systems, activities such as washing, …
m number of the iteration step

2.1.3 Conservation of energy
In the air the heat transfer equation can be written as:
t
(ρC)t+∆t,m Ti t+∆t,m −(ρC)t Ti
Vi
∆t
= Qs + Qlay + Qcur + Qb (7)
Where Vi volume of layer i, m³

Qs heat sources or sinks, W
Q lay the heat flow due to mass transport from layer i with layer i-1 and layer i+1, W
Q cur the heat flow from or to the wall currents, W
Qb the heat flow from inversion between layers expressed as
qb= Cb Ab (T(i−1) − T(i) ), W
Cb the value obtained when a stable (2.3W/m²°C) or unstable (112W/m²°C)
temperature stratification is formed
𝐴𝑏 cross section area of the top or bottom layer, m²
1 layer: height dependent

on the number of layers
Wall cell: fictive cell

without dimensions
FIG 1. Schematic representation of the wall currents and air flows in one layer
2.2 Coupling of the zonal model with TRNSYS

TRNSYS (Solar Energy Laboratory 2010) is a TRaNsient SYStems Simulation program with a
modular structure. Next to its own library of components, this software allows for implementing an
own written component as a dynamic link library in the simulation environment, which is called ‘the
Simulation Studio’ (SEL 2013). The multizone building model of TRNSYS is represented by ‘Type
56’ and calculates the heat and moisture balance of a building for given boundary conditions such as
temperature, solar radiation, heat and moisture gains. The building can contain different zones. Each
zone represents one or more rooms in which T and RH is typically assumed to be well-mixed.
The zonal model is written as a new component for TRNSYS v17. This version of TRNSYS has the
possibility to divide one zone into several air nodes (Solar Energy Laboratory 2012). Each air node of
the building model must be connected with a layer in the zonal model. FIG 2 shows a schematic
overview of the procedure used to couple the thermal zonal model with the building energy simulation
program TRNSYS. The interior surface temperature (si) and the heat transfer coefficient (his) act as
boundary conditions for the zonal model. Initial values of the temperature and relative humidity from
each air node are used as an input of the zonal model, together with the vapour mass flux of the walls.
This can be done by the implementation of a HAM-model for TRNSYS(Steeman 2010) or by using a
simplified model provided by TRNSYS.

Once the boundary conditions are given to the zonal model, the latter calculates the wall currents and
the heat and mass conservation equations for all the layers until convergence is reached. As a
convergence criterion, the maximum temperature difference in the layers between successive
iterations is chosen. Afterwards, the net heat and moisture gain for every layer is passed on to the
multizone building model (TRNBuild). So the coupling of the BES-software with the zonal model
consists of a convective heat and water vapour gain for every air node in TRNSYS calculated by the
zonal model. Looking closer to the heat conservation in TRNSYS, TRNSYS calculates in every air
node a convective heat gain. To avoid to count in twice this heat gain (one by the wall currents in the
zonal model and one by TRNSYS), the convective surface heat flows calculated by TRNSYS may not
be taken into account. Therefore, this value also has to be passed to the zonal model, in which it will
be taken into account. The radiative heat gain, which is decoupled of the convective heat gain, will be
calculated by TRNSYS.
INPUT Coupling OUTPUT
ZONAL Model
- Initial values: T, RH -additional
- Twall wall currents convective heat gain
-Vapour flux of the Mass balance -additional moisture
wall Twall Temperature balance Heat gain gain
Tair node Moisture
RHair node gain
Qconv,air node
TRNSYS
-Geometric model Air node heat and Air node temp
-Boundary conditions moisture balance Air node RH
-radiation model
FIG 2. Schematic overview of the coupling between the multizone building model in TRNSYS and the
developed thermal zonal model.
3. Initial results
3.1 Test case
The case studied in this paper was the case that can be found in the report of Togari et al. (Togari et al.
1993) and that of Arai et al.(Arai et al. 1994) which was used for evaluating the implemented model.
The geometrically simple test room had a ground plane of 3m x 3m and measures 2,5m in height. The
room consisted of insulation boards (three vertical walls, a ceiling and a floor) and one glass wall. In
the wall opposite to the glass wall, two openings were foreseen in the symmetry plane: a supply inlet
at 0,625 m above the floor and a return outlet at 0,250m above the floor. Several configurations were
measured in this test room e.g. air heating or cooling and natural convection.
To validate the calculated results of the newly implemented model, the natural convection case (N10)
has been used in this paper. In this case, the temperature outside the room was kept at about 12°C and
was then raised to 42°C, while 24 hours later it was again lowered to 12°C. The response of the test
room was monitored, while no air was supplied into the room. The moisture transfer between the air
wall surface and the air was not considered in this case study.
The first calculation with the zonal model was a steady-state calculation of the layers temperature
using the measured interior surface temperatures at two hours of testing as boundary conditions.

Measured surface temperatures according to TABLE 1 were used together with constant heat transfer
coefficients. The coefficient was 3.5 W/m²K for the glass wall and the insulated wall. For the floor
and the ceiling, the coefficient was 4.6W/m²°K when heat flow was upward, and 2.3W/m²K when the
heat flow was downward. The calculated temperatures are given in TABLE 1. These results show that
the average room temperature was slightly overpredicted. This could be due to the constant heat
transfer coefficients and the use of a Cb coefficient for stable or unstable stratification.
TABLE 1.Measured temperature for the wall and glass surface and the zone temperature at different
heights (0,25m – 0,75m – 1,25m 1,75m 2,25m) in the test case presented by Togari and the calculated
temperature in the zonal model.
height Tzone0.25m Tzone0.75m Tzone1.25m Tzone1.75m Tzone2.25m Twall Tglass Tcalculated

(m) [°C] [°C] [°C] [°C] [°C] [°C] [°C]
0.25 16.7 16.9 17.1 17.4 17.6 16.3 28.4 17.4

0.75 17.4 17.5 17.6 18 18.5 17.2 27.7 18.8
1.25 18.2 18.1 18.4 18.9 19.1 17.9 27.6 19.5
1.75 19.9 19.2 19.2 19.4 19.7 18.1 26.2 20.1
2.25 19.5 19.5 19.9 20.1 20.1 17.9 27.2 20.8
The next step was to perform a dynamic calculation taking transient terms in eq.6 and eq.7 into
account. For that, 48 hours were simulated with coupled TRNSYS - zonal model. The goal of this
study was to examine the effect of physical and numerical parameters such as the thermal resistance of
the walls, the value of the heat convection coefficient, the number of layers and the value for the time
step. In this case, the full test room was modelled.
Because material parameters were unknown, several compositions for the walls were tested. Good
agreement was found for the thermal resistance R=1.6 m²K/W for the insulated walls and with the
glass wall as a massless layer. However, the surface temperature of glass was slightly under predicted.
On FIG 3, results were depicted for the first seven hours in which outdoor temperature was raised
from 12 to 42°C. The measured outdoor temperature served as boundary conditions, while the
temperatures for the insulated walls and the glass surface were calculated by TRNSYS. This was
opposite to the calculated results of Togari in which all interior surface temperatures at each time step
(one hour) were given by the experiment (steady-state calculation for every hour). The calculated
results were compared to the measurements and the calculations of Togari (FIG 3). The zone
temperature was lower than was seen in the calculations of Togari, which was probably due to the
lower surface temperatures of glass and the dynamic character of the calculation (Togari adapted the
boundary conditions every hour based on his measured results).
On FIG 4, the effect of the number of layers on the zone temperature and relative humidity was
visualised. The temperature and the relative humidity in the zone represented the values for whole the
volume. The more layers, the more the stratification was detailed. Further, it could be noticed that the
number of layers has an impact on the zone temperature and the surface temperature of the floor and
the ceiling. The more layers, the higher the temperature and the lower the relative humidity became. In
future, further research needs to be done on the allowed maximal number of layers. Next, the effect of
the time step has also been investigated, which had no effect on the calculated results, and the value of
the heat convection coefficients (FIG 5). In the first calculation, heat transfer coefficients were
considered as simple constant values as described earlier. In the second calculation, heat transfer
coefficients were adapted for every time step based on the temperature in the layer. In first instance

the algorithms for natural convection proposed by TRNSYS were used. Using these algorithms led to
a slower heating up and a slower cooling down.
2,5
Height above floor level [m]
measured
2
1,5
zonal model
1 + TRNSYS
0,5 Model
Togari
0
10 15 20 25 30 35
Temperature [°C]
FIG 3:Comparisons between calculated values, measured values and the calculations of Togari in
case where outside was heated for every hour, during seven hours(N10).
2,5 2,5
L3
Height above floor level [m]
2,0 2,0
L5
1,5 1,5
L7
1,0 1,0
L10
0,5 0,5
0,0 0,0
30,0
32,0
34,0
36,0
38,0
40,0
16,0
16,5
17,0
17,5
18,0
18,5
19,0
19,5
20,0
20,5
21,0
T[°C] RH[%]
FIG 4: Comparison between the calculated temperature and relative humidity after two hours of
heating for a different number of layers.
45
40 Outside temperature
Temperature [°C]
35
Ti 0,01 H fixed
30
25 Ti 0,15 H fixed
20
Ti 0,01 H TRNSYS
15
10
4
0
12
16
20
24
28
32
36
40
44
48
Time [h]
FIG 5. Comparison between the calculated values with a time step of 0,01h and 0,15h and between a
constant heat transfer coefficients and the heat transfer coefficients calculated by the algorithms from
TRNSYS
4. Conclusion and further work

Currently the thermal-zonal model is operational in the version of TRNSYS v17. This version of
TRNSYS has the possibility to divide one zone into several air nodes. For each air node, the zonal

model defines a convective heat gain and a gain in vapour mass. To validate the zonal model, the case
of natural convection described by Togari was used. The initial testing has been positive in order of
the prediction of the temperature stratification. However, some fundamental work still must be done.
As shown, the results are dependent of the choice of heat transfer coefficient and the number of layers.
Also the choice for the Cb coefficients will have an impact. For these parameters, a more detailed
sensitivity analysis will be done. Further must it be kept in mind that the model is a simplification, in
which some assumptions were made such as the value of the Cb coefficient. Also, if the number of
layers increase, one can question the physical background accepted in the simplifications. So the
allowable number of layers and the choice of the Cb coefficient need further study.
In the next stage the model will also be further developed. The equations for the jet flows will be
implemented and there will also be looked at the possibility to combine the model with TRNFLOW in
order to make more complex models.
References
Anon (2011) 'ASHRAE handbook: Heating, ventilating, and air-conditioning applications, SI edition'
in, American Society of Heating, Refrigerating and Air-Conditioning Engineers, 23.1-23.22.
Arai, Y., Togari, S. and Miura, K. (1994) 'Unsteady-state thermal analysis of a large space with vertical
temperature distribution', ASHRAE Transactions, 100(part 2), 396-411.
Heiselberg, P., Murakami, S. and Rulet, C.-A. (1998) Annex 26: Ventilation of large spaces in
buildings. Part 3. Analysis and Prediction Techniques, Denmark, Aalborg, Aalborg
University
Megri, A. C. and Haghighat, F. (2007) 'Zonal Modeling for Simulating Indoor Environment of
Buildings: Review, Recent Developments, and Applications', HVAC&R Research, 13(6), 887-
905.
Schlichting, H. and Gersten, K. (1979) Boundary-Layer Theory, Seventh Edition ed., New York and
London: MacGraw-Hill.
SEL, U. o. W.-M. (2013) 'A TRaNsient SYstems Simulation Program TRNSYS 17', [online],
available: http://sel.me.wisc.edu/
Solar Energy Laboratory, U. o. W.-M. (2010) TRNSYS 17: A Transient System Simulation Program,
Solar Energy Laboratory, U. o. W.-M. (2012) Mult izone Building modeling with Type56 and
TRNBuild, Volume 5, Solar Energy Laboratory, University of Wisconsin-Madison.
Steeman, M. (2010) Hygrothermal modelling for building energy simulation applications, unpublished
thesis Ghent University.
Takemasa, Y., Togari, S. and Arai, Y. (1996) 'Application of an unsteady-state model for predicting
vertical temperature distributions to an existing atrium', ASHRAE Transaction, 102(part 1),
239-247.
Togari, S., Arai, Y. and Milura, K. (1993) 'A Simplified Model for Predicting Vertical Temperature
Distribution in a Large Space', ASHRAE Transaction, 99(part 1), 84-90.
Wang, X., Huang, C. and Cao, W. (2009) 'Mathematical modeling and experimental study on vertical
temperature distribution of hybrid ventilation in an atrium building', Energy and Buildings,
41(9), 907-914.

Full paper no: 10
A heat-airflow model for simulating the effects of air leakage on
the temperature field in porous insulation
Clément Belleudy, Ph.D. Candidate 1,2,3
Ahmad Kayello, M.A.Sc. Candidate 2
Monika Woloszyn, Professor 1
Hua Ge, Assistant Professor 2
Paul Fazio, Professor 2
Marx Chhay, Ph.D. 1
Daniel Quenard, Ph.D. 3
1
LOCIE, CNRS UMR 5271, Université de Savoie, Chambéry, France
2
CZEBS, BCEE, Concordia University, Montréal, Canada
3
CSTB, Saint-Martin-d'Hères, France
KEYWORDS: air leakage, building envelope, numerical model, permeability, heat loss, defect
SUMMARY:
Air leakage through the building envelope can result in an increase in energy consumption and in
potential moisture damages. In this paper, the development of a simplified coupled heat-airflow model
is presented. The model is applied to a ceiling section insulated with blown-in cellulose, separating an
attic space from a heated indoor space. This ceiling section, part of a full scale test-hut built in an
environmental chamber, is tested experimentally with and without air leakage, where temperature
sensors in the cellulose provide a map of the temperature profile at various locations about the air
leakage point. A comparison of the experimental data and the simulation results is performed.
1. Introduction
Ineffective design and poor workmanship can both lead to a dramatic decrease in building’s air
tightness. Occupants can also inadvertently create additional openings in the building envelope after
the building is delivered. Excessive air leakage can result in significant heat losses, as well as
condensation formation and mould growth in the building envelope. Statistical surveys carried out in
France over the past decade identified the areas of the building envelope where air leakage is more
often observed (Litvak, 2005). Excluding electrical conduits and thin gaps between window frames
and the walls, airflow due to air leakage through building defects generally flows through the
insulation material. The model developed here aims to assess the effect of the airflow on the
temperature profile in porous insulation materials. Following a brief overview of the existing research
work on heat-airflow coupling, the governing equations, geometry, and boundary conditions of the
model are presented. A comparison between the experimental data and the simulation results
demonstrates the ability of the model to predict the temperature conditions in the insulation.
2. Numerical model
2.1 Current state
The influence of air leakage on the performance of porous insulation has been investigated since the
late eighties. An experiment conducted at the Fraunhofer Institute of Holzkirchen showed that a 1 mm
crack in an air membrane placed on fibrous insulating material could reduce its effective R-value by a
factor of 4.8 (Wagner, 1989). Langlais et al. (1990) presented a numerical model to assess natural

convection within horizontally placed mineral wool insulation caused by temperature gradients
between the indoor and outdoor air.
Follow-up research has focused on different wall configurations with cracks on the interior and
exterior sheathings to investigate different air paths through insulation materials (Buchanan &
Sherman, 2000; Abadie et al., 2002). Svoboda (1999) developed a 2D model which accounts for heat
and air transfers, introducing a convective linear thermal transmittance ψΔP similar to the linear heat
loss coefficient of a thermal bridge. Similar work has been done by Barhoun (2006), who developed a
simplified model of a wall subjected to air leakage (inlet at the bottom, oulet at the top) and calculates
an effective U-value accounting for this defect. More recently, several heat-air-moisture (HAM)
models have been developed, some with a uniform flow rate on the whole wall section (van Schijndel,
2008; Li et al., 2009; Tariku et al., 2010), and some modeling air fluxes through cracks between
different materials of wall assemblies (Saber et al., 2012). Moreover, Langmans (2013) developed a
2D HAM model, implemented in the software Delphin simulating airlfow entering and exiting cracks
located in a wall’s vapour/air barrier, flowing through fibrous insulation material in the wall, and
assessing its impact on the wall drying capacity.
The objective of this research is to develop a HAM model versatile enough to deal with two-
dimensional airflow through both porous media and through thin air gaps, while implementing
suitable boundary conditions with respect to the defect configuration. This could aid in development
of guidelines for modeling various permeability defect configurations and can lead to a catalog similar
to those for thermal bridges. The flow through a thin crack will be dealt the same way as through a
porous medium, using an equivalent permeability based on Poiseuille law. The first step of our work,
presented in this paper, is a heat-air (HA) model applied to a permeability defect where air enters a
porous medium via an opening in the air barrier and exits through an open interface.
2.2 Governing equations

To describe coupled air and heat transfer in porous insulation, classical conservation laws are used. All
properties are averaged on representative elementary volumes (REV) within the continuous medium
approximation (Bories et al., 2008). This allows the local mass, momentum and heat conservations
equations to be written using partial differential equations (PDE). The continuity equation on a REV is
∂ρ air
= −∇ ⋅ ( ρ air × v) which reduces to ∇ ⋅ v = 0 for incompressible flow (1)
∂t
where ρair density of dry air (kg/m3), assumed incompressible
v intrinsic velocity of air (m/s)
In the case of a porous medium, the REV contains matter in both solid and fluid states. Therefore, an
averaged velocity over this volume, called the Darcy velocity (u), is introduced. The Dupuis-
Forchheimer relationship provides the link between v and u:
u =ε ×v (2)
where ε porosity of the material
Therefore, (1) becomes:
∇⋅u = 0 (3)
The momentum conservation for laminar airflow is described by the Navier-Stokes equation, which is
highly nonlinear and often solvable only with numerical methods.
∂v
ρ air + ρ air (v ⋅ ∇) v = −∇P + µ Δv − ρ air g (4)
∂t

where P air pressure (Pa)
µ dynamic viscosity of air (Pa·s)
g gravity acceleration (m/s2)
For a flow through a porous medium, an additional drag due to the resistance of porous matrix against
the flow appears. This drag is described by a volume force in the Navier-Stokes equation. For a pore
Reynolds number of order unity or smaller and for incompressible steady-state flow, neglecting edge
effects, the momentum equation is called Darcy law, which is widely used:
µ
0 = −∇P − u (5)
kmat
where kmat permeability of the material (m2)

To deal with interfaces between a porous material and air, taking into account the shear induced
momentum transfer is required (Nield and Bejan, 2006; Vafai and Thiyagaraja, 1987). As in Navier-
Stokes equation, a Laplacian term is added. The transient Darcy-Brinkman equation is written as
follows:
ρ air $ ∂u u' µ
& + (u ⋅ ∇) ) = −∇P − u + µ Δu (6)
ε % ∂t ε( kmat
The pore Reynolds number provides information on the flow regime in the porous medium and is
expressed as
Lc ρ air u
Re p = (7)
µ air
The characteristic length (Lc) for the flow can be approximated by the square root of the material
permeability (Nield and Bejan, 2006).
Lc = kmat (8)
In building physics, the infiltration/exfiltration air velocity driven by pressure differences typically do
not exceed 0.5 m/s (Chan et al., 2003). For our experiment, given that the permeability of cellulose
insulation is 7.67 × 10-10 m2 (Trechsel, 2011) and the inlet airflow rate is 2 or 5 LPM (see section 3):
Re p = 4.8 → 1 < Re p < 10 (9)
For Rep>10, the flow regime is still laminar but the friction drag becomes quadratic as described in the
Forchheimer equation:
ρ air $ ∂u u' µ
& + (u ⋅ ∇) ) = −∇P −
−1/2
u − cF kmat ρ air u u + µ Δu (10)
ε % ∂t ε( kmat
where cF dimensionless constant

Because this transition is quite smooth, and our velocity is much smaller in the vicinity of the inlet
because of widening of the air section, we will use (6) in our model (Nield and Bejan, 2006). The
diffusion-convection equation is written in the general form:
∂h
= −∇ ⋅ (qconduction + qconvection ) (11)
∂t

Where h volume enthalpy (J/m3)
qconduction heat conduction flux density (W/m2)
qconvection heat convection flux density (W/m2)
The first step is to determine whether thermal equilibrium can be considered between fluid and solid
phases of the insulation. If not, two coupled equations with a sink and source term must be considered
to describe the heat transfer between the two phases. The temperature of the solid matrix, Ts, and the
temperature of the fluid, Tf, would be the dependent variables. For an airflow velocity of 0.08 m/s at
Tf= 0ºC passing through glass fiber insulation at 25°C, it has been proven (Buchanan and Sherman,
2000) that the air reaches the temperature of insulation at a distance of 1.5 × 10-5 m from the inlet.
This assumption is made for this study. The heat balance equation is a follows:
∂T
ρ mat cmat = −∇ ⋅ (−λmat ∇T ) − ∇ ⋅ ( ρ air c pair T u) (12)
∂t
Where ρmat density of the material (kg/m3)
cmat specific heat of the material (J/kg·K)
λmat thermal conductivity of the material (W/m·K)
3. Experimental Setup
A test hut built inside an environmental chamber at Concordia University provides the experimental
setup for this study. The test hut contains an unvented attic space insulated at the ceiling level with
380 mm of blown-in cellulose above 38 mm of rigid polyisocyanurate (PIR). The PIR also acts as the
air and vapor barrier of the ceiling. To simulate air leakage, sampler pumps installed in the indoor
space are used to continuously deliver indoor air into the attic space at a controlled rate. The air is
supplied at the bottom of the cellulose insulation through an orifice in the PIR using a tube with a 6.4
mm inner diameter. In the case where there is no airflow into the attic, the orifice is sealed with tape at
bottom of the PIR.
FIG 1. Location of temperature sensors FIG 2. Temperature sensors held in place with a
across the ceiling low-profile wooden support system
Figure 1 shows the location of the temperature sensors in the indoor, ceiling, and attic space. Tatt and
Tint are the attic and indoor space temperature sensors, respectively. A way to indirectly map air
leakage through building components is to measure temperature in the vicinity of the hole (Desmarais
et al., 2000), so the remaining sensors are located within the cellulose insulation at three different
heights and at various horizontal distances away from the air leakage orifice in the PIR. Tatt, Tint, T3A
and T6 are resistive temperature detectors embedded in relative humidity sensors with an accuracy of
±0.3ºC. The other temperatures are measured by thermocouples (30AWG, NSB special limits of error

±0.5ºC). The temperature sensors in the cellulose insulation are supported by a thin wooden structure
to ensure proper placement of the sensors without risk of displacement (Figure 2). Horizontal wood
members less than 10 mm in diameter support the sensors at each of the three heights in the cellulose
insulation; the horizontal members are themselves supported by vertical members at least 100 mm
away from the sensors. The indoor and outdoor temperature conditions were maintained at 22ºC and
5ºC, respectively. The air leakage rates used are 0, 2 and 5 LPM. At each air leakage rate, thermal
steady-state conditions are attained.
4. Numerical simulation
The model in this study uses COMSOL's built-in physics of "heat transfer in porous media" and
"Brinkman equations" to implement the conservation equations in the different domains as described
in section 2. To reduce computational time, the transfer phenomena are considered to be rotationally
symmetric about the longitudinal axis that starts at the air orifice and points upward, neglecting
therefore the thermal bridge induced by the truss and wooden structures supporting the sensors (Figure
2). The simulation is therefore performed on a 2D axisymmetric plane, creating the 3D rotational polar
coordinate system.
TABLE 1. Input Parameters
Parameter Symbol Units Value Source/Note
Thermal conductivity of cellulose λcel W/(m·K) 0.038 Manufacturer
Thermal conductivity of PIR λpoly W/(m·K) 0.022 ASHRAE
Thermal conductivity of air λair W/(m·K) 0.026 At 20°C
Dynamic viscosity air µair Pa·s 1.8 × 10-5 At 20°C
Surface film coefficient int. hint W/(m2·K) 9.26 ASHRAE
Surface film coefficient attic hattic W/(m2·K) 4.32 ASHRAE
Permeability of cellulose kcel m2 7.67 × 10-10 (Trechsel, 2011)
Porosity of cellulose εcel - 0.95 CSTB (Optimob)
Air density ρair kg/m3 1.2 At 20°C
Thermal capacity air cpair J/(kg·K) 1006 Assumed constant
Boundary conditions for air and heat must be chosen on each boundary. The measured temperatures
Tint and Tatt are used as boundary conditions for the model. To simplify the geometry, polyisocyanurate
insulation has been integrated as thermal boundary condition with the interior surface film coefficient
in an equivalent global heat transfer coefficient (Figure 3):
1
heq = (13)
1
Rpolyiso +
hint
The interior temperature is imposed at the entrance of the hole, and both open boundary and natural
convection are assigned to the top of the cellulose. The heat flux brought by air can both enter and
leave the domain on interfaces with the thermal open boundary condition (COMSOL, 2011):
%T = Tatt if n ⋅ u < 0
& (14)
'−∇T ⋅ n = 0 if n ⋅ u ≥ 0
where n unit vector normal to the surface

For the cut section 1 m away from the axis, adiabatic conditions are considered. The air inflow is
imposed at the inlet. A no-slip condition is applied to the interface between PIR and cellulose. An
open boundary with no viscous stress is imposed at the upper interface. COMSOL’s built-in meshing

is used; the meshing is refined in narrow regions and in regions where high temperature/velocity
gradients are expected. The mesh size is refined until the flux and temperature calculated become
stable. The geometry has a total of 233,220 meshes and the fully-coupled model requires
approximately 7 minutes to converge using an Intel® Core i7-2600 CPU at 3.4 GHz and 8 GB RAM.
FIG 3. Domain physics and boundary FIG 4. 3D plot of the temperature field for a flow
conditions of simulation rate of 5 LPM
5. Results and discussion
FIG 5. Isotherms obtained from simulation, values are in ºC
TABLE 2. Temperature data from the experiment and simulation, values are in ºC
Experiment Simulation
Data Point No airflow 2 LPM 5 LPM No airflow 2 LPM 5 LPM
Tint 21.9 21.8 21.8 21.9 21.8 21.8
T1 16.6 20.3 20.8 17.7 21.5 21.8
T2A 12.1 16.1 19.4 13.1 15.9 18.9
T2B 12.1 15.7 19.0 13.1 15.5 18.3
T3A 6.4 6.8 7.9 7.2 7.5 8.0
T3B 6.4 6.6 7.5 7.2 7.5 8.0
T3C 6.4 6.6 7.2 7.2 7.4 7.8
T3D 6.3 6.4 6.8 7.2 7.4 7.6
T4 16.3 16.9 17.3 17.7 17.9 18.2
T5 12.0 12.6 13.3 13.1 13.4 13.9
T6 6.4 6.5 6.6 7.2 7.3 7.4
Tatt 6.7 6.8 6.8 6.7 6.8 6.8

The results from the simulation are shown in isotherm graphs in figure 5; the temperature values from
both the simulation and the experiment are tabulated in table 2. Figure 4 shows the temperature field
of the 3D section with an air leakage rate of 5 LPM.
Compared to the case with no air leakage where the isotherms are horizontal and equidistant in the
cellulose insulation, imposing air leakage shifts the isotherms upwards at the axis of the orifice,
creating bell-shaped curves. The increase in temperature in the cellulose is more pronounced with
higher flow rates, as expected. The simulation results compare well with the temperature readings
from the experiment. With no air leakage, the simulated temperatures in the cellulose are roughly 1ºC
higher than the measured ones. This discrepancy may be partly due to variations in the material
properties, the simplified boundary conditions in the model, as well as a possible lack of homogeneity
of the cellulose in the test setup. Also, the temperature of the top layer of the cellulose in the
experiment is lower than the attic air temperature, which may be attributed to heat loss by radiation to
colder surfaces in the attic such as the roof sheathing.
When simulating air leakage through the section, the flow rate is calculated on several horizontal
planes at different heights to verify mass conservation. The simulation and the experimental results are
in good agreement with regards to the altered temperature field in the cellulose caused by air leakage.
The simulation tends to slightly overestimate the temperature increase by up to 1ºC, except at T2A and
T2B where it is underestimated by up to 0.5ºC. Again, the non-homogeneity of the cellulose can partly
be the cause. Accounting for moisture conditions and moisture transport can improve the accuracy of
the model; for example, the moisture content of the insulation has an effect on its thermal
conductivity.
6. Conclusion
Air leakage is a major contributor to moisture problems and energy losses in buildings. Airflow
through the building envelope often occurs through porous insulation materials, and understanding its
effect on the enclosure performance is of importance. This study presents a simplified, coupled heat-
airflow model, applied to an experimental setup. The experiment focuses on the temperature profile of
cellulose insulation in an attic of test hut, subjected to a 15ºC temperature difference and various air
leakage rates. The results from the experiment and simulation show good agreement. Ongoing
research aims to expand this model to couple moisture transport, which can further increase its
applicability, especially for greater temperature differences and transient processes. Parametric and
sensitivity analyses with multiple orifice areas, thicknesses of insulation, and flow rates could also be
performed.
7. Acknowledgements
This work is financially supported by ADEME (Agence De l'Environement et de la Maîtrise de
l'Énergie), CSTB (Centre Scientifique et Technique du Bâtiment), and the Région Rhône-Alpes. The
experimental data is taken from a project under the NSERC Smart Net-Zero Energy Buildings
Strategic Research Network (SNEBRN), sponsored by the Natural Sciences and Engineering Research
Council of Canada (NSERC) and 14 industrial partners including KOTT Group. The project is also
supported through a NSERC discovery grant. The materials and installation of the test hut are supplied
by KOTT Group.

8. References
Abadie, M.O., Finlayson, E.U., Gadgil, A.J., 2002. Infiltration heat recovery in building walls:
Computational fluid dynamics investigations results. Lawrence Berkeley National Laboratory.
Barhoun, H., 2006. Influence des transferts aérauliques dans les parois sur leurs performances
thermiques. Institut National des Sciences Appliquées de Lyon.
Bories, S., Mojtabi, A., Prat, M., Quintard, M., 2008. Transferts de chaleur dans les milieux poreux:
Conduction, convection, rayonnement. Tech. Ing. Génie Énergétique.
Buchanan, C.R., Sherman, M.H., 2000. A mathematical model for infiltration heat recovery. Lawrence
Berkeley National Laboratory.
Chan, W.R., Price, P.N., Sohn, M.D., Gadgil, A.J., 2003. Analysis of US residential air leakage
database.
COMSOL, 2011. Comsol Multiphysics User ́s Guide - Version 4.2.
Desmarais, G., Derome, D., Fazio, P., 2000. Mapping of Air Leakage in Exterior Wall Assemblies. J.
Therm. Envel. Build. Sci. 24, 132–154. doi:10.1106/BWH8-9D3J-R939-957E
Langlais, C., Arquis, E., McCaa, D.J., 1990. A theoretical and experimental study of convective effect
in loose-fill thermal insulation.
Langmans, J., 2013. Feasibility of Exterior Air Barriers in Timber Frame Construction. KU Leuven,
Belgique.
Li, Q., Rao, J., Fazio, P., 2009. Development of HAM tool for building envelope analysis. Build.
Environ. 44, 1065–1073. doi:10.1016/j.buildenv.2008.07.017
Litvak, A., 2005. Campagne de mesure de l’étanchéité à l’air de 123 logements RT2000 - Rapport
final. CETE de Lyon.
Nield, D.A., Bejan, A., 2006. Convection in porous media. Springer, New York.
Saber, H., Maref, W., Elmahdy, H., Swinton, M., Glatzer, R., 2012. 3D heat and air transport model
for predicting the thermal resistances of insulated wall assemblies. J. Perform. Simul. 75–91.
Svoboda, Z., 1999. The analysis of the convective-conductive heat transfert in the building
constructions. Presented at the 6th Building Simulation Conference, Kyoto, pp. 329–335.
Tariku, F., Kumaran, K., Fazio, P., 2010. Transient model for coupled heat, air and moisture transfer
through multilayered porous media. Int. J. Heat Mass Transf.
Trechsel, H.R., 2011. Moisture analysis and condensation control in building envelopes. ASTM
International, West Conshohocken, Pa.
Vafai, K., Thiyagaraja, R., 1987. Analysis of flow and heat transfer at the interface region of a porous
medium. Int. J. Heat Mass Transf. 30, 1391–1405.
Van Schijndel, A.W.M. van, 2008. Heat and Moisture Modeling Benchmarks using COMSOL.
Presented at the COMSOL conference, Hannover.
Wagner, H., 1989. Luftdichtkeit und Feuchteschutz (beim Steildach mit Dämmung zwischen den
Sparren). Dtsch. Bauz. Heft 1289 1639–1646.

Full paper no: 11
Airtightness of office and school buildings in Sweden –

measurements and analyses of the implication for the energy use
Åke Blomsterberg, Assistant Professor 1
Stephen Burke, Ph.D. 2
1
WSP and Lund University, Sweden
2
NCC Construction, Sweden
KEYWORDS: Airtightness, blower door, energy, measurement, office building, pressurisation,

school.
SUMMARY:
The airtightness of office and school buildings influences energy use and thermal comfort. A leaky
building is likely to have a high use of energy and thermal discomfort. The knowledge of real
airtightness levels of entire buildings and their impact on the energy use is very low, except for a study
carried out in the USA. Therefore two different methods of airtightness testing were applied to six
entire Swedish office and school buildings built since 2000. The first method involves using the
ventilation system of the building and the second one to use a number of blower doors. Information on
30 other airtightness tests of similar buildings was collected. During the airtightness testing the air
leakage paths were detected using infrared scanning and smoke sticks. Most of the tested buildings
showed a very good airtightness level. All previously tested office buildings in the USA, Canada and
the UK are much leakier. The tested buildings had some leakage paths, which could easily have been
taken care of during construction, but are rather difficult to stop now. The infiltration rate for these
buildings was estimated by a simplified calculation procedure to be low, implying a low energy use
caused by infiltration. For most of the buildings stringent airtightness requirements had been applied.
1. Introduction
It is well-known that the building sector plays an important role in the work towards sustainable
development. The sector represents extensive economic, social and cultural values, at the same time as
it causes extensive environmental impact due to its high use of energy and materials. An important
part of the energy use within the building sector is related to office and school buildings. The total
energy use of an average Swedish office building is 220 kWh/(m²year) (heated usable floor area) of
which electricity stands for 108 kWh/(m²year). Of this 108 kWh/(m²year), 57 kWh/(m²year) is due to
office equipment, of which 23 kWh/(m²year) is lighting. This was shown in a study of 123 office and
administrative buildings of different ages (Persson 2007). Of the floor area in all office buildings, 69
% is heated by district heating and the average use of district heating energy is 110 kWh/(m²year)
(SCB 2006). Both new and old office buildings have a substantial potential for energy savings and
improvement of indoor climate. While many new office buildings may have a low energy use for
heating compared with older office buildings, they may have a higher electricity use. This is due to a
high use of electricity for ventilation, cooling, lighting and office equipment. The situation for school
buildings is similar, with the exception of cooling which is rare in school buildings. An important
parameter affecting the energy use for space heating and cooling, and thus the indoor climate, is the
airtightness of the building envelope. In a leaky building the energy use increases due to uncontrolled
infiltration/exfiltration. The air leaking in and out through the building envelope increases the energy
use as it, for example, does not pass through a heat recovery unit. The uncontrolled air leakage can
contribute to discomfort such as draught, which can result in the indoor temperature being raised to
improve the comfort, causing a further increase in energy use from the user’s behaviour.

Unfortunately, there is no simple and accurate method of relating the airtightness of a building to the
air leakage for an office or school building in operation. This is due to difficulties in determining the
location and characteristics of all leakage paths and determining the wind pressure coefficients
(Blomsterberg 1990).
The aim of this project (Blomsterberg 2012) was:
• to use different measuring methods for determining the airtightness of school and office
buildings,
• to determine the airtightness for modern school and office buildings,
• to estimate the influence of airtightness on the energy use for space heating.
2. Method
The hypothesis is that, in many cases, the airtightness can be measured using the ventilation system of
the building instead of using a number of blower door units. Two different methods were used:
- Airtightness testing using the ventilation system of the building. Canadian standards were
applied (CGSB 1986, CGSB 1996).
- Airtightness testing using a number of blower doors (portable fans),
www.energyconservatory.com. European standard 13829, Method B was applied (CEN
2000).
The measurements involve pressurizing and depressurizing the entire building and measuring the
corresponding air flow to maintain the different pressure differences between inside and outside.
Ventilation openings and lead-throughs are sealed before the measurements. Thus the airtightness of
the building envelope is determined. The location of leakage paths are determined using thermography
and smoke.
When using the ventilation system of the building the following has to be investigated before
proceeding:
- Assess the building automation system to ensure that the ventilation air flows can be
controlled and that it has the capacity to provide the necessary air flows. It is usually easier if
the building has a demand controlled ventilation system.
- Ensuring that the air flows can be measured and that it can be done with adequate accuracy.
Within this project three schools and three office buildings were tested. Additional tests had been
carried out before by other Swedish organizations.
To determine the air infiltration/exfiltration rate from the results of pressurization tests there are
different ventilation models. The ventilation models can be divided into: “air change” methods,
reduction of pressurization test data, regression techniques, theoretical network methods, simplified
theoretical methods (Liddament 1986). The first three models are empirical techniques, which tend to
be loosely based on the physical principles of air flow. The other models are theoretical models, which
are based on a much more fundamental approach involving the solution of the equations of flow for air
movement through openings in the building envelope. Empirical methods are usually straightforward
to use, but tend to be unreliable and have a limited field of application. On the other hand, theoretical
models have a potentially unrestricted applicability but are often demanding in terms of data and
computer execution time. Theoretical calculation techniques can be divided into: single zone network
models, multi zone network models and simplified theoretical techniques. These models require a lot
of information e.g. wind pressure coefficients, air leakage distribution for the building envelope, local
wind speed, geometry of the building. Due to the limited amount of information on the tested

buildings the method using reduction of pressurization test data was chosen in order to determine an
order of magnitude for the average infiltration/exfiltration rate.
The reduction of pressurization test data method does nevertheless provide valuable information
concerning the average infiltration performance of the building. The artificial
pressurisation/depressurisation of a building to determine air leakage performance is now fairly
common practice. The test only provides data regarding the “leakiness” of the building. The result
provides no information on the distribution of openings or on how infiltration will be affected by
wind, temperature, terrain, or shielding. However, several experimental results have shown that the
approximate air infiltration rate in practice will be in the order of one twentieth of the measured air
change rate at 50 Pa (ASHRAE 2009), i.e.:
Qinf = Q50 /20 (1)
Where Qinf infiltration rate (h-1)
Q50 air change rate at 50 Pa.
Calculations have shown that the ratio can vary between 6 and 40 depending upon the house, the
climate and the shielding (Blomsterberg 1990).
To determine the energy use caused by air infiltration/exfiltration the infiltration/exfiltration rate was
first calculated from the pressurization tests and then the energy use was calculated using degree days
for Stockholm. For most of the buildings the only available information was the floor area, the
volume, type of ventilation system and type of building technology, the results of a pressurization test.
3. Tested building
The aim was to test school and office buildings built after the year 2000 with a floor area preferably
larger than 1000 m². They should be a mix of buildings both with and without specific airtightness
requirements.
3.1 Outside this project

Thirty buildings had previously been tested by different organisations e.g. SP - the Technical Research
Institute of Sweden, Akademiska Hus, Skanska, WSP. All the buildings were built between 2007 and
2012. The buildings are mainly schools and offices, but also include homes for the elderly, shops, and
sports centres. The smallest building had a floor area of 800 m² and the biggest 17 000 m². All
buildings had balanced mechanical ventilation with heat recovery. The building envelopes varied,
ranging from prefabricated concrete to stud walls.
3.2 Within this project

Six buildings were tested for the purpose of this project, three office and three school buildings. All
the buildings were built between 2007 and 2011. The smallest building had a floor area of 1,030 m²
and the biggest 20,000 m². All buildings had balanced mechanical ventilation with heat recovery. The
building envelopes varied, ranging from prefabricated concrete, prefabricated glass facade to stud
walls.
4. Results
All the buildings tested outside the project were very airtight (see table 1). The average airtightness
was 0.3 l/sm² @ 50 Pa which is equivalent to the voluntary Swedish requirement for passive houses
(Nollhus 2012). The best building had a value of 0.1. For most of the buildings airtightness
requirements ranged from 0.2 to 0.8 l/sm² @ 50 Pa, which can be compared with the requirement of
the previous Swedish building code (before year 2006), of 1.6 l/sm² @ 50 Pa. The current building
code does not have any specific requirement. Only two buildings did not meet their requirement. All

previously tested office buildings in the USA, Canada and the UK are much leakier (Blomsterberg
2009). Common leakage paths were exterior doors and connections between façade elements and
floors/roofs, most of which would be difficult to tighten afterwards. Most buildings were tested with
blower doors. Some were tested with the ventilation system.
TABLE 1. Measured air leakage and leakage paths.

Type of Year of Envelope Airtight- Measured Main leakage paths
building con- area, m² ness airtight-
struction require- ness, l/sm²
ment, l/sm² @ 50 Pa
@ 50 Pa
Shop 2011 18,721 - 0.18 Concrete element joints, exterior
doors
Sport 2011 6,616 0.4 0.44 Exterior doors etc.
Centre
Office 2008 2,580 - 0.34 Entrance parts/windows/exterior
doors
Office 2010 0.4 0.27
Office 2010 4,237 0.5 0.43 Connections between floor and
wall
Office 2010 14,610 0.6 0.55
Office 2007 - 0.8 0.7 Connection between facade
elements, facade and roof
elements
Office/i 2009 4,560 0.25 0.26 Connection between ceiling and
ndustry wall/workshop – exterior doors
Storage/ 2011 10,034 0.3 0.29 Exterior doors
worksho
p/office
Food 2011 3,995 0.8 0.62 TRP/expanded clay, windows,
store Entrance parts
School 2009 4,912 0.5 0.36
School 2011 2,607 0.2 0.13
School 2008 3,335 0.45 0.41
School 2008 5,180 0.4 0.21
School 2009 2,832 0.6 0.27 Exterior doors
School 2008 2,414 0.3 0.26
School 2010 2,460 0.6 0.23 Exterior doors and windows
School 2010 2,460 0.6 0.19 Exterior doors and windows
School 2010 2,182 0.6 0.57
School 2010 5,513 0.2 0.09 No major leakage paths
School 2007 3,941 0.4 0.45 Exterior doors etc.
School 2011 2,261 0.6 0.48 Roof, windows and doors
School 2010 2,295 0.3 0.4 Connection wall-ceiling,

exterior door
School 2011 4,822 0.2 0.16
School 2010 5,641 0.8 0.88
Home 2012 4,081 0.3 0.20 Exterior doors
for the
elderly
Home 2011 3,900 0.2 0.14
for the
elderly
Average 0.44 0.30
For twelve of the buildings information on the volume was available and the airtightness could be
recalculated to ach @ 50 Pa (see table 2). A comparison of the buildings is now different due to
different ratios between volume and envelope area. Using a simple method of calculating the
infiltration (see Method) an average infiltration rate was estimated. The result was an average air
infiltration rate during the heating season of 0.03 ach (air changes per hour), varying between 0.01 and
0.06. This is equivalent to an energy use for space heating of 4 kWh/m²year. If the buildings would
have only met the requirements of the previous building code the energy use might have been five
times higher i.e. 20 kWh/ m²year.
TABLE 2. Measured air leakage and calculated energy use for heating infiltrating air.
Type of Year Measured Measured Infiltration/e Energy use for heating
building airtightness airtightness xfiltration, infiltration,
l/sm² @ 50 ach @ 50 ach kWh/m²year
Pa Pa
Shop 2011 0.18 0.20 0.01 2
Office 2008 0.34 0.60 0.03 3
Office 2010 0.43 0.43 0.02 3
Food store 2011 0.62 1.11 0.06 10
School 2010 0.09 0.13 0.01 1
School 2011 0.17 0.34 0.02 2
School 2011 0.48 1.18 0.06 7
School 2010 0.4 1.04 0.05 7
School 2011 0.16 0.24 0.01 2
School 2010 0.88 1.28 0.06 9
Home for the 2012 0.20 0.20 0.01 1
elderly
Home for the 2011 0.14 0.18 0.01 1
elderly
Average 0.34 0.58 0.03 4
Also the recently tested six buildings were fairly airtight, but not as airtight as the previously tested
buildings (see table 3). One contributing factor might be that there were only two buildings which had
a specified airtightness requirement.

TABLE 3. Measured air leakage and leakage paths.
Type of Year of Test method Envelope Airtightnes Measured Main leakage
building construction area, m² s require- air paths
ment, l/sm² tightness,
@ 50 Pa l/sm² @ 50
Pa
Exhibition 2011 Ventilation 40,400 0.4 0.39 Connection
/office system, the between
whole façade
building elements and
columns,
between
facade and
roof, exterior
doors.
Office 2009 Ventilation 5,600 0.85 Connection
system, between
storey 3, infill walls
back and steel
pressure columns,
storey 2, 4, windows
atrium and
staircase
School 2007 Blower 3,923 - 0.87 Lead-
Door throughs,
windows
School 2011 Blower 2,775 - 0.45 Doors,
Door windows
School 2009 Blower Door 4,307 - 0.62 Lead-
throughs,
windows,
doors.
Average 0,64
For the recently tested buildings, information on the volume was available and the airtightness could
be recalculated to ach @ 50 Pa (see table 4). The comparison of the buildings is now different due to
different ratios between volume and envelope area. Using a simple method of calculating the
infiltration (see Method) an average infiltration rate was estimated. The result was an average air
infiltration rate during the heating season of 0.06 ach (air changes per hour), varying between 0.01 and
0.14. This is equivalent to an energy use for space heating of 6 kWh/m²year. If the buildings would
have only met the requirements of the previous building code, the energy use might have been three
times higher at about 20 kWh/ m²year.

TABLE 4. Measured air leakage and calculated energy use for heating infiltrating air.
Type of building Year Measured Measured Infiltration/e Energy use for heating
airtightness airtightness xfiltration, infiltration,
l/sm² @ 50 ach @ 50 ach kWh/m²year
Pa Pa
Exhibition/office 2011 0.39 0.28 0.01 5
Office 2009 0.85 0.36 0.02 2
Office 2009 0.68 2.77 0.14 4
Education 2007 0.87 1.44 0.07 8
Education 2011 0.45 1.51 0.08 7
Education 2009 0.62 1.34 0.07 8
0.64 1.28 0.06 6
5. Conclusion
Two different methods of measuring the airtightness of entire buildings have been used, using the
building’s ventilation systems or using a number of blower doors. Both methods can be used
separately or combined. Choice of method depends on the prerequisites of the individual building. For
large buildings using the ventilation system can be preferable. This presupposes that the air flow can
be controlled within the desired range and measured accurately enough. Usually this is the case for
demand controlled ventilation. Tests during construction, which are recommended to ensure good
airtightness, can often only be carried out using blower doors. The two methods can be applied to
office buildings, apartment buildings, industrial buildings and other premises. For apartment buildings
the blower door technique is often the only method as the ventilation system often has insufficient
capacity, unless the building is very airtight. Complete testing includes determination of the location
of leakage paths.
This study clearly shows that it is possible to build very airtight school and office buildings i.e. which
fulfils the Swedish criterion for passive houses at 0.3 l/sm² @ 50 Pa. This seems to be possible with a
building envelope which is prefabricated or framework construction or curtain wall. Most likely, the
energy use for infiltration in these buildings is almost negligible i.e. in the order of magnitude of a
couple of kWh/m²year. This number can be compared with the total energy use for space heating for a
typical average Swedish office building of 110 kWh/m²year, where infiltration might account for 10-
20 kWh/m²year if only the airtightness requirement of the previous building code is fulfilled, which is
likely.
An important prerequisite for arriving at a very good airtightness is obviously that requirements
regarding airtightness are set at an early stage. These requirements have to be specified as a maximum
air leakage value, which should be tested during the building process.
6. Acknowledgements
The project was funded by SBUF (Development Fund of the Swedish Construction Industry), NCC
and WSP.
7. References
ASHRAE 2009. ASHRAE Handbook Fundamentals. ASHRAE, Atlanta, USA.
Blomsterberg, Å., 1990. Ventilation and airtightness in low-rise residential buildings - Analyses and
full-scale measurements. Swedish *Council for Building Research, D10:1990, ph.d. thesis.

Blomsterberg, Å. 2009. Airtightness and energy use of an Swedish office building – measurements
and calculations. Presented at the 30th AIVC- and the 4th BUILDAIR-Conference in Berlin.
Blomsterberg, Å., Burke, S. 2012. Real airtightness of large buildings – Measurements and
calculations (in Swedish). SBUF (Development Fund of the Swedish Construction Industry), NCC
and WSP.
CEN 2000. Thermal performance of buildings - Determination of air permeability of buildings – Fan
pressurization. European Committee for Standardization, EN 13829.
CGSB 1986. Determination of the Airtightness of Buildings envelopes by the Fan Depressurization
Method Using the Buildings’s Air Handling Systems, Canadian General Standards Board, Standard
149.10-M86, National Standard of Canada.
CGSB 1996. Determination of the Overall Envelope Airtightness of Buildings by the Fan
Pressurization Method Using the Buildings’s Air Handling Systems, Canadian General Standards
Board, National Standard of Canada.
Liddament, M. 1986. Air Infiltration Calculation Techniques – An Application Guide. AIVC.
Nollhus, 2012, Kriterier för passivhus m.m. (Criteria for passive houses etc.), www.nollhus.se .
Persson, A. 2007. Improved energy statistics for premises – Report for year 1 – Inventory of offices
and administrative buildings (in Swedish). The Swedish Energy Agency, rapport ER 2007:34.
SCB 2006. Energy statistics for premises 2005. Swedish Statistics, EN16SM0603, www.scb.se.

Full paper no: 12
Manageable supervision of air tightness for large buildings

Mikael Kläth 1
Peter Sunvisson 1
1
Dry-IT, Sweden
KEYWORDS: Air tightness, air leakage, blower door, pressure test, building envelope
SUMMARY:
It is important to approach issues concerning building air tightness systematically and implement this
throughout the building process to ensure that air leakage requirements are met for large buildings.
Air leakage testing needs to be applied early on during construction to better understand how to
increase the quality of the building and efficiency of the construction process. This paper explains how
to work with air tightness management in the different construction phases and gives a practical
example of a large building that met high ambitions for air tightness.
1. Introduction
Several benefits can be derived from testing air tightness in the early stages of constructing large
buildings. The key to making this possible is to use a systematic approach to managing issues
concerning air leakage throughout the construction process.
Difficulties that occur concerning air leakage testing during the production phase of large buildings
are foremost of the practical types. Examples include limitations of the equipment capacity, the scale
of the provisional sealing methods and time consumption of the capsuling techniques and other
operation steps. Different parties may participate in different ways to contribute towards an air-tight
end result, through defining requirements, technical collaboration meetings or workshops,
coordination for difficulties in practice, storage and use of sustainable materials and well thought-out
modifications.
This paper aims to answer how to efficiently manage air tightness issues from experience of working
in projects for the construction of large buildings. First we present the authors’ general method of
practice for implementing air-tightness management in the different stages of the building process.
Furthermore a project with a high ambition level is presented where the actions taken concerning air
tightness are mapped out and how the issued air leakage requirements were met.
2. Air tightness management in the building process

Regrettably, air tightness (or building physics in general) engineers and consultants are rarely involved
early on in the construction process, during the time where many decisions of importance to air
tightness issues are set. This chapter describes how air tightness consultants should be utilized in the
different stages of the building process.
2.1 Air tightness management during the design and development phases
A project developer can set an ambition level for air tightness that can be regulated as a part of the
company’s standard for construction or it can be project specific. Air tightness in this phase is often
related to energy efficiency, which has a strong connection to the attractiveness of the building as a
product, i.e. if it can be promoted as a “green” building by itself or with an environmental
classification system.

It is important that relevant information concerning air tightness issues is updated and spread through
the project organisation during the course of the design phase. A dialogue between the air tightness
consultant and the designers (as well as other consultants in other fields of expertise) needs to be
managed and documented systematically. This can be achieved by creating an Air Tightness
Description, which is a document that describes project specific air tightness aspects that needs
attention so as to ensure that requirements and regulations from the developer and authorities can be
met. The contents of this document are dynamic and may be revised if any conditions are changed.
There are many disciplines that need coordinating in this phase so as to strive towards the same goal.
An example of this is when the architect is weighing on how far to recede the windows in the wall
construction, then the construction engineer needs coordination on how this affects the choice of
sealing method between the walls’ air barrier and window.
The consequence of choices made early on in the design phase may be beneficial in one aspect but
negative in another, which for example is sometimes the case for detail designs in relation to
insulation thickness. It is therefore advantageous if the coordinating role has experience in building
physics (building physics coordinator) and has enough knowledge in different fields to be able to
identify if a design may lead to problems for the indoor environment.
2.2 Air tightness management during the construction phase

Air leakage restrictions that are set by the developer or client (often a result of requirements set by
environmental classification systems such as Miljöbyggnad, BREEAM or LEAD) need to be related to
the contractors’ process for construction. During the construction phase it is therefore important to
involve air tightness engineers that can lift these issues within the project’s organisation early on in the
construction phase. An effective method to assist this process is to create a Plan for Air Tightness,
which is a document that establishes how issues and secondary aspects related to air leakage are to be
dealt with, so as to create efficient models to monitor and achieve the requirements governed by the
Air Tightness Description. Within the documentation an organisational chart is set, that for instance
points out specific personnel that carry responsibility for different aspects to make sure that
regulations are met; eventual planning that affects procurement, scheduling, testing and monitoring,
education and quality assurance.
Necessary for large buildings, provisional encapsulations are used for sectioning off building parts
during the ongoing construction to enable pressurised testing. The air leakage index is measured with
computer operated fan equipment, e.g. Blower-door, attached to customised tents or makeshift walls
that enclose the area of interest. Improvement measures for air leakages sought out during the test are
prescribed for the following building parts of the same type to be installed. Furthermore, problems that
need to be solved concerning practice methods are informed to the involved carpenters and other
craftsmen so that lessons can be learnt from previous mistakes. It is important that experiences from
the ground level are documented and that the information is quickly distributed to the right people so
that the organisation has a chance to adapt.
Suitable parts of the building for air leakage testing are identified together with the contractor during a
start-up meeting, preferably before the procurement process of materials and subcontractors. The scale
of the testing areas may vary due to the complexity of the building and the level of prefabrication.
Sometimes it is more suitable to test samples of for example window connections, prefabricated wall
sections, prefabricated bathroom modules, a fire cell, a whole floor, two floors to check the wall knee,
material joints, wall corners, stairwells, before a commissioning test is relevant. For some buildings a
full scale pressurised test isn’t technically possible before the ventilation system is functional or that
several buildings are internally connected and individual air leakage indexes are irrelevant. Common
tools for localising discrepancies in the building envelope or air barrier include an anemometer,
thermography imaging (during the colder period of the year) and smoke machines (note that the choice
of gas mixture affects the natural smoke movement due to its density).

2.3 Air tightness management during the operational phase
During the operational phase it is important to spread knowledge to the involved parties concerning air
tightness issues and to do so in a systematic approach toward the property manager and tenant of the
building. This can be achieved by developing routines for sustainable air tightness management
related to renovation maintenance, outer refurbishment, internal repairs and tenant modifications,
aging of materials, technical lifespans and unexpected changes to the buildings air permeability
conditions.
3. A practical example of air tightness management in the construction

phase of a large building
3.1 Project description
A good example of how air tightness management can be implemented for the construction of a large
building is Uarda 5, see figure 1, located in Solna, Stockholm. The building’s floor area (Afloor)
amounts to roughly 55,000 m2 and area for the building envelope to nearly 28,500 m2(Aenv). The
property received the highest Miljöbyggnad certification grade: Gold. The maximum air leakage
allowed was set to 0.3 l/s,m2(Aenv) @ 50 Pa differential pressure over the building envelope. The
authors’ company was involved in the project during the design stage, mainly for consultation in the
area of moisture safety, where there are concerns that are closely related to air leakage. Further
regulations for air tightness management were expressed under a chapter in the Moisture Safety
Description. The resulting air leakage index was measured to 0.27 l/s,m2(Aenv) @ 50 Pa.
FIG 1. Illustration of Uarda 5.
3.2 Work methodology

Management of the issues concerning air tightness to achieve the requirements was initially planned
through a start-up meeting and executed through continuous discussions with the contractor. An Air
Tightness Control Plan was established in accordance with the projects conditions and the critical
parts of the construction were identified for further examination. Technical meetings were set when
the subcontractors had been procured to discuss the steps that had to be taken to achieve an airtight
result.
Some of the critical building parts that were identified included the roofing, skylights, exterior walls,
windows, outer glazing, ventilation aggregates, garage doors, airlocks, basement walls and details
around the foundation. It was concluded that pressurised air leakage testing (SS-EN 13829) was to be
conducted to a skylight, a few sections of the exterior walls and a larger coherent section of the outer
glazing. After testing the first of each building section to be installed, experiences were documented
and improvement measures were taken for the following building sections that were in line to be
installed. The ventilation aggregates had tougher requirements and were subdued to a more vigorous

testing procedure (VVS-AMA Täthetsklass B). The rest of the critical building parts were surveyed to
find air leakage paths through thermography during the commissioning pressurised test.
FIG 2. Sketch of the floor plan.
The areas of interest for testing were marked out on overview plans and the contractor was given
instructions on the method for constructing the provisional encapsulations, which were also explained
in detail beforehand on site together with the carpenters.
FIG 3. Blower-door equipment attached to a provisional wall with a smoke machine placed at the fan
inlet.
3.3 Exterior walls

Initially the exterior walls were air leakage tested in smaller sections, for example some of the wall
joints on a few floors. Later on whole flights in one of the four building corners was subdued to
pressurise testing. The provisional capsuling consisted of a wooden framework c/c 400 - 600 mm,
MDF or plywood sheeting and an exterior polyethylene foil with taped joints and clamped with screws
to the framework with battens. An acrylic sealant is applied between the polyethylene foil and
concrete to fill any voids that may occur due to imperfections in the batten, joist or the concrete. The
provisional capsulations sectioned off the testing area between the stairwell and to the adjacent
building corners, see “Hus D” in figure 2. The Blower-door equipment was attached to the capsulation
and the smoke generator was placed at the fan inlet at the time that pressure was applied to the testing
volume, see figure 3, or used inside the testing area by applying smoke directly to the building
envelope, see figure 4. Personnel on the inside and outside of the testing volume monitored the
envelope for smoke leakage during the test.

FIG 4. Air leakage testing of window details with smoke.
3.4 Ventilation aggregates

The building’s four ventilation aggregates were each approximately the size of an apartment and had
tough air tightness requirements to fulfil. The aggregates were depressurised with a pressure
difference of 50 Pa. The result was extrapolated using the air leakage coefficient and exponent to
estimate a value at 200 Pa, which was the specified output. Later the aggregates were pressurised at
200 Pa to compare the extrapolated air flow result to the actual measurement.
3.5 Skylights
The skylights are situated over the atrium in the core of the building with a vast open space
underneath, see figure 5. The provisional encapsulation that was constructed around one of the
skylights was made up of a customised polyethylene tent that was sealed to the floor of the
scaffolding. The air tightness engineer and representatives from the contractor and subcontractors
enclosed themselves inside the testing area during the testing. The fan was mounted to the
polyethylene foil and the test area was pressurised and inspected for air leakage with the use of
thermography, anemometer and a smoke machine. It turned out that the attachment of the capsulation
to the scaffolding could not ensure an air tight test result during pressurisation but air leakages could
still be found and repair work was prescribed.
FIG 5. Rendered illustration of the skylight construction.
3.6 Commissioning air tightness test

The buildings own air-handling equipment was employed for the commissioning air tightness test
(CAN/CGSB 149.15-96). Simultaneously air leakage paths were exposed, especially for details
around the drains, doors, windows and installations through the façade using thermography (SS-EN
13187). The few parts of the building that had yet not been completed at the time of the
commissioning test were applied with complementary provisional encapsulations. To ensure that the

measured air leakage index was reasonable, the test was redone by a two separate ventilation
aggregates.
The set requirements for maximum air leakage index (0.3 l/s,m2) was met and measured to 0.27 l/s,m2
for the project Uarda 5. The property was awarded the title of SGBC Green Building Award 2012 and
the highest classification Miljöbyggnad Guld. We are at this point unable to present a calculation of
the benefits concerning reduced energy consumption for this building due to concerns of the buildings
safety regulations set by its current occupants. The extra energy consumption for changing the design
air leakage index from 0.3 to 0.5 l/s,m2(Aenv) has been calculated for a similar project of this scale,
which amounts to a difference of 2 kWh/m2(Afloor). If applied to Uarda 5, the consequence would be an
extra energy demand of 110 MWh/y.
4. Conclusions
There are several benefits for air tightness management throughout the building phases and to scale
down planning for manageable testing methods to be applied during the course of the construction.
One of the foremost reasons for these benefits is that the air tightness engineer or consultant may
implement systematic routines for spreading information to the different organisational parts of the
building process, so that the end installer or carpenter is up to date with accurate knowledge on how to
effectively work with these issues and why they are important. There are vast possibilities to engage
these issues early on to affect the building process and have an impact on the end result.
Work that is needed to ensure a comprehensive grip on air tightness is implemented during the design
phase, when construction designs are made that are critical to air tightness issues and need to be
identified and analysed in an Air Tightness Description, for issues that concern both the construction
and operational phase of the building. Before commencing the production, a Plan for Air Tightness
needs to be established, which describes the execution of critical steps that are practically explained
through technical meetings. The choice of suitable building sections to be tested is of great
importance, especially for buildings where it is not applicable to pressurise the total area of the
building envelope at once, for results from these sections will represent the whole building. It is
important to work with air leakage surveillance and apply pressurised testing to strategically chosen
parts of the building during the course of construction, to assure the quality of execution so that the
end result amounts to less than the design value. Through a systematic approach to air tightness
management throughout the construction process and documenting the path taken, lessons may be
learnt and a more widespread knowledge of the importance of air tightness may infiltrate the building
industry.
5. Acknowledgements
The authors wish to thank Fabege for involving us early in the project Uarda 5 and Peab Sverige AB
for the partnership surrounding air tightness testing.
References
CAN/CGSB-149.15-96. Determination of the Overall Envelope Airtightness of Buildings by the Fan
Pressurization Method Using the Building’s Air Handling System, Canadian General Standards
Board, Ottawa, Canada, 1996.
SS-EN 13187. Thermal performance of buildings - Qualitative detection of thermal irregularities in
building envelopes - Infrared method. Swedish Standards Institute SIS, 1999.
SS-EN 13829. Thermal performance of buildings - Determination of air permeability of buildings -
Fan pressurization method (ISO 9972:1996, modified). Swedish Standards Institute SIS, 2000.

AMA VVS & Kyl 09. Allmän material- och arbetsbeskrivning för vvs- och kyltekniska arbeten,
Svensk Byggtjänst, 2010.

Full paper no: 13
Method for measuring the air-tightness of facing formworks

Paul Wegerer 1
Thomas Bednar, Professor 1
1
Vienna University of Technology, Austria
KEYWORDS: air-tightness, durability, facing formwork, in-situ measurement, tracer-gas
SUMMARY:
Air-tightness of constructions is essential when evaluating a building’s energy efficiency and
durability. Even in connection with refurbishment air-tightness is becoming increasingly important.
Especially when using facing formworks with vapour barriers as interior insulation to thermally
renovate an existing building, air-tightness is an essential parameter. It has a significant influence on
the proper functioning of the construction because convective moisture entry can lead to considerable
damage.
In this paper an in-situ measuring carried out on a real object is being presented, which allows to
assess and to proof the air-tightness of facing formworks. This method was developed on a testing
wall in the laboratory of the Research Centre for Building Physics and Sound Protection of Vienna
The measurement starts after the completion of the construction by using a tracer-gas (CO2) and by
adjusting excess pressure in the testing room. At the same time the increase of CO2 in the facing
formwork is recorded. Measurements are carried out with at least two different pressure ranges. From
the results gained, the air flow volume coefficient C and the leakage exponent n are calculated in
accordance to EN 12114.
1. Introduction
A construction’s air-tightness substantially influences its functionality as convective moisture entries
contain a high risk of causing damage. The quality of the airtight layer often does not meet the planned
requirements. This has been pointed out in many publications. For example, Sandberg (2005)
described typical problems caused by poor air-tightness of constructions and specified experience
from site visits. Wahlgren (2010) presented a method detecting air leakages at an early construction
stage by using negative pressure. However, the given examples of joints and connections solely refer
to the entire building envelope as an airtight structure.
A building’s air-tightness can be determined by negative- or excess-pressure-measurement using
standardised measuring methods (Blower-Door-measurement according to EN 13829). Thereby, the
total air-tightness at a difference in pressure of 50 Pa is given as a n50-value in air exchange rate per
hour or as a q50-value in m³/(m²h) according to the formulas (1). There is, however, no differentiation
whether the leakiness occurs in the area of the windows, doors or joints and connections. Because of
this, size and form of leakages cannot be quantified with this method. Furthermore, a judgement
concerning a construction’s durability cannot be drawn from these measurement results.
V50 V50
n50  q50  (1)
V AE
If, on the other hand, the air-tightness of a single building component or a single construction layer is
to be determined, a different measurement concept must be used. In this case the focus is not on

determining the whole building’s air-tightness but a single component layer, e.g. gypsum board
planking, the vapour retarder or a foil in general is observed. This is especially relevant when using a
facing formwork as interior insulation as with this method of construction leakiness of the vapour
retarder may lead to damage. In this paper, a method to measure the air-tightness of a facing formwork
as interior insulation will be introduced.
2. Air flow through facing formwork constructions

The calculation of air flows in building components has already been validated in several publications.
In Bednar et al. (2010) the convective moisture entry into a flat roof is shown. In this case, two holes
of the same size are assumed as entry- and exit-leakages. Wegerer et al. (2012) demonstrated the
influence of leakiness of inside insulated walls using the example of a wooden beam bearing. He
showed that untight facing formworks used as interior insulation bear a huge potential for damage.
Facing formwork constructions normally consist of a post-and-beam-construction which is covered
with panels, commonly gypsum boards. To implement an airtight layer and to restrict moisture
entering the construction a vapour barrier is fitted between the post-construction and the planking. The
quality of this component layer is directly linked to the functionality of the whole component.
Leakages resulting from improper bonding as well as penetrating the vapour retarder with screws lead
to an air flow through the facing formwork. This may result in moisture entering the component from
the air of the room.
In the following two depictions (FIG.1) possible paths of air flows are shown schematically. In the
picture on the left, you can see a one-dimensional structure in the form of a typical facing formwork
with a possible air-path. The picture on the right shows possible air-paths using the example of a
ceiling-to-wall connection.
FIG. 1 left: reveal with facing formwork as interior insulation; right: ceiling-to-wall connection with
facing formwork as interior insulation
It is noticeable that the air-tightness of the facing formwork depends on at least two layers. In the one-
dimensional case these are the vapour barrier within the construction and the exterior load-bearing
wall, e.g. the masonry. With multi-dimensional component connections flow-paths to neighbouring
rooms or storeys may exist. As a prerequisite for a convective moisture entry into the facing formwork
there has to be an air-path from the facing formwork to the outside or a neighbouring room
additionally to the crack in the vapour retarder. Thus, to determine the airflow through the facing
formwork, all leakages in the construction’s interior and exterior cladding must be scrutinized. The
volume flow rate into the construction in essence depends on the leakage size and the difference in

pressure between the room and the facing formwork and can be expressed according to the following
equation:
V  C  p n (2)
Where V airflow in m³/(m²s)

C airflow coefficient in m³/(sPan)
p pressure difference between the two spaces in Pa
n leakage exponent
Hall (2003) has examined the connection between difference in pressure and volume flow rate through
certain leakages in laboratory tests. Thereby, various practice-oriented situations such as slashes or
distortions in the vapour barrier and holes and tears in various materials were measured. This study
serves as the basis for the following laboratory tests concerning an in-situ method of measurement.
3. Measuring convective moisture entry in facing formworks

In the following a test set-up which serves to measure convective moisture entry in a facing formwork
will be introduced. This in-situ method of measurement may be conducted on existing constructions,
and thus provides information about the construction’s actual air-tightness.
3.1 Laboratory measurements and analysis

The method of measurement was developed using a facing formwork in a laboratory test rig. Mass
entry into the construction was determined using CO2 as a tracer gas. The facing formwork (5 cm
thick) consists of three fields separated by vertical wooden uprights. In addition, each field is divided
into three horizontal segments. In total, there are nine fields, with fields 2, 5 and 8 being filled with
mineral wool as shown in FIG. 2 on the right. The remaining six fields are without mineral wool and
thus have a closed plenum filled with air. Furthermore, these six fields are fitted with tiny fans to
ensure permanent and even airflow. All nine fields are fitted with CO2 sensors. In the measuring room
there are two sensors, one at the bottom and one in the upper region of the ventilated room.
FIG. 2 facing formwork in laboratory test rig; left without foil, right with foil and indication of fields
The facing formwork is lined with a foil which is stuck to the existing wall around the test
construction in a manner as to provide an airtight space. Furthermore, the foil is attached to the facing
formwork’s vertical uprights with wooden strips to simulate planking with plasterboard. Deliberately,
no full planking was provided and see-through foil was used in the laboratory test rig for us to be able

to closely observe the processes in the facing formwork. The CO2 sensors‘ and fans‘ cables were led
out via an airtight cable gland in field 7.
Additionally, an air hose is mounted to the vapour barrier in field 3 as shown in FIG. 3. It connects the
air space in the facing formwork to the outside air. This air hose thus simulates the outer leakage of the
whole construction.
FIG. 3 left: hose leading from facing formwork to the outside; right: schematic depiction of the test
set-up (layout)
During the test excess pressure is set up in the measuring room using a Blower-Door device. At the
same time, CO2 is released. The concentration of CO2 in the room is set to approximately 2.000 ppm
and kept on a constant level. Because of the difference in pressure between the test room and the
facing formwork resp. the exterior the CO2-air-mixture runs into the facing formwork through leaks in
the foil, thus increasing CO2-concentration in all nine measuring fields. In this test, the connection
hose attached to the facing formwork simulates the leakage towards the exterior and thus favours the
mass through cracks on the inside.
If one looks at a period of time with approximately constant CO2-concentration in the test room, the
mass entry ṁ into the facing formwork can be calculated in kg/d. This is effected by determining a
linear regression from the CO2-concentration measured in the facing formwork and by averaging over
the period of time with approximately constant CO2-concentration in the test room.
c
m   V      C   p 
b
cR  cc (3)
Where ṁ massflow rate of air into cavity in kg/s
c rise of CO2-concentration in test field in facing formwork in ppm/d
cR CO2-concentration in test room in ppm
cc CO2-concentration of test field in facing formwork construction in ppm
V volume of the air space in test field in facing formwork in m³
 air density in kg/m³
p Pressure difference in Pa
C air flow coefficient non-dimensional
b air flow exponent non-dimensional

Should the measurements be conducted for at least two pressure stages, the volume flow rates can be
plotted against the related difference in pressure and the air flow coefficient and the air flow exponent
can be determined. This will be presented in chapter 3.2 on the example of an in-situ measurement.
Results of measurements in the laboratory

Various parameters were examined during the measurements on the test wall in the laboratory. At first,
the influence of standing and mixed air was examined in a hollow test field. At that, small fans were
mounted in fields 1, 4, 7, which are not filled with insulating material, and 3, 6 and 9. These fans
ensure the air’s thorough-mixing. Then a number of trials with and without fans were conducted.
Furthermore, artificial leakages at various points were examined in both variants (mixed/not mixed).
The following two diagrams show the influence of the air’s thorough-mixing in the facing formwork’s
cavity using the example of an artificial leakage in field 7.
FIG. 4. CO2-concentration in facing formwork at pressure difference of 20 Pa and leakage in field 7;

left: with thorough-mixing; right: without air’s thorough-mixing in facing formwork’s cavity
It is clearly visible in the left diagram that the CO2-concentration in fields 1, 4 and 7 rises markedly as
CO2 entering through the leakage in field 7 is mixed with the surrounding cavities. The right diagram
also shows a steep rise of the CO2-concentration in field 7 right behind the artificial leakage. Because
of the low mixing of the air within the construction the remaining sensors react later to the air
entering. The following two depictions FIG. 5 and FIG. 6 show measurement alternatives with
insulation in the facing formwork (left) compared with alternatives without insulation in the facing
formwork (right).

FIG. 5. Leakage in field 7; left with insulation in field 5; right without insulation in field 5
FIG. 6. Leakage in field 5; left with insulation in field 5; right without insulation in field 5
Based on the CO2-progressions it becomes clear that the measurement of hollow facing formworks
provides a solid mean regarding the actual airtightness of a wall-area. When measuring facing
formworks with insulation, however, heavily varying results are gained. This is presumably due to
flowpaths and cavities within a construction filled with insulation material causing an uneven spread
of the tracer gas.
3.2 In-situ measurements of facing formworks on the building site

Based on the method described the air-tightness of an existing construction will now be determined.
The measurement is prepared with the following steps:
1. The planking (most commonly gypsum board) is opened carefully between two vertical uprights
without damaging the foil underneath. Possible damage to the foil caused by opening must be
repaired carefully to restore the initial condition.
2. The vapour barrier is opened at two positions. One aperture is to be made just under the ceiling,
the other near to the floor as a cross section. The filling material – usually the insulation
material made of mineral wool – can be removed through these openings. It is crucial for the
test that the facing formwork’s entire test field is empty and has a defined air-volume. The
dimensions of this air space must be measured as the air-volume is needed for the evaluation.

3. Two fans are mounted in the cavity to guarantee thorough-mixing of the air in the facing
formwork’s examined segment.
4. Two to three CO2-sensors are inserted into the facing formwork through the upper aperture and
hung at different heights in the cavity. The cable passage through the vapour retarded must be
sealed airtight.
5. An air hose connecting the facing formwork to the outside air is mounted to the lower aperture
in the vapour retarder. This hose represents the outer leakage. The connection to the vapour
barrier also must be airtight.
6. After checking the cable passages for air-tightness, the gypsum board planking must be
restored. This is done by sticking on an additional film to the existing planking in the area of the
test field.
The measurement is conducted according to the laboratory test. At least two pressure stages are
examined. Furthermore, the air-tightness of the entire inner layer (gypsum board and vapour barrier)
as well as just the vapour barrier’s air-tightness are measured. This is done by removing the foil
attached according to the description in 6. Thus, the build quality of the vapour barrier is quantified.
Not the entire facing formwork of a room is examined, but particular fields between two vertical
construction components. In FIG. 7 a wall is depicted schematically. For instance, three of the facing
formwork’s fields were chosen arbitrarily, the insulation material was taken out and the measuring
equipment was built in. Thus, all existing leakages in the entire construction are statistically factored
in. In FIG. 7 all possible alternatives of flowpaths are mapped. The black arrows show leakages and
air-flowpaths recorded in the measurement. At that, the facing formwork’s inner leakage is either
located in the area of a test field or an air-flowpath crosses a test field creating a rise of tracer-gas-
concentration in this area. The dotted arrows represent air-flowpaths not taken into account during the
measurement as they do not run through a test field. As all test fields of a test room are measured at the
same time, the mean airtightness per square metre of wall area can be stated.
FIG. 7. Schematic view of wall area with three test fields and possible flowpaths both taken and not
taken into account. Each test field contains multiple sensors to detect flowpaths.
In FIG. 8 some results of the in-situ measurements are presented. In this case three test fields – one in
the reveal and two in the wall left and right to the window – were analysed. The results show that the
increase of CO2 in the window reveal (field 3) and in the neighbouring test field 1 are approximately
the same. It stands to reason that a connected plenum exists between these two test fields, whereas test
field 2 shows no connecting flowpath to the other test fields. The measurements were done at three
stages of pressure differences. Thus, the coefficients C and n could be determined.

C n
testfield 1 2 E -5 0,694
testfield 2
5 E -7 1,587
testfield 3
2 E -6 1,786
FIG. 8. CO2-concentration in three fields of facing formwork (left); Coefficients C and n (right)
4. Conclusions
Measurements in the laboratory test rig were conducted with and without insulation in the facing
formwork. The results clearly show that the insulation material must be removed from the construction
to obtain a defined volume of air which then serves as basis for evaluation. Furthermore, it was shown
the air-volume in the test field examined in the facing formwork has to be thorough mixed with small
fans to achieve an average airtightness. The measurements in the laboratory test rig have also shown
that CO2 entering the facing formwork does not only depend on leakage size and difference in
pressure, but also on the dimension of the connection hose between facing formwork and outside air
(i.e. the outer leakage). Because of this connection the mass-flow into the construction can be depicted
in dependence of the relation between inner and outer leakage. This thesis will be the topic of future
laboratory tests.
The in-situ measurement on a real object has shown that all test fields of a cohesive facing formwork
provide similar results. It may thus be assumed that the tracer gas does spread immediately after
entering the construction and forms flowpaths to neighbouring rooms. This means that more test fields
must be arranged for more air-tight constructions to gather a statistic figure for the construction’s air-
tightness. The sizing of the artificial outer leakage in the form of a hose makes for further need for
research.
References
Bednar T. & Deseyve C. & Jung M. & Nusser B. & Teibinger M. 2010. Impact of Airflow on the Risk Assessment of Flat
Roofs. in: 7th International Conference on Indoor Air Quality, Ventilation & Energy Conservation in Buildings -
IAQVEC 2010, Syracuse University, (2010)
Hall M. & Hauser G. 2003. In situ Quantifizierung von Leckagen bei Gebäuden in Holzbauart.
Sandberg P.I. & Sikander E. 2005. Airtightness issues in the building process. Proceedings of the 7th Nordic Symposium on
Building Physics NSB 2005
Wahlgren P. & Sikander E. 2010. Methods and Materials for an Airtight Building. Buildings XI, 2010
Wegerer P. & Neusser M. & Bednar T. 2012. Auswirkungen der Luft(un)dichtheit auf die Feuchtebelastung von
Konstruktionen mit Innendämmsystemen. 7th International Symposium on Building and Ductwork Air Tightness in
Practice May 11 – 12, 2012, Stuttgart, Germany
EN 12114:2000 Thermal performance of buildings – Air permeability of building components and building elements –
Laboratory test method
EN 13829:2001 Thermal performance of buildings – Determination of air permeability of buildings – Fan pressurization
method

Full paper no: 14
Airtightness variation over the year

Paula Wahlgren, Senior lecturer 1
1
Chalmers University of Technology, Sweden
KEYWORDS: Airtightness, season, fan pressurization method, air leakage
SUMMARY:
An airtight thermal envelope is important to achieve buildings with high energy efficiency and
moisture safety. The airtightness is commonly measured using the fan pressurization method, and
international studies have shown large variations in the measured airtightness at different seasons. In
this paper the airtightness variation at different seasons is studied using measurements on two
wooden frame, one-family residential buildings. Numerical simulations are presented that show the
variation in airtightness due to varying air densities of the air.
The airtightness measurements in the two buildings have been performed during seven months, from
summer to winter. The trend in the measurements is that the airtightness is lower (more leaky
envelope) when the indoor air is drier (low relative humidity). Consequently, the air leakage is
largest during the winter measurements. The decrease in airtightness during the measurement period
(from summer to winter) is in the order of 8-10%.
1. Introduction
The airtightness of a building has an impact on the energy use and on the moisture safety of a
building. It also affects the thermal comfort, the air quality in a building, sound insulation and the
spread of fire gases (Sandberg et al. 2007a and 2007b). Measuring the airtightness of a building has
become more common lately, much due to the increased energy use in leakier buildings. The
airtightness of a building can be measured in order to attain a certification or on demand from a
developer. The consequences if failing the target can sometimes be severe. Therefore, it is of great
importance to obtain a correct and representative measure of the airtightness.
The airtightness in a building is created by a continuous and airtight thermal envelope. The airtight
layer in a thermal envelope can be either a thin layer, such as a polyethylene foil, a board, such as
plywood, a homogeneous construction (e.g. a concrete component) or an outer coating, such as
rendering. In all examples it is of great importance that the joints are properly sealed.
Airtightness measurements are usually performed in accordance with EN 13829:2000 (Fan
pressurization method). In this standard there are limitations with respect to the climatic conditions
during measurements. There is for example a limit on the maximum allowed wind speed and the
maximum allowed temperature difference over the thermal envelope. The purpose of the limitations is
to assure a correct measured airtightness. Nevertheless, measurements have shown that there is a
variation in the measured airtightness with respect to the time of year for the measurement. Yoshino
(2012) described variations of ±20% over the year. Boorsboom et al. (2012) analyzed airtightness
measurements, from the 80ies, made on 21 window frames mounted in masonry or concrete walls.
The average difference in air tightness between summer and winter was about 30% (higher leakage
during winter) and the maximum seasonal difference was 120%. To be noted, some window frames
showed a lower leakage rate during winter. Boorsboom et al. suggest measurements during three
subsequent seasons in order to obtain correct values. Also Kim and Shaw (1986) showed increased
leakages during winter time. The highest leakages occurred in winter and early spring, and the lowest

in late summer and fall. Two wood frame constructions were studied and the effect was more
pronounced in the leakier building. The measurements indicate that there is a correlation between
indoor humidity and envelope leakage.
There are also measurements showing a higher leakage during the summer. An example is Dickinson
et al. (1986) who, in one out of three residential, wooden frame houses, measured a lower air leakage
in winter time. They do however speculate in the influence of snow and ice on the airtightness.
Bracke et al. (2013) measured the airtightness in two new buildings during almost four months
(December to April). In the masonry building, there was an increase in air leakage over time. A
possible explanation is the different thermal expansion of the masonry/concrete structure and the
plaster that assures most of the airtightness in these buildings. This difference in thermal expansion
could create cracks in the plaster. Another possible explanation is a gradual deterioration of the
ventilation ductwork due to repeated dismantling for the preparation of the pressurization tests.
In order to investigate the possible variations in airtightness at different seasons and the relation to
climate, full scale measurements and numerical simulations have been be performed.
2. Methods
The air tightness’ variation over season and climate will be studied by measuring the air tightness of
three buildings, during one year, and by performing numerical simulations on the climate and the
effect on airtightness. The first measurements, on two one-family wooden buildings, started in June
2013. Measurements on a multi-family concrete building will start March 2014. The measurements
have been performed by SP, Technical Research Institute of Sweden. Initial numerical calculations
have been made on the influence of wind, and on air properties.
The first airtightness measurements were performed on two residential one family houses, both
located in the south west part of Sweden, one house in Landvetter and one in Sevred, located outside
Borås. The houses are light weight wooden houses in two floors (plain wood in Landvetter and light
weight wooden beams/joists in Sevred) and they both have slab on ground and cold attics. Both
buildings have mineral wool insulation and polyethylene foil on the inside (between insulation and
board) as air barrier and moisture barrier. The house in Landvetter is built in 2004 and the house in
Sevred is built in 1993. Both houses have mechanical exhaust ventilation systems.
The airtightness quantity used is air permeability, q50 (l/sm²). It is the amount of air that passes
through the thermal envelope at a pressure difference of 50 Pa, per area of thermal envelope (l/sm²).
The air flow is measured both when the building is pressurized and depressurized and the mean value
is used. The airtightness measurements are made according to EN 13829:2000, using a Minneapolis
BlowerDoor. Temperature and relative humidity, indoor and outdoor, is measured at each
measurement occasion, as well as the outdoor wind speed. Air leakage search was performed during
the first airtightness measurements (June 2013) using air velocity meter and thermal camera, and will
be performed again during winter conditions. The airtightness measurements are performed
approximately every 1.5 month in order to study the different seasons and climate conditions.
3. Numerical simulations
Simulations have been performed in order to study how the measured air flow is affected by different
densities of the air. The density of the air is different due to variations in temperature and relative

humidity over the year. Both high temperatures and high relative humidities result in low densities.
Numerical investigations have also been performed on the effect of wind.
During a pressurization test, the air is drawn into the building through leakages in the air barrier of the
thermal envelope. These leakages can have different geometry and surface roughness and the air flow
can also pass through an air permeable material. The magnitude of the air flow depends on the
pressure difference over the leakage path, of the characteristics of the leakage path, but also on the
characteristics of the air. The density of the air changes due to temperature and relative humidity and
is thus not the same for all measurement conditions. When measuring according to standard
EN 13829:2000, the density change that affects the measurement equipment is corrected for.
However, there is also the air flow through the leakages, which is slightly different at different air
densities. This is investigated with numerical simulations.
The geometry of the building in the calculations is a box with an equal amount of leakages on all
sides (including the roof). The effect of air temperature and relative humidity (RH) is determined by
investigating two temperature conditions, -20°C and 30°C (RH=40%), and two relative humidity
conditions, 0 and 100% (Tmean=10°C). A pressure of 50 Pa is used for the calculations and the
resulting airtightness of the building is approximately 0.8 l/sm² (varying slightly in the different
cases).
Assuming simple leakage geometry with a gap between two plates, the expression for the air flow, Ra
(m³/s), as a function of pressure difference, ∆P (Pa) and leakage geometry is
∆P
Ra = (1)
Sg
12 µ ⋅ L
Sg = (2)
b2 ⋅ A
Here S g (Pa∙s/m³) is the air flow resistance, L is the length of the air gap (m), µ (Ns/m²) is the
dynamic viscosity, b (m) is the height (or width) of the gap and A (m²) is the entrance (or exit) area. In
order to include the entrance and exit pressure losses, the air flow resistance at entrance and exit S e
needs to be known. The air flow resistance S e depends on the air flow which requires the use of a
parameter S e' (Pa/(m³/s)²).
1.8 ⋅ δ a
S e' = (3)
2 ⋅ A2
The density of the air δ a (kg/m³) is included. Total air flow through a crack is determined by
1
Ra = ( S g2 + 4 ⋅ ∆P ⋅ S e' − S g ) (4)
2 ⋅ Se
'
As previously mentioned, the leakages are evenly distributed over the building. By investigating
extreme temperatures and relative humidities, the factors that possibly influence the airtightness
measurements are determined.
The investigation shows that the temperature of the air can affect the measurements by affecting the
air flow through the air gaps, see Table 1. Using -20°C, the measured airtightness is 0.76 l/m²s and at
a temperature of 30°C the measured airtightness is 0.81 l/m²s. Consequently, the difference between
the two extreme measurement situations is 5.7%.

Different relative humidities, however, have minor importance. At 0% relative humidity the
airtightness is 0.789 l/m²s and at 100% it is 0.791 l/m²s. The difference is 0.14%, thus negligible. The
effect of the air density on the measurement equipment (not leakages as above) is compensated for
when measuring according to EN 13823:2000.
TABLE 1. Airtightness at different temperatures and relative humidities.
Air tightness at Air tightness at Air tightness at Air tightness at

minimum temperature maximum temperature minimum relative maximum relative
(l/sm²) (l/sm²) humidity (l/sm²) humidity (l/sm²)
0.763 0.807 0.789 0.791
5.7% 0.14%
The simulations on wind are made to investigate if different wind speeds give different airtightness
results. Since the average wind speed can be different during different seasons of the year, this could
be a part in explaining why different seasons have different airtightness.
In the standard, it is noted that if the meteorological wind speed exceeds 6 m/s or reaches 3 on the
Beaufort scale, it is unlikely that a satisfactory zero flow pressure difference will be obtained. Three
kinds of zero flow pressure is measured before and after the pressurization test and if either of these
zero flow pressures is over 5 Pa, the test does not meet test conditions according to EN 13823:2000.
The shape of the simulated building is quadratic, with a flat roof. The shape factor Cp (-), that
determines the pressure difference over a wall at a certain wind influence, is for the windward wall,
0.4, of the leeward wall, -0.2, and for the other walls, -0.3. The roof is -0.6 and the shape factor of the
inside of the building, Cpi, is determined by a mass air flow balance. The pressure difference, ΔP (Pa),
over a wall subjected to wind is
v2
∆P = (C p − C pi ) ⋅ (5)
2
The building is first simulated with equally distributed leakages, and then with a windward side that is
twice as leaky as the other sides.
The results from the simulations show that there in many cases is a small difference in the measured
airtightness values for pressurization and depressurization when wind is present. However, the
average value is not affected until the wind speed increases. For example, at a wind speed of 9 m/s,
the building is estimated 2% more airtight with wind than without wind. At a wind speed of 9 m/s, the
zero flow pressure difference is most likely exceeding the value accepted in EN13823:2000. In the
simulations, a higher wind speed resulted in lower calculated air permeability (more airtight building).
For the case of a non-uniform air leakage distribution, simulations were made for a wind speeds up to
9 m/s. The results are similar to those of the equal leakage distribution, i.e. unless the wind speeds are
high there is little error due to wind.
4. Airtightness measurement results

Full scale airtightness measurements have been performed in the buildings described in Section 2.
The families living in the buildings are not at home at the time of measurement. Consequently, there
is only little moisture production in the buildings at the time of measurement. The moisture supply is
in all cases but one below 2 g/m³. The lowest measured relative humidity outdoor is 54% (July,
Landvetter) and it reaches 100% in November in Landvetter. Indoor, the relative humidity ranges
from 23% (January, Landvetter) to 60% (September, Landvetter).

In the house in Sevred, there is a constant decrease in relative humidity indoor from July to
November, while the house in Landvetter has the highest indoor relative humidity at the measurement
in September. Both buildings have the lowest relative humidity indoor at the winter measurements.
The wind speed is low, at most 3.5 m/s during the measurements.
Both buildings have the lowest airtightness at the winter measurement (January-February). This
coincides with the lowest indoor air relative humidity. The measured airtightness (expressed as air
permeability) as a function of time is shown in Figure 1, and as a function of relative humidity indoor
in Figure 2.
FIGURE 1. Measured airtightness as a function of time for the two buildings.

FIGURE 2. Measured airtightness as a function of relative humidity indoor.
There is an increase in air leakage over time. The increase from the first measurement is 10% for the
building in Sevred and 7.7% for the building in Landvetter. For both buildings, the airtightness is less
when the indoor relative humidity is the lowest. There is no clear correlation with indoor or outdoor
temperature, or outdoor relative humidity. The variations in airtightness can be noticeable for
stakeholders aiming for a certification.
The measurements will continue and will also include a concrete building (from March 2014). In the
two measured wooden buildings shown above, the reason for the increase in permeability could be
that the wooden construction dries when the relative humidity indoor decreases. The main leakages in
the leakiest building (Landvetter) are found around the windows in the bottom floor, around the attic
hatch (see Figure 3) and at the connection between the top and bottom floor (see Figure 4). The
reason is probably a poor connection between window and polyethylene for the windows, a poor
connection between hatch and polyethylene foil plus a leaky hatch (poor seal) for the attic hatch and a
discontinuous polyethylene foil in the exterior wall.

FIGURE 3. Thermographic image of a leaking attic hatch.
FIGURE 4. Thermographic image of the connection between the upper and lower floor, the brighter
diagonal part in front is the hand rail in the stairs.
5. Conclusions
Airtightness has been measured in two residential wooden buildings. The trend in the airtightness
measurements is that the airtightness is lower when indoor air is drier. The winter measurements (last
reported measurements) have the lowest airtightness of all measurements during the seven months
that the measurements have been ongoing. The decrease in airtightness during the measurement
period is in the order of 8-10% (from July to February).
The simulations show a small change in air flow through the leakages due to high or low air
temperatures (affecting the air density of the air that flow in the leakages), but no change in air flow
due to different relative humidities. The effect of the air density on the measurement equipment (not
leakages) is compensated for when measuring according to EN 13823:2000.
The measurements will continue to cover a whole year, and in March 2014 a concrete building will be
included.

6. Acknowledgements
The research and the measurements have been financed by SBUF, the Development Fund of the
Swedish Construction Industry, and supported by FoU-väst (regional comittee), which is greatly
appreciated.
References
Borsboom W. and de Gids W. 2012. Seasonal variation of facade airtightness: field observations and
potential impact in NZEB. Proceedings of the AIVC-TightVent International Workshop, Bryssel,
Belgium, March 2012
Bracke W. Laverge J. Van Den Bossche N. Janssens A. 2013, Durability and measurement uncertainty
of airtightness in extremely airtight dwellings. Proceedings of the AIVC-TightVent International
Conference, Athens, Greece, September 2013
Dickinson J.B, Feustel H.E. 1986. Seasonal variations in effective leakage area, Thermal performance
of the exterior envelops of buildings III, Atlanta, ASHRAE, 144-160.
Kim A.K. Shaw C.Y. 1986. Seasonal Variation in Airtightness of Two Detached Houses, Measured Air
Leakage of Buildings, ASTM STP 904, 1986
Sandberg P-I. Sikander E. Wahlgren P. Larsson B. 2007a. Lufttäthetsfrågorna i byggprocessen- Etapp
B. Tekniska konsekvenser och lönsamhetskalkyler, SP Report 2007:23 (in Swedish only)
Sandberg P-I. Bankvall C. Sikander E. Wahlgren P. Larsson B. 2007b. The effects and cost impact of
poor airtightness- Information for developers and clients. Proceedings of the Thermal Performance
of the Exterior Envelopes of Whole Buildings X, Florida, USA, December 2007
EN 13829:2000 (Thermal performance of buildings - Determination of air permeability of buildings -
Fan pressurization method (ISO 9972:1996, modified))
Yoshino. H. 2012. System for ensuring reliable airtightness level in Japan. Proceedings of the AIVC-
TightVent International Workshop, Bryssel, Belgium, March 2012

TOPIC
Building Envelope Systems
Page....................................119-320

Full paper no: 15
Performance of internal wall insulation systems - experimental

test for the validation of a hygrothermal simulation tool
Valentina Marincioni, M.Sc. 1
Hector Altamirano-Medina, Ph.D. 1
Ian Ridley, Ph.D. 2
1
University College London, UK
2
RMIT University, Australia
KEYWORDS: internal wall insulation, hygrothermal simulation

SUMMARY: In the UK, transient models of heat, air and moisture transport (HAMT) are common
tools used by building practitioners to better understand moisture movement within building elements
and construction systems. Enforced by BS 5250:2011, hygrothermal simulations are also used for
condensation risk analysis and to estimate the likelihood of mould growth and fabric decay. This
paper describes the methodology applied in the validation of a hygrothermal-modelling tool used in
the evaluation of internal wall insulation. Wall assemblies typically constructed for internal insulation
were exposed to transient boundary conditions derived from vapour pressure profiles and their
response to step changes and fluctuations were analysed. The wall assemblies were constructed using
one wall substrate (aerated clay blocks and gypsum plaster) and eight commonly used internal
insulation systems. Relative humidity and temperature levels measured at the interface between the
wall substrate and each insulation system were used to assess the hygrothermal performance of each
insulation system. As a result, the wall assemblies were clustered in three subgroups; dense capillary-
active insulation, lightweight vapour-permeable insulation and synthetic vapour-closed insulation,
and the hygrothermal performance of the proposed clusters compared with the results provided by the
simulation tool. It was found that simulated assemblies have similar hygrothermal performance as
those monitored.
1 Introduction
In England, approximately 6.5 million homes are built of solid wall – 31% of the total housing stock,
of which around 60% have been built before 1920. Solid wall dwellings are considered “hard-to-treat-
homes”, since they cannot be upgraded with easy or cost-effective fabric energy efficiency measures
(BRE, 2008). Improving the energy efficiency of these dwellings becomes even harder in conservation
areas, listed buildings, or building with decorative façades where the only feasible solution is internal
wall insulation (IWI); planning permission for external wall insulation (EWI) is often denied.
However, the installation of IWI may affect the interstitial temperature and vapour permeability of the
building envelope leading to moisture accumulation and the reduction of the building durability; high
interstitial relative humidity is ideal for mould growth and timber decay.
This paper describes the experimental test carried out for the validation of a numerical tool for heat,
air and moisture transport. The experiment was designed to help understand the hygrothermal
behaviour of internal wall insulation exposed to transient boundary conditions of relative humidity and
temperature, and to validate a simulation tool commonly used to estimate moisture movement within
building elements and the likelihood of mould growth and fabric decay in buildings. Two walk-in
environmental chambers are utilised for the experiment; wall samples were exposed to climate
conditions set independently in each chamber.

2 Methodology
2.1 Experimental method
Eight internal wall insulation systems were built and assessed under transient boundary conditions of
temperature and relative humidity; these were controlled to define specific vapour pressure levels, to
trigger vapour diffusion (varying direction and magnitude) and enhance moisture transfer within the
construction assemblies. Similar methodologies have been used for the analysis of the hygrothermal
behaviour of internal wall insulation systems exposed to a winter condition, combined with X-ray
tomography on moisture distribution in samples (Vereecken and Roels 2011) and for the analysis of
timber frame wall samples under external vapour pressure excess (Carmeliet and Derome 2012).
Temperature and relative humidity at the critical interface between the masonry substrate and the
insulation system were measured and the data were used in the validation of a numerical tool for the
evaluation of capillary active internal wall insulation.
2.1.1 Wall assembly

A partition wall between the two environmental chambers was constructed considering assemblies of
one wall substrate and eight different internal insulation systems (Figure 1). The wall substrate
consisted of 175mm-thick aerated clay block and 10mm-thick gypsum plaster (outside to inside). The
eight wall samples are described in Figures 2-3 and considered capillary active and conventional
insulation technologies. The assemblies were constructed as individual units to avoid moisture
movement between them; each side of the assemblies was sealed using a polyethylene membrane,
leaving only the surface of the aerated clay block and the surfaces of each insulation system in contact
to the set exterior and interior climate conditions.
a) b)
FIG 1. Test wall built between the environmental chambers a) view of eight insulation systems exposed
to internal boundary conditions b) view of clay block wall exposed to external boundary conditions

Lime plaster 8 mm Gypsum plasterboard 12.5 mm
Dense woodfibre 100 mm Dense woodfibre 100 mm
Lime plaster 5 mm Lime plaster 5 mm
Gypsum plaster 10 mm Gypsum plaster 10 mm
Aerated clay blocks 175 mm Aerated clay blocks 175 mm
1. 2.
Gypsum plasterboard 12.5 mm Gypsum plasterboard 12.5 mm
Light woodfibre 80 mm Vapour Control Layer (VCL 1) 1 mm
Wood-clay board 20 mm Light woodfibre 100 mm
Lime plaster 5 mm Gypsum plaster 10 mm
Gypsum plaster 10 mm Aerated clay blocks 175 mm
Aerated clay blocks 175 mm
3. 4.
FIG 2. Construction assemblies of the capillary active insulation systems, samples 1 to 4

Aluminium foil 0.1 mm Phenolic foam 60 mm
Phenolic foam 60 mm Aluminium foil 0.1 mm
Air layer 5 mm Air layer 25 mm
Gypsum plaster 10 mm Gypsum plaster 10 mm
Aerated clay blocks 175 mm Aerated clay blocks 175 mm
5. 6.
Air layer 25 mm Vapour Control Layer (VCL 2) 1 mm
Aluminium foil 0.1 mm Mineral wool 90 mm
Polyisocyanurate (PIR) 60 mm Gypsum plaster 10 mm
Aluminium foil 0.1 mm Aerated clay blocks 175 mm
Gypsum plaster 10 mm
Aerated clay blocks 175 mm
7. 8.
FIG 3. Construction assemblies of the conventional insulation systems, samples 5 to 8
2.1.2 Boundary conditions

Profiles of temperature and relative humidity (Table 1) were used to create two vapour pressure
gradients between the environmental chambers and within the insulation samples. The first set of
boundary conditions (Set 2) considered an internal vapour pressure excess of 400 Pa (pv,int – pv,ext =
400 Pa), while the second set (Set 3) an external vapour pressure excess of 400 Pa,(pv,ext – pv,int = 400
Pa). Vapour pressure pv (Pa), was calculated using equation 1, where T and ϕ represent temperature
(ºC) and relative humidity (%) respectively.
2 . 62
p =ϕ ∙ exp (22.565- - .5 ) (1)

Hygrothermal equilibrium within the wall was achieved by setting similar profiles of temperature and
relative humidity (initial conditions) in both chambers. Samples were exposed to T=15 ºC and ϕ=80 %
for a period of 30 days and to Set 2 and Set 3 for 30 days and 13 days respectively.
TABLE 1.Boundary conditions for experimental test
Internal External Average
T (ºC) ϕ (%) pv, int (Pa) T (ºC) ϕ (%) pv, ext (Pa) Δpv = pv,int
– pv,ext (Pa)
Set 1 (initial 15 80 1364 15 80 1364 0
conditions)
Set 2 18.4 ± 1.5 58 ± 2 1228 ± 158 6.6 ±1.9 85 828±109 400
Set 3 18.4 ± 1.5 29 ± 5 614 ± 164 11.1 ± 3.7 75 991 ± 245 -377
2.1.3 Monitoring
The relative humidity at the interface between the insulation and the substrate wall was monitored
using six temperature and relative humidity sensors applied to the gypsum plaster and subsequently
covered by the insulation system (total of 96 sensors). Thermocouples and capacitive sensors were
used for temperature and relative humidity respectively. Also, room temperature and relative humidity
were monitored in each environmental chamber. Data were collected every 5 minutes, for 43 days, and
averaged every hour. The collected data were then used as the boundary conditions input in the
simulations for the tool validation.
2.2 Simulation method

The paper presents a validation of a heat, air and moisture transport (HAMT) tool; the tool used for the
one-dimensional hygrothermal simulations is WUFI® Pro, developed at Fraunhofer IBP and compliant
with EN 15026:2007. The material properties, boundary conditions and method of simulation are
described below.
2.2.1 Wall assembly

The wall samples analysed in the experiment were reproduced in the simulation tool using material
data provided by the respective manufacturers and properties taken from the simulation tool database
(Table 2).
TABLE 2. Material properties for simulation
ρ ψo Cp λ (W/mK) μ (-) at 0% RH
(kg/m ) (m3/m3)
3
(J/kg K) at 0% RH
Aerated clay block 1400 0.74 850 0.58 10
Gypsum 850 0.65 850 0.2 8.3
Lime 1600 0.3 850 0.7 7
Dense woodfibre 155 0.981 2000 0.042 3 (1.5 at 60%RH)
Light woodfibre 53 0.96 2100 0.039 1.35 (1.58 at 72%RH)
VCL 1 130 0.001 2300 2.3 3500
VCL 2 130 0.001 2300 2.3 100000
Air 1.3 0.999 1000 0.155 0.51
Aluminium foil 130 0.001 2300 2.3 1500000
Phenolic Foam 43 0.95 1500 0.04 30

2.2.2 Boundary conditions
The boundary conditions considered in the simulation were those monitored in the experimental test
and differed slightly from those set in the experiment. Boundary conditions used in the simulation are
described in Table 3.
Initial conditions used in the simulation were those measured at the interface between the insulation
systems and the wall substrate when the samples were in equilibrium and Set 2 was introduced (time t
= 0) as shown in Table 4.
TABLE 3. Boundary conditions for simulation
Internal External Average
Δpv = pv,int –
pv,ext (Pa)
T (ºC) ϕ(%) pv (Pa) T (ºC) ϕ (%) pv (Pa)
Set 1 (initial 15 80 1364 15 80 1364 0
conditions)
Set 2 17 ± 1.5 60 ± 2 1163 ± 150 6.9 ± 1.9 81 ± 5 806 ± 156 357
Set 3 17 ± 1.5 31 ± 4 601 ± 135 10.8 ± 3.4 76 ± 1 984 ± 237 -383
TABLE 4. Interstitial initial conditions for simulation
T (ºC) ϕ (%)
Sample 1 15.5 83.8
Sample 4 15.5 74.36
Sample 5 15.7 84.11
Sample 6 15.5 86
2.2.3 Monitoring methodology

Similar to the experimental test, the simulation was set to last for around 42 days. Temperature and
relative humidity levels were monitored using virtual monitoring sensors positioned between the wall
substrate and the insulation, in the layer representing gypsum plaster.
3 Results and discussion

The experimental test was designed to generate data for the validation of the heat, air and moisture
transport model as well as to understand the effect of environmental conditions on the moisture levels
at the interface between wall substrate and insulation. The relative humidity at the substrate-insulation
interface of 8 internal wall insulation systems was measured and analysed considering the level of
humidity and the fluctuation amplitude of the humidity curves after the steps change between the two
sets of boundary conditions.
An increase in the relative humidity of sample 5, sample 8 and capillary active insulation systems,
(Δϕ/Δt = 0.283 to 0.779 %/h) was observed in the first 12-hour period after the step change between
Set 2 and Set 3. On the other hand, the variation of relative humidity in sample 6 and sample 7 was
found to be minimal (Δϕ/Δt = 0.049 and -0.108 %/h respectively). Similar results were observed after
the step change between Set 2 and Set 3; the relative humidity of capillary active systems, sample 5
and sample 8 decreased with a rate of Δϕ/Δt = -0.124 to -0.048 %/h, whereas sample 6 and sample 7
show negligible variations (Δϕ/Δt = -0.007 and 0.015 %/h).
Samples 1, 2, 3, 6, 7 showed a daily fluctuation of negligible amplitude. Sample 5 on the other hand,
showed a small fluctuation under Set 2, more visible during Set 3. Samples 4 and 8 presented visible
fluctuations throughout the test.

TABLE 5. Results
Set 2 Set 3
Δϕ/Δt (%/h) Peak-to-peak Δϕ/Δt Peak-to-peak
amplitude (%) (%/h) amplitude (%)
Sample 1 0.283 0.48 -0.105 0.39
Sample 2 0.309 0.32 -0.065 0.46
Sample 3 0.308 0.40 -0.048 0.42
Sample 4 0.613 1.18 -0.199 1.34
Sample 5 0.291 0.76 -0.078 1.06
Sample 6 0.049 0.32 -0.007 0.48
Sample 7 -0.108 0.55 0.015 0.42
Sample 8 0.779 1.78 -0.124 1.53
FIG 4. Hygrothermal response to a step change

The results showed that the hygrothermal performance of the insulation systems does not relate
exclusively to their generic materials (e.g. wood). For instance, samples 1, 2, 3 – all based on
woodfibre – show a comparable hygrothermal performance, while samples 4 and 8 have a similar
performance but different materials (woodfibre and mineral wool respectively). For this reason, the
samples were clustered in three groups according to their hygrothermal behaviour: Cluster A (samples
1, 2, 3) included samples largely affected by the step change between boundary conditions but
showing a negligible amplitude of the daily fluctuations; these samples feature woodfibre-based dense
boards with low vapour diffusion resistance coefficients. Cluster B (sample 4 and sample 8) included
samples affected by the step changes and with significant daily variations; these samples are made of
lightweight fibrous material (light woodfibre batt and mineral wool respectively) and feature a vapour
control layer with low to medium vapour diffusion resistances (sd=3.5 m and sd=100 m are the
respective equivalent air layer thicknesses). Cluster C (sample 6 and sample 7) included samples
which presented minor correlations between boundary conditions and interstitial relative humidity;
common characteristics of these samples are the presence of foam insulation (preventing moisture
movement) and the use of a highly resistant vapour barrier, with μ=1500000. Sample 5, which has
similar material properties to sample 6 (see Figure 3), presented a hygrothermal behaviour comparable
to samples 4 and 8; this is due to undesired air convection occurring in the sample.

A comparison between the experimental test and the simulations was carried out taking into account
the measured and calculated profiles of relative humidity. A sample of each cluster was selected and
simulated. Results of the hygrothermal tool were considered acceptable if the profile of predicted
relative humidity fell within the data intervals monitored, including a variation/error (εϕ) in the
measured relative humidity of ± 3.5 %. Sample 1, sample 4 and sample 6 were used as representative
of Cluster A, Cluster B and Cluster C respectively. Sample 5 was also simulated and compared to the
measured data.
Results of the simulations showed that the predicted profiles of relative humidity were most of the
time in agreement with the measured data. Relative humidities of sample 6, fell the entire period
within the range of measured data, whereas relative humidities of samples1 and 4 were at times
slightly off from the range of the monitored data, yet following a similar trend. It might be possible
that there was a higher actual error of the measured relative humidity; at low temperatures the error
(εϕ) of the relative humidity sensors could be as high as ± 30 % (Fossa and Petagna, 2004). In contrast,
the trend and the levels of the relative humidity modelled for sample 5 were completely different to
the monitored data. These results suggest a likely inaccuracy when constructing the sample; samples 5
and 6 have similar materials and construction assembly, yet dissimilar measured and modelled data.
FIG 5. Comparison between experimental data (solid line) and simulation results (dashed line).

4 Conclusion
For certain building typologies, internal wall insulation is the most likely measure for increasing the
walls thermal resistance. However, it may have other effects; internal wall insulation may as well help
to reduce the temperatures and increase the vapour diffusion resistance of the retrofitted walls, raising
the relative humidity at the existing wall-insulation interface, therefore increasing the risk of mould
growth and timber decay.
This paper has presented the methodology and results of a test developed to help understand the
hygrothermal behaviour of internal wall insulation and to provide data for the validation of numerical
simulation tools. The hygrothermal behaviour of eight internal wall insulation assemblies exposed to
transient boundary conditions of relative humidity and temperature were analysed by measuring the
temperature and relative humidity at the interface between the masonry substrate and the insulation
systems. Results of the experimental work showed that the insulation systems had a different response
to same boundary conditions and the interstitial relative humidity varied considerably. The difference
in hygrothermal behaviour was not only related to the generic materials used in the insulation
assemblies, but to the specific properties of the insulation system. Similar differences were observed
when the assemblies were modelled and the hygrothermal simulation tool validated.
Acknowledgments
The authors would like to thank Natural Building Technologies and Technology Strategy Board who
funded the research project.
References
BRE. 2008. A Study of Hard-to-treat Homes Using the English House Condition Survey. Part 1 —
Dwelling and Household Characteristics of Hard-to-treat Homes. Watford, UK, Building Research
Establishment.
Carmeliet J. & Derome D. 2012. Temperature driven inward vapor diffusion under constant and cyclic
loading in small-scale wall assemblies: Part 1" Building and Environment. 48-56.
Fossa M. & Petagna P. 2004. Humidity measurements inside Atlas and CMS: notes on sensor
calibration. [http://proj-jcov.web.cern.ch/proj-
JCOV/EBmeeting_22/040422_Rh_Sensor_calibration.pdf accessed on 13/12/2013]
Vereecken E. & Roels S. 2011. Hygric performance of different interior insulation systems: an
experimental comparison. 12th International conference on Durability of Building Materials and
Components, Porto, Portugal.

Full paper no: 16
Watertightness of window-wall interfaces in wood-frame

construction
Nathan Van Den Bossche, Assistant Professor1
Glenn De Meersman, M.Sc. 1
Lynn Devos, M.Sc. 1
Silke Maertens, M.Sc. 1
Arnold Janssens, Full Professor1
1
Ghent University, Belgium
KEYWORDS: Watertightness, airtightness, window-wall interface, wood-frame construction.
SUMMARY:
In recent years there has been an increased use of wood-frame construction in Belgium, which
likewise requires an increased focus on both hygrothermal analysis as well as best practice
guidelines to ensure the longevity of this construction type in a temperate maritime climate. This
experimental research investigates both the airtightness and watertightness of window-wall
interfaces, and the interrelation between both. Three different approaches to secure the
watertightness on the exterior side were evaluated, more specifically the use of a watertight layer,
spray-in-place polyurethane foam, and self-expanding tapes. The sensitivity of these installation
methods was analyzed in respect to the airtightness of the window-wall interface. Based on 3
airtightness performance levels, it was found that the airtightness, and hence the pressure
equalization, has a significant effect on the risk for water ingress. Furthermore, the amount of water
ingress was also quantified under a range of pressure differences under both static and dynamic
boundary conditions. This quantitative approach is also of importance to evaluate the moisture
sensitivity and moisture tolerance of wood-frame constructions. Finally, the paper highlights typical
installation errors, and provides practical guidelines for airtight window installation in wood-frame
construction.
1. Introduction
The market share of wood-frame construction in Belgium is increasing. Studies show that the market
share of wood-frame construction for newly built residential dwellings was 5.6% in 2004, it rose to
11% in 2009, and it is expected to increase up to a market share of 20% by 2020 (BBRI, 2020). In
Belgium, residential buildings are typically constructed with brick cavity walls with about 8 to 20cm
of insulation in the cavity. Consequently, the majority of technical guidelines and sector documents
refer to that building practice, and little to no information is available on wood frame construction.
Furthermore, Viitanen et al. (2010) report the modelled mass loss of pine sapwood for all of Europe,
which shows that the climatic conditions in Belgium are among the worst in respect to wood decay.
In Belgium, the BBRI (2007) reported that 24% of all interventions for buildings relate to moisture,
and water infiltration is the most important cause of building damage (condensation, absorption due to
capillary action and initial moisture content are excluded). Brick cavity walls are the predominant
construction type in Belgium, and this type of wall has the specific capacity of absorbing significant
amounts of water without leading to premature deterioration or failure. Commonly, there is an exterior
brick leaf with solid bricks, a ventilated air cavity, insulation, an extruded and perforated interior
brick wall, and a stucco finish. A limited amount of inadvertent water ingress past the insulation layer
may be stored by the interior brick wall, and due to the high vapour permeability of the construction

on both sides, the water may dry out without causing any damage. In general, water ingress will only
cause damage when the storage and drying capacity of the construction are superseded, from which it
may be evident that wood frame construction will be more prone to risk of failure when occasional
water ingress occurs.
Wood frame construction is a relative new typology in Belgium, the Belgian climate shows to be
rather severe in respect to wood conservation, moisture problems are the predominant type of building
mitigation, and wood frame construction is probably less robust when considering occasional water
ingress. Consequently, more research is required to evaluate to what extent the introduction of a new
construction typology in a county with a history of moisture problems may lead to excessive damage.
Based on a review of 50 buildings in Alberta, 57 out of 105 rainwater penetrations were situated
around windows and doors (Ruest, 2000). RDH (2000a, 2000b) found 15 failing window-wall
interfaces in 127 field tests in British Columbia. For the same area, a different report on 37 buildings
which experienced envelope failures showed that at least 25% of all problems were related to the
windows or the window-wall interface (Rousseau, 1999). Similar conclusions were drawn in Florida
(Lstiburek, 2005). Even though no scientific data were found for Europe, building practitioners often
report that the window-wall interface is prone to water ingress problems, and hence, the window-wall
interface was selected for this research paper. In this paper, the watertightness of several types of
window installation methods are evaluated both qualitatively as quantitatively, under static and
dynamic conditions, and taking into account a range of typical airtightness performance levels of
window-wall interfaces. In the subsequent chapters, an overview of literature on the watertightness of
window-wall interfaces is provided, followed by the description of the experimental setup, test results
and conclusions.
2. Literature on the watertightness of window-wall interfaces.

A query in the database of the European Committee for Standardization yields over 2000 standards
solely related to the performance of building components and materials (excluding products related to
building services). When an architect draws a window-wall interface for a masonry brick wall, 148
standards may apply to the window, 73 to the glass, 85 to the masonry, 94 to the insulation, 70 to fire
safety, 167 to concrete, 55 to acoustics… Additionally, ISO standards, national and regional codes can
add to the complexity. Only nine standards relate to watertightness: concrete, windows, doors, curtain
walls and roof coverings are addressed (query executed 04/12/2013). For the watertightness of
window-wall interfaces, there are no specific test standards.
Nelson and Norris (2010) evaluated the watertightness of window-wall interfaces, for storefront
windows installed in stucco walls (storefront windows: US terminology for windows without flanges,
i.e. similar to European practice). Interestingly, although all perfect barrier systems failed, the authors
do not question the concept of perfect barrier systems, and only insist making it more perfect. In New
Zealand a new window installation method was advocated by the national window association
(Knowles, 2002). Here a double-stage joint is applied at the window-wall interface, but drainage is
poorly designed. The weather resistive barrier underneath the sill is not sloped, certain locations
prone to wetting are not drained, and the design of the front seal is typical for perfect barrier systems.
In 2003 it was even suggested in the New Zealand parliament to forbid the use of silicone sealant as
primary means of weathersealing systems (2003).
Lacasse et al. (2008) compared the performance of a face-sealed window installation with a rain
screen joint. It was concluded that the most airtight plane should not coincide with the rain screen.
The exterior side of the joint should not be airtight, to reduce the driving force for water to enter.
Consider that the location of the drainage plane is not clear at the top and sides. Consequently water
ingress was reported due to insufficient drainage design (Lacasse et al. 2009). Salzano et al. (2010)

also found that drained systems perform better than perfect barrier installation methods in a rain
screen wall. Conversely, for a concrete masonry wall, designed as perfect barrier system, the
installation method with a single barrier performed better. Consider that for these test results the
leakage paths were not described, the installation method in the concrete wall is not described in
detail, and the sill flashing is level instead of sloped. Why the single-joint performs poorly in one
situation and reasonably well in a different situation is not explained.
Note that the necessity to apply the rain screen principles to window-wall interfaces was already
described explicitly and in great detail in 1957 by Birkeland (1957), who ends his paper with the
sentence “Tightness is not increased by caulking the joint”. Even though the use of drained joints is
often advocated (Edgar, 1999; Finn, 1991), it seems that currently it still cannot be considered as
common building practice in the US.
3. Experimental setup
A non-operable wooden window (1.55m high by 1.23m wide) was installed in a typical wood frame
wall of 2.28m high and 1.94m wide. This wall comprised an interior airtight sheeting of taped
oriented strand boards, an air-open exterior impregnated fibre board, but for reasons of visual
inspection by means of an endoscope, no insulation was installed between the studs. On the exterior
side, the cladding consists of slender horizontal planks, which was assumed the be rather permeable
for wind driven rain. At the sill and top of the window, cross cavity flashing was installed by means of
an aluminium profile, with end dams at the sill. The wood frame wall was installed perfectly airtight
in a steel test box.
At the bottom side of the rough window opening a water collection tray and drainage system was
installed, to measure the amount of water ingress into the window wall interface. Infiltration water
from the drainage was directed to a hermetically sealed gravimetric trough and weighing system. The
drainage system was designed in such a way that the water that was drained did not occlude the
drainage tube, and in order to avoid artificial pressure gradients caused by the weighing system itself,
the collection through was connected to the window wall interface with a pressure tube. That allowed
for perfect pressure equalization, omitting any inadvertent pressure gradients due to perturbations in
the collection system.
3.1 Test protocol
There is no specific test standard to evaluate the watertightness of joints or interfaces. For windows, a
static test sequence is provided in EN 1027. The European standard EN 12865:2001 describes a test
protocol to determine the resistance of external wall systems to driving rain under pulsating air
pressure. According to this standard, water is sprayed at two locations: 72l/h-m at the top as run-off,
and 90L/h-m² on the whole area to simulate direct Wind driven rain. The pressure difference is
applied in a step-wise approach to determine at which level water entry occurs. Furthermore, the
standard offers two procedures: one for qualitative testing (A - steps of 10 minutes after 20 minutes of
initial wetting), and the other for quantitative testing (B - steps of 60 minutes). Although there is no
standard for window-wall interfaces, the interface is perhaps expected to perform well according to
both test sequences (windows and walls). Consider that the EN 12865 only provides a test sequence,
without information on the required performance levels as a function of exposure to climate
conditions. Consequently, the samples will be tested according to the static test protocol up to 750Pa
(EN 1027) and the cyclic test protocol up to 600Pa (EN 12865).

Figure 1. Dynamic test sequence according to EN 12865.
To assess the effect of the airtightness of the window-wall interface on the risk for water ingress, the
airtightness of every setup was also determined based on EN 12114. Please refer to (Van Den Bossche
et al., 2012) for more information on the airtightness measurements and error analysis.
3.2 Test parameters
Four different parameters were varied during the test sequence: mechanical installation of the window
to the wall, installation method to ensure the watertightness, position of the window in respect to the
wall, and the level of airtightness of the interface.
3.2.1 Mechanical installation
In Belgian brick construction practice, windows are typically fixed to the wall using steel brackets.
Direct fixation with screws through the window frame itself – typical for German window installation
in ETICS – is not advocated in Belgium, and flanged windows are not available. The brackets allow
for a practical and flexible installation, but conversely hamper the easy installation of airtightness
membranes at the window perimeter. The alternative is a plywood casing around the window frame,
which in turn replaces the brackets. The plywood provides a solid ground for the finish of the window
reveal, and allows to secure the airtightness easily.
3.2.2 Position of the window in respect to the wall
The position of the window frame might depend on the alignment with the thermal insulation,
esthetical or mechanical boundary conditions, but by consequence will also affect the exposure to
wind driven rain. Four positions are considered (see figure 2): behind the impregnated fibre board (a),
onto the back of the fibre board (b), aligned with the fibre board (c), or projected outwards (d). Mind
that the critical pathways for water ingress also shift from the bottom to the top as the position of the
window shifts to the exterior. The highest risk for infiltration is located at the corners, where –
depending on the position of the window – complex folding may be required to ensure the continuity
of the foil. When the window is aligned with the fibre board, the foil can easily be installed with a low
risk for installation errors.

a b c d
Figure 2. Position of the window in respect to the wall.

3.2.3 Installation method
The watertightness of the interface is primarily determined by the joint between the window frame
and the impregnated fibre board. For this joint, three options were evaluated: foil, self-expanding
sealant tape, and spray in place polyurethane foam. On one side the foil is adhered to the window
frame with caulking, on the other side there is an adhesive with release liner to adhere it to the wall.
The self-expanding sealant tape consists of an open cell impregnated foam, that is designed to act as a
single stage joint for watertightness (type BG1). The spray in place polyurethane foam is a one-
component flexible and low expansion foam. The main advantage of flexible foam is that it is capable
of following the movements of the window frame induced by wind pressures, thermal gradients and
forces due to opening and closing the sash.
Figure 3. Installation with self-expanding tape (left) and foil (middle and right)

3.2.4 Airtightness
The amount of water that enters into the cavity will depend upon the pressure gradient over the
exterior, that in turn will be affected by the pressure equalisation and the airtightness of the interface.
Based on stochastic analysis of the Belgian building stock, Van Den Bossche et al. (2012) derived
three performance levels for the airtightness of window wall interfaces. Good installations entail an
air leakage below 0.33m³/h.m; poor installations cause air flows above 3.3m³/h.m, and average
performance is characterized by an air flow rates between the threshold levels reported above.
4. Results
4.1 Mechanical installation
The tests indicate that the type of mechanical installation did not affect the watertightness
performance when self-expanding sealant tape or watertightness foil was applied. This was most
evident for installations with a good airtightness, whereas installations with significant air leakages
tend to show small discrepancies. Nevertheless, differences in water ingress could be attributed to
executing performance. In general, installations using a plywood frame around the window frame
proved to be less susceptible to errors in the airtightness: the use of metal brackets typically
introduces more difficulties to ensure the airtightness, but for equal airtightness similar performances
were found. When spray in place polyurethane foam was applied, a difference in failure behaviour
was noticed. The foam itself did not show any leakage, but local deficiencies at obstacles may lead to
water ingress. When installing metal brackets, it is important to make sure that the bracket itself does
to protrude to the exterior, where it is exposed to runoff. Even though the foam was installed
carefully, it does not provide a continuous barrier. For the installation with plywood frames, small
wood blocks are used as spacer to position the window frame in respect to the rough window opening.
At these blocks, the continuity of the foam is interrupted, which also gave rise to water ingress.
4.2 Position of the window in respect to the wall
Only for the case where the watertightness was ensured by a water resistive barrier as flashing, the
effect of the position of the window was evaluated. As the window is located more inwards, the
flashing at the bottom is more exposed, and inadequate drainage may lead to local accumulation,
hydrostatic pressure and infiltration. Likewise, as the window is located more towards the exterior,
the flashing at the top is similarly more exposed. However, the results show that more problems can
be expected at the sill, as the window itself causes a high runoff rate at the sill, whereas at the top
most of the water is diverted by the cross-cavity flashing.
4.3 Installation method and airtightness
The installation with sealant tape did not perform well. At the sill, the installation of the sealant tape
apparently led to accumulation of water, which caused a hydrostatic pressure. Next to that, additional
repetition tests showed that the specific sealant tape used in this study displayed shrinkage and
swelling behaviour under wetting and drying cycles. With each cycle the sealant tape shifted slightly
downwards due to gravity. Consequently, horizontal joints should be avoided with this type of tape,
and otherwise these must be shielded from rain and runoff as much as possible. The compression of
the tape should be adequate to prevent any sinkage of the tape. For a good airtightness the amount of
water infiltrating to the interior was rather limited, whereas the water ingress increased as the
airtightness decreased. The results reported in table 1 are those for window location b. The
installation with a self-adhering flashing membrane to the window frame proved to be very
susceptible to execution errors. The adhesion was found inadequate, and the joint between window
and flashing was henceforth sealed by means of caulking (reported in the table below). Without
caulking, water ingress rates of 14,7 up to 34,6 gram/min were measured under static conditions, and
also under dynamic conditions the infiltration rates were higher. Table 1 again indicates that a good

airtightness has a beneficial effect on the occurrence and amount of water ingress. Even though the
average airtightness seems to entail a better watertightness compared to the airtight situation, this can
be attributed to the premature failure of the flashing adhesion: during the cyclic loading the membrane
failed, leading to high air flow speeds and water ingress. Furthermore, it may be evident that the
amount of water ingress is also affected by the exposure to direct rain impingement and runoff. By
installing an additional shielding that diverts most of the water away from the flashing, the infiltration
rate reduced by as much as 63%. Additionally, tests were done on a different damage pattern as well:
during installation of the sill, local forces may induce excessive stress in the adhesion of the flashing.
A representative installation error increased the infiltration rate from about zero to 23.2, 14.2 and 12.3
gram/min under static test conditions for a poor, average and good airtightness respectively, and
likewise an increase for dynamic conditions to 27.8, 18.5 and 18.5 gram/min.
Table 1.Overview of a selection of test results
Water ingress (gram/min) Pressure at infiltration (Pa)
Rain barrier Airtightness
Static Dynamic Static Dynamic
Poor X X 0 0
Sealant tape Average 13,33 9,9 0 0
Good 0,48 0,15 300 450
Poor 6,6 9,5 150 0
Flashing Average 0 0 n.f. n.f.
Good 0 0,6 n.f. 450
PUR Good X X 150 300
X: water ingress could not be measured; n.f.: no failure up to at least 600 Pa. The results reported in
this table are the average infiltration rate monitored at 0, 150, 300, 450 and 600Pa for both static
and dynamic pressure conditions.
When polyurethane was installed, the cavity between the window and the rough window opening was
filled and thus blocked any infiltrating water to run down towards the collection through. Next to that,
the airtightness of the setup could not be altered easily as the polyurethane itself secured the
airtightness, and consequently only measurements were done for installations with good airtight
sealing. Table 1 does not comprise measured infiltration rates, but water infiltrated at rather low
pressure differences: 150 and 300 Pa for static and dynamic conditions respectively. Mind that the
accompanying infiltration rates were very limited. The leakage pathways that did arise could be
attributed to the presence of oversized spacers between window and rough opening. When no spacers
are present, no infiltration was evident up to 600Pa.
5. Conclusions
The watertightness of window-wall interfaces in wood-frame constructions was evaluated in lab
conditions, under static and cyclic wind pressure loadings. Two types of mechanical fixing to the
interior wall were evaluated: steel brackets and plywood casing around the window. The steel
brackets require more attention when installing the airtightness foil, but no significant effect was
found on the level of watertightness. When polyurethane foam is used, one should ensure a
continuous layer of foam, avoiding any crossing brackets or other obstructions. Tests on the position
of the window were not reported in detail, but for all positions the sill was most prone to water
ingress probably due to higher local runoff rates. The use of sealant tape led to high infiltration rates

at lower pressure differences. Even though reasonable results were found when the airtightness was
good, this installation showed to be susceptible to errors in execution as well. The results for the
water resistive barrier show a clear correlation between water infiltration rate and air leakage.
Polyurethane foam only showed infiltrations when to foam was breached by means of steel brackets
or spacers. Note that the static and cyclic watertightness tests do not lead to identical results in terms
of infiltration rates and pressure difference at moment of failure. Next to that, it is unclear which of
both is most severe: the failure initiation phenomena differ from installation to installation, leading to
different interpretation under different test sequences. Note that all results are highly sensible to
installation, materials used in the setup and craftsmanship.
References
BBRI, 2010. Research on new building concepts, and the impact on natural resources. (in dutch).
Belgian Building Research Institute, VITO, KULeuven.
Viitanen H., Toratti T., Makkonen L., Peuhkuri R., Ojanen T., Ruokolainen L., Räjsänen J., 2010.
Towards modelling of decay of wooden materials. European Journal of Wood and Wood Products
68, 3: 303-313.
BBRI, 2007. Annual report. www.wtcb.be
Ruest K., Wall moisture problems in Alberta dwellings. CMHC Research Highlights (2000) CMHC,
Ottowa, Canada.
RDH, 2000a Water penetration resistance of windows – study of manufacturing, building design,
installation and maintenance factors, RDH Building Engineering Limited, Vancouver, 2002.
RDH, 2000b Water penetration resistance of windows –study of codes, standards, testing and
certification. RDH Building Engineering Limited, Vancouver, 2002.
Rousseau M., An Overview of the Survey of Building Envelope Failures in the Coastal Climate of
British Columbia, performed by Morrisson-Hershfield Limited for CMHC (1996). Journal of
Building Physics (1999) 22: 364-367.
Lstiburek J.W., Rainwater management performance of newly constructed residential building
enclosures during august and September 2004. (2005) Building Science Corporation, MA.
Nelson C., Norris R.E., Mock-Up Water Test Results of Sample Flashing Systems for Storefront
Windows in Stucco Walls. Interface (2010) RCI inc., Raleigh, NC.
Knowles S., WANZ Window Installation System. Window Association of New Zealand. Wellington,
NZ, 2002.
House of Representatives, Weathertightness of Buildings in New Zealand. Forty-Seventh Parliamant,
March 2003.
Lacasse M.A., Rousseau M.Z., Cornick S.M., Manning M.M., Nicholls M., Nunes S., Performance
evaluation of Wall-Window Interface Details Phase 1 – Watertightness, Air Leakage and Rainwater
Management of CMHC Specified Assemblies. Report B-1229.1. National Research Council, 2008.
Lacasse M.A., Rousseau M.Z., Cornick S.M., Manning M.M., Ganapathy G., Nicholls M., Williams
M.F., Assessing the effectiveness of window-wall interface details to manage rainwater – selected
results from US window installation practice. Journal of ASTM International (2009) 6, 8: 1-35.
Salzano C.T., Masters F.J., Katsaros J.D., Water penetration resistance of residential window
installation options for hurricane-prone areas. Building and Environment (2010) 45: 1373-1388.
Birkeland Ø. The design of multi-layer walls. Report No 24, Norges Byggforskningsinstitutt. Oslo,
1957.
Edgar J., Performance of Source Drainage Exterior Insulation Finish System at the window/wall
junction. Journal of Building Physics (1999) 23: 57-77.
Finn D., Rain penetration control of the window-wall joint. Institute for Research in Construction,
NRC. NRCC-38747, 1991.
Van Den Bossche N., Huyghe W., Moens J., Janssens A., Depaepe M., 2012. Airtigthness of the
window-wall interface in brick cavity walls. Energy and Buildings 45: 32-42.

Full paper no: 17
An Outdoor Hut Testing Platform for

Evaluating Advanced Building Enclosures
Bryan Urban, S.M.1
Nitin Shukla, Ph.D.1
Anthony Fontanini, Ph.D. Candidate2
Jan Kosny, Ph.D.1
1
Fraunhofer Center for Sustainable Energy Systems, USA. Email: [email protected]
2
Iowa State University, USA
KEYWORDS: thermal testing, heat flux, insulation, R-value, measurement
SUMMARY:
Evaluating the thermal performance of novel advanced building enclosures is challenging, and there is
a need for better and more cost effective methods for doing so. Laboratory testing of components
provides accurate results under special conditions, however, these may fail to represent complex
dynamic interactions between system components. Field testing can offer more realistic performance
evaluation, however, studies involving occupied buildings suffer from many confounding factors, such
as occupant interactions and unknown building material properties that make it difficult to generalize
results. Whole building energy modeling is the ultimate goal for evaluating and designing new
building systems, however, without proper validation, the models cannot be trusted. In this paper we
present a flexible approach for cost effective evaluation field testing of building advanced enclosures.
1. Introduction
New building components or systems must be evaluated and tested before going to market. From an
energy and durability standpoint, it is desirable to measure performance on several dimensions,
including thermal performance (heat gain or loss through components), occupant comfort (temperature
and humidity in occupied zones), moisture performance (moisture accumulation and material
durability), air tightness (and its impact on energy consumption), and energy performance (HVAC
energy reduction, peak load shifting). Various testing approaches have been developed and tried in the
last century. This paper is not meant to be a comprehensive review of such methods; however, it does
illustrate some different philosophies and why the proposed approach may be appropriate for specific
testing goals.
We present a flexible, nearly-calorimetric testing platform that combines the advantages of outdoor
field testing with the controllability of a laboratory experiment. The platform is comprised of multiple
small testing structures, or huts. One hut is built to standard construction practices (e.g., typical
residential construction) and acts as a control or baseline case. The other hut(s) are constructed
identically, with the exception of the novel building materials or components. All huts are conditioned
with identical HVAC systems and control strategies, and HVAC energy consumption is metered
during the experiment. Air tightness of each hut is measured prior to each experiment. Detailed
temperature and heat flux measurements are made across assemblies of each hut. A weather station
records outdoor conditions. Resulting data may be used to support the validation of energy models for
advanced systems or to rapidly evaluate prototype building systems or components.
This paper presents the results of testing measurements made during the fall in three huts used to
evaluate an advanced insulation material. Here we focus on description of the platform, construction
and experimental technique, and the measurements obtained during testing. Through a six month

experiment, we have determined that this platform may be used to effectively test and compare
thermal performance of insulation systems.
2. Description of Testing Platform

2.1 Testing Platform Architectures
In the testing platforms we consider, test structures are used to represent a specific type of building
construction. Unlike experiments that completely isolate all components except for the one of interest,
this platform seeks to obtain results that are more representative of a fully constructed system. Since
real systems will have measurable air leakage and 3D conduction effects, it is important to test new
products in situations where these effects may be non-negligible. The dynamic interaction of heat flow
and thermal storage between various building components during the daily heating and cooling cycle
is difficult to match in a laboratory experiment.
Two main approaches (single-structure and multi-structure) and two testing methods (serial and
parallel) exist for performing direct-comparison in-situ building tests, each with distinct advantages
depending on the situation and goals of the test.
2.1.1 Serial vs. Parallel Testing

In serial testing experiments, the performance measurements of components are made in sequence.
This is akin to “before and after” testing or pre- and post- retrofit comparison. Serial testing is
challenging since environmental conditions are often not identical or equivalent during both testing
periods and it can be difficult (though not impossible) to control or adjust for variations by using
simulation or engineering calculations.
In parallel testing, all systems are tested simultaneously. In this way, uncontrollable environmental
factors are kept consistent during the test, eliminating an important source of error. When practical,
parallel testing is preferred for its higher accuracy and more consistent results for making direct
comparisons between systems.
When direct comparisons between systems are desired, parallel testing is often preferred. Parallel
testing requires more infrastructure than serial testing, since it is necessary to have enough
instrumentation for each product that will be tested. Serial testing, conversely, requires less equipment
but more labor and time to remove and replace sensors before and after each reconstruction phase. The
cost/benefit tradeoffs depend on the goals and resources of the experiment.
2.1.2 Single-Structure vs. Multi-Structure Testing

Single-structure testing, naturally, involves only one test structure. In the case of a serial single-
structure test, this single structure is used to measure both baseline and retrofit conditions through a
pre- and post- retrofit measurement study. This option suffers from uncertainty driven by
environmental variability between measurement phases. In a parallel testing single-structure
experiment, the single structure is made large enough to accommodate several different test elements
within the same building so testing can be done simultaneously. For instance, one wall could be
divided into several subsection sections, each one used to test a different type of insulation material. It
may be easier to maintain consistent and uniform internal hut conditions in a single-structure test than
in a multi-structure test, however this depends on what HVAC system types are used to control the
space temperature. If multiple systems are tested simultaneously in a single-structure test, it is
necessary to provide sufficient spacing between test elements so that lateral conduction does not
influence the measurement results.
Multi-structure testing involves constructing multiple, independent, yet identical test structures. At
least one structure represents a control or baseline condition, built to standard or typical construction

practices. The other structures are modified in some way, perhaps by with a new kind of insulation or
roofing system.
2.1.3 Examples
Two example of a single-structure test is the test room of (Athienitis et al. 1997), which compared
phase change material board temperature with that of ordinary wall board and developed mathematical
models of material performance, and the test room of (Shilei et al. 2007), which tested phase change
wallboard energy performance by measuring electric heating energy and temperature.
A more sophisticated, but slightly different type of single-structure test is a calorimetric testing
chamber that exposes one side of a test element to the exterior environment and controls the interior
side precisely. All remaining sides of the test chamber are heavily insulated and instrumented with
heat flux sensors to obtain a careful heat balance of the space. This kind of apparatus has been used
and tested extensively throughout research laboratories in Europe under various names DYNASTEE,
PASLINK, and PASYS (e.g., Baker 2008). This approach is useful for obtaining assembly-specific
parameters that can be used in simulation, but it may not capture the full system-level interactions
between multi-zone spaces.
Examples of multi-structure testing include (Szymocha et al. 2005), which tested an advanced thermal
solar system with heat storage and compared surface temperature, surface irradiance, and space
heating with three test structures; (Medina et al 2007) tested a phase change material structural
insulated panel using arrays of thermistors and heat flux sensors with two test structures; (Khudhair
and Farid 2007) tested phase change material load shifting in two test structures
In some (intentionally unnamed) cases the researchers have mistakenly relied on insufficient data to
produce conclusions about energy savings. For instance, measuring only interior temperature and
cumulative heating or cooling energy consumption is normally not sufficient to convincingly show
that (1) savings do exist, and (2) savings originate from the element in question. Accurate and
sufficient measurement is the cornerstone of any evaluation. Since, in many cases, the effect of
improving a building component can be small (say 5% improvement), it is essential to use sensitive
equipment and sample frequently enough to measure these effects reliably.
2.2 Example Multi-Structure Test Huts
2.2.1 Experiment
In this experiment, we set out to measure the thermal and energy performance of two proprietary
insulation systems against an uninsulated baseline structure. Each hut was conditioned to an identical
setpoint temperature, and allowed to run for several months during the summer and fall of 2013 in the
desert climate of Albuquerque, NM. We monitored temperature and heat flux through the assemblies,
energy consumption of the cooling equipment, and climate conditions. We used the results to analyze
energy reduction potential, peak load shifts and reductions, and zone temperature behavior.
The insulation systems were identical other than the presence of a reflective surface foil. Ordinarily,
the baseline condition would include some standard insulation materials; however, in this case, the
uninsulated case was more representative of the particular end-use application.
2.2.2 Hut Structures

We constructed three identical structures composed of uninsulated 8’x12’ wood framing with 2x4
studs spaced 16” on center. The exterior walls and roof deck are composed of oriented strand board
(OSB). A hinged door on the north-facing side of each hut is the only means of access to the interior.
The unvented gable roof consists of a layer OSB beneath a dark green colored metal roof. The huts
were positioned on the test site to minimize solar shading.

The floors are made of tongue-in-groove plywood and rest above 2x4 floor joists with the floor
cavities insulated with R-3.3 fiberglass insulation. The floors were insulated with a further R-1.8 foam
board above the plywood layer. Initially, the huts were built without a ceiling to permit the installation
of insulation on the interior side of the roof deck. In this state, the huts were tested for initial air
tightness to ensure similar air leakage rates in the baseline condition.
Figure 1: Three identical test huts, pictured from the North-east corner of the lot.
Figure 2: Left to right: uninsulated hut interior without ceiling, non-reflective insulation installed on
the roof interior of Huts 2 and 3, reflective insulation installed on the walls of Hut 3 with ceiling
installed.
The test insulation was then nailed to the wall and roof deck studs of Huts 2 and 3, while Hut 1
remained uninsulated. Afterwards, ceilings were installed in each hut to separate the interior
conditioned zone from the unconditioned attic space. Hut 2 received the insulation packaged in a non-
reflective facing surface, while Hut 3 received insulation packaged with a reflective foil surface on
both sides of the batts. The purpose of the reflective facing on the Hut 3 insulation was to reduce
radiative heat transfer across the stud cavities created by the insulation. After the insulation was
installed in Huts 2 and 3, OSB ceilings were installed on all three huts to create a separate,
unconditioned attic space.
2.2.3 Space Conditioning

One of the most important and challenging aspects of running a multi-structure experiment is in
maintaining similar interior conditions through controlled space conditioning. Several strategies exist,
each with strengths and weaknesses.
For cooling, one can select from a number of air conditioners on the market, including portable or
fixed room or window AC units, central AC units, ductless mini-split units, chillers with fan coil units,
and others. For cost and simplicity reasons, we have experimented with fixed through-the-wall style
and portable AC units. These are relatively inexpensive, and fairly straightforward to install and
replace. Disadvantages include inability to use external thermostat control and the introduction of
some additional air leakage as a consequence of the hot exhaust. In the case of a through-the-wall unit,
there will be some air leakage through the unit itself. Some portable AC units have one or two venting
ducts. Systems with two ducts (one for intake, one for exhaust) reduce pressure build-up within the
huts and result in more controlled air exhaust. Systems with one exhaust duct force air out of the hut,
creating a low pressure interior space. Exhausted air is replenished through uncontrolled infiltration.
Depending on the importance of air tightness in the experiment, there may be advantages of one of
these approaches. Ductless systems, meanwhile, do not need to exhaust air, as heat is carried away in

the refrigerant line; however these systems tend to be more expensive and it is more difficult to
monitor refrigerant temperature and flow rate.
For heating, the easiest and most straightforward approach is to use a simple fan-assisted electric
resistance space heater. Some portable air conditioners include a heating mode that works via a heat
pump or through resistance heat.
Since individual space conditioning units may have slightly different performance characteristics,
despite being of the same make and model, one should not rely on power consumption values alone
when comparing systems. It is also essential to monitor the space temperatures of all huts (of a multi-
structure test) during an initial testing period to ensure they are being cooled to a similar temperature.
If a single-structure test is used, it is important to ensure the zone temperature distribution is uniform.
In this experiment, we were principally concerned with cooling loads, so we installed a portable air
conditioning unit (4.1 kW cooling capacity) in each test hut. These units have two vents, an intake and
exhaust vent that are used to cool the condenser and expel the heat from the huts. We sealed the
perimeter of the vent penetration with caulk and expanding foam insulation to minimize air leakage
from the gaps and cracks. Additionally, we wrapped the hot air exhaust duct with fiberglass insulation
to prevent heat from conducting or leaking back into the huts. A moisture drain was installed on each
air conditioner to expel any condensed moisture outside of the hut.
3. Instrumentation and Measurement

To monitor material performance, we installed a profile of temperature and heat flux sensors across all
elements of key assembly components, including all six walls of the conditioned zones, the roof
assembly, and the exterior surface of all walls. We sampled data at 20 second intervals, providing a
high time-resolution of thermal performance.
3.1 Weather Station

As with most building experiments, we measure the usual climatic conditions, including solar
radiation, outdoor air temperature and relative humidity, and wind speed and direction. Since this work
was completed at our outdoor research facility, we made use of the existing permanent weather
stations used for photovoltaic research. For many projects, a basic logging weather station is
sufficient. If detailed calibrated simulation will be performed, more accurate measurements may be
preferred.
3.2 Temperature
We measured temperature across all key wall assembly components, as well as the interior zone
temperatures of all huts. By measuring temperature across lightweight construction materials of
known properties, it becomes possible to estimate the heat flux through those materials. This provides
a backup method for calculating heat flux through surfaces where sensors fail or where cost prohibits
their use.
The temperature measurements were obtained using NIST Calibrated type T thermocouples with
specified accuracy of ±0.5°C or 0.4%. Further laboratory tests were conducted to determine the
accuracy by measuring thermocouples in conjunction with a calibrated thermistor in a metrology well
from a range of 0°C to 60°C. To more accurately measure temperatures with our data logger, an
external reference junction box was used. The isolated isothermal external reference junction box
(zone box) is used instead of the data logger’s internal reference junction, minimizing fluctuations that
could be caused by rapid changes in the outdoor temperature.

3.3 Heat Flux
As with the temperature sensors, we installed heat flux transducers on all six walls of the conditioned
zone, and on one roof deck surface of each hut. This allowed us to construct a virtual energy balance
on the huts by estimating the heat flow through each component surface. We calibrated the sensors in a
heat flow meter, with the transducers sandwiched between layers of insulation sheets or smart boards
siding (ASTM 2012).
To keep a flat contact surface between two gypsum boards without any air gap, shallow grooves were
cut into the 6x6” board panels to make room for the heat flow transducer and its wires, and the sensor
was then secured with thermal epoxy. This method allowed good thermal contact between the sensor
and its surrounding components. When installing sensors on a reflective surface, we covered the
sensors with matching reflective tape to ensure the sensors have the same radiative properties as the
surrounding surface (ASTM 2007).
Figure 3: Installing temperature and heat flux sensors. Left to right: temperature sensor on exterior
wall, heat flux sensor on interior side of OSB and thermocouple array through assembly, and
installing heat flux sensor on reflective insulation with reflective tape.
3.4 Electricity
We monitored electricity consumption of each hut in one-minute resolution to approximate the cooling
energy required by each hut. In this experiment, the instrumentation power was very low (<5W), and
there were no additional internal building loads.
3.5 Air Tightness

Test hut structures are often too small and too air tight to use standard, full-sized blower door
equipment to measure air tightness accurately. Instead, we used a duct-blaster system with a blower-
door frame system that was capable of measuring air flow in the appropriate pressure range. In this
experiment, there were no windows in the hut. In prior experiments, it was possible to perform air
tightness measurements using the window as the test orifice to measure air leakage through the door.
In our experiments, air tightness was similar among all huts, and within typical ranges for U.S.
residential construction (3-10 ACH), depending on the system. Because the huts have a high surface to
volume ratio, the ACH value may be larger than normal without causing significant concern. If the air
tightness of the huts is appreciably different, it may be necessary to perform additional air sealing, or,
if that is not possible or effective, control holes may be drilled into the envelope of the tightest
building until a similar tightness value are achieved.
4. Sample Results
The following results from a sample week illustrate the sample data results and relative differences in
performance among the three test huts.

4.1 Temperature
Figure 4 illustrates the air temperature in the conditioned zone and in the unconditioned attic among
the three huts, and Figure 5 indicates the interior wall temperatures.
Figure 4: Interior zone temperatures.
Figure 5: Interior wall temperatures.

During the daytime, the temperature oscillates due to air conditioner cycling. The air conditioner in the
uninsulated Hut 1 struggles to supply enough cooling to maintain a matched temperature. Huts 2 and 3
maintain similar indoor temperatures during the experiment. Slight adjustments to the setpoint
temperatures in all three huts were made during the initial testing phase of the experiment to bring
interior space temperatures as near as possible. The occasional night-time temperature spikes were
identified as a predictable source of electronic noise, which were later filtered out.

4.2 Heat Flux
Figure 6 shows the measured heat flux across each principal vertical wall. Here we can see that the
magnitudes of the heat flux was similar for Huts 2 and 3, while Hut 1 displays daytime heat fluxes at
least twice as high, and often exceeding 100 W/m2.
Figure 6: Wall heat flux from different surfaces: EW=east wall, WW=west wall, etc.
4.3 Energy Balance and Power Consumption

Using the heat flux from each hut surface and multiplying by the facing area of the corresponding
wall, we are able to construct an estimation of the instantaneous zone load for each hut. We also show
the air conditioning power, which follows the calculated load curves fairly closely. In the uninsulated
hut, peak loads are highest, onset of cooling occurs hours earlier in the day, and nightly cooling occurs
most quickly. The peak load shift was 44 min. between Huts 1 and 2, and 51 min. between Huts 1 and
3.
Figure 7: Zone Load (calculated from surface heat flux measurements) and AC Power draw.

By calculating the cumulative electricity usage from the power curves of Figure 7, we determined that
that Hut 2 used only about 22% of the energy as Hut 1 for space conditioning, while Hut 3 used only
12%. On the surface, this makes it seem like Huts 3 performs considerably better than Hut 2, owing to
the reflective insulation. On further investigation, however, the analysis of zone loads indicates that
Huts 2 and 3 behave similarly, with 19-20% the cumulative heat gain of the uninsulated hut. This may
be at least partially due to internal differences in the AC systems. For this reason, it is not advisable to
base performance conclusions strictly on energy comparisons in multi-structure experiments.
5. Conclusion
We have described and demonstrated a flexible platform for accurately testing the thermal
performance of advanced building systems through the use of small outdoor test structures. By
carefully measuring key building parameters, we have found that is possible to study and characterize
the performance of building systems in scaled-down versions of realistic building constructions.
In the materials we tested we found that insulated structures consumed about 20% of the energy
needed by the uninsulated structure. Future work will use the detailed thermal measurements to
develop calibrated energy models for the test huts, and energy savings will be simulated with buildings
that have similar characteristics.
6. Acknowledgements
This material is based upon work supported by The Small Business Innovative Research (SBIR)
program and the Engineer Research and Development Center – Construction Engineering Research
Laboratory (ERDC-CERL) under Contract No. W9132T-12-C-0012. Special thanks to Bob Andrews
and Alliston Watts for their many important contributions to this project.
References
ASTM. 2007. C1046-95(2007). “Standard Practice for In-Situ Measurement of Heat Flux and
Temperature on Building Envelope Components.” American Society for Testing and Materials.
ASTM. 2012. C1130-07(2012). “Standard Practice for Calibrating Thin Heat Flux Transducers.”
American Society for Testing and Materials.
Athienitis, A.K., C. Liu, D. Hawes, D. Banu and D. Feldman. 1997. “Investigation of the Thermal
Performance of a Passive Solar Test-room with Wall Latent Heat Storage.” Building and
Environment, Vol. 32, No.5, pp. 405-410.
Baker, P.H. and H.A.L. van Dijk. 2008. “PASLINK and dynamic outdoor testing of building
components.” Building and Environment. 43(2008)143-51.
Khudhair, A.M. and M.M. Farid. 2007. “Use of Phase Change Materials for Thermal Comfort and
Electrical Energy Peak Load Shifting: Experimental Investigations. In: GOSWAMI, D. Y. &
ZHAO, Y., eds. Solar World Congress 2007, September 18-21. Beijing, China. Solar Energy and
Human Settlement.
Medina, M.A., J.B. King, and M. Zhang. 2008. “On the Heat Transfer Rate Reduction of Structural
Insulated Panels Outfitted with Phase-change Materials” Energy. Vol. 33, Issue 4:667-678.
Shilei Lv., F. Guohui, Z. Neng, and D. Li. 2007. “Experimental Study and Evaluation of Latent Heat
Storage in Phase Change Materials Wallboards.” Energy and Buildings. Vol. 39 (2007), pp 1088-
91.
Szymocha, K. 2005. “Advanced Thermal Solar System with Heat Storage for Residential House Space
Heating.” Proceedings of SESCI 2005 Conference British Columbia Institute of Technology,
Burnaby, British Columbia, Canada, August 20-24.

Full paper no: 18
Stochastic Modelling of Hygrothermal Performance of
Wood-frame Wall Assemblies: the Influence of Cladding
Ventilation
1
Lin Wang, M.A.Sc., Ph.D Student
1
Hua Ge, Ph.D., P. Eng. Assistant Professor
1
Department of Building, Civil and Environmental Engineering, Concordia University, Canada
KAYWORDS: Wood-Framed Wall, Hygrothermal Performance, Cladding Ventilation, Stochastic

Modelling
Summary
The modeling of hygrothermal performance using computer programs has increasingly become a
common task in building envelope design. Typically, the deterministic approach is used in
hygrothermal modeling to assess the Heat, Air and Moisture responses and moisture risks of various
designs. However, in reality factors influencing the hygrothermal responses are stochastic in nature
such as the variability of material properties, boundary conditions, construction quality, and
environmental loads, especially the wind-driven rain and runoff loads. To improve the reliability of
hygrothermal modeling, stochastic modeling approach can be used to establish the correlation between
input uncertainties with the output uncertainty and identify the most influential parameters. However,
the importance of these parameters is not constant and varies with wall configurations, boundary
conditions, and moisture loads. To investigate the importance of various design parameters, stochastic
modeling approach is applied to a brick veneer wood-frame wall assembly. Four scenarios with
different cavity ventilation rates and rain leakage deposition positions are tested. For each scenario,
Monte Carlo simulation and multiple regression analysis are performed to rank the influential
parameters. The results indicate that the importance of material properties is largely influenced by the
cavity ventilation and the deposition location of rain leakage.
1. Introduction
Hygrothermal performance modeling has increasingly become a common tool used in building
envelope design in recent years, especially for innovative and energy efficient buildings. Typically, the
deterministic approach is used in hygrothermal modeling to assess the Heat, Air and Moisture
responses and moisture risks of various designs. However, in reality factors influencing the
hygrothermal responses are stochastic in nature such as the variability of material properties, boundary
conditions, construction quality, and environmental loads, especially the wind-driven rain and runoff
loads. The stochastic approach has been applied in several studies in hygrothermal performance
simulations to investigate the influence of material properties (Salonvaara et al. 2001; Holm and
Kuenzel, 2002; Defraeye et al. 2012). Zhao et. al. (2011) also looked into the influence of boundary
conditions and ranked the importance of each parameter using multiple regression analysis. The
influence of environmental conditions including rain loads and internal moisture gains on the
hygrothermal performance of wall assemblies was also investigated by parametric studies and scenario
analysis (Cornick, et al. 2009 and Pallin, 2013).
Previous studies mainly focused on the influence of the stochastic nature of material properties. As
indicated in the study by Zhao et al. (2011), the importance of each influential parameter was not
constant and varied with the weather conditions. To provide a reliable hygrothermal performance
assessment of building envelopes, especially for those highly insulated or with innovative materials, a
stochastic approach taking into account the combined influence of material properties, boundary
conditions, and moisture loads is necessary. The objective of this paper is to develop a methodology to
assess the moisture damage risks of wood-frame construction using a stochastic approach. A brick

veneer wood-frame wall, which was tested under the coastal climate of British Columbia (Simpson,
2010), is selected as an example. The influence of cavity ventilation, rain leakage, and material
properties is investigated using the stochastic approach.
2. Background of the Field Test

To investigate the drying and wetting effect of cladding ventilation on rainscreen walls in the coastal
climate of British Columbia, twelve test walls with two types of cladding, i.e. brick veneer and fiber
cement panels, were installed on the southeast side of a Building Envelope Test Facility. All of the
wood-frame back wall assemblies were identical and built using common building materials and
construction method. The size of the back wall was 1.22m by 2.44m. The assembly is consisted of:
 12mm unpainted gypsum board with a vapor barrier of 6 mil polyethylene
 38mm x 140mm wood frame with R20 fiber glass batt insulation
 12mm plywood (Douglas fir) sheathing
 Spun bonded polyolefin (SBP) sheathing membrane
The test period was from the beginning of January 2008 to the end of June 2008. The change of the
moisture content (MC) in plywood was used as an indicator to evaluate the impact of cavity
ventilation. Further information can be found in Simpson (2010).
3. Stochastic Hygrothermal Modeling

3.1 Material Properties
The stochastic hygrothermal model is constructed based on the typical wood framed wall with red clay
matt brick as the rainscreen, as tested. The configuration of the wall assembly is presented as follows:
Red Matt Clay Brick
Air Cavity
Spun Bound Polyolefin Membrane
Plywood
Glass Fiber
Polyethylene
Gypsum Board
Figure1 Wall Configuration
According to Zhao et al. (2011), the thermal properties seem having insignificant influence when MC
is taken as the performance indicator. Therefore, only the hygric properties are considered as
stochastic variables. To examine the effect of cavity ventilation and rain leakage, the material
properties of red brick, SBP sheathing membrane and plywood are selected for the stochastic analysis.
Both the deterministic and stochastic parameters are listed in Table 1. The material properties are
taken from the database in WUFI Pro 5.0. The properties of stochastic variables, mean value with a
standard deviation in the bracket, are determined based on the measurement results reported by
Kumaran (2002). All the stochastic variables are assumed to follow a normal distribution.

3.2 Environmental Conditions
The weather data collected at the test facility are used as inputs in the hygroghermal simulations,
including outdoor temperature, relative humidity, solar radiation, rainfall, wind speed and wind
direction. The indoor conditions are kept at 21oC and 40% RH. The wall faces southeast, the
prevailing wind-driven rain direction at the test site. The initial moisture content of each layer is
specified based on the measurements of RH and MC readings taken in the experiment. They are: Brick
(2.9kg/m3), Plywood (121.96kg/m3), Glass Fiber (0.17kg/m3), and Gypsum Board (4.6kg/m3).
Table 1 Material Properties
Unit Red Air SBP Plywood Glass Polyethylene Gypsum
Matt Layer Membrane Fiber Board
Clay
Brick
Thickness mm 90 25 0.2 12.7 140 0.15 12.5
Density kg/m3 1935 1.3 448 432 88 130 625
Porosity - 0.217 0.999 0.001 0.69 0.999 0.001 0.706
Heat J/kg.K 800 1000 1500 1880 840 2300 870
Capacity
Heat W/m.K 0.495 0.155 2.4 0.084 0.043 2.3 0.16
Conductivity
Water Vapor - 137.8 328.4(22.5) 730.8 1.21 50000 7.03
Diffusion (9.11) (115.8)
Resistance
Factor
Reference Kg/m3 1.25 64.4
Water (0.085) (4.5)
Content
Free Water Kg/m3 56.11 550
Saturation (3.8) (36.2)
Water Kg/m2s0.5 0.0268 0.0042
Absorption (0.0018) (0.00028)
Coefficient
3.3 Scenario description

In order to examine the effect of cavity ventilation and rain leakage, four scenarios are investigated:
Scinario1: Cavity ventilation, no rain leakage, which is the same as the tested scenario.
Scinario2: Cavity ventilation with rain leakage deposited on the SBP membrane.
Scenario3: Cavity ventilation with rain leakage deposited on the plywood sheathing.
Scinario4: No cavity ventilation but with rain leakage deposited on the SBP membrane.
The range of air cavity ventilation is between 1 and 11ACH, which was predicted based on measured
wind pressure differential and buoyance induced pressure differential between the ambient and inside
of the air cavity (Simpson, 2010). Therefore, an air source with ACH 1~11 is added in the air space in
WUFI simulations to account for cavity ventilation. The cavity ventilation rate is assumed to follow a
normal distribution. According to ASHRAE160-2009 (ASHRAE, 2009), the default value for water
penetration through exterior surface should be 1%, which deposits on the exterior surface of the
water-resistive barrier. Therefore, a moisture source with 1% rain penetration is added on the surface
of the SBP membrane to simulate the rain leakage. A 1% rain leakage deposited directly on the
plywood sheathing is also investigated, i.e. scenario 3. Driving rain load is also considered as an

uncertainty source for the stochastic analysis. According to ASHRAE160, the driven rain load can be
calculated as following:
rbv=FE∙FD∙FL∙U∙cosθ∙rh (1)
where
FE rain exposure factor
FD rain deposition factor
FL empirical constant, 0.2kg.s/(m3.mm)
U hourly average wind speed at 10m (m/s)
θ angle between wind direction and normal
rh rainfall intensity, horizontal surface (mm/h)
rbv rain deposition on vertical wall (kg/(mm2.h))
The rain deposition factor is considered as a stochastic variable. As suggested by ASRHAE 160P, the
range of the deposition rain factor is 0.35~1.0
4. Monte Carlo simulations

There are four steps in Monte Carlo simulations.
Step 1: Random number generation
A group of random numbers following a normal distribution is generated. In this paper, the hygric
parameters, ventilation rates, and wind driven rain deposition factor are considered as stochastic
variables. According to Lomas (1992), the improvement of accuracy becomes insignificant after 60-80
simulations. Therefore, 80 samples of each parameter have been generated.
Step 2: Sampling
Random sampling is commonly used in Monte Carlo analysis and in this paper the generated
parameters are combined randomly. Each combination represents an individual hygrothermal
simulation.
Step 3: Simulations
For each randomly generated sample, a group of parameters are input into the hygrothermal modeling
program and a set of hygrothermal simulation results are obtained.
Step 4: Regression analysis
This step examines the relationship between the inputs and the outputs. In this paper, partial
correlation coefficient (PCC), a measure of the linear dependence of a pair of random variables from a
collection of random variables in the case where the influence of the remaining variables is eliminated,
has been used as an indicator for the sensitivity of MC of plywood to each influencing parameter.
5. Result analysis and discussion

Figures 2-5 show the stochastic results of MC of the plywood for the four scenarios. The blue line is
the mean value of the moisture contents and the grey lines are the stochastic results. The initial MC of
the plywood is 28% and it stays at that level for over one month due to the wet weather. For Scenario
1, where no simulated rain leakage, the mean MC of plywood picks up a bit during the month of Feb.
with a very slow drying rate until mid-April. Starting from mid-April quick drying occurs due to the
warm up of the weather and less rain and the MC level decreases to below 10% in June. In some cases
the plywood picks up moisture from the beginning up to mid-April and then slowly dries to below 10%
in June. These cases typically have a higher rain deposition factor. In some cases, the plywood starts
drying from mid-Feb. These cases typically have a lower rain deposition factor. For scenario 2, in
which both cavity ventilation and rain leakage are simulated, the results are almost the same as that of
Scenario 1, which indicates that the simulated rain leakage does not have much influence on the MC

level of the plywood. This may be because the rain leakage was specified at the surface of SBP
membrane, which is supposed to shed any leakage water away. As expected, the MC of plywood is
higher in Scenario 3, where the rain leakage is added on the layer of plywood. Without the protection
of the SBP weather resistive barrier, the MC of plywood is more sensitive to the rain deposition factor.
In scenario 4, the MC of plywood starts to pick up at the beginning of February, when a high rainfall
occurred as shown in the weather data. This is followed by a gradual increase with the MC of plywood
reaching the peak around mid-April. Without the cladding ventilation, the drying effect starts from
mid-April, when there was less rain. The drying starts almost two months later than the scenarios with
cavity ventilation and the peak of the MC of plywood is about 6% higher than that of scenario 1 and
scenario 2
Figure 2 Stochastic Results of Scenario 1 (no rain leakage with cavity ventilation)
Figure 3 Stochastic Results of Scenario 2 (1% rain leakage on SBP membrane with cavity ventilation)
Figure 4 Stochastic Result of Scenario 3 (1% rain leakage on Plywood with cavity ventilation)

Figure 6 and Figure 7 show the PCC of each parameter under different scenarios at specific dates. The
typical winter rainny day of Feb. 5th and a summer sunny day of June 30th are selected for the analysis.
Figure 6 shows that the most influential factor in winter is the reference moisture content of plywood,
which is used to generate the moisture storage curve in WUFI. It has a positive effect on the MC of
plywood. The PCC of mr_plywood under scenario 1, scenario 2 and scenario 3 is slightly lower than
that under scenario 4, which has no cavity vantilation. This reflects that cavity ventilation does
promote drying, therefore, the MC of plywood is relatively less sensitive to the variation of moisture
storage function. The free water saturation of plywood has the similar effect on the MC of plywood.
When comparing scenario 3 with senario 2, the PCC for the most influential material parameters is
lower, which indicates that with rain leakage deposited directly on the plywood, the influence of rain
deposition factor increases while the influence of material properties slightly decreases. Figure 7
indicats that during summer conditions the reference moisture content of plywood is also the most
influential factor, while the influence of free water saturation become less significant. This reflects that
the moisture transport is mainly in the diffusion region. Because the rainfall in the summer is lower
than that in the winter, the moisture accumulated in the plywood may mainly come from the moisture
in the air. The cladding ventilation may bring more moisture from the ambient air into the cavity,
therefore, wet the plywood. Therefore in the summer time the PCC of the influential parameter with
ventilation is higher than that without ventilation.
Figure 5 Stochastic Results of Scenario 4 (1% rain leakage on SBP membrane without cavity
ventilation)
PCC in Feb. 5th
1
0.5
PCC
0
-0.5
-1
PCC NO Rain Leakage with Cavity Ventilation PCC 1% Rain Leakage on SBP with Cavity Ventilation
PCC 1% Rain Leakage on Plywood with Cavity Ventilation PCC 1% Rain Leakage on SBP without Cavity Ventilation
u_brick-vapour diffusion resistance factor of brick
mr_brick-reference moisture content of brick
mf_brick-free water saturation of brick
A_brick-water absorption coefficient of brick
u_SBP-vapour diffusion resistance factor of Spun Bound Polyolefin
u_plywood-vapour diffusion resistance factor of plywood
mr_plywood-reference moisture content of plywood
mf_plywood-free water saturation of plywood
A_plywood-water absorption coefficient of plywood
FD-rain deposition factor
VR-air change rate
Figure 6 PCC of Each Parameter for February 5th

PCC in Jun. 30th
1
0.5
PCC
0
-0.5
-1
PCC No Rain Leakage with Cavity Ventilation PCC 1% Rain Leakage on SBP with Cavity Ventilation
PCC 1% Rain Leakage on Plywood with Cavity Ventilation PCC 1% Rain Leakage on SBP without Cavity Ventilation
u_brick-vapour diffusion resistance factor of brick
mr_brick-reference moisture content of brick
mf_brick-free water saturation of brick
A_brick-water absorption coefficient of brick
u_SBP-vapour diffusion resistance factor of Spun Bound Polyolefin
u_plywood-vapour diffusion resistance factor of plywood
mr_plywood-reference moisture content of plywood
mf_plywood-free water saturation of plywood
A_plywood-water absorption coefficient of plywood
FD-rain deposition factor
VR-air change rate
Figure 7 PCC of Each Parameter for June30th

What worth mentioning is the discrepancy between the stochastic results and the measured MC. As
shown in Figure 8, the simulated MCs are much higher than that measured MC between February and
June. This difference may be attributed to the approximation of material hygric properties. To consider
the influence of moisture storage function, the reference moisture content and free water saturation has
been considered as the stochastic variables. While the approximated moisture storage function cannot
reflect the material property exactly, the difference between the approximated moisture storage
function and the actual moisture storage function may have been amplified by the stochastic
simulation process. As the results of sensitive analysis have shown, the reference moisture content and
free water saturation are significant parameters that influence the MC of plywood. This means that the
moisture storage function is an important material property that should be determined as accurate as
possible for the hygrothermal simulations. The discrete isotherm data instead of approximation, which
are included in WUFI material database, is used in the deterministic simulation. The simulated results
are compared to measurements. As shown in Figure 9, the simulations have a better agreement with
measurements.
Figure 8 Comparison of Measured Data and Stochastic Modeling Results

Figure 9 Comparison of Measured Data with simulation results with discrete Isotherm Data
6. Conclusion
Stochastic modeling can be used in the sensitivity analysis of hygrothermal performance of building
envelopes. The sensitivity of the performance indicator in relation to input parameters often varies
with wall configurations, environmental conditions, and moisture loads. To examine the combined
effect of material properties and environmental conditions, both stochastic analysis and scenario
analysis are performed using hygrothermal simulations. A typical wood framed building envelope
with cladding ventilation is used for the stochastic modeling. The influence of ventilation rates and
deposition rain factor is investigated by performing stochastic simulations under four different
scenarios. The results indicate that the importance of the material properties is largely influenced by
the cavity ventilation and the deposition location of rain leakage. Further research should be focused
on developing wall configurations that are less sensitive to the variation of local environmental
conditions and rain loads.
References:
ASHREAE. 2009. ASHRAE Standard 160-2009 Criteria for Moisture-Control Design Analysis in Building.
American Society of Heating, Refrigerating and Air-Conditioning Engineers, Inc.
Cornick, S.M., Dalgliesh, W.A. & Maref, W. 2009. Sensitivity of Hygrothermal Analysis to Uncertainty in Rain
Data. Journal of ASTM International, 6 (4): 1-17.
Defraeye, T., Blocken, B. & Carmeliet, J. 2013. Influence of uncertainty in heat-moisture transport properties on
convective drying of porous materials by numerical modelling. Chemical Engineering Research and Design ,
91(1): 36-42.
Holm, A. & Kunzel, H.M. 2002. Practical application of an uncertainty approach for hygrothermal building
simulation-drying of an AAC flat roof. Building and Environment. 37(8-9): 883-889.
Kumaran, K., Lackey, J., Normandin, N., van Reenen, D. & Tariku, F. 2002. Summary Report from Task3 of
MEWS Project. Institute for Research in Construction, National Research Council Canada. Ottawa, Canada,
(NRCC-45369), pp.1-68.
Lomas. K.L., & Eppel, H. 1992. Sensitivity Analysis Techniques for Building Thermal Simulation Programs.
Energy and Building, 19(1): 21-44.
Pallin, S. 2013. Risk Assessment of Hygrothermal Performance-Building Envelope Retrofit. Ph.D Thesis.
Chalmers University of Technology. Goteborg, Sweden.
Salonvaara, M., Karagiozis, A., & Holm, A. 2002. Stochastic Building Envelope Modelling-The Influence of
Material Properties. Proceedings for Performance of Exterior Envelopes of Whole Building Ⅷ: Integration of
Building Envelope, Clearwater Beach, Florida.
Simpson, Y. 2010. Field evaluation of ventilation wetting and drying of rainscreen walls in coastal British
Columbia. MASc. Thesis. Concordia University. Montreal, Canada.
Zhao, J., Plagge, R., Nicolai, A. Grunewald, J. & Zhang, J. 2011. Stochastic Study of Hygrothermal Performance
of a Wall Assembly- The influence of Material Properties and Boundary Coefficients. HVAC&R Research,
17(4): 591-601.

Full paper no: 19
LOW CARBON ECONOMY IN CITIES OF CHINA

POSSIBILITIES TO ESTIMATE THE POTENTIAL OF CO2-EMISSIONS
Hans-Peter Leimer, Professor Dr.-Ing.1,*
University of Applied Sciences and Arts – HAWK Hildesheim, Germany

Hefei University, Anhui, China
Nanchang University, Jiangxi, China
BBS INSTITUT, Wolfenbuettel, Germany
KEYWORDS:
CDM, Cities, Thermal Energy Simulation, CO2-Emissions, Energy Demand
SUMMARY:
Increasing urbanisation and climate change are one of the greatest challenges in the 21st Century.
Many regions already face different negative impacts on cities such as growing local pollution, solid
and liquid waste, traffic congestion and noise. A growing consumption of fossil fuels in cities leads to
increasing CO2-emissions accelerating climate change. All over the world and also in China, cities
are trying to reduce their carbon footprint. In that context, the concept of so called “low carbon
cities” is promoted. Currently, 79 cities in China are considering low carbon concepts. To build a low
carbon city is not an easy task as there are various barriers to be removed, such as financial,
traditional and institutional barriers. Building belong to the most traditional and slow-changing
innovative areas in industry. Further solutions are different for hot and warm climates than for cold
climates. Due to a very fragmented and divided structure in smaller enterprises and in different areas, a
holistic approach is hardly to get.
1. Introduction1
The Chinese government decided to reduce the domestic emission reduction goals in the 12th Five
Year Plan (2011-2015) according to the goals of international emission reduction. The reduction of
carbon intensity per unit of GDP (Gross Domestic Product) in China is supposed to be reduced by 40-
45% in 2020 against the intensity of 2005 levels.
The implementation of CDM2 in the United Nations is in the beginning; therefore China has the
chance to implement these systems as the first and the most important Nation worldwide in the
building sector for LOW Carbon Cities. Germany, as one of the leaders in Low Carbon projects, has
the chance, in collaboration with China, to increase the leadership in this working field and to become
the master in the implementation of CDM projects in buildings worldwide. An excellent possibility to
reach this goal is therefore the implementation on financial incentives by pCDM (Programmatic
1
Corresponding author email: [email protected]
1
This report is part of the final report of the study for the German Federal Ministry for the Environment, Nature
Conservation and Nuclear Safety, Chapter 1602 Title 896 05, 2009 as granted by the notification dated
December 3rd, 2009
2
The Clean Development Mechanism (CDM), defined in Article 12 of the Protocol, allows a country with an
emission-reduction or emission-limitation commitment under the Kyoto Protocol (Annex B Party) to implement
an emission-reduction project in developing countries. Such projects can earn saleable certified emission
reduction (CER) credits, each equivalent to one tonne of CO2, which can be counted towards meeting Kyoto
targets. [http://unfccc.int/2860.php]

CDM) or other advanced methods with this project. In consequence of the development of the city
planning in china during the next years and the increasing building industry, the Chinese government
needs a set of guideline and reliable prediction methods for the regional LOW Carbon potentials and
their application. This project offers China the possibility to receive a forecast of the Low Carbon
potentials for their NEW LOW CARBON CITIES. With these tools China can influence and lead the
market under the aim of LOW CARBON reduction during the next years to full fill the domestic
emission reduction goals. The project will provide the methodology for the implementation of pCDM
for the different climate regions in China as well as the guideline during the planning and realisation
phase of Low Carbon Cities. Only with both approaches it will be possible to find solutions for the
practical implementation for different climate situations (heating, cooling, heating and cooling) and for
suitable requirements.
2. Calculation of CO2 - Emission-Reduction for the most important

Building Types – Example for Xiamen - China
This chapter describes the basic approach to evaluate the energetic behaviour of buildings as well as to
provide a calculation methodology for the CO2-emission reductions in the PoADDs (Program of
Activities Design Documents). In order to detect and to investigate the CO2 saving potential in the
different building types, the following steps of work are necessary:
 Numeric simulation of the range of the energy demand from the selected building types for
heating, cooling and dehumidification
 Setting a baseline for each building type
 Calculation of the energy saving-/ CO2 emissions saving potential of each building type
 Forecast of the CO2 emissions saving potential for a whole pilot-region
In scope of this investigation, the available values from measurements of buildings/building categories
will be compared to numeric calculations of energy requirement.
3. Identification of the baseline

The detection of the CO2 saving potential is based on baselines, which are set by the energy standard
of national requirements and/or regulations.
4. Thermal energetic simulations of buildings (TES)

In this project, three simulation computer models are used:
 TRNSYS
 LEC
 China GB-50189-2005 by the computer program BEED developed by MoHURD
TRNSYS delivers the calculation of the final energy demand of each building type, LEC will deliver
an easy method to estimate the primary energy demand taking into account the primary energy and
CO2-emissions used to produce and transmit the final energy for the use in buildings.
4.1 TRNSYS
TRNSYS is a complete and extensible simulation environment for the transient simulation of thermal
systems including multi-zone buildings.

It is used by engineers and researchers around the world to validate new energy concepts, from simple
solar domestic hot water systems to the design and simulation of buildings and their equipment,
including control strategies, occupant behaviour, alternative energy systems (wind, solar,
photovoltaics, hydrogen systems), etc. Throughout its thirty year history, TRNSYS has been under
continual enhancement by an international group. TRNSYS is based on a model developed by the
Solar Energy Laboratory at University of Wisconsin-Madison, USA, 1992.
4.2 LEC
LEC (Low Energy Certificate) is the result of two years of research to develop a planning tool to
evaluate the energy performance of buildings. Thanks to the evaluation programme it is possible to
evaluate nearly all building types and parts of a building with regard to their energetic quality
separately after the heating period and after the cooling period. The examination of the buildings with
regard to the regional climate conditions is based on a pure physical basis. In this context it is
important to mention that the calculations are exclusively based on results of analysis of the thermal
equations.
The basis for the evaluation of the heating period is a comparison with similar buildings (Reference–
building) which is in accordance with a building according the standard of the 80ies. As far as the
cooling periods are concerned, the evaluation is based on comparisons with an “optimal” envelope of a
building. The evaluation of the energy for cooling and heating are re-evaluated, classified and shown
with regard to certain criteria. The result of the energetic verification is presented with a simple star
system. An increase of stars clearly shows the energetic quality of the building, which means that the
user can immediately recognize the energetic quality of the building thanks to a simple illustration.
4.3 BEED
Building Energy Efficiency Design, Calculation and Economy Analysis Software was developed by
the Centre of Science & Technology of Construction Minister of Construction; P.R. China cooperates
with Beijing E-house building science and technology development Co. Ltd. to finish the task of the
World Bank "economy analysis about building energy efficiency measures and stimulation calculation
module in BEED". The result has passed the MOC evaluation and the level is in the top in China.
BEED was designed according as "Thermal design code for civil building" and "Energy conservation
design standard for new heating residential building.
4.4 Calculation results

The simulation by TRNSYS delivers the Energy demand (kWh/m²) and the Maximum energy load
(kW/m²) for heating, cooling and dehumidification.
LEC calculates the energy level considering the outer surface only. Other methods such as China GB-
50189-2005 calculate a more general use of energy, based not only on the building surface, but also on
system engineering and on political conversion factors. The evaluation tool offers the possibility to
evaluate almost every building and construction part regarding to its energetic quality, separated by
cooling period and heating period. The evaluation of the buildings under consideration of regional
climatic circumstances is strictly based on fundamentals of physics. The calculations are based on
results of thermo technical equations.
5. Calculation of the CO2-emissions of buildings

Basis of the classification are the 18 types of buildings according to the MoHURD-Ministry of
Housing and Urban-Rural Development typology. For these types of buildings the energy demand and
other relevant data are available in 23 cities of China. For the Xiamen project, however, the analysis of

18 building types is too detailed. Therefore, there was made a first selection of 7 main categories
divided in
 Residential Buildings
 Non Residential Buildings
For residential buildings, a classification of the buildings was made on the basis of the building
geometry. Here, the ratio of outer wall and roof to volume (A/V ratio) will be taken and the
residential buildings will be differentiated in
a) single family houses up to 3 stories;
b) multi-storage houses for multiple families.
Non-residential buildings will be differentiated in

c) office buildings; d) Shopping centres;
e) Hotels;
f) Buildings for trade and fair;
g) Schools and other buildings.
The simulation of energy demand is calculated by TRNSYS and based on a single-zone-model
according to the standard room. Therefore, the standard room is calculated with different building
components, different window to wall ratio and different utilization. Furthermore, the orientation and
position of the room is varied.
The properties of the building parts fulfil the minimal requirement of the GB 50189-2005 and JGJ 75-
2003. At this, the window to wall ratio fits the classification of the current Chinese requirements.
Different user profiles have been provided for the utilization in accordance with the applicable
standards in China.
Based on the determination of the final energy demand for cooling in Xiamen, CO2-emissions of
primary energy demand will be calculated by taking into account the primary energy structure, energy
efficiency of energy transformation and the losses during transmission of electricity. For the PoA, to
be conservative, additional reductions may be taken into account for user behaviour, planning and
construction quality.
pCDM are based on the real energy consumption / real CO2-emissions of a building or a side area. The
problem is, that at the beginning of a Low Carbon building City planning phase (town planning phase)
no detailed values of the CO2-emissions are existent. In this phase, pCDM can be based on a
calculation of the energy consumption / CO2-emissions. Calculation methods offer the following
advantages:
 considering the energetic behaviour of the building under repeatable boundary conditions
 provide the possibility to directly compare buildings (under the same boundary conditions)
 provide a forecast on the savings potential for CO2 during the planning period
Summing up, they provide the possibility to compare the economical savings potentials from pCDM
plus the saved amount of energy to the costs of the investment. Based on the variety investigation, an
economical investigation is herewith possible.

6. Energy Demand of the 7 different building types
The aim was the calculation of the energy consumption and the CO2 emissions of the 7 characteristic
building types for new buildings in Xiamen, China under the climate conditions according to
GB50189 for the region “hot summers and warm winters” by implementing a building energy
efficiency simulation.
Table 1. End Energy Demand of the building types (a-g)
No. Building type Simulated energy demand (mean value)
for cooling, heating, dehumidification
current Energy Standard China
kWh/m² kg CO2/m²
a Single family houses 43.10 36,59
b Multi-storage/families houses 54.28 46,64
c Office buildings 101.29 87,03
d Shopping centres 114.02 97,97
e Hotels 164.93 141,71
f. Congress/Fair buildings 112.60 96,74
g Schools 76.37 65,61
In this case, the possible variables, including building type, size, height, orientation according to
azimuth of the buildings, the different ratios of the wall-/window areas of the facades, different uses
and quality standards of the building services installations for energy distribution and energy
production, are taken into account. Furthermore, the range of influence on planning- / arithmetic errors
and execution mistakes had to be described.
7. Application of the models for the calculation of the CO2-potential for

Xiamen
The Chinese government decided the domestic emission reduction goals in the 12th Five Year Plan
(2011-2015) according the international emission reduction goals the reduction of carbon intensity per
unit of GDP in China by 40-45% in 2020 against 2005 levels. To fulfil these requirements the CO2-
potential for China has to be known. One of the parts in this research project is to know well the
energy demand of buildings according to the CDM standard. With this information it will be possible,
to get a forecast of the CO2-potential for new buildings for whole China. To get this forecast based on
the level of the baseline CDM the following information has to put into account in a bottom-up
calculation process (see Figure 1):
The transfer of the results of the single buildings to complex new LOW Carbon Cities was reviewed
by the city of Xiamen/Jimei.

Figure 1. Calculation of the CO2 emissions’ potential for the new LOW Carbon Building City
Xiamen/Jimei
Table 2. – CO2 emissions Jimei, different building standards according to the building standard LEC
(t CO2/a)
Building type Floor area (m²) Baseline 3* 4* 5*
Single family houses 125969 4624 3081 2977 2664
Multi storage houses 1142439 47464 29545 27139 25180
Office buildings 946974 68184 61800 54332 46194
Multi storage houses /
Office buildings (mix) 1247028 64288 43795 39312 36264
Shopping Centers 647986 52158 43418 40041 37955
Hotels 142016 16831 8503 7854 7083
Congress / Fair 11192 1078 834 772 733
Schools 68966 3493 1476 1249 1187
Other buildings 535384 40473 35351 31715 28292
TOTAL 4867954 298594 227805 205390 191837
Total (in % of baseline) 100% 76% 68% 64%
CO2-emissions for heating, cooling and dehumidification MIN values.
Standards according to: Baseline = applicable Chinese standard
3* acc. Chinese Green Building Standard; 4* German ENEV 2001; 5* German ENEV 2009
Even considering regional characteristics, it is shown that there is a possibility to determine the CO2
potential for new „Low Carbon Cities” of China. The results will deliver the base to make decisions by
implementing new CDM-Baselines to limit the energy demand in a special „Low Carbon Standard“, a
standard quite above the actual energy standard of China.

8. Outlook / Conclusions
The possible CO2 savings potential of the new Low Carbon Cities currently being planned (see Figure
2) shows that an economic and ecological building process in China will be realisable once the
requirements of the energetic quality of the buildings are reasonable increasing as well as the
requirements for efficient energy distribution and energy production systems. This building process
may lead the way to accomplish the global restriction of CO2 emissions in the construction sector. To
establish China as a pioneer for CDM in the building sector, actions have to be done in a short term.
Figure 2. Bottom-up-Calculation for the expected CO2-Emissions in China
References
PROPOSED NEW BASELINE AND MONITORING METHODOLOGIES - (CDM-NM)
CDM project activity categories – III.AE. Energy efficiency and renewable energy measures in new
residential buildings
ASHRAE5 Guideline 14-2002, Measurement of Energy and Demand Savings, Whole Building
Calibrated Simulation Performance Path6
GB 50189-2005 Chinese Guideline - public buildings
GB/T 50378-2006 Chinese Guideline - Green Building Standard
GB 50176-93 Chinese Guideline - residential buildings
Energieeinsparverordnung–EnEV2009 and DIN V 18599–Energetische Bewertung von Gebäuden
(07/2005)
TRNSYS - Transient Energy System Simulation Tool; University of Wisconsin, Madison
LEC- Low Energy Certificate - http://www.lowenergycertificate.com

Full paper no: 20
Innovative façade refurbishment with integrated air ducting for

the existing building stock
Matthias Ziegler, M.Sc. 1
Michael Krause, Dr.-Ing. 1
1
Fraunhofer Institute for Building Physics, Kassel, Germany
KEYWORDS: Energetic retrofitting, façade integration, central ventilation system, air ducting
SUMMARY:
For successful transition of the energy system being sustainable, the energetic retrofitting of the
building stock is a crucial factor. About 40% of the current final energy consumption in Germany is
caused by the building sector for heating purposes. Energetic retrofitting can reduce the heat demand
of buildings considerably but extensive refurbishments often fail due to high costs, leading to only
individual measures on the façade being carried out. New retrofitting concepts are required which
are both energy efficient and cost-effective to increase the rate of retrofitting significantly.
A promising solution, developed by the Fraunhofer Institute for Building Physics, provides the
implementation of the air duct system in the building envelope. Hereby, the air ducts for the
installation of a central residential ventilation system are integrated directly in the mounted ETICS
plate. Such a system has already been successfully tested in one and two family houses. However, the
challenge of implementing the system in a multi-storey building is much larger due to structural
dependencies on the façade, fire protection regulations and other limiting factors. These constraints
have been identified to develop appropriate solutions for the planning of the air ducting on the
façade.
1. Introduction
1.1 Importance of the building sector in our future energy system
In Germany, the building sector causes 40 % of the current final energy consumption. The main part
of that energy is used in buildings for space heating and domestic hot water obtained from fossil
energy sources. The energy concept of the German Federal Government (BMWi/BMU, 2010) assigns
the energetic retrofitting of the existing building stock a significant role for the upcoming energy
transition. Until the year 2050, the entire energy supply in Germany which is today mainly supplied
by fossil and nuclear energy sources should be replaced by renewable energies. Furthermore these
enormous efforts to rehabilitate the existing building stock are crucial to achieve the climate change
objectives of the European Union.
Currently, the legislature in European countries has already reached a very high level of energy
efficiency for new building, like the directive on the energy performance of buildings (EPBD) in 2003
or national ordinances like the Energy Conservation Act (EnEV) in Germany. New energy standards
for buildings have been developed like the passive house concept, a building without a conventional
heating system due to a high level of thermal insulation and optimized passive solar energy gains or
the plus energy house, which produces throughout the year due to a large photovoltaic system more
energy than it consumes. These concepts are proven, technically mature and well-established in
building practice (Erhorn, 2012). Despite these technical processing, throughout all European
countries the existing building stock contributes about 80 % of the overall energy consumption. The

possibilities for governmental requirements for energetic retrofitting are limited because of economic
reasons like cost effectiveness and manageable payback periods. There are several ways to overcome
these restraints for example with new innovative concepts for retrofitting like minimally invasive and
multifunctional concepts to reduce the costs for renovations. These innovations are urgently needed to
increase the rate of retrofitting significantly which is currently about 1% per year.
1.2 State-of-the art

Extensive renovations are costly, so that normally only single measures like the replacement of the old
windows or the refurbishment of the building envelope with an external thermal insulation composite
system (ETICS) are being carried out. Single measures are less energy efficient and one of the reasons
why no more retrofitting is realized at towards a significant period on a certain building. Building
retrofitting is traditionally planned and executed separately by trade with lack of coordination. This
can often lead to problems because basic building physics principles, like thermal bridges or air
tightness are not considered in an optimal way (Stiegel 2012). Structural enhancements like ETIC
system or new windows can reduce the transmission loss of a building and the associated energy
demand significantly but the increased sealing of the building envelope can lead to moisture related
problems due to improper or insufficient ventilation. The installation of a user-independent central
ventilation system with heat recovery can ensure the necessary hygienic air change and provides many
advantages. So the indoor air quality can be improved and saving at the same time additional energy
due to the recovery of the thermal energy contained in the exhaust air. Retrofitting barriers, such as
the high cost of the internal ductwork and the impairments caused by dirt and noise for the tenants
often prevent the installation of a central ventilation system. Other solutions with independent and
decentralized ventilation units are cheaper but have a lower heat recovery and require more
maintenance.
2. Innovative Retrofitting with façade integrated air ducting

The Fraunhofer Institute for Building Physics in Kassel developed in cooperation with an industry
partner a promising solution for the subsequently installation of a central ventilation system in ETICS
façades – “the FreshAirWall” system (FAW). The air ducts are integrated in the insulation panels, so
that the functions thermal insulation and air ducting are combined in one construction. In figure 1 an
ETICS plate with the integrated air duct is shown. The air duct is separated by a divider due to
structural reasons and has a cross section of about 30 mm x 300 mm. For better processing on the
building site and enhanced connection the plate features a groove and tongue system around the duct.
Figure 2 shows an example how the panels are arranged on the façade. The panels are connected
alternately to form the complete vertical duct.
FIG 1: ETICS panel with integrated air duct (FAW) FIG 2: Air ducting on the façade

2.1 Current design and functionality
The current design of the insulation panels is specified for the ventilation of one or two rooms with
fresh air. For this reason the cross section of the air duct corresponds hydraulically to a round duct
with a nominal pipe size of DN90. The air ducting on the façade should be exclusively vertical, to
avoid deflections and necessary fittings which would complicate the positioning on the building
envelope. A wall bore hole connects the room with the air duct and all of the distribution pipes are
located in the basement or the attic depending where the ventilation unit stands.
2.2 Previous investigations

The FAW system was examined by simulation, in-situ measurements and experimental investigation
in the flow laboratory. These examinations are necessary to determine the basic functionality of the
retrofitting system and to identify possible optimization for the system.
First, the System was implemented in the dynamic simulations program TRNSYS regarding the heat
losses of the air ducting on the façade and the achievable heat recovery for a certain building. The air
flows in all rooms provided by the ventilation system were calculated with a multi-zone air flow
model and the results are already published (Ziegler, 2012). During a research project, the innovative
retrofitting system was carried out in a two-storey residential house in Kassel (GER) and monitored
for three years with long-term measurements to determine the heat recovery, the achievable air
temperatures as well as an energy efficient operation with low pressure losses and minimal electrical
energy demand for the fans. The hydraulic performance of the air ducts integrated in the insulation
panels was mostly unknown, that is why experimental investigations in the flow laboratory to
determine the pressure drop of the straight air ducts were carried out. The results of the examinations
showed that the measured pressure drop of the air ducts corresponds to calculated pressure drop of a
sheet metal duct with the same geometry (Ziegler, 2013). The pressure drop of the fittings like the 90°
deflection and the wall connection to the room were comparatively higher.
3. Application in different types of residential buildings

The results of the investigations on the two-storey building described above demonstrate that an
energy efficient and cost effective system is technically feasible and simple to plan. The application of
this retrofitting system in larger buildings like multi-storey buildings has not been investigated yet.
Buildings from the post-war period constructed from the 1950s to 1970s in Germany are quite
suitable for the renovation with the FAW system. These buildings are often in a poor condition
because most of them are made of debris or substitute materials. Their high energy demand promises a
high potential of energy saving. Especially, when the renovation of the building envelope is pending, a
new ventilation concept must be evolved, e.g. the planning of a mechanical ventilation. The external
appearance of the post-war buildings is mainly dominated by plaster façades without any ornaments
or decorating elements. Due to this, constraints caused by preservation order are rarely found. Cubic
structure and plain cubature simplify the arrangement of the air ducting on the façade.
Based on the example of several buildings the fundamental dependences for the planning of the air
ducting on the façade should be identified to develop appropriate solutions. The selection of the
buildings is aligned to the building classes of the German model building code (MBO, 2002) since it
contains the relevant requirements concerning fire protection like building material classes or fire
resistance classes compared to the pertinent building typologies. In table 1, the building classes for
residential buildings according to German Model Building Code are listed and the image of a typical
building is attached. High rise buildings and special structures are not considered. The height,
mentioned in table 1, is measured from ground floor level to the finished level of the highest storey
suitable for accommodation.

TABLE 1:Residential building classes according to Model Building Code (MBO) in Germany
Building class Description Example
Building class 1 (GK1) detached buildings up to a height of
7 m and not more than two units with a
max. area of 400m²
Building class 2 (GK2) buildings up to a height of 7 m and not

more than two units with a max. area
of 400m²
Building class 3 (GK3) other buildings up to a height of 7 m
Building class 4 (GK4) Buildings up to a height of 13 m and

each unit not more than 400 m²
For every building class, a typical building was selected and examined regarding possible solution for
air ducting on the façade with the FAW system. Class 1 building is a detached residential house with
two dwellings and class 2 building is a terraced house with one dwelling. Furthermore, the example
building for class 3 is a three-storey residential building which you can find mostly in larger cities.
Also class 4 building is a typical multi-storey building in urban area with four floors.
3.1 Fire protection

For the planning of air duct systems on the façade the observance of the applicable fire protection
regulations and their impacts are important. In general, in Germany there is a difference in
requirements regarding building materials and building components.
Building materials are characterized according their fire behavior in different building material
classes. There is a distinction between nonflammable materials A, flame resistant materials B1 and
normal flammable materials B2. Air ducts integrated in the thermal insulation made of polystyrene
can be produced as flame resistant by chemical additives. Flammable materials are allowed to use in
building of class 1-3 but in class 4 buildings only flame resistant materials are permitted. For
energetic retrofitting of building façades with ETIC systems, the thickness of thermal insulation must
extend 100 mm due to regulations. These ETIC systems (thickness ≥ 100 mm) need further fire
protection measures to obtain the flame resistance (FVWDVS, 2009). One possible solution is the
installation of a fire barrier above every second floor with a minimum height of 200 mm. Another
solution is a window lintel protection above every window to prevent the fire from spreading
vertically to the next storey. Both additional fire protection measures are displayed in figure 3. As
shown, these measures have a great influence on the planning of the air ducting because of less space
on the facade. Either the air ducts must be arranged around the window linter protection or
alternatives to match fire protection and air ducting on the façade need to be developed.

FIG 3: Additional fire protection for class 4 buildings with fire barrier above every second floor (left)
and with window lintel protection (right)
When connecting two rooms of different units via one air duct, further fire protection measures are
needed to prevent fire and smoke from spreading to another unit. The installation of maintenance free
fire dampers for example can eliminate that danger.
3.2 Structual dependences on the façade

The arrangement of the air ducting on the façade is difficult in larger buildings like multi-storey
buildings because of structural dependences on the façade and the layout of the floor plan. Below,
these constraints and restriction are analyzed using the example buildings listed in table 1.
3.2.1 Floor plan design

The floor plan design can have a significant influence on the planning of air duct systems on the
facade. Especially the position of the supply and exhaust air areas affects the air ducting. In most
buildings the floor plans in the individual storeys do not differ, so similar rooms are always located
above each other. This rather simplifies the design of air ducting. Many buildings of the 1950s and
1960s contain old chimneys which were used earlier for single fired ovens. Here, it would be possible
to install subsequently several ducts in the chimneys for one room in each floor to minimize the
number of ducts on the façade. The ducts could be made of flame resistant materials and the space
between them in the chimney could be filled with an intumescent substance which foams and seals the
duct in case of fire. An image of this concept can be seen in figure 4.
FIG 4: Subsequently installation of air ducts in existing chimneys

The ventilation unit can be positioned either in an unused or not extended attic or the basement of the
building. For maintenance reasons, the ventilation device should be easily accessible. Usually, in the
basement is not enough space for the installation of the central ventilation station because of rented
cellars. Another limiting factor is the low ceiling height in the basement, so that the horizontal
distribution air ducts are difficult to arrange. Possible solutions for the connection of the air duct to

the attic and the basement are illustrated in figure 5. The jamb wall and the cellar wall are suitable for
linking the ducts together. In range of the basement window, the air ducting needs to be planned very
carefully.
FIG 5: Connection between the inner distribution channels and the air ducting integrated in the
thermal insulation in the attic through the jamb wall (left) and in the basement through the cellar wall
(right)
3.2.2 Façade design

The available wall area between two windows in multi-storey buildings is limited assuming an
exclusively vertical air ducting. This should be explained by the following example in the class 3
building. In figure 6 a front view of the class 3 building with the adjacent inner and outer walls is
illustrated. The location of the lateral walls is important because the air ducts are connected to the
room via a wall breakthrough. So, the usable wall surfaces in certain rooms are partially less than
50 cm wide, which is not sufficient to put multiple air ducts next to each other.
FIG 6: Front view of class 3 building with lateral inner and outer walls in dark lines (left) and the
associated floor plan (right)
Buildings from the reconstruction years in Germany rarely have projections and recesses on the
façade but on the rear side often recessed balconies or loggias can be found. The rooms located
behind the balconies cannot be connected to a vertical air duct without using deflections on the
façade. These constraints are shown in figure 6 (right side) to illustrate that challenge. Therefore, the
air ducting in the area of the balcony can be realized for example by breaking through the ceiling
selective. However, the sealing of the balcony should be maintained and the process of retrofitting
should be preferably with little effort. Another possibility is to provide the lateral surfaces for air
ducting, when they are not needed otherwise.

4. Possible solutions for multi-storey buildings
The constraints and structural dependencies in multi-storey buildings demonstrated above leads to
significant challenges for the planning of the air duct system on the façade. Now, in this part
appropriate solutions are being developed.
The arrangement of the air ducts is to a decisive extent dependent on the available wall surfaces
between the windows. Usually, it is not possible to connect all the rooms on each floor with its own
vertical duct. Using the example of the three-storey buildings of building class 3, two options are
shown to integrate the air ducting on the façade. First option provides the connection of two rooms
with one duct. In this case, the dispersion of fire and smoke from one dwelling to another must be
prevented for a certain time period. The second option would be to separate the first two floors from
the third floor. Two ventilation devices are required, one in the basement and one in the attic. Both
possible solutions can be seen in figure 7.
FIG 7: Solution for the air ducting on the façade with limited space between the windows: connection
of two rooms with one duct (left) and with air ducting through the attic and the basement (right)
For large buildings (building class 4) additional fire protection measures are required e.g. fire barriers.
The concept of separation of floors, described above, can be applied (figure 8). The fire barrier then
splits two floors and the rooms are either supplied via air ducts from the attic or the basement. If only
one ventilation device is desired, the fire barrier will be crossed by air ducts. This is technically
possible with nonflammable ducts made of metal and laminated with an intumescent coating. For this
purpose, further investigations are necessary to evaluate the performance in case of fire and smoke.
FIG 8: Fire barrier and separated supply air ducting through the attic and the basement, exhaust air
ducts installed subsequently in existing chimneys

5. Conclusion and outlook
The retrofitting system with air ducts integrated in the thermal insulation (FAW system) is energy
efficient and cost effective and has been tested both with theoretical and experimental investigations
in one and two family houses. There are almost no limiting factors for the air ducting for class 1 and 2
buildings regarding structural dependencies or fire protection regulations. The planning of the air
ducts on the façade for multi-storey buildings, however, is subject to much more constraints by
obliged fire protection measures, flanking balconies and narrow available wall surfaces. For these
limitations suitable solutions have been developed in order to realize the air ducting on the façade.
In further investigations an appropriate planning tool for the air ducting on the façade should be
developed. In addition to that an enhanced design of the air duct in the insulation plates with same
fluid dynamic properties and minimal heat losses can improve the system.
6. Acknowledgements
The author wants to thank the German Federal Environmental Foundation (DBU) for the PhD
scholarship and the associated promoting of the research project. Furthermore, the authors are grateful
to ABG Nova for the providing of the demonstration building that is to be renovated within the EU
Project (EU RetroKit, 2012).
References
Erhorn H. 2012. Das Effizienzhaus Plus. Fachtagung „Die neue Plus-Energie-Welt“. Berlin
http://www.forschungsinitiative.de/PDF/Plusenergie_IBP.pdf
European Union. 2002. Directive on the energy performance of buildings (EPBD). http://eur-
lex.europa.eu/LexUriServ/LexUriServ.do?uri=OJ:L:2003:001:0065:0065:EN:PDF
European Union. 2012. RetroKit project. http://www.retrokitproject.eu/web/guest
Federal Government of Germany. Energiekonzept für eine umweltschonende, zuverlässige und
bezahlbare Energieversorgung. Press information Office of the Federal Government. Berlin.
Germany
Fraunhofer Institute for Building Physics. 2010. Innovativ saniert! – mit integrierten Lüftungskanälen
in der Außendämmung, press release. Kassel. Germany
Hauser G. &Kaiser J. 2013. Dämmstoffintegrierte Kanäle für zentrale Lüftungsanlagen mit
Wärmerückgewinnung. Article Bauphysik 6/2013. Ernst & Sohn. Berlin.
IS-Argebau. 2002. Musterbauordnung (MBO). Model Building Code. Berlin. Germany
Stiegel H. & Krause M. 2012. Minimalinvasives Sanierungssystem mit vorgefertigtem,
multifunktionalem WDVS-Fassadenmodul. GI Gesundheits-Ingenieur. 133(2012) Heft 6.pp. 290-
302
WDVSysteme. Fachverband Wärmedämmverbundsysteme e.V. 2009. Technische Systeminfo –
„Brandschutz“. specialized information. Baden-Baden. Gemany
Ziegler M. 2012. Thermische und strömungstechnische Simulation einer Fassadendämmung mit
integrierter Luftführung für die Bestandssanierung. conference BauSim 2012. Berlin. Germany
Ziegler M. & Krause M. 2013. Innovative façade refurbishment with integrated air ducting for the
existing building stock. conference SB13. Munich. Germany

Full paper no: 21
Achieving Sustainable Roof Systems through Innovative Design

Thomas W. Hutchinson, AIA, CSI, FRCI, RRC, RRP 1
1
Hutchinson Design Group, USA
KEYWORDS: Long-Term Performance, Roof System Design, Durability, Sustainability
SUMMARY:
The discussion on sustainability is now over a decade long. How this concept relates to roof systems is
unique in that the variety of potential solutions is almost endless. If one accepts the conclusion of the
CIB/RILEM Joint Committee on Sustainable Low Slope Roofing (2001), that established the 21 Tenets
of Low Slope Roofing (which have become the international standard bearer for sustainable low slope
roofing), that the essence of sustainability is ‘Long Term Service Life’, then designers must move
beyond current fad and use empirical experience and innovative methods to achieve sustainable roof
systems. This paper will present two roof system design case studies in which the author called upon
his 25 years of roof system design experience, empirical in-field service performance, and
architectural background to design roof systems whose long-term service life expectations are 50
years or more. Design concepts, details and photographs from these projects will be presented.
1. Introduction
Prior to any new concept implementation, such a Sustainable Roof Systems, the ultimate end goal
should be clearly and definitively stated and understood through the definition of terms. Make no
mistake about it: Sustainability is about long-term performance. Additionally, in combination with the
concept that roofs are systems that are only as good as their weakest component, the idea of
sustainability must be holistic in approach, involving the entire roof system. The inclusion of green
parameters, which are only one part of sustainability, and without the other key parameters will not
result in the realization of long-term performance. Thus, the keys to sustainability are not independent,
but must act synergistically so that they result in a system that results in long-term performance,
achieving the owner’s and designer’s desired sustainability goals.
Keys to sustainability that have been developed over years and have been proven by the test of time
are as follows (Hutchinson 2007):
1. Achieving the owner’s buy-in.
2. Be aware of budget appropriateness. Quality cannot be achieved with under-developed budgets.
3. A roofing system designed for long-term performance.
4. Use of quality materials that support long-term performance.
5. Inclusion of thermal insulation performance appropriate for the region in which the roof is
installed.
6. Environmental considerations.
7. Use of qualified contractors.
8. Quality assurance.
9. Management of rooftop equipment.
10. Maintenance.
The following is a discussion of each of the above-mentioned 10 keys to sustainability.
1.1 Achieving owner buy-in

Designers and contractors work with a variety of clients who have differing performance goals, needs,
and budgets. All clients must be educated in the development of their new roof system design and

enlightened not only as to the importance of long-term performance, but also to its benefits. Clients
and building owners are becoming more sophisticated, requiring and demanding greater performance
from their building systems: roofing systems are no less. Many clients assume warranties are
justification for long-term performance; on the contrary, warranties are issued by manufacturers as a
minimum standard, and are a promise to render a leak watertight, not a promise of a leak-free roof.
Additionally, all too often this author has had to discuss owner concerns such as, “I only want to hold
the building for 10 years”; “It’s not financially responsible to do so”; and “It’s just a roof”. All
concerns suggested above have been dissuaded when resale value, altruistic values, return on
investment, environmental benefit, etc. are reviewed and explained in a factual manner to those who
doubt.
1.2 Roof systems designed for long-term performance

In the quest for long-term performance, roof system design requires specialized and concerned
individuals. Long-term performance is defined as a 30-year minimum service life. However, as an
industry we should be striving to achieve even greater performance duration. Achieving this goal
encompasses much more than just roofing system design; it is giving consideration to how the,
building’s architectural elements impinge upon the roofing system; such as masonry, EIFS, HVAC,
and telecommunications equipment, as well as roof top access and traffic, and incorporating them into
the roof system design. Qualified roofing system designers are mandated. They must be educated,
knowledgeable, and in many instances licensed to perform design in the locality in which the
construction project is to take place. The roof system must be appropriate not only for the building,
building use, and anticipated roof top use, but also for the climatic and geographical conditions where
it will be installed. Additionally, minimizing penetrations and foot traffic upon a roof go a long way to
assist in the roof system’s performance.
This author is convinced that quality long-term roof system performance cannot be achieved without
detailed building to roof specific details. The use of only specifications describing installations, and
manufacturer’s details, is inappropriate and unacceptable for the attainment of long-term performance.
1.3 Materials
It continues to amaze this author that new materials (note: new material are not only those coming on
the market, but also those whose formulation are revised, changed and modified in an attempt to
correct in field performance issues). In the U.S.A. the primary example is TPO (thermoplastic
polyolefin membrane), which are so readily and easily accepted by the building design community,
when there is no proven history of performance for these materials in the roofing industry, nor in the
region or locality in which the materials are being installed. Therefore, a key of all materials
implemented on a new roof system looking to achieve sustainability is that each material has an
established history of long-term in situ performance for the system in which it will be installed. The
roof installation methodology should also be appropriate for the roof system involved.
Additionally, if possible, roof system materials should also be restorable, recyclable or reusable after
their service life. What is the availability of roof system materials and their recycling programs in the
locale in which the system it is being installed? Can the roof be restored? One fine example is the
restoration of EPDM roofs after they are either out of warranty or their service life, resulting in an
extended service life.
1.4 Thermal performance

The U.S. Department of Energy recently released data that suggests that it is the heating load that
dominates, not cooling (U.S. Department of Energy 2011). Therefore, rendering an opinion that is
diametrically opposed to that of the cool roofing proponents. Consequently, most buildings should be

designed for their heating loads rather than their cooling loads, as energy costs have historically risen
over time; appropriate consideration of potential energy savings options is prudent.
There is a number of parameters that will affect thermal performance that should be given
consideration. One such parameter affecting thermal performance to consider is the use of vapor and
air retarders. Not only should consideration be given to these retarders during the installation process
as a method for achieving a more homogeneous installation, but also to reduce construction induced
moisture drive; they will thus protect the roof system through the new construction and/or re-roofing
construction process. They will also reduce the amount of air transport into the roof system. Air
transport is one of the greatest causes of heat-loss, and the movement of moisture into the roof system.
Through this author’s experience, it is mandatory that multiple layers of insulation be installed to
achieve the highest level of energy conservation. Recent inspections by this author on a multitude of
roofs have shown that the allowable dimensional change of insulation (typical with polyisocyanurate
or EPS) results in open gaps between insulation boards of up to ¾ inch (1.9 cm) in thick layers of
insulation, resulting in numerous concerns including heat loss, unsupported roof covers, and the
potential for moisture condensation at the roof cover. While this is not necessarily a performance
concern of the material, but actually a characteristic of the material that is not properly described in
product literature or communicated to the building designer. Thus, this author has required that all
insulation voids at perimeters, penetrations, roof curbs, and drains be sealed with spray foam insulation
to eliminate these thermal shorts.
Recent studies by Desjarlais et al. (2008) at the Oak Ridge National Laboratories (ORNL) have shown
that the use of ASTM #4 ballast and/or concrete pavers can provide the same benefit as reflective roof
surfaces, while additionally moving heat load away from peak periods.
1.5 Environmental considerations

Consideration of the environmental concerns is only one of the keys, not THE key, to sustainability.
The goal of environmental consideration should include:
• Minimizing the burden on the environment;
• Using products with known long-term service lives; and
• Recognizing regional and climatic factors in the design of the roof system.
The practitioner must remember that it is often more appropriate to use a material that at first review
appears less environmentally friendly than others, but who’s long-term benefit is superior. For
example, if a superior material is manufactured at a greater distance from the construction site, it may
be worth the transport rather than using a local material that may not achieve long-term system
performance.
Garden roofing has become a popular environmental treatment for a variety of roof surfaces. While the
potential for extending a roof’s service life is high when incorporating an overburden that protects a
roof covering from thermal shock and UV degradation, thin overburdens and improper or reduced
quality system design are resulting in systems of less than anticipated performance.
1.6 Contractor qualifications

This author cannot emphasize enough the necessity of pre-qualifying contractors. This is important not
only for installation of the roof system and its related components, but also for value with regard to
how business opportunities are viewed.
1.7 Quality assurance

Twenty-five years of roof system design and construction experience have led this author to realize
that even with high quality drawings, specifications, budget appropriations, and quality contractors;
nothing can supplant the continuous on-site observation of the roof system installation by qualified

roof consultants. On-site observation provides the opportunity to address questions, to verify
compliance with the contract documents, and at times, to enforce quality related concerns. Even the
best contractors occasionally have installation quality issues, such as joints in the insulation that are
greater than ¼ inch (0.6 cm), or joints that are not offset as specified. Field observation will allow for
the clarification of unknown or difficult conditions that may need resolution. Pre-construction
meetings, weekly project coordination meetings, a view of the clear-cut definition and goal of the
project should all be included under the auspices of “quality control”.
1.8 Pro-active maintenance

All roof systems require maintenance. This bears repeating: all roof systems require maintenance! It is
required by the warranty. Good roofing practices require it. It is key to achieving sustainability. All
roof systems require maintenance. Keeping roof drains clean and free of debris; minimizing
unnecessary maintenance crew’s (HVAC and cable techs, etc.) access to the roof will go a long way to
achieving a long-term service life and preventing unwanted damage to the roof.
2. Case Studies
In the past 5 years this authors has been involved with educational, healthcare and fortune 50
companies whose building stock is substantial, and for whom the concept of long-term service life is
especially appealing. The decided upon long-term service life goal was to exceed 30 years. Design,
detailing, material selection, contactor qualifications, and on site observation raised the authors
prediction of long-term service life to 50 years.
A couple of the main parameters (concerns) of designing roof systems for 50 years are: 1.) How will
the material perform under UV radiation, and its potential deleterious effects on the roofing material,
and 2.) What type of in situ performance does the material have. In regard to the case study projects,
the material selected was EPDM. This author’s experience with the material is excellent, with
numerous projects installed over 25 years ago still performing. In fact, studies by SKZ (Süddeutsche
Kunststoff Zentrum) in Würzburg, Deutschland (Zahn 2003), declared that EPDM should have a
service life of 50 years. Furthermore, this author and the EPDM Roofing Association (ERA) undertook
a study looking at EPDM roofs over 30 years old, tested the material for the physical characteristics,
and found that 90% held characteristics at levels above the minimum level for new materials. A second
study heat aged the 30 year in situ material for an additional 30 year equivalency, and then performed
physical characteristics tests. Not surprisingly the results were positive (a paper on the results and
conclusion of these findings entitled Characteristics of In-Service, 30-Year Old EPDM Roof
Membrane will be presented at the 2014 Durability of Building Materials and Components Conference
in Sao Paulo, Brasil, in October 2014).
2.1 Case Study 1: Hawthorn School District 73: Middle School North
Client: Hawthorn School District 73
Building: Middle School North
Location: Vernon Hills, IL (20 miles north of Chicago)
Building Use: Middle School for grades 6-8
Size: 145,000 ft2 (13,471 m2)
Height: 1 and 2 Story: Max height: 28 feet (8.5 m)
Project Goals:
• 30 Year Service Life: Increased to 50 by author after design
• Increase thermal value by 200%
• Budget: $1,650,000 (approx. 1,218,200 € / 10,816,500 kr)
• Roofing to be completed during summer break: June to mid-August

Design Concept and Innovation: Use the existing roof membrane as a vapor retarder and install a new
roof system over the top.
The existing Middle School had its original roof replaced in 1986 with a fully adhered 60-mil EPDM
system installed over two layers of 1.5 inches (3.8 cm) polyisocyanurate insulation (Pier) with a ½
inch (1.27 cm) fiber cover board. The budget at the time was minimal; consequently, the roof design
was basic, as was the installation. Subsequent roof top equipment installations and unabated foot
traffic resulted in some minor physical damage, and the low level of insulation did little for a school
that is heated 7 or more months a year. After several repeating roof leaks, as well as high energy costs,
the board budgeted for a new roof system.
2.1.1 Investigation:
Initial investigation of the existing roof system found that while there were several deficiencies, they
were not wholesale, and that the roof system as it was, was in good structural condition: It wasn’t
deteriorating below the membrane, and its attachment to the structure was intact. As a consequence of
this investigational conclusion it was decided to use the existing roof system as the base for a new,
highly thermally sustainable roof system. A review with the roof membrane manufacturer of the
concept found them to be a willing partner.
2.1.2 Design:
Utilizing the existing roof system required special attention to the attachment of the new roof system.
As the concept was to utilize the existing roof system as a vapor retarder, penetration of the existing
membrane would not be prudent. Thus we looked to adhesives. Following a review of potential
materials it was decided to use a low rise insulating, two component spray polyurethane adhesive with
enhanced plasticity. Test mock-ups in the field prior to the finalization of design and consequential
construction documents were undertaken with the finding that the proper preparation (cleaning) of the
existing membrane and compression of new insulation board were required to achieved the required
attachment and uplift resistance.
FIG 1. Typical Roof Section Detail

In order for the existing roof membrane to be used as a vapor retarder the new insulation above needed
to be of an R value great enough to keep the membrane below the dew point. Dew point calculations
determined that 2 layers of 2” (5.08 cm) polyisocyanurate insulation with R 24 were required. As this
roof design called for long-term service life, a double coated fiberglass insulation facer, that does not
support mold growth and provides a superior surface for adhesive attachment, was specified. To
prevent foot traffic damage to the insulation, and insulation and membrane adhesive bond, the top
layer of insulation was specified as a composite board with the top ½ inch (1.27 cm) being a high
density polyisocyanurate of 100 psi.

A 90-mil fleece backed EPDM set in full coverage spray foam was specified for its superior wind
uplift resistance, as well as its tenacity in deflecting foot traffic concerns. A section of the roof system
design can be observed in Fig. 1 and Fig. 2. To enhance the roof system, 6” seam tape lap seams with
self adhering cover strips were designed.
A B
FIG 2A -B. Roofing crews set the 90-mil EPDM into the full coverage two component spray
polyurethane adhesive, which will be rolled into place with a 90 # (41 kg) water roller. Note the two
layers of insulation with offset joints and the roof edge sheet metal installation on the roof edges
beyond. In the right photo the 45-mil vapor retarder has been installed, and the raising of the roof
curbs and edges, replacing roof drains, and modifying mechanical equipment can take place.
Additional roof system design features included:

1. The roof edge had to be raised to accommodate the additional insulation. To reduce the number
of horizontal layers of roof edge sheet metal, a vertical prefinished aluminum flush panel and
coping were designed.
2. All roof drains were raised with extensions and new overflow roof drains added.
3. Abandoned equipment was removed, and the roof deck in-filled.
4. Smoke hatches were removed and replaced with new.
5. Roof hatches were removed and replaced with new.
6. Ladder and stair access to varying roof levels were added for maintenance.
7. Roof top mechanical equipment and piping was disconnected, curbs raised, and the integration
with the new roof system coordinated.
8. Deteriorated masonry above the roof was restored.
9. Deteriorated areas of roof were removed and replaced with new to match.
2.1.3 Construction:
The project was bid in early spring 2012 with the main construction commencing in early June. The
roof installation was observed by a Registered Roof Consultant (RRC) and took approximately 4
months. Roof edge sheet metal took another 2 months.
Contentedly, the owner reported an approximate drop in the total building energy usage in the first
year.
2.2 Case Study 2: Moraine Valley Community College: Building L

Client: Moraine Valley Community College
Building: Building L
Location: Palos Heights, IL (10 miles southwest of Chicago)
Building Use: Educational
Size: 67,500 ft2 (6,271 m2)
Height: 2 Story: Max height: 26.0 ft (8.0 m)

Project Goals:
• 50 Year Service Life.
• Improve roof drainage.
• Recycle the existing ballast, PVC membrane, insulation, and sheet metal.
• Budget $1,056,000 US (approx. 777,500 € / 6,889,340 kr)
Design Concept and Innovation: Use 45-mil EPDM as vapor retarder, mold resistant insulation, 90-mil
ballasted EPDM with filter fabric protection.
The existing ballasted PVC membrane served the college well for the first 10 years of its service life,
after which the PVC migration into the ballast resulted in a rigid membrane prone to splits, especially
under foot traffic. Numerous roof leaks into the interior disrupted the educational process and created
unsafe slip conditions, as well as manifesting staff concerns with mold.
This project was second in a series of three required to complete this large multiple building complex.
Prior to the projects HDG met with the college Building and Grounds Director, ‘Green’ Consultant,
and Vice President to review the concepts of sustainability. The Green consultant immediately asked
for a cool roof membrane, armed with incorrect precepts: It will save energy and reduces the urban
heat island. After a review of the facts, that ballasted roof systems are a greater energy saver than cool
roofing, the reflective roofs actually increase the ambient air temperature above the roof surface, and
that with the exception of one commercially available in the US market, reflective roofing has an
unproven track record in the field; with formula’s being revised almost yearly they could even be
considered “experimental”! A ballasted 90-mil EPDM system with enhancements was decided upon
for the colleges preferred roof system.
FIG 3. Typical Roof Section Profile

2.2.1 Investigation:
Roof cores revealed expanded polystyrene insulation below the 45-mil PVC membrane on a 2 ply
built-up roofing vapor retarder on a concrete roof deck. The existing vapor retarder was found to be in
good condition and very well adhered to the concrete. Removal of the vapor retarder would be costly
and without merit, so it was decided to be allowed to remain.
2.2.2 Design:
While ballasted roof systems have been around for decades, and this author has observed numerous
ones over 25 years of age and still performing, the goal for this roof system was 50 years of service
life. As this project involved the removal of the existing roofing, roof edge raising, roof drain removal
and replacement, as well as renovation, it was decided to enhance the existing vapor retarder with fully
adhered 45-mil EPDM (see Fig. 2B) so that the entire existing roof could be removed, the ancillary
details performed, and the new roof system installed in a more monolithic manner.
Additionally, high PSI polyisocyanurate, which is more stable, and with double coated fiberglass
facers, which are mold resistive, was selected. A 90-mil EPDM membrane with 6” seam tape, and with
all lap seams covered with self-adhering seam tape was designed. Looking forward 25 to 30 years

when the ballast might start to shatter and create shapes, a protective filter fabric was installed over the
EPDM as a buffer between the EPDM and ballast (see Fig. 3). As the thickness of 90-mil EPDM, even
with its flexibility, is difficult to install tight into the angle changes from horizontal to vertical, and
heavy to extend up and over roof edges, this condition was designed and detailed with anchoring the
field sheet to the wall, and using a 60-mil base flashing sheet with 6” seam tape and cover strip. To
raise the roof edge a pre-fabricated metal roof curb in lieu of wood blocking was designed.
Additional roof system design features included:
1. Sealing of all voids between the insulation and roof edge perimeters, pipe and roof curb
penetrations, and roof drains to eliminate thermal shorts and avenues for air transport.
2. Heavy gauge prefinished roof edge sheet metal copings.
3. Pipe penetration and roof curb flashing enhancements to resist anticipated abuse.
4. Protective 2’-0” x 2’-0” x 2” thick rubber walkway pads used to access roof top equipment.
2.2.3 Construction:
The construction commenced in mid-March 2013 and was completed in late June. The author observed
the construction and several anomalies were observed. The existing roof edge wood blocking was
found to be in good condition and was allowed to remain. Even with a high caliber contractor, their
project organization and crew management was less than desirable and lengthened the construction
time. Undulating concrete roof decks challenged the tapered insulation, and the required detailing on
what many would consider the simplest of roof systems took considerable time to install.
3. Conclusion
In over twenty-five years of roof system design, this author has had the opportunity to design
numerous long-term performance roof systems, even before the concept of sustainability was popular.
To this author’s knowledge, 90% of all the roof systems that he has designed are still in place and
functioning properly. Through empirical observations and discussions with roofing contractors,
building scientists, and others in the building industry such as manufacturers and owners, the keys to
sustainability have actually been known intuitively by all of us for years. Now they are stated and need
to be put into universal practice. Building owners, architects, the roofing industry, ‘Green Activists’
need to move forward with enthusiasm, to embrace the concept of long-term service life and thus
sustainability (long-term performance), and to accept the challenges to achieving roof systems that
will perform for half a century: for our clients, for ourselves, and for our children.
References
CIB W083 / RILEM 166-MRS Joint Committee on Roofing Materials and Systems. 2001. Towards
sustainable roofing. CIB Publication No. 271. http://cibworld.xs4all.nl/dl/publications/Publi271.pdf
Desjarlais A.O. et al. 2008. Evaluating the energy performance of ballasted roof systems. ORNL
Report Number UF-04-396
Hutchinson T. W. 2007. Keys to sustainable roofing. Roofing Canada, p. 16-30.
Hutchinson T.W. 2014. Characteristics of in-service, 30 year old EPDM roof membrane. Proceedings
of 2014 Durability of Building Materials and Components Conference in Sao Paulo, Brasil.
U.S. Department of Energy. 2011. Building Energy Data Book.
Zahn A. et al. 2003. Final Results of Study to Estimate the Service Life of Hertlan EPDM roof
sheeting. Life expectancy of EPDM roofing membrane over 50 years. SKZ (Süddeutsche
Kunststoff Zentrum) Final report no. 37236/99-IV. SKZ - TeConA GmbH, Würzburg, Germany.

Full paper no: 22
External insulation with cellular plastic materials – thermal

properties, long term stability and fire properties
Lars Schiøtt Sørensen, Senior Researcher, Ph.D. 1
Anker Nielsen, Professor 1
1
Danish Building Research Institute, Aalborg University, Denmark
KEYWORDS: External insulation, Cellular plastic material, Fire properties, Thermal insulation,
Moisture, Long term stability, ETICS
SUMMARY:
External thermal insulation composite systems (ETICS) can be used as extra insulation of existing
buildings. The system can be made of cellular plastic materials or mineral wool. There is a European
Technical guideline, ETAG 004, that describe the tests that shall be conducted on such systems. This
paper gives a comparison of systems with mineral wool and cellular plastic, based on experience
from practice and literature. It is important to look at the details in the system and at long time
stability of the properties such as thermal insulation, moisture and fire. Investigation of fire
properties must be done before utilisation of the system, including the risk of fire spread from one
storey to the next for practical solutions. An elaboration of fire spread risks require thermo physic
knowledge about ignition temperatures, critical radiation, upward flame spread velocities etc. of the
actual insulation.
1. Introduction
External insulation is a good solution for many buildings as it reduces the effect of cold bridges and in
most cases does not need to disturb the inhabitants in the building. It can be used for both single
family houses and blocks of flats. In most cases it will change the architecture of the building as we
get a new outer layer. This can be a problem for old houses, where we will keep the exterior look. The
external insulation can be mineral wool or cellular plastic material covered of an external layer of
plate material or plaster. There is a European Technical guideline (ETAG 004 from 2011) for external
insulation systems. The systems are called ETICS, an abbreviation for External Thermal Insulation
Composite Systems. The guideline describes the tests for each component and the system that has to
be done to get a CE-mark. The acceptance of the system is based on the whole system, so it is not
allowed just to change for instance the insulation material without a new approval. The European
Association for External thermal insulation (EAE) has made a report (EAE 2011) that describes a
quality system for external extra insulation. In Denmark it has not been allowed to use plastic
insulation as external insulation and an overview of the experience from other countries is interesting.
We will specially focus on the use of cellular plastic material and the problems related to fire. But we
will also discuss the use of mineral wool and compare systems with plastics and mineral wool, as they
have different properties on some important parameters.
2. Components
The systems consist of:
 Adhesive (binds the insulation to the existing wall surface)

 Insulation material
 Anchors (to fix the insulation if the adhesive is not sufficient)
 Base coat of plaster
 Reinforcement (glass fibre mess)
 Top coat of plaster
 Accessories (strips, corner and bottom profiles, expansion profiles etc.)
For anchors a European approval (ETAG 014) defines the properties needed for plastic anchors. For
each of the insulation materials are also found European standards that define the tests that have to be
done.
2.1 Existing wall
An external insulation can be performed on existing walls. The outer surface material of the wall will
influence the solution.
 Heavy constructions. This includes walls of lightweight concrete, concrete and masonry of
different types. In these cases it will normally be easy to fix the insulation systems with anchors
and/or glue.
 Light constructions. This can be wood, plywood and gypsum. The mechanical strength of the
material can be limited and there is a higher risk for movements of the surface. So the fixing is
more complicated. In this case it is important to keep water away from the light construction.
In ETAG 004 are described test methods that must be used to evaluate the system for different outer
surfaces. In the EAE guide is a check list to avoid mistakes in design and execution.
2.2 Insulation material
The thermal insulation material must have a CE-mark and the most important factor is the thickness
and the thermal conductivity. The thermal conductivity shall be declared by the producer. For external
insulation we can normally use the declared value as the insulation should be kept dry. The declared
values for mineral wool are 0.034 -0.040 W/mK. For cellular plastic materials there is more variation
depending on the type – polystyrene, polyurethane, polyisocyanurate or phenol foam. For polystyrene
the value will also depend on the production method – expanded polystyrene (EPS) 0.034 – 0.041
W/mK and extruded polystyrene (XPS) 0.025 – 0.038 W/mK. For phenol, polyurethane and
polyisocyanurate foam, the values can go down to 0.022 W/mK. This is caused by gases used in the
production process. The declared value takes into account the change in gas concentration during the
lifetime.
2.3 Fixing system and surface
The insulation must be fixed to the wall without air gab between wall and insulation material, as this
will reduce the heat resistance of the system. For EPS or mineral wool it is normal to glue all over the
surface or along the borders of each plate and some points in the centre. It is very important to
consider details as corners and connections, so thermal bridges are reduced. The insulation layer must
be fixed with anchors if the glue is not sufficient. In the description of the ETICS it must be described
which type of anchors (and how many) that are needed. Insulation of plastics does not tolerate
ultraviolet-radiation as it can degrade the material, so is it important that the surface plaster is put on

as soon as possible. For the outermost plaster layer it is important to take into account all the details
with windows, doors, overhang, roof connections and all pipes or cables going through the façade.
Examples are fixings for outdoor light and ventilation pipes. These points are critical for the risk of
driving rain penetrate into the insulation material and hits the surface of the old wall. The EAE guide
includes examples of these details, but each producers of a system must show how these details are
solved.
3. Thermal properties
The systems must keep the thermal insulation properties over time, so we still get the energy savings
we calculated for the system, also after 10-20 years. We will normally expect that if we calculate the
transmission heat loss after the European standards then this is a good estimate for the energy savings.
It is possible to measure U-values before and after the installation, but it is very difficult in real
buildings to be sure that the indoor (and outdoor) climate is the same and the inhabitants behave in the
same way. So a theoretical calculation is sufficient, but we have to look at possible problems.
3.1 Change of U-value over time
For most materials there will be no change in thermal conductivity over time. Only cellular plastics
with a foam gas that can diffuse out will increase the thermal conductivity over time, but this should
be included in the declared conductivity.
3.2 Gaps in the insulation
With air gaps between the insulation material and the wall or between the insulation panels, then cold
outer air can flow into the insulation layer, and we do not achieve the expected U-value. This is
important to avoid in the construction phase. Furthermore we must ensure that the fixing of the panel
not loosen during the lifetime of the system.
3.3 Moisture in the insulation
Moisture in the insulation will increase the U-value. The moisture can come from the outside in the
form of driving rain or from the inside, if the moisture barrier is poor. Most critical is driving rain. If
the extra moisture only comes in short periods it will probably dry out.
4. Fire properties
Cellular plastics are made of plastic which is foamed by injecting a gas. The final foam material
consists of approximately 95-98% gas. Cellular plastic insulations are flammable, unlike mineral
wools which are non-combustible. Fire characteristics for cellular plastics are, of this reason,
important to investigate. The risk of fire spread, especially vertically from one storey up to the next, is
important for practical solutions. An elaboration of fire spread risks requires thermo physic
knowledge about ignition temperatures, critical radiation, flame spread velocities etc. Additives as fire
retardants could be relevant to slow down the fire speed or reducing the flammability. This is
discussed further below.
Cellular plastic insulation materials have very different fire characteristics, and there exist a number
of opportunities to improve these through the addition of varying amounts of fire retardants. The

plastic-based insulation materials are found in a variety of configurations, which provide different fire
properties.
We do not have one common European test method for facades. Some countries apply their own
national tests, other utilize test methods, which in principle are developed for other purposes. This
applies to the use of the EN 13823 single burning item (SBI) test for construction products` reaction
to fire. Sweden and Norway use a Swedish full-scale test (SP Fire 105).
Cellular plastic has a large surface area relative to body mass, i.e. a large specific surface, and the
material has great access to oxygen. These factors result in a fast and clean combustion of the
material. The cell structure results in a low density of the material and consequently a relatively low
energy production by combustion, measured by volume unit (Troitzsch 2004). (DBI 2012) provides an
overview of the typical cellular plastic type density, thermoplastic properties and energy, and is
summarized below in Table 1.
Material Thermo plastic or Density Energy content Heat conductivity

thermal curing
(Davies Per Mass Per Volume (EST, 2004)
(Rakic 2003) 2001) (Davies
2001)
EPS Thermo plastic 20 kg/m3 40 MJ/kg 800 MJ/m3 0.035-0.045 W/mK
XPS Thermo plastic 35 kg/m3 40 MJ/kg 1400 MJ/m3 0.025-0.038 W/mK
PUR Thermal curing 45 kg/m3 26 MJ/kg 1170 MJ/m3 0.022-0.035 W/mK
PIR Thermal curing 45 kg/m3 24 MJ/kg 1080 MJ/m3 0.021-0.022 W/mK
PF Thermal curing 45 kg/m3 29 MJ/kg 1305 MJ/m3 0.025 W/mK
Table 1. Overview of selected physical properties of typical plastic-based insulation materials (DBI,
2012).
4.1 Ignition temperatures
The ignition temperature is dependent on whether there is a pilot flame present or not. If the foam
only is affected by heat radiation, the possibility of ignition depends on the exposure time, the
radiation intensity and the size of the exposed area.
Table 2 from (DBI 2012) shows examples of ignition temperatures and critical thermal radiation of
various cellular plastic insulation materials. The critical thermal radiation is the lowest thermal
radiation which may cause ignition. Due to the materials very product specific properties, there may
be a discrepancy between the table and some product values.

Material Ignition temperatures Critical radiation
(auto ignition)
With pilot flame Auto ignition Auto ignition (Babrauskas 2003)
(Davies 2001) (Davies 2001) (Babrauskas 2003)
EPS 245-345 °C 490 °C 440-448 °C 27 kW/m2
XPS 245-345 °C 490 °C - -
PUR 285-310 °C 415-500 °C 457-494 °C 22-26 kW/m2
PIR 415 °C 510 °C - 23-24 kW/m2
PF 490 °C 450 °C - -
Table 2. Examples of ignition temperatures and critical thermal radiation of the typical cellular
plastic insulation materials (DBI 2012).
4.2 Fire retardant
By the addition of a fire retardant (additive) during production, the fire performance of the foam is
changed. An additive can be used to alter cellular plastics ability to shrink or melt, char, slow down
the fire speed or to reduce the flammability (DBI 2012). For example, the bromide flame retardants
are widely used.
It may be noted that, for example ISOBYG`s EPS burn different than normal EPS, due to adding of
flame retardant substances, which make the material does not burn in a self-sustaining reaction. That
is, it turns off automatically when there is no direct flame exposure (ISOBYG 2012).
There are a large variety of additives. This coupled with a myriad of different combinations and
proportions of different types of cellular plastics, makes it very difficult to accurately describe the
additives effect on cellular plastics fire performance. Therefore, a fire-related classification of a foam
material must be determined on the basis of a test of the exact composition of the foam and
additive(s) (DBI 2012).
Fire and flame retardant additives have the greatest impact by initial heating, from small heat sources,
i.e. corresponding to fire effects in the early course of the fire. In a fully developed fire, flame
retardants have no effect on the fire performance (NFPA 2003).
4.3 Recommendations on the fire domain
The fire risk is not yet fully documented for cellular plastic insulation on facades. The risk of external
flame spread seems to be the most critical parameter in this context. It is recommended that flame
spread from one storey to the one above, should be further investigated in order to prove a fire safety
level according to the Building Regulations. At the moment, there exists only the Nordic SP Fire 105
test, which in fact is developed for other types of facades. By the way the test is not suitable for
handling molten material dripping from the burning facade.

5. Long term experience
The Fraunhofer institute of building physics has from 1975 to 2006 made a survey of ETICS systems
in Germany, Austria and Switzerland (Künzel et al 2006). This is an investigation of outside plaster
systems mounted on buildings from 18 to 35 years ago. The results show that there can be algae on
part of the façade, but not very many mistakes in the form of peeling or cracks. The conclusions are
that damage of the plaster is less than plaster direct on masonry. There are more algae on ETICS
houses. The cost of repair and maintenance is similar to buildings with plaster on masonry. These
systems are with 30 to 60 mm insulation on bricks. In Scandinavia it is important with a thicker
insulation layer to improve the U-value.
The lifetime for plaster is important as it will give the lifetime for the ETICS systems. The
information from literature is that the lifetime is similar to a masonry wall or a wooden panel. In all
cases it is assumed that ordinary maintenance is carried out.
In Sweden it has been popular to build high insulated wooden wall with insulation and plaster on the
outside as seen in Figure 1. An investigation of 800 buildings (Samuelson and Jansson 2009) shows
that moisture will come into the construction. There is damage in 55 % of the buildings, mostly in
houses with EPS as insulation. Using mineral wool as insulation reduces the damage to 32%. This is
still a very high damage percentage. The problems are related to details as around windows, doors,
shading roofs, balconies and cracks in the plaster at fixation of light or sunblinds. Testing the systems
before it is used is important and also to reduce the number of special details, where there is a risk of
damage. Later reports (Jansson 2011) confirm the results with EPS giving a high risk of moisture
problems for light walls.
FIG 1. Sketch of system on a light wall from Samuelson 2009, 1- plaster layer 2- external insulation 3
gypsum plate 4 internal heat insulation 5 vapour barrier 6 gypsum plate
The Swedish problems are related to water in the insulation. Drying out of moisture in the external
insulation layer is important. The most efficient way is to use a ventilated space behind the plaster or
outer plate in the system. Swedish calculations from Falk and Sandin 2013 has shown large variations

in drying-out time. The relative drying out time for mineral wool with external plaster layer is 4 times
longer than for a ventilated space. For plastic insulation with EPS and external plaster layer the factor
is 20-30. So the plastic material increases the drying out time significantly.
6. Conclusion
External insulation of facades has, in Denmark, typically been done with systems of mineral wool as
insulation material. This has worked very well, but it is interesting to look at the use of systems with
cellular plastic material. Danish experience with plastic material in external insulation system is
rather limited, primarily due to limiting building regulations, especially in relation to fire. Experience
is taken from Sweden and Germany where it is widespread.
The fire risk is not yet fully documented for cellular plastic insulation on facades. The risk of external
flame spread is found to be the most critical parameter in this context. It is recommended that flame
spread from one storey to the one above, should be further investigated in order to prove a fire safety
level that fulfils the building regulations.
There is no extensive documentation showing that houses with external placed insulation should have
a higher general fire risk than buildings without external insulation.
References
Babrauskas, V. 2003. Ignition Handbook. Interscience Communications Ltd. 2003.
Davies, J.M. 2001. Lightweight Sandwich Construction. Blackwell Science Ltd. 2001.
DBI 2012. Dansk Brand- og sikringsteknisk Institut. Sandwichpaneler med celleplastisolering-

Brandmæssige egenskaber. Rapport rekvireret af Beredskabsstyrelsen. Januar 2012.
EAE 2011; European guidelines for the application of ETICS (external thermal insulation composite
system) - Quality with system, EAE – European association for External thermal insulation
composite systems, March 2011.
EN 13823:2010 Reaction to fire tests for building products. Building products excluding floorings
exposed to the thermal attack by a single burning item. 2010
ETAG 004 2011. Guideline for European technical approval of external thermal insulation composite
systems with rendering, European Organisation for Technical Approvals, Brussels, 2011.
ETAG 014 2011. Guideline for European technical approval of plastic anchors for fixing of external
thermal insulation composite systems with rendering. European Organisation for Technical
Approvals, Brussels, 2011.
EST 2004; Information om brandteknisk dimensionering. Erhvervs- og byggestyrelsen. 2004
Falk, J. and Sandin, K.; 2013. Ventilated rain screen cladding: Measurements of cavity air velocities,
estimation of air change rates and evaluation of driving forces, Building and Environment vol. 59
164-176. 2013
ISOBYG 2012. http://www.isobyg.dk fire information from 6. juli 2012.

Jansson, A. 2011. Putsade regelväggar - Erfarenheter från undersökningar som SP har utfört, Sveriges
Tekniske Forskningsinstitut, SP rapport 2011:61.
Künzel, H., Künzel, H.M. and Sedlbauer, K. 2006; Long-term performance of external thermal
insulation systems (ETICS), Architectura 5 (1) 2006, 11-24.
NFPA 2003. Fire Protection Handbook, 19. edition, Volume II. National Fire Protection Association.
2003.
Rakic, J. 2003. Fire rated insulated (sandwich) panels. Fire Australia. May 2003.
Samuelson, I. og Jansson, A. 2009; Pudsede regelväggar, Byggnadsfysik, SP Sveriges Tekniske

forskningsinstitut, SP rapport 2009:16.
SP Fire 105. Swedish standard for large scale testing of facades system. SP Technical Research
Institute of Sweden, Borås, Sweden.
Sørensen, L.S., Nielsen, A. and Wittchen, K.B. 2013. Udvendig efterisolering med plastbaserede
materialer (external insulation with plastic based materials). Statens Byggeforskningsinstitut,
Aalborg Universitet, København.
Troitzsch, J. 2004. Plastics Flammability Handbook. Published by Carl Hanser Publishers, Munich.
2004

Full paper no: 23
U-value determination of masonry walls with complex shape units

and thermal insulating mortars applied as rendering
Nuno M. M. Ramos, Assistant Professor 1
Vasco P. de Freitas, Full Professor 1
Pedro F. Pereira, M.Sc. 1
Hipólito de Sousa, Associate Professor 1
Rui Sousa, M.Sc. 1
Ângela Nunes, M.Sc. 2
Dina Frade, M.Sc. 3
1
University of Porto - Faculty of Engineering, Portugal
2
Secil, SA
3
Secil Argamassas, SA
KEYWORDS: U-value, masonry, simulation, laboratory measurements
SUMMARY:
Single leaf masonry walls with the application of thermal insulating mortar can be an interesting
solution for façades, providing the performance of its components can lead to a competitive thermal
resistance. The determination of the walls U-value will therefore be of most importance.
Laboratory tests were conducted in a wall composed of a thermal insulating mortar applied on a
single leaf masonry of complex shape units. The objective was the evaluation of the U-value and the
validation of calculation methodologies. The tests included material properties evaluation and U-
value determination of the wall using a climate chamber and heat flow meters.
Numerical simulations were performed with the focus on the calculation of the thermal resistance of
masonry using two approaches. One was a simplified 2D model based on EN ISO 6946 calculation
methodology, and the other one a 3D model based on the finite element method (FEM).
The numerical results obtained from the simulations were compared with laboratory measurements
leading to a difference of about 5%. The differences observed and the results obtained in the material
characterization phase supported recommended strategies that can be used to improve the feasibility
of U-value calculations for the studied wall, including manufacturer declared moisture dependence of
key material properties.
1. Introduction
A growing interest in envelope solutions based on thermal insulating mortars applied on masonry
walls of optimized geometry can be observed nowadays. The effective U-value of these solutions
should however be carefully evaluated. Practitioners often use reference values available in literature
or declared by manufacturers to determine U-values that will be computed in building thermal
calculations. The case of masonry walls is typically sensitive in the sense that manufacturers
frequently base the declared values on calculations using (EN-ISO-6946, 2007) methodology. For the
case of masonry using complex shape units and thermal rendering the definition of the expected U-
value raises additional problems that have to be tackled so that manufacturers declared values can be
trusted.

This paper studies a façade system of the type described above. The study starts with the measurement
of material properties that are used in the numerical model. The U-value was determined following
the principles outlined in the standards (EN-1745, 2012) and (EN-ISO-6946, 2007). A wall specimen
was tested in a climatic chamber according to (EN-1934, 1998) and (ISO-9869, 1994) for comparison
with the calculated values.
2. Studied System
A specimen representing the wall system under analysis was installed in a climatic chamber. It
includes a thermal insulation mortar applied as external rendering. The specimen is 3.61 m2 (1.9 m x
1.9 m) and its constitution is described in TABLE 1 and FIG. 1
TABLE 1 - Materials applied in the test specimens.

Component Description Thickness [cm]
Commercially available lightweight concrete 28.5
Masonry units
masonry units, vertically perforated
Cork aggregates incorporating hydraulic binders, 1
Laying mortar
limestone and siliceous aggregates and additions.
Thermal insulating mortar Dry mortar with EPS aggregates. Density 400 4
(FIG. 2) kg/m3.
Final coat applied above Dry mortar, pre-dosed, formulated from hydraulic 0.15
the thermal insulating binders, aggregates, adjuvants, fibers and resins.
mortar (FIG. 2) Density 1500 kg/m3
Dry mortar, pre-dosed, formulated from hydraulic 1.5
Coat applied above the
binders, aggregates, adjuvants, fibers and resins.
substrate (FIG. 2)
Density 1700 kg/m3
Final coat applied above Dry mortar, pre-dosed, formulated from mixed 0.2
the substrate (FIG. 2) binders, fillers and limestone additions.
FIG. 1 – Wall system.

FIG. 2 – Application of renderings.
3. Material Properties
The thermal conductivity (λ) of the masonry units base materials and of the thermal rendering was
determined. Conductivity can be affected by different factors. The λ is moisture dependent (Freitas,
1994), especially for these materials. Also, according to (Stahl et al., 2012) the pressurization used on
the plastering machine can affect the actual performance of the insulating mortar. Hence, the actual
insulating mortar applied on the wall was studied. Two 30 x 30 cm2 panels were extracted from the
specimen (FIG. 4). The λ of the masonry units material was determined from shell samples removed
from the units. The determination of the thermal conductivity was been carried out using a thermal
conductivimeter by Hot Wire “CT-mètre”. The dry value was calibrated with measurements using the
guarded hot plate according to (EN-12664, 2001).
The λ values of the samples of thermal insulating mortar (A) (bulk density ~400 kg/m3) and masonry
units (B) (bulk density ~440 kg/m3) was obtained for different moisture content values, with results
displayed in TABLE 2 and FIG. 3. In FIG. 3, the thermal conductivity as a function of moisture
content of two types of lightweight concrete (Kumaran, 1996) is also included for comparison. The
lightweight concrete and aerated concrete values from literature demonstrate an upgraded
performance of the material applied in the masonry units of the studied system. The thermal insulating
mortar performance is similar to the polystyrene concrete from literature, although its moisture
dependence is higher.
TABLE 2 – Thermal conductivity values of the samples in various conditions.
W λ
Sample Sample conditions
[%-kg/kg] [W/m.ºC]
A Dried at 70ºC 0% 0.095
A Laboratory conditions (RH ≈ 70%) 2% 0.107
A Intermediate moisture content 28% 0.188
A Saturated 47% 0.273
B Dried at 70ºC 0% 0.178
B Laboratory conditions (RH ≈ 70%) 1% 0.221
B Intermediate moisture content 12% 0.320
B Saturated 19% 0.360

1.0
Thermal insulating mortar
0.9 Masonry unit
Lightweight concrete
0.8
Aerated concrete
0.7 Polystyrene concrete ρ 300
0.6 Polystyrene concrete ρ 370

λ [W/m.ºC]
Polystyrene concrete ρ 430

0.5
0.4
0.3
0.2
0.1
0.0
0% 10% 20% 30% 40% 50%
W [kg/kg - %)
FIG. 3 –Thermal conductivity values obtained for the thermal insulating rendering sample.
FIG. 4 – Extraction of samples of insulating mortar from the specimen and shells from the masonry
units
4. Numerical Simulations
Numerical simulations were performed to determine the thermal resistance of the masonry and the
resulting U-value of the wall system, with two different procedures:
• 2D model based on a simplified geometry of the unit and vertical joints, and on the
calculation methodology of (EN-ISO-6946, 2007) for the determination of the thermal
resistance of the masonry unit with air spaces/voids and (EN-1745, 2012) used on the
calculation of the overall resistance of the wall, i.e. with the influence of the mortar bed joints
and rendering layers;
• 3D model based on the finite element method (Simulia), in which micro modelling of the
masonry system was performed with an equivalent thermal conductivity for the air spaces
(voids) calculated trough the method indicated in (EN-ISO-6946, 2007).

For practical reasons, in particular for the 3D model, the simulations were performed with a
representative portion of the masonry system. The accuracy of both approaches was verified
according to (EN-1745, 2012).
Two different conditions were considered in the simulations:
• Simulation 1: the constituent materials considered with a moisture content (W=2%) to
reproduce the moisture equilibrium with the ambience of the laboratory (HR ≈ 70%);
• Simulation 2: the constituent materials considered in dry conditions (W=0%).
The thermal conductivity values, λ, of the constituent materials were adopted according to section 2
results. Except for the laying mortar (joints) and undercoat mortar (rendering), where conductivity
values were obtained from the manufacturer, and were corrected according to (EN-1745, 2012) in
order to consider a moisture content of 2%. Surface transfer coefficients were not a part of this
specific problem as conventional Rsi=0.13 and Rse=0.04 were added to both numerical and
experimental results.
The masonry was simulated with a temperature variation of 10ºC, and with a surface thermal
resistance of 0.13 m2.ºC/W and 0.04 m2.ºC/W. The thermal conductivity values used for the materials,
and results obtained for two situations are shown in TABLE 3.
The study by (Ghazi Wakili and Tanner, 2003) found a relevant effect of laying mortar penetration on
the final U-value. In the present study, the masonry units in the zone of the laying mortar were probed
and it was possible to see the penetration of the mortar into the unit voids. Because of the masonry
unit geometry and the configuration of the laying mortar used in this study (FIG. 1), the penetration of
laying mortar was only found on the two voids near each surface of the unit. On the bottom voids the
penetration fills the voids up to a height of 1.3 cm. On the top voids of the unit the penetration was
only visible in approximately half of the voids with a height of about 1.3cm. The simulations were
repeated to include the mortar penetration effect.
TABLE 3 – Results obtained in the numerical simulations of the masonry system
U U
without laying with laying
λ mortar mortar
Simulation Materials
[W/m.ºC] penetration penetration
(2D/3D) (2D/3D)
[W/m2.ºC] [W/m2.ºC]
Thermal insulating mortar (rendering) 0.107
Simulation 1
(materials in Lightweight concrete (units) 0.221
0.426 / 0.433 0.429 / 0.438
equilibrium with Laying mortar (joints) 0.177
70% RH)
Undercoat mortar (rendering) 0.573
Thermal insulating mortar (rendering) 0.095
Simulation 2
(materials dried at Lightweight concrete (units) 0.178
0.384 / 0.390 0.387 / 0.393
70ºC and 10% Laying mortar (joints) 0.167
RH)
Undercoat mortar (rendering) 0.540
The U-value obtained from 2D and 3D approaches presented small differences. Although the masonry
units are complex shaped, the voids are based on a simple rectangular shape. Moreover, one
dimensional heat flux was considered (perpendicular to the thicknesses of the wall), since in
stationary regimes other flux directions are usually too low, hence neglected for practical reasons.
Therefore, the results obtained from simplified 2D models or 3D models are expected to be similar.

Nevertheless, the 3D models are expected to be more accurate than 2D models, since they can
represent more realistically the geometry (e.g. irregular shaped voids/air spaces, thickness variation of
the air spaces, amongst others aspects) and thermal behaviour of the walls.
5. Laboratory Tests
The specimen was conditioned by the climatic chamber Fitoclima EDTU 1000 (FIG. 5). The
temperature provided by the device ranges from -50 to 180° C with an accuracy of ± 0.5°C while the
relative humidity range is set from 10 to 98% with an accuracy of ± 2%. The test specimens were
constructed in a detachable part of the climatic chamber (FIG. 5). The wall surface with thermal
rendering is conditioned by the climatic chamber and the opposite surface is exposed to laboratory
conditions. For the determination of the heat flow through the specimen, the equipment presented on
the FIG. 6 was used. This equipment has an accuracy of 5%. The temperatures on the surfaces were
determined using thermocouples of the type T as shown on FIG. 6. The temperature and the relative
humidity near the specimen inside and outside of the climatic chamber was measured with Onset
HOBO data loggers with a temperature range from -20 to 70 ºC with an accuracy of 0.35ºC while the
relative humidity range is from 10 to 90% with an accuracy of 3% (FIG. 6).
FIG. 5 – Climatic chamber Fitoclima EDTU 1000.
FIG. 6 – Thermocouple, HOBO and Heat flow meter.

In order to evaluate the influence of the mean temperature of the specimen and the moisture content
value of the wall, the ambiance of the climatic chamber was changed and thus the conditioning of the
test specimen. The climatic chamber was programed until nearly steady state conditions were
achieved with set-points presented in TABLE 4, along with the test results. In the table, Ti is the
climatic chamber temperature, RHi is the climatic chamber relative humidity, Tmean is the average
temperature of the wall specimen and ΔT (Ti – Te) is the temperature difference between the climatic
chamber and the laboratory. To ensure that the rendering moisture content was in equilibrium with the
climatic chamber RH a sample was collected and weighed.

TABLE 4 – Specimen conditions and thermal transmittance values of the specimen.
Ti RHi Tmean. ΔT (Ti – Te) U
Specimen conditioning
(ºC) (%) (ºC) (ºC) (W/ m2.ºC)
Equilibrium with 70% RH 35.0 70.0 29.5 10.7 0.458
Dried at 70ºC and 10% RH 35.0 26.2 30.1 11.4 0.410
6. Conclusions
The main conclusions of this study are the following:
• The 2D and 3D numerical simulations demonstrated good accuracy with the experimental
results, although 3D gave a better approach;
• A difference of about 5% was found between the numerical simulations (3D) and the
experimental determination of the U value;
• Considering the penetration of laying mortar into the voids of the masonry units, reduces the
difference between measured and calculated (3D) U-Value to 4%.
• A numerical procedure based on a 3D FEM model, in which the thermal conductivity values
used in the model are obtained from laboratory tests made with samples removed from the
masonry components, can be useful to determine the U values for several moisture conditions
without using too many laboratory resources;
• The U-value of this wall system is very sensitive to moisture content even in the hygroscopic
range and the manufacturers should clearly state it;
• This study is part of a work in progress and additional analysis will be performed on the
hygrothermal behaviour of these systems.
7. Acknowledgements
The authors would like to thank Secil and Secil Argamassas for their technical support.
References
EN-1745 2012. Masonry and masonry products. Methods for determining thermal properties.
EN-1934 1998. Thermal performance of buildings. Determination of thermal resistance by hot box
method using heat flow meter. Masonry. 1998-09-30: CEN/TC 89.
EN-12664 2001. Thermal performance of building materials and products - Determination of thermal
resistance by means of guarded hot plate and heat flow meter methods - Dry and moist
products of medium and low thermal resistance.
EN-ISO-6946 2007. Building components and building elements -- Thermal resistance and thermal
transmittance - Calculation method.
FREITAS, V. P. 1994. A influência da humidade na resistência térmica de elementos de construção.
GHAZI WAKILI, K. & TANNER, C. 2003. U-value of a dried wall made of perforated porous clay
bricks: Hot box measurement versus numerical analysis. Energy and Buildings, 35, 675-680.
ISO-9869 1994. Thermal insulation -- Building elements -- In-situ measurement of thermal resistance
and thermal transmittance. ISO/TC 163/SC 1.
KUMARAN, M. K. 1996. IEA ANNEX 24 - Heat, Air and Moisture Tranfer Through New and
Retrofitted Insulated Envelope Parts (Hamtie) - Task 3: Material Properties, Laboratorium
Bouwfysica - Department Burgerlijke Bouwkunde - K.U. Leuven Belgium.

SIMULIA. Abaqus/CAE User's Manual [Online]. Available:
https://www.sharcnet.ca/Software/Abaqus610/Documentation/docs/v6.10/books/usi/default.ht
m [Accessed 01/11/2013.
STAHL, T., BRUNNER, S., ZIMMERMANN, M. & GHAZI WAKILI, K. 2012. Thermo-hygric
properties of a newly developed aerogel based insulation rendering for both exterior and
interior applications. Energy and Buildings, 44, 114-117.

Full paper no: 24
Experimental investigations of highly hygroscopic and vapour

permeable walls exposed to real climate
Yannick-Ariel KEDOWIDE 1
Amandine PIOT, PhD 2
Monika WOLOSZYN, Professor 1
Nolwenn LE PIERRES, Associate Professor 1
Timea BEJAT, PhD 2
1
LOCIE, UMR 5271 Université de Savoie, France
2
CEA, LITEN, Laboratoire d'Energétique du Bâtiment, at INES, Le Bourget Du Lac, France
KEYWORDS: Heat and moisture transfer, outdoor experiment
SUMMARY
An experimental facility consisting of a wall exposed to the external environment and installed on a
cell with indoor controlled conditions is studied. The wall is timber framed with wood fibre panels for
insulation. The first experimental results are presented. The results show the predominant effect in the
short term of the temperature and insolation on the distribution of moisture in the walls, when for the
long term it is rather the water content of inside and outside atmospheres that directs the walls’
moisture distribution.
NOMENCLATURE
Symbol Definition Unit
λ Thermal conductivity mW/(m.K)
θm Moisture content kg/kg
HR Relative Humidity %
µ Vapour diffusion equivalent air layer thickness -
Cp Specific heat capacity J/(kg.K)
T Temperature °C
C Water adsorption coefficient kg/(m².min0.5)
1. Introduction
To study Heat, Air and Moisture (HAM) coupled transfers in building envelopes, many numerical
studies have been done and many experimental facilities under controlled conditions or outdoor
climate have been established [Kêdowidé et al., 2012]. Most of these studies concern realistic
constructions, with high air and water tightness together with a vapour barrier, thereby limiting the
impact of mass transfers on energy performance of the envelope. Yet moisture movements influence
the temperature fields inside the walls [Labat et al., 2012] and indoor comfort [Teodosiu et al., 2003].
There is also a persisting need for precise assessment and understanding of the impact of mass transfer
on the performance of highly hygroscopic wall assemblies, as well as for detailed validation of
dedicated simulation tools. In order to better assess the effects of coupled transfers within wall
assemblies, a new experimental work with highly vapour-permeable and very hygroscopic materials
was used in the work presented below.

2. Experiment description
2.1 PASSYS cells
The PASSYS test cells were originally designed to test passive solar components under real weather
conditions. Two of them were used in the present experiment in order to precisely control indoor
conditions on one side of the tested façade. Both cells are located on the CEA-INES experiment
platform in Le Bourget du Lac, south-eastern France. They are 8.44m long, 3.61m wide and 3.8m
high. The cells are made of a metallic frame insulated by 48 cm of polystyrene and mineral wool to
have 5 highly insulated (U=0.09W/m².K), water and vapour proof walls. The 6th face is reserved for a
wall of maximum 3.6*3.3 m² to be tested. Each cell is placed on a dedicated rotating platform, for a
free choice of orientation of the tested wall. Each cell has a HVAC system to control indoor
temperature and relative humidity. Indoor temperature can be controlled from 15 to 35°C with +/-1°C
tolerance; and indoor air moisture content from 6 to 14 g/kg of dry air with +/- 10% tolerance.
2.2 Walls layout

The tested walls are wooden framed. To have a highly-hygroscopic and vapour-permeable wall, the
insulation was made using two layers of 80mm wood fibre panels. Those panels have 120-140 kg/m3
density. Other characteristics of the panels are in the table below.
TABLE 1: Physical characteristics of the wood fibre panels

Property Value
Thermal conductivity λ = 0.28 * θm + (0.108*T+38)
Sorption isotherm
θm = 5 * 10-5 * HR3 - 0.0053 * HR2 + 0.2925 * HR
Vapour permeability µ = 1.26 (dry cup) - 4.97 (wet cup)
Specific heat capacity Cp = 1551
The walls are then coated with a 3mm roughcast to protect the insulation from liquid loads such as
rain (C<0.20 kg/m².min0.5) [EN 998-1] without being a barrier to vapour diffusion (µ<25).
During the first period of the experiment, reported in the present paper, the 2 walls are identical, and
exposed to the same indoor climate.
2.3 Monitoring system

83 sensors were implemented per wall, together with 8 sensors in rooms’ volumes, as detailed in table
2. Temperature and relative humidity sensors have been placed in different specific areas within the
wall: current part, close to wooden frame and within thermal bridges due to PASSYS cells frame. The
different locations are presented in Figure 1. There are also different locations within the thickness:
between the layers of insulation, in the middle of each layer (a dedicated hole was made), and on the
two sides of the roughcast. For the sensors close to the timber frame, they are placed on the surface
and inside the wooden beam.
The rooms’ volumes were instrumented with seven Pt100 temperature sensors, in order to check the
uniformity of indoor conditions. A thermo hygrometer has been added in the middle of each room,
next to the central Pt100 sensor, to monitor the relative humidity inside the cell.
For exterior conditions, a weather station, located 50m from the cells, recorded meteorological data
such as temperature, relative humidity, wind speed and direction, pluviometry as well as solar (direct,
global) and infrared radiation. Also, a pyranometer was placed vertically on each façade, in order to
anticipate errors due to calculation of the global radiation on each wall taking into account
surrounding buildings, by recording that radiation.

The data was recorded every 5 minutes.
TABLE 2: Overview of sensors used on each wall of the experimental facility

Sensor Number Specifications Accuracy
Thermocouple 36 Type T +/- 1°C
Thermohygrometer 42 Manufacturer : Sensirion +/- 0.3°C; +/1.8% RH
Model : SHT 75
Heat flux sensor 3 Manufacturer : Captec 0.06 W/m² + 0.003% * Measured Value
Dimensions : 100*100mm
Sensibility : 40µV/(W/m²)
Differential 1 Manufacturer : Furness Controls +/- 0.5% of Measured Value
pressure sensor Model : 332
Pyranometer 1 Manufacturer : Kipp & Zonen 7 W/m² (Zero Offset)
Model : CMP 11
Location : On the wall
Tilt : 90°
Cell insulation Coating
Interior
FIG 1: Position of the sensors: General view of the tested façade from the outside (left); Sensors in the
thickness of the wall (right), dimensions in cm
3. Implemented tests
Tests on presented experimental facility began in March 2013, which corresponds to the end of winter
- early spring. The figure 2 shows an overview of the experimental indoor (int) and outdoor (Ext)
conditions over 5 months: temperature, relative humidity and vapour pressure. According to the
weather data, recorded results can be divided into two main sessions.
3.1 Spring session

First, the two facades were oriented to the North. That objective was twofold: first to check whether
both facades had identical behaviour while exposed to exactly the same conditions; and second, to
study their responses to indoor hygric loads with a minimal solar radiation.

The results presented in the following focus on a two-weeks period (May, 18th to June, 1st) when the
indoor temperature was set to 25°C, and a step-rise of relative humidity was made (from 40 to 70%
RH); the duration of the step was one week.
3.2 Summer session

During summer (July, 1st to September, 15th), one of the walls (PASSYS cell n°2) has been turned to
the south. Here the objective was to investigate the effect of sun on the behaviour of the wall. After 6
days, the indoor temperature has been decreased to 16°C (cooling mode) and as the vapour pressure
was maintained constant; relative humidity increased from 40 to 70%. Those conditions were kept on
10 days and then RH was brought back to 40%, which dropped the inner vapour pressure.
FIG 2: Overview of indoor/outdoor Temperature, Relative humidity and Vapour pressure
4. Results
4.1 Spring session
Here will be presented the first relevant test done with the experimental facility. Comparing pairs of
sensors placed at the same position in the two tested wall showed that the responses were very similar;
therefore, only one wall’s results will be presented.
4.1.1 Before the step

The Figure 3 shows the measurements in the wall at the beginning of the session. A monotonous
temperature gradient is observed within the insulation layer, except the temperature just under the
coating (16 cm) because of the non-negligible solar radiation. The temperature difference between the
inside and outside varies from 19°C to 9°C. A time lag for the highest temperatures during the day is
also observed, from the outside towards the inside.
The vapour pressures within the wall intersect twice a day and the moisture flow is inverted from days
to nights. The average values of vapour pressures inside the insulation are fairly close, but their
amplitudes are bigger from the outside to the inside, and time lags similar to the ones recorded in
temperature values are noticed. Daily variations of vapour pressure within the walls seem to be
correlated with the variations of the temperatures.

FIG 3: Temperatures, Vapour pressures and radiation in the wall
4.1.2 Step period

The 24th of May, the relative humidity inside the room was risen up to 70%. The Figure 4 shows the
temperatures, vapour pressures in the wall and global solar radiation at the wall’s surface during the
entire step experiment. Just after the step from 40 to 70% of RH, a small increase in temperature
values can be noticed on the inner surface (0 cm) and in the first four centimetres inside the wall. The
difference between the surface temperature and the indoor air decreases simultaneously to the rise in
the relative humidity. Those temperatures stayed higher than before the step (1 °C on the surface, 1.3
°C at 4cm) until the step back to 40% RH on the 31st of May. This effect is small, but could be seen at
several places of the wall, and on the two test cells, although the indoor air temperature has not
changed. The higher RH has also led to a redistribution of the vapour pressure, and there is a
permanent gradient from inside to outside until the return to initial conditions. The disturbance on
temperature and moisture distribution due to the steps made seems to be really short-term. A dynamic
steady-state is reached after 5 days.
FIG 4: Temperatures, Vapour pressures and radiation on the walls during the steps experiment

4.2 Summer session
During that session, each condition (Temperature and Vapour pressure) was changed separately.
4.2.1 Cell turned to the south

The façade was turned to face South on the 25th of July. These results in temperatures in the wall
higher than the outside air temperature during the day (see Figure 5). We can notice the same effect on
the vapour pressure.
The internal temperature is maintained at 25°C, which is approximately the average temperature of the
outside air; the heat flux therefore changes of direction between day and night. This is not true for the
vapour pressure: the inner relative humidity being of 40% (see Figure 2), the associated vapour
pressure is always lower than the outside vapour pressure.
FIG 5: Temperatures, Vapour pressures in the wall and radiation on it when turned to the south
4.2.2 Change of inner temperature: Cooling mode

The temperature was changed on the 31st July (figure 6). The temperature was dropped to 16°C, while
the relative humidity was raised from 40% to 70%, in order to keep constant the vapour pressure.
This leads to a heat flux being always from outside to inside, as average temperature is no longer the
same. However it does not disturb the distribution of vapour pressure in the wall; the only disturbances
are due to regulation of the HVAC system.

FIG 6: Temperatures, Vapour pressures and RH on the walls when temperature dropped
4.2.3 Change of inner vapour pressure

On the 13rd of August, the indoor relative humidity was decreased to 40% (see Figure 7). Vapour
pressure drop causes a slight decrease of temperature at the inner surface (0.8°C) and inside the
insulation of the wall, as could be seen in the step during the spring session. Yet the temperature
decrease at the surface is close to the measurement accuracy of the sensor, and the decrease at 4cm
inside the wall is under that accuracy, and could not be considered as relevant.
FIG 7: Temperatures, Vapour pressures and RH on the walls when vapour pressure dropped
4.3 Discussion of the results

Moisture content distributions represented here by vapour pressures variations from days to nights
show us that the short time distribution of moisture inside the wall are mostly influenced by
temperature. In fact, amplitude and time of variation are fairly correlated between vapour pressure and
temperature. Thus, most short term moisture movements in the wall for that type of assembly are

thermal driven. Otherwise, the vapour pressure leads long term moisture distribution in walls,
referring to the different distributions of vapour pressure during the test. An influence of vapour
transfers on temperatures is also resorted. It could be explained by sorption (or desorption) for the
short time moves, during the changing of internal conditions from 40 to 70% RH (or back from 70 to
40% RH). This shows up the hygroscopic characteristics of the wood fibre used for the walls. Long-
time moves of temperatures could be explained by the change of heat fluxes due to latent heat or the
change of internal convection coefficient and thermal conductivity in the thickness of the wall with
adsorbed moisture due to the step; or it could be a mix of all those physical effects. Those hypotheses
will be investigated through numerical simulations.
That experimental data may be used for the validation of HAM models, and presents the advantage of
less problems for modelling external conditions due to air layer between claddings and the rest of the
wall, such as encountered in (Desta et al. 2011).
5. Conclusions
A test made on a wall exposed to external environment was studied. The wall is a timber frame with
wood fibre insulation, and a vapour permeable outer coating. The results show that for a very
hygroscopic and vapour permeable wall, on the one hand the movement of vapour are directed by
temperature fluctuations in the short term and the vapour pressure gradient between the exterior and
interior in the long term, and secondly that the vapour pressure variation reciprocally influence the
temperature of the hygroscopic material in the wall. Numerical studies should help better understand
and assess these effects.
6. Acknowledgments
This study is done for the HYGROBAT (ANR-10-HABI-0005) project, financed by the National
Agency of Research of France (ANR).
References
Desta, T. Z., Langmans, J., & Roels, S. , 2011. Experimental data set for validation of heat, air and
moisture transport models of building envelopes. Building and Environment, 46(5), 1038-1046.
Kêdowidé Y., Woloszyn M., Le Pierrès N., 2012, Synthèse sur les dispositifs expérimentaux d'étude
des transferts couplés de chaleur et d'humidité dans les parois des bâtiments ; Colloque International
Francophone d’Energétique et Mécanique (CIFEM), Ouagadougou.
Labat M., Noel M., Woloszyn M., Piot A., Garnier G., Roux J.-J.. , 2012, Comparison of moisture and
temperature measurements for six envelope types with different moisture transfer properties. 5th
International Building Physics Conference. Kyoto, Japan, 28-31 May, 2012.
NF EN 998-1 (Décembre 2010) : Définitions et spécifications des mortiers pour maçonnerie – Partie 1
Teodosiu C., Hohota C., Rusaouën G., Woloszyn M., 2003, Numerical prediction of indoor air
humidity and its effect on indoor environment, Building and Environment, Volume 38, Issue 5, Pages
655–664.

Full paper no: 25
Co-heating test – state-of-the-art and application challenges

Geert Bauwens, Ir. Arch. 1
Staf Roels, Prof. Dr. Ir. Arch. 1
1
KU Leuven, Building Physics Section, Department of Civil Engineering, Belgium
KEYWORDS: Co-heating, reliability, state-of-the-art, heat loss coefficient, thermal performance
SUMMARY:
Several studies show that the actual ‘as-built’ thermal performance of the building envelope can differ
significantly from the theoretical, calculated value. Characterisation of building envelope
performances based on in situ dynamic measurements can help to bridge this gap between ‘designed’
and ‘as-built’ performances. A common method to evaluate the thermal performance of a building in
situ is the co-heating test, which is a quasi-stationary method based on linear regression analysis of
dynamic measurement data.
After a short state-of-the-art on the co-heating test methodology, the limitations and opportunities
associated with the use of the co-heating test method to characterise the thermal performance of
buildings are investigated on the basis of a real full-scale experiment: a co-heating test performed on
a terraced house in Herstal, Belgium. Renovation induced drops in the overall heat loss coefficient of
the dwellings are characterised by the test method and compared with calculated values.
1. Introduction
In order to reduce the energy use of buildings, several countries have put forward more stringent
demands on energy performance of new and renovated buildings. Without exception, these supervised
buildings are characterised or awarded a label in the design phase. A theoretical energy use calculated
on the basis of building plans and specifications determines the performance category. An important
distinction needs to be made, however, between this theoretical energy performance and the actual as-
built performance. Several studies have shown that these can differ rather significantly (Bell et al.
2010, Lowe et al. 2007).
The energy performance of a building is essentially determined by the (1) thermal characteristics of
the building envelope, (2) installed services and (3) building usage. As the latter is not easily
predicted nor controlled, the first two are decisive in achieving the envisaged building energy
performance, both for new buildings and renovations. Hence, the thermal performance
characterisation of a building envelope represents a crucial first step to bridge the gap between its
designed and as-built energy performance. A common method to evaluate the thermal performance of
a building envelope in situ is the co-heating test.
To the knowledge of the authors, (Sonderegger and Modera 1979) is first to mention the use of
thermostatically controlled electric heating experiments to assess a building’s energy performance.
Real full-scale dwellings were alternately heated using the building’s own services and electric heaters
with known efficiency. Hence the name co-heating. Ever since it’s conception, the co-heating test
method has been used to (Sonderegger and Modera 1979):

 assess thermal efficiency of the installed services, e.g. distribution efficiency of duct systems
(Sonderegger and Modera 1979);
 estimate thermal characteristics of the building envelope, e.g. overall heat loss coefficient and
solar aperture (Bell et al. 2010, Lowe et al. 2007, Bauwens et al. 2012);
 load localisation (Sonderegger and Modera 1979)
The method was further explored during the 1980’s and applied sporadically throughout the 1990’s.
Several transient and steady-state derived test procedures were proposed along the way. Recently,
renewed interest in the characterisation of the thermal performance of buildings (Roels 2011) is
apparent and has brought about a revival and further development of the co-heating test method
(Bauwens 2012).
This paper tries to crystallise the current state-of-the-art on the co-heating test, as it is applied to assess
the thermal characteristics of the building envelope. The co-heating test methodology is defined, with
a clear focus on its data analysis part: basic heat balances are set up and applied simplifications are
discussed. The analysis procedure is applied to measurement data collected during a co-heating test
performed on a terraced house in Herstal, Belgium. During the test period multiple renovation steps
were performed, allowing to, aside from a benchmark characterisation, characterise the renovation
induced improvement of the fabrics’ thermal performance.
2. Co-heating test methodology

During a co-heating test, the investigated dwelling is homogeneously heated to an elevated steady-
state interior temperature (e.g. 25°C), using electric heaters. The electrical energy use necessary to
retain this elevated temperature, the indoor and outdoor temperatures, wind speed and direction, and
solar radiation are monitored throughout the test. Using regression analysis on averaged data, the
monitored indoor and outdoor conditions are related to the electrical heating energy needed to sustain
a constant indoor air temperature (Eq. 1(1).
Qh + ∑(Asw,*qsw,*) = HLC(Ta - Ti) = HLCΔT (1)
where Qh energy supplied by heaters and dissipated by ventilators (W)
Asw,* solar aperture coefficient of surface * (m2)
qsw,* global solar radiation, normal to surface orientation * (W/m2)
HLC overall heat loss coefficient (W/K)
Ta - Ti =ΔT indoor –outdoor temperature differential (K)
The coefficients describing the stationary heat balance in Equation 1 represent building thermal
performance characteristics of interest: the overall Heat Loss Coefficient (HLC), in W/K, and one or
more characteristics relating the heating energy to e.g. solar radiation. The overall HLC, on its turn,
constitutes a combined transmission and ventilation heat loss. To decouple both, a co-heating test can
be combined with a building air leakage test, i.e. blowerdoor or tracer gas test.
Notwithstanding a stationary heat balance is assumed, the measurement data are intrinsically dynamic,
due to weather conditions. To diminish thermal dynamics resulting from charging and discharging of
the buildings’ thermal mass, the experiment period needs to be chosen sensibly and the collected
measurement data averaged over a sufficient time span (mostly days, but possibly also several days or
weeks). Hence, the co-heating test essentially represents a quasi-stationary test method based on linear
regression analysis of building performance data.

3. Co-heating test data analysis
3.1 Revisiting the stationary heat balance: applied simplifications
3.1.1 Correlation between solar radiation projections

As a consequence of practical and statistical issues, Equation 1 is challenging to solve. When
considering measurement data averaged over a larger time span, for instance 1 day, the global solar
radiation normal to different orientations * naturally exhibit a strong mutual correlation. For example,
on a sunny day, we are likely to monitor, on average, high global solar radiation values, regardless of
how our pyranometer is orentied. Hence, qsw,* are linear dependent vectors and factors Asw,* in Eq. 1
cannot be estimated separately. The stationary heat balance equation is simplified accordingly:
Qh + ∑(Asw,*qsw,*) = Qh + Asw,*qsw,* = HLCΔT (2)
where Asw,* now represents the solar aperture coefficient corresponding to the sole global solar
radiation component selected as input. For instance, in a next section, measurements of global solar
radiation on a horizontal surface, qsw,hor, is selected.
3.1.2 Disaggregation of transmission and ventilation heat loss

As mentioned earlier, the overall heat loss coefficient HLC actually comprises transmission heat
losses, ∑(AU), and ventilation heat losses, caGa, as illustrated in Equation 3. On the basis of a separate
estimate of the actual air change rate occurring over the course of the measurement period (Bell et al.
2010), both can be decoupled. The actual air change rate can be estimated on the basis of a
blowerdoor tests or tracer gas tests. The latter comes with greater accuracy, but also with greater cost.
Qh + Asw,*qsw,* = HLC ΔT = (∑(AU) + caGa) ΔT (3)
where ∑(AU) transmission heat loss (W/K);
caGa ventilation heat loss (W/K)
The blowerdoor test yields an estimate of the air change rate occurring at a pressure difference inside-
outside of 50 Pa (n50-value). Evidently, this pressure difference is not representative for real scenarios.
Using a rule of thumb, following Kronvall and Persily (Sherman 1987), the n50-value can be related to
the average actual air change rate taking place under real pressure difference scenarios: nactual = n50/20.
The corresponding average ventilation heat loss, caGa, can then be calculated as:
caGa = ca(1/3600)ρanactualV (1/3)n50/20 (4)
where ρa density of air (kg/m³)
V air volume of dwelling (m³)
3.2 Estimating parameters using linear regression analysis

Assuming the heat balance in Equation 3 to hold, the parameters of interest, framed in Eq. 5, are
generally determined by applying simple or multiple linear regression techniques on co-heating
measurement data:
Qh = HLCΔT - Asw, * qsw,* + c (5)
Where c describes the discrepancy between the regression model fit and the actual aggregated
measurement data.
Essentially, three options can be discerned:

1. The energy supplied to the interior under the form of electrical energy can, on a daily average
basis, be corrected for solar gains and plotted as a function of ΔT. This correction implies that
an assumption is made for the solar aperture parameter Asw,*. As illustrated in Fig. 1 (a), the
slope of the regression line resulting from a simple linear regression on this corrected
measurement data set yields an indication of the overall heat loss coefficient (Bell et al.,
2010);
2. An alternative to the method described above is to, aside from ΔT, consider qsw,* as an
additional independent variable explaining the variability of Qh. Multiple regression
techniques allow to determine both HLC and Asw,* in Eq. 5 (Lowe et al., 2007; Everett, 1985).
3. A third method is based on dividing all terms in Eq. 5 by ΔT. An equation is obtained on
which a simple linear regression can be performed, assuming Qh/ΔT as dependent variable and
qsw, */ΔT as independent or explanatory variable, as in Eq. 6.
Qh/ΔT = HLC - Asw, * qsw,*/ ΔT (6)
As illustrated in Figure 1 (b), an estimate of HLC is then given by the intercept. Note that this
mathematical transformation implicitly forces the above described multiple linear regression
through zero. In both of the earlier mentioned methods, a non-zero intercept is possible due to
discrepancies between the measurement data and the assumed stationary model to which it is
fitted. In the third method, these errors are included in the HLC-estimate.
(a) Simple linear regression (b) Simple linear regression transformed

equation
FIG 1. Estimation of HLC and Asw,* by applying simple linear regression
3.3 Thermal lag

From a harmonic analysis, it can be derived that the building fabric typically introduces a phase shift,
between the internal heating power Qh and the external temperature Ta and solar radiation qsw,*. Due to
this phase shift, ΔT at time t will not be representative for the needed energy supply Qh at that time.
To cope with this, Qh at time t can be correlated with ΔT and/or qsw,* averaged over time step t and t-1:
Qh,t = HLC (0.5ΔTt + 0.5ΔTt-1) - Asw,* (0.5qsw,*,t + 0.5qsw,*,t-1) + c (7)
3.4 Reliability of co-heating test methodology

Regardless of the effort made to diminish the transient effects introduced by weather conditions and
the investigated buildings’ thermal mass, stationarity will never fully hold during a co-heating test.
Averaging measurement data and taking into account thermal lags, as described in Section 3.3, only

partly addresses this issue. Hence, the estimates for HLC, Asw,* and possibly c, resulting from linear
regression analysis are necessarily associated with a certain bias (e.g. if solar radiation is not taken
into account) and error. In general, the reliability of co-heating test analysis results will depend on the
investigated building, the imposed regression model and the period (start date and duration) in which
the experiment is performed. To illustrate this, regression analysis can be performed on various
subsets of the available co-heating measurement data, with each subset having a different starting date
and a different duration. The reliability of the co-heating method can then be visualised by plotting
the resulting collection of estimates. This is exemplified in Figure 2 (Bauwens et al. 2012). As the
duration of the experiment increases, the regression result converges, hence the influence of the
experiment starting date diminishes. For shorter measurement durations, however, the starting date is
shown to be crucial. Fig. 2 (b) collects the results of applying multiple linear regression on
measurement data, collected around winter. It can be seen that the results are generally reliable, and
more so for high average Qh monitored during the experiment.
(a) Basic multiple linear regression: ΔT as (b) Multiple linear regression when only
descriptive variables, thermal lag not considering winter months, ΔT and
taken into account, intercept c. qsw,* as descriptive variables, taking
into account thermal lags and forcing
intercept through zero (c = 0).
FIG 2. U-value estimation through multiple linear regressions applied on data acquired during a
simulated co-heating test on an insulated cavity wall component. Different subsets of the
measurement data are considered: with different starting dates (vertically aligned points) and
various durations (along the x-axis). The data points are coloured according to the average Qh
during the considered measurement data subset: red points indicate a high average Qh, blue
points indicate a low average Qh.
FIG 3. Investigated terraced house in Herstal, FIG 4. Co-heating test equipment: heat sources
Belgium. (a) Front façade; (b) Garden façade. controlled by thermostats, ventilators,
temperature sensors and pulse meters spread
throughout the investigated dwelling. Sensor
data (including weather station data) is
transmitted to a central logger located on site.

4. Co-heating in practice: terraced house in Herstal, Belgium
From the 2nd of February to the beginning of May, an extensive co-heating measurement campaign
was performed on a social row house in Herstal, north of Liège, Belgium. During the experiment, two
renovation steps were executed:
 Renovation step 1: blowing in of insulation in the façade wall and party wall cavities and
insulating the attic floor slab;
 Renovation step 2: insulating the floor above basement from underneath.
As depicted in Figure 4, the co-heating measurement equipment was fully deployed. A weather station
was placed in the garden to monitor the outdoor air temperature, solar radiation, wind speed and wind
direction. The indoor air temperatures in every room and the heat input necessary to elevate the indoor
air temperature to 25°C were monitored throughout the test. Two blowerdoor tests were performed,
before and after renovation step 1 (insulating wall cavities and attic floor slab).
4.1 Estimates of renovation induced performance improvement of fabric components

For verification purposes, the U-value reductions for the outside walls, cavity walls and ceilings are
calculated on the basis of the applied insulation thicknesses and their respective λ-values (Table 1).
TABLE 1. Predicted thermal performance improvement induced by renovation step. d=thickness of
applied insulation (cm); λ=heat conductivity (W/(mK)); R=thermal resistance (m2K/W); hi = indoor
surface heat transfer coefficient (W/m2K), he=outdoor surface heat transfer coefficient (W/m2K);
U=heat transfer coefficient (W/m2K), A=component surface (m2).
d λ hi he ΔU A Δ(UA)
2 2 2 2
(cm) (W/mK) (W/m K) (W/m K) (W/m K) (m ) (W/K)
roof insulation 25 0.045 7.70 7.70 -2.00 40.32 -80.77
cavity wall insulation 5 0.034 7.70 23.00 -0.68 51.03 -34.94
party wall cavity insulation 5 0.034 7.70 7.70 -0.74 79.1 -58.40
Table 1 does not take into account certain aspects that could lead to an underestimation or
overestimation of the corresponding reduction of the overall Heat Loss Coefficient HLC:
 Some areas of the cavity are not filled with insulation. The reason for this lies in the way the
building was originally constructed. At the time of construction, it was common practice to
close the cavity around the windows to facilitate installing and stabilising window frames.
Where the cavity is closed no insulation will be added and thermal bridges are created. In
other words, locally the thermal performance will not be improved. Other impurities in the
cavity wall, e.g. interconnections of both masonry wall leafs, rubble in the cavity, …, further
reduce the potential HLC reduction;
 The neighbours moved out shortly after the first renovation step, hence the party wall heat loss
is expected to increase, leading to a possible overestimation.
4.2 Estimates of renovation induced HLC reduction

Figure 5 shows the HLC’s estimated on the basis of a multiple linear regression – Qh as a function of
ΔT and qsw,hor, taking into account thermal lags and forcing the regression surface through the origin –
applied on the co-heating measurement data before the renovation step, after renovation step 1 and
after renovation step 2, respectively. Similar to Fig. 2, discussed in Section 3.4, different data subsets
(different durations, different start dates) were considered in each case, effectively showing how the
estimates for HLC converge towards the values reported in Table 2. As expected, renovation step 1,
which included insulating the cavity walls and the attic floor slab, induces a far more significant

reduction of the HLC (110.92 W/K) than renovation step 2, during which the floor slab above
basement was insulated (reduction of 27.79 W/K). It is seen that renovation step 1 only attains about
63.7% of the projected performance improvement. On the basis of the aspects listed in Section 4.1,
however, this falls within the expected range.
FIG 5. HLC estimates through multiple linear regressions with Qh as a function of ΔT and qsw,hor,
taking into account thermal lags and forcing regression surface through the origin, applied on
measurement data acquired during co-heating test on a terraced house in Herstal, Belgium. Different
subsets of the measurement data are considered: with (1) different starting dates and (2) different
durations. The data points are coloured according to the average Qh over the course of the considered
measurement data subset: red points indicate a high average Qh, blue points indicate a low average
Qh.
TABLE 2. Results of applying simple and multiple linear regression on daily averaged co-heating test
data, before and after renovation step 1 and 2: coefficient of determination R2, estimates for HLC and
Asw,*, p-value Pr(>|t|). The p-value indicates the probability the variable is not significant.
R2 HLC Std.Error Pr(> |t|) Asw,* Std.Error Pr(> |t|)
- (W/K) (W/K) - (m2) (m2) - ΔHLC
Before 0.9983 296.55 13.51 3.98e-09 3.15 3.85 0.433
After step 1 0.9935 185.63 6.27 2.14e-15 4.76 0.69 3.70e-06 110.92
After step 2 0.9899 157.84 9.62 4.54e-10 2.48 0.68 3.06e-03 27.79
Both ΔT and qsw,* prove to be significant variables in all cases but one. Before renovation step 1, the
average solar radiation was rather limited, which explains the fact that the solar radiation as a
descriptive variable proves to be less significant in that case (p-value > 0.05).
4.3 Decoupling HLC into transmission and ventilation heat loss

Before renovation step 1, the measured n50-value was 3.58 h-1, corresponding to a ventilation heat loss
of approximately (1/3) * (n50/20) * V = (1/3) *(3.58/20) * 270.91 = 16.16 W/K. After the first
renovation step, the n50-value has risen slightly to 3.92 h-1, corresponding to a ventilation heat loss of
17.70 W/K. During part of the measurement campaign following renovation step 1 (26th of March to
2nd of May), detailed air change rate measurements were performed, using a tracer gas test with
constant tracer gas pressure. The average air change rate was found to be 0.48 h-1, which is
significantly higher than the air change rate estimated on the basis of the blowerdoor test result and the
rule-of-thumb.

5. Conclusions
This paper investigated the applicability of a co-heating test to determine the overall Heat Loss
Coefficient (HLC) of a dwelling. The co-heating test was introduced and the corresponding data
analysis methodology developed.
A co-heating test was performed on a terraced house in Herstal, Belgium, over an extended period of
time (three months). During the measurement period, two renovation steps were performed. The co-
heating test proves to be a valuable tool here, not only to assess the HLC benchmark value, but also its
reduction as a result of renovations. Multiple linear regression analysis was performed, with the
heating power as a function of indoor-outdoor air temperature difference and global horizontal solar
radiation, taking into account thermal lags introduced on both and forcing the intercept through zero.
Confidence in the HLC-estimates is backed by the fast and rather steady convergence seen in scatter
plots of HLC-estimates resulting from applying the regression analysis on measurement data subsets
with different (increasing) duration and different start dates.
The blowerdoor test, generally used to decouple the HLC into its transmission and ventilation heat loss
parts needs to be used with caution. Tracer gas tests showed that the average actual air change rate far
outpassed the value estimated by reducing the measured n50-value by a factor 20. The considerable
cost associated with a tracer gas test, however, prevents its use on a frequent basis.
6. Acknowledgements
The authors wish to thank Knauf Insulation, for setting up an expert network concerned with
investigating the gap between the ‘designed’ and ‘as-built’ thermal performances of building
envelopes
References
Sonderegger, R. C., Modera, M.P. (1979). Electric Co-heating: A method for Evaluating Seasonal
Heating Efficiencies and Heat Loss Rates in Dwellings. In Second CIB Symposium on Energy
Conservation in the Built Environment. Lawrence Berkeley National Laboratory, March 1979.
Everett, R. (1985). Rapid thermal calibration of houses, ERG, Open University Energy Research
Group, Milton Keynes UK.
Roels, S. (2011). Reliable Building Energy Performance Characterisation Based on Full Scale
Dynamic Measurements. Annex text, November 2011.
(http://www.kuleuven.be/bwf/projects/annex58/index.htm)
Bell, M., Wingfield, M. and Miles-Shenton D. (2010). Low carbon Housing. Lessons from Elm
Treed Mews. Technical report, Joseph Rowntree Foundation.
Lowe, R.J., Wingfield, J., Bell, M. and Bell J.M. Evidence for significant heat losses through party
wall cavities in load-bearing masonry construction. Building Services Engineering Research and
Technology, 28 2:161-181, 2007.
Bauwens, G., Standaert, P., Delcuve, F., and Roels, S. (2012). Reliability of co-heating measurements.
First Building Simulation and Optimization Conference. IBPSA, September 2012.
Max Sherman, Estimation of Infiltration for Leakage and Climate Indicators, in Energy and Buildings,
10, 1987, p.81.

Full paper no: 26
The performance of unventilated wooden roofs with smart vapour

barriers during winter conditions
Stig Geving, Professor 1
Eirik Thorsrud, M.Sc 2
1
Norwegian University of Science and Technology, Norway
KEYWORDS: smart vapour barrier, roof, moisture
SUMMARY:
Unventilated wooden roofs typically need a vapor barrier at the warm side to avoid interstitial
condensation due to vapor diffusion and air leakages from the interior. So-called smart vapour
barriers (SVB) or retarders, could allow condensed moisture, built-in-moisture or moisture from
minor leakages to dry to the interior. The concept of SVBs or retarders have been known for some
time, consisting of a material that changes its vapour resistance according to the level of relative
humidity in the surrounding air. SVB’s should be relatively vapour open during summer conditions to
allow drying to the interior. During winter conditions however, the vapour resistance should
preferably be as high as possible to avoid condensation problems. Various commercial SVB’s have
different properties in regard to the level of their vapour resistance and variation with relative
humidity.
In this laboratory study the level of moisture accumulation during winter conditions have been
investigated for unventilated wooden roofs applying four various types of SVB’s and compared with a
standard roof with polyethylene foil as vapour barrier. Four levels of indoor air humidity were
applied during the test period of three months. The measurement results were compared with
hygrothermal simulations. The results showed that the various SVB’s had very different robustness
against condensation problems during winter conditions, and thus the type of SVB should be chosen in
regard to the expected level of indoor air humidity.
1. Introduction
In Nordic climates a vapour barrier is typically used on the warm side of the building envelope to
avoid interstitial condensation during the heating season due to vapour diffusion and air leakages from
the interior. Often a polyethylene foil is used, which has a very high vapour resistance. Such a vapour
tight product does however not allow moisture to dry to the interior. A more vapour open product
could allow condensed moisture, built-in moisture or moisture from minor leakages to dry to the
interior. This could be especially useful for constructions with reduced or no possibility to dry to the
exterior, such as unventilated flat roofs. It might allow cheaper solutions, making it possible to use
organic (wooden) materials in such constructions or skip the use of wooden preservatives.
The term “vapour retarder”, as opposed to “vapour barrier”, is often used on products that have a
lower vapour resistance than recommended for vapour barriers. Such products may allow for some
drying to the interior. While a vapour retarder has a given constant vapour resistance, some vapour
barriers or retarders are sold (Note: “barrier” is used in the rest of this paper for simplification) on the
European and North American market with adaptable vapour resistance in regard to what is actually
needed. Popular terms for these products are “smart”, “intelligent”, “moisture adaptive” or “humidity
dependent” vapour barriers. The physical behaviour of these products varies, but the main principle is
that the vapour barrier should function as an ordinary vapour tight vapour barrier most of the time,
preventing vapour diffusion into the construction from the indoor air – especially during winter
conditions. If, on the other hand, the construction is wet, for example due to built-in moisture or

leakages, so that the relative humidity (RH) on the exterior side of the vapour retarder gets high, the
vapour resistance will be reduced so that there may be possibilities for drying inwards - especially
during summer conditions. Another possibility may be that the barrier layer has capillary properties
with possibilities to transport condensed water to the interior surface.
Probably the first commercial product that can be put under the term “Smart Vapour Barrier” (SVB)
was Hygrodiode that was developed in Denmark during the 80’s (Korsgaard, 1985). It consisted of
synthetic fibres sandwiched between stripes of polyethylene, giving it a constant very high vapour
resistance similar to the polyethylene film. If on the other hand moisture condensed on the exterior
side of the vapour barrier, the water would be transported sideways through the fibres by the wicking
action and be allowed to evaporate to the interior side.
In the mid 90’s another type of SVB was developed in Germany (Künzel, 1996). The product is named
Difunorm Vario, but is also manufactured under other names, including MemBrain in the US, and
Delta-Novaflexx in some parts of Europe. It consists of a polyamide film with no capillary prop-erties,
but with a pronounced difference in vapour resistance for high and low ambient RH. The Sd-value is
approximately 4-5 m when the RH is below about 40 % and Sd = 0,1-0,2 when the RH is above 80%
(Künzel 1996). As for Hygrodiode, also this product was first introduced as beneficial for unvented
wooden roofs. Most research from then and till today has been focused on use of this product.
Since then and up till today there have been developed various similar products with RH-dependent
vapour resistance. Some of these new products may have a higher resulting vapour resistance for
winter conditions, compared to Difunorm Vario, reducing the risk for interstitial conditions. And some
may have a lower resulting vapour resistance for summer conditions, resulting in higher drying rate to
the interior. Documented scientific studies of applications for these new products are however limited
although they may have some advantages compared to Hygrodiode and Difunorm Vario.
This study is investigating and comparing the performance of two newer products (Intello and
AirGuard Smart) with Delta-Novaflexx (i.e. Difunorm Vario) and Hygrodiode – applied in
unventilated wooden roofs under winter conditions where condensation due to vapour diffusion from
interior side may occur. In previous studies in this project the same products applied in the same
construction has been investigated in regard to the drying potential for built-in moisture during
summer conditions (Geving, Stellander and Uvsløkk 2013; Geving, Thorsrud and Uvsløkk 2013).
2. Laboratory tests
2.1 General
The purpose of these tests was to investigate the risk for undesirable vapour diffusion from interior air
during winter conditions in unventilated wooden roofs with SVB’s, possibly leading to condensation,
moisture accumulation or other moisture problems. SVB’s should be relatively vapour open during
summer conditions to allow drying to the interior. During winter conditions however, the vapour
resistance should preferably be as high as possible to avoid condensation problems.
standard roof with polyethylene foil as vapour barrier. Four levels of indoor air humidity were applied
during the test period of three months. The effect of having some part of the insulation at the interior
of the SVB was also investigated.
The following type of compact unventilated wood frame roof was tested (from exterior side):
 Roofing membrane
 200 mm glass wool
 SVB or PE-foil
 13 mm gypsum board

Even though the construction considered is an unventilated compact roof, the results may also be
applicable to assess the risk for moisture problems in walls and ventilated roofs with SVB’s – although
these constructions have drying possibilities to the exterior.
The measurements took place in the laboratories of Department of Civil and Transport Engineering,
Norwegian University of Science and Technology, during spring 2013.
2.2 Experimental set-up

Five different configurations of the wood frame roof were tested, where the only difference was the
type of vapour barrier used, i.e. four SVB’s and PE-foil. Usually such constructions have a plywood
sheathing or similar below the roofing membrane, but this was of practical reasons omitted in this test,
and instead a thin cotton cloth was used to hold any condensed water below the membrane in place.
This was considered acceptable since there was a constant vapour gradient directed outwards, and the
roofing membrane was practically vapour impermeable. The five configurations were tested each with
two specimens to get a better accuracy, i.e. a total of 10 specimens.
The roof specimens were built up in storage boxes of polypropylene; where the bottom of the boxes
was used to imitate the roofing membrane and the rest of the materials were adjusted to fit the box, see
Figure 1. The boxes had interior dimensions b = 0,28 m, l = 0,40 m and h = 0,22 m, and flanges onto
which the SVBs and gypsum board were taped. The corners of the boxes were curved so the resulting
“light-opening” of the boxes had an area of 0,105 m2. The SVB was taped with double sided tape to
the flanges of the box while the gypsum board afterwards was taped with vapour tight aluminium tape
to the flanges.
FIG 1. Test box with RH/T-sensor before mounting of SVB (left) and after mounting of SVB and
gypsum board.
To simulate a cold external climate on one side and a warm indoor climate on the other side the boxes
were installed in a wall separating two different climate chambers with controlled temperature and
relative humidity, as shown in Figure 2. A horizontal section through two of the boxes as installed in
the wall is shown in Figure 3. This means the specimens were installed vertical instead of horisontal,
that may potentially give some internal convection in the insulation layer. The insulation was however
installed with no air gaps, so it is assumed that the effect of internal convection is insignificant. Since
the main focus of this test was on the total moisture accumulation, any internal convection is anyhow
considered of little importance. On the cold side a continuous wood fibre board layer was installed to

ensure airtightness of the test-wall. On all sides of the test boxes a glass wool and EPS layer was
installed, see Figure 3.
FIG 2. The experimental set-up showing the 10 test boxes installed in the wall, seen from the warm
side.
FIG 3. Horizontal section through two of the test boxes installed in the wall.
2.3 Materials
The four types of SVBs are;
 Delta-Novaflexx
 Intello
 AirGuard Smart
 Hygrodiode 200A
Delta-Novaflexx is a polyamide with an Sd-value ranging from 0,2–5 m. This product is manufactured
under many different names, including MemBrain in the US, and Difunorm Vario in Germany. It was
originally developed in Germany in the late 90’s (Künzel, 1996), and was the first commercial
available relative humidity dependent vapour barrier product for building applications. Most scientific
studies on use of SVBs in building applications are limited to this product only. Intello is a
polyethylene copolymer, where the copolymer is an acrylic with hygroscopic properties. The Sd-value
ranges from 0,25-26 m (Pro Clima, 2006). AirGuard Smart is a polyvinyl alcohol film (with spun bond
polypropylene as reinforcement and protecting layer) with an Sd-value ranging from 0,05 - 102 m
(DuPont, 2012). Hygrodiode 200A is composed of a fabric of synthetic fibres with good capillary
properties laminated with stripes of perforated PE-foil and polyamide-foil. It should be noted that this
product is a further development of the original Hygrodiode mentioned earlier, i.e. the polyamide film
giving the product drying possibilities also for conditions below 100% RH. The Sd-value ranges from

1-20 m (Icopal, 2005). The variation of Sd as a function of RH (data from manufacturer) is given for
three of the SVBs in Figure 4 (detailed measurements have not been available for Hygrodiode). As a
reference one configuration had 0,15 mm PE-foil with Sd = 70 m.The gypsum board has an Sd-value of
approximately 0,10 m.
120
Air Guard
100 Smart
80 Intello
Sd (m)
60
Novaflexx
40
20
0
0 20 40 60 80 100
Relative humidity (%)
FIG 4. Sd-value as a function of RH
2.4 Boundary and initial conditions

The intention of the test was to study the risk for moisture accumulation during winter conditions, with
varying indoor air humidity levels. It was chosen to do four cycles, each lasting three weeks, i.e. the
test lasted a total of 12 weeks (84 days). The real measured average temperatures and RH, and the
moisture supply Δv, for each cycle are given in Table 1. No built-in moisture was added, the initial
moisture content of the materials used were according to a hygroscopic level of approximately 30%
RH.
TAB 1. Measured average T and RH for each cycle
Cycle # Tcold (Cº) RHcold (%) Twarm (Cº) RHwarm (%) Δv (g/m3)
1 -5,1 77,2 21,4 22,4 2
2 -5,1 77,5 23,0 32,8 4
3 -5,0 76,6 23,3 42,6 6
4 -5,0 77,0 23,2 51,7 8
2.5 Measurements
The rig and test boxes were constructed so that the boxes could be dismounted and weighed regularly
to follow the total drying of the configurations. In addition the temperature and RH at the interface
between the SVB and the 200 mm insulation were measured (logged each sixth minute) in all test
boxes. The RH and temperature of the indoor climate were also continuously logged. The RH/T-
sensors were controlled before and after the test, and the results were adjusted according to this
calibration.
3. Hygrothermal simulations
The laboratory test setup were simulated for all the described configurations, except for the
configuration with Hygrodiode where detailed material properties were missing, using WUFI 1D Pro
5.2 (WUFI 2013). WUFI 1D Pro 5.2 is a simulation program for coupled non-stationary, one-
dimensional heat and moisture transport. The program includes the moisture transport by vapour
diffusion and capillary transport, and takes into account the moisture capacity of the materials. The

boundary conditions and material properties given in chapter 2.3 and 2.4 were used. The sorption
curve of the gypsum board is obviously important for these simulations with a focus on moisture
accumulation. Swedish data for interior gypsum board (“Gipsskiva, innvendig”) with density 625
kg/m3, included in the material database of WUFI were used, with the following sorption curve up to
50% RH; 35%-3kg/m3, 50%-4kg/m3.

The accumulation of water in the five roof configurations are shown in Figure 5, to the left is shown
all specimens to give an impression of the variance. We see that the two specimens show a similar
moisture uptake for all the five roof configurations, and very similar for the Intello and PE-foil
configurations. This indicate that the repeatability of the experimental set-up is good. When it comes
to the effect of type of vapour barrier on the moisture accumulation we see that the Novaflexx
configuration accumulates the most, followed by Hygrodiode, Intello, AirGuard Smart and PE-foil. In
general that is what could be expected when comparing the Sd-values for the various vapour barriers
given in chapter 2.3. The average RH across the SVB’s ranges from approximately 19 % for Cycle 1
till 36% for Cycle 4, i.e. according to Figure 4 Air Guard Smart has higher Sd-value than Intello during
all cycles and Novaflexx has the lowest Sd-value. Regarding Hygrodiode we did miss a detailed
description of Sd-variation as a function of RH, but it is interesting to observe that the moisture
accumulation curve in Figure 5 lies between Intello and Novaflexx. The producer of Hygrodiode
gives a dry Sd-value of 20 m, which seems reasonable when we know that the dry value of Intello and
Novaflexx is 26 and 5 m respectively.
35 300
Total accumulated water (g/m2)
30 PE (1) 250
Total accumulated water (g)
Intello (1)
25
Hygrodiode (1) 200
PE
20 Novaflexx (1)
150 Intello
15 AirGuard Smart (1)
Hygrodiode
PE (2)
100 Novaflexx
10 Intello (2)
AirGuard Smart
5 Hygrodiode (2) 50
Novaflexx (2)
0 AirGuard Smart (2) 0
0 21 42 63 84 0 21 42 63 84
Days Days
FIG 5. Total accumulated water in the five roof configurations, all 10 specimens to the left and the
average of two specimens to the right.
In Figure 6 is shown the comparison of measured moisture accumulation with the hygrothermal
simulations with WUFI. For the Novaflexx configuration the simulation compare very well with the
measurements through all cycles. Also for the PE-foil and AirGuard Smart configurations the
simulations and measurements compare very well, with the exception for some deviation for the last
cycle. For AirGuard Smart the simulations give higher moisture accumulations than the measurements
during the last cycle, while the opposite applies for the PE-foil. One explanation for the deviation in
regard to AirGuard Smart could be that the Sd-value in reality is higher for the given RH levels than
we see in Figure 4. New cup-measurements made for AirGuard Smart seem to confirm that
explanation. For Intello the deviation between the simulation and measurements is however bigger, the
simulation showing much lower moisture accumulation than the measurements. Due to the little
deviation between the two test specimens of the Intello configurations we observe in Figure 5, the
most obvious explanation may be that the real Sd-value for Intello is lower than given in Figure 4. It

should be noted that the Sd-value for Intello given in Figure 4 is linearly interpolated for measurement
values between 25 and 50% RH, and thus the real Sd-value could be lower in this interval.
300
250
Total accumulated water (g/m2)
200 PE
Intello
Novaflexx
150
AirGuard Smart
PE‐sim
100 Novaflexx‐sim
Intello‐sim
50 AirGuard Smart‐sim
0
0 21 42 63 84
Days
FIG 6. Comparison of measured and simulated (“sim”) total accumulated water in four of the roof
configurations.
22
20
PE (1)
Intello (1)
Relative humidity (%)
18
Hygrodiode (1)
Novaflexx (1)
16
AirGuard Smart (1)
PE (2)
14 Intello (2)
Hygrodiode (2)
12 Novaflexx (2)
AirGuard Smart (2)
10
0 21 42 63 84
Days
FIG 7. Measured RH at the interface between the glass wool and SVB/PE-foil for all test specimens.
The RH at the interface between the glass wool and SVB/PE-foils were measured and is shown in
Figure 7. We observe that even if the various vapour barriers have very different vapour resistances,
this do not influence the RH at the external side of the vapour barrier too much. For the PE-foil
configuration the RH varies between approximately 12 – 14 %, while for the other SVB’s the RH
varies between approximately 14 – 21% - increasing somewhat as the indoor air humidity and water
accumulation in the test specimens increases during the test. According to Figure 7 the average RH at
the interface between glass wool and the SVS are approximately 16%, 17%, 19% and 20% for Cycle

1, 2, 3 and 4 respectively. This means that the average RH across the SVB’s are approximately 19%,
25%, 31% and 36% for Cycle 1, 2, 3 and 4 respectively. To be able to make high quality
hygrothermal simulations for winter conditions, we need good values/measurements of the Sd-value
for the SVB in the regime 25-40%. Using linear interpolation of the Sd-value between for instance
25% and 50% as is the case for Intello (see Figure 4) is probably not satisfactory.
5. Conclusions
standard roof with polyethylene foil as vapour barrier. Four levels of indoor air humidity were applied
during the test period of three months. The measurement results were compared with hygrothermal
simulations.
The results showed that the various SVB’s had very different robustness against condensation
problems during winter conditions, and thus the type of SVB should be chosen in regard to the
expected level of indoor air humidity. However, whether a certain level of moisture accumulation
during winter conditions is a problem, also depends on the drying during summer conditions. It is for
example interesting to observe that while built-in moisture drying experiments for summer conditions
(Geving, Stellander and Uvsløkk 2013; Geving, Thorsrud and Uvsløkk 2013) show faster drying for
Novaflexx and AirGuard Smart compared to Intello and Hygrodiode, Novaflexx show the highest
moisture accumulation and AirGuard Smart the lowest for this experiment for winter conditions.
The simulations of the test conditions compared well with the measurements, except for a few cases –
where one explanation for the deviations could be that the given Sd-values for those products may not
be sufficiently correct or detailed.
6. References
DuPont. 2012. Tyvek ©. Product Brochure. http://construction.tyvek.co.uk.
Fraunhofer IBP. 2010. WUFI 1D (Version 5.1) [Computer Program]. Fraunhofer IBP, Holzkirchen,
Germany.
Geving, S., Stellander, M. and Uvsløkk, S. 2013. Smart vapour barriers in compact wood frame roofs.
Proceedings of Buildings XII Conference – Thermal Performance of Ecterior Envelopes of Whole
Buildings, December 1-5, 2013, Clearwater Beach, Florida.
Geving, S., Thorsrud, E. & Uvsløkk, S. 2013. Smart vapour barriers in unventilated wooden roofs in a
Nordic climate – laboratory study of drying effect under shaded conditions. Proceedings of 2nd
Central European Symposium on Building Physics, September 9-11, 327-334.
Icopal, 2005. Hygrodiode ©. Product Brochure. www.icopal.dk.
Korsgaard, V. 1985. Hygro diode membrane: A new vapor barrier. ASHRAE/DOE/BTECC
Conference: Thermal Performance of the Exterior Envelopes, Clearwater Beach, Florida,
December 2-5.
Künzel, H.M. 1996. Humidity controlled vapour retarders reduce risk of moisture damages.
Proceedings of the 4th Symposium on Building Physics in the Nordic Countries, Espoo, Finland,
September 9-10.,447-454.
WUFI. 2013. http://www.wufi-pro.com

Full paper no: 27
On the hygrothermal performance of applying exterior air

barriers in Northern Climates.
Jelle Langmans, Ph.D.1
Staf Roels, Professor 1
1
Building Physics Section, KU Leuven, Belgium
KEYWORDS: Exterior air barrier, HAM-simulation, DELPHIN 5, air flow, natural convection
SUMMARY:
This paper studies the hygrothermal performance of timber frame wall configurations with exterior air
barrier systems in various European climates. First, the impact of different insulation materials will
be discussed for Belgium climate conditions. The simulation results demonstrate that the application
of mineral wool insulated timber frame walls in combination with exterior air barriers result in
increased moisture loads. Moreover, unavoidable (small) air channels between the mineral wool and
the adjacent exterior air barrier significantly increase natural convection and add up to harmful
moisture levels. Yet the use of cellulose insulation can avoid these issues. The second part of the paper
studies the performance of these wall elements for 12 European climates, focussing on Nordic
countries in specific. The simulation results indicate that the technique of exterior air barrier is more
suitable for continental climates rather than for cold and moderate sea climates in Europe.
1. Introduction
The execution of exterior air barrier systems in timber frame construction has gained increasing
interest in the aim to meet the severe airtightness requirements of Passive houses (n50<0.6 ACH). In
practice, fewer joints and perforations of exterior air barriers are an advantage compared to the
traditional interior air barrier systems (Langmans, 2010). Also in Norway the practical advantages of
exterior air barriers are mentioned (Myhre & Aurlien, 2005 and Holøs & Relander, 2010). Yet for cold
and moderate climates, such as the Northern European countries, the hygrothermal impact of exterior
air barriers remains unstudied. Consequently, the execution of wall systems with an exterior air barrier
only is currently not recommended for these climate types.
The present paper performs numerical heat, air and moisture (HAM)-simulations on light weight walls
with exterior air barrier systems to explore the risks involved. Yearly simulations under realistic
climate conditions were conducted with an adjusted version of DELPHIN 5, capable to model forced
and natural convection in interaction with detailed heat and moisture transport ( Nicolai, 2007 and
Langmans et al. 2012). The model applies a two-domain approach; (1) porous building materials with
(2) adjacent air channels assuming fully-developed laminar flow. The applied model is extensively
documented and evaluated with experimental results in previous work (Langmans, 2013).
The current article performs numerical simulations to investigate the hygric response of exterior air
barrier systems in timber frame construction in a European climates. In a first step, interest lies in the
impact of different insulation materials. Thereafter, simulation are conducted for various European
climate conditions. Herein, focus is on the performance in Nordic countries. The transient simulations
were conducted for yearly climate conditions and the mould index (M) (Hukka & Viitanen, 1999) and
accumulated interstitial condensation (AC) were applied to assess the performance of these timber
frame elements (Mmax < 3 and ACmax < 0.1 kg/m²).
The following section will discuss the used wall configurations and climate conditions.

2. Wall configuration and climate conditions
2.1 Wall configuration
The simulations throughout this paper are restricted to wall elements only. Fig. 1 shows the general
configuration which is in accordance with the current building practice of highly insulated timber
frame construction in Belgium. Herein, the ventilated cavity of 25 mm contains top and bottom
ventilation openings. The material properties of the cladding system are based on (Nore, 2009); the
wooden cladding layer is modelled according to (Zillig, 2009) and its exterior paint layer has an sd-
value of 0.1m. The wooden top and bottom plates have thickness of 40 mm and extend the full width
of the insulation layer (30cm). A service cavity of 40 mm is applied at the inside of the wall element,
which is covered with a gypsum board finishing layer. It is important to the note that both the interior
vapour retarding layer (OSB) and the gypsum board layer are not continuously airtight, so, they will
not act as an interior air barrier. Slots of 1cm are left at both 20 cm from the top and bottom of the
wall. The air barrier, however, is positioned at the outer side of the insulation layer by sealing the
joints in the exterior sheathing layer.
FIG 1.: Applied wall configuration and simulation grid.

The material properties of the applied materials are summarised in Table 1. Herein, a bituminous
impregnated wood fibreboard (BIFB), a spunbonded foil (FOIL), oriented strand board (OSB),
mineral wool of 20kg/m³ (MW20) and cellulose insulation of 60kg/m³ (CL60) are included. In
addition, the simulations apply non-standard mineral wool with a density of 30 kg/m³ and cellulose
insulation of 40 and 50 kg/m³. Yet these materials are excluded from Table 1 for brevity but can be
found in (Langmans, 2013). Further, the properties of wood are adopted from (Zillig, 2009).
2.2 Climate conditions

The walls are oriented to the North and the applied climatic data is retrieved from the software
package Meteonorm (TRY). The air pressure difference across the wall is a combination of (1) stack
pressure, (2) wind pressure and (3) mechanical ventilation. For the simulations at hand, the stack
pressure is calculated for a two-storey building, assuming that the main leakages are located near the
foundation, wind pressures are calculated based on surface-averaged pressure in which the wind
pressure coefficients are adopted from Swami & Chandra (1987) and a 5 Pa overpressure is assigned
for mechanical ventilation. The inner temperature is assumed constant (20°C) throughout the year. Yet
the inner humidity conditions are determined by a single zone model:
(1)

TABLE 1: limited overview of applied material properties (d (thickness), ρ (density), cp (heat
capacity), λ (thermal conductivity), K (air permeability) and sd (equivalent vapour diffusion).
BIFBI FOIL OSB MW20 CL60
d (mm) 18 0.2 15 300 300
ρ (kg/m³) 285 - 630 21.3 60
cp (J/(kgK) 2068 - 1880 840 2544
λ (W/m/K) 0.045 - 0.1 0.031 0.034
K┴ (m²) 4E-14 airtight 8E-15 1.30E-09 8.8E-10
K║(m²) - - - 3.80E-09 8.8E-10
sd90 (m) 0.1 0.05 0.3 0.3 0.6
sd50 (m) 0.23 0.1 4.2 0.3 0.6
in which V (m³) corresponds to the volume of the room and HIR* (kg/m3%RV) to its hygric inertia, n
(1/h) is the ventilation rate and Gvp(kg/s) is the vapour production in the room. The used parameters
for the single zone model are chosen rather conservative: a small volume of 50 m3 with a high
moisture load of two active persons (120 gram/h) between 8h-22h. The nominal ventilation rate in
living spaces is 3.6m3/m2/h according to the Belgium standard, corresponding to a ventilation rate of
1.51/h for this room. Several studies, however, indicated that the actual ventilation rate is often much
lower in reality. Yet to include the effect of a realistic ventilation system in operation, the nominal
ventilation rate was reduced to 0.5 1/h. Finally, a value of 1.5gram/m3%RV was chosen for the hygric
inertia.
3. Simulation results
3.1 Insulation material
First, the impact of the insulation material is investigated. Herein, two commonly applied groups of
insulation materials are selected for the numerical investigation: (1) mineral wool and (2) cellulose
insulation. Most often mineral wool insulation with a density of 20 kg/m3 (MW20) is used in Belgium.
Densities of 30 kg/m3 (MW30) have only a limited market as they are more expensive. Nevertheless,
both densities (MW20 and MW30) are included in the investigation here. For inflated loose-fill
cellulose insulation, however, densities of around 60 kg/m3, commonly applied in practice, are
adopted here. In addition, cellulose insulations with lower densities (40 kg/m3 and 50 kg/m3) are
examined as well. Table 2 presents the maximum mould index and accumulated condensate on the
exterior air barrier (BIFB) for the reference configuration (Fig.1) with the discussed insulation
materials. The predicted maximum mould index never exceeds the proposed limit of 3 and no
condensation occurred for this type of insulation for the wall elements insulated with cellulose. For the
mineral wool insulation, however, mould growth problems are predicted for the standard density of 20
kg/m3. A mould index of 4.3 was found for this density, corresponding to a visual mould covering
percentage between 10-50%. Mineral wool with increased density (30 kg/m3), in contrast, meets the
proposed moisture limit state. Finally, this table shows that no condensation on the BIFB exterior air
barrier occurred for both the mineral wool and cellulose.

TABLE 2: Maximum mould index (Mmax(-)) and accumulated condensate (ACmax (kg/m²).
Cellulose Mineral wool
ρ (kg/m³) 40 50 60 20 30
Mmax (-)/ ACmax (kg/m²) 2.3/- 1.8/- 1.6/- 4.3/- 2.3/-
The results of Table 2 correspond to an ’ideal’ installation of the insulation layer, in that, perfect
contact between the insulation and the adjacent layers is assumed. On the contrary, several
experimental studies proved the importance of small discontinuities along the interface of mineral
wool insulation layers (Brown et al., 1993). Such small air cavities are responsible for increased levels
of natural convection and consequently higher moisture loads. As a result the presence of small air
cavities around mineral wool insulation should be considered. Therefore, the simulations with mineral
wool insulated elements were repeated, introducing imperfections along the interface for the insulation
and the adjacent layers. Herein, five configurations are considered, varying the position of these air
cavities: an air channel along the (a) cold side of the insulation, (b) on top of the insulation layer (c)
both cold and warm sides, (d) both the cold and warm side and at the top of the insulation, and (e) at
all four sides of the insulation. For these configurations the impact of channels of 1 mm, 3 mm and 5
mm have been studied. All combinations are simulated for both mineral wool with a density of 20
kg/m3 and 30 kg/m3 for which the corresponding simulation results are summarised in Table 3. This
table confirms that the risk for moisture problems is highly influenced by the presence of small air
channels along the mineral wool layer. Even a small air channel of 1mm at both sides of the insulation
increases the mould index from 2.3 to 3.7 for MW30. As a consequence, the above-mentioned
conclusion that higher mineral wool densities decrease the risk for moisture problems only holds for
perfect contact conditions. From the moment imperfections around the mineral wool insulation layer
(which are unavoidable in practice) are considered, the density of the mineral wool becomes of minor
importance. In addition, Table 3 shows that most critical moisture limit state again corresponds to
mould growth.
TABLE 3:Iimpact of air cavities (1mm, 3mm and 5mm) around insulation layer on the maximum
mould index M (-) and maximum accumulated condensation MC (kg/m²) for mineral wool.
MC (kg/m²)
M (-)
Density (kg/m³) 20 30 20 30 20 30 20 30 20 30
1 mm 4.5/- 2.8/- 4.5/- 2.9/- 4.6/- 3.7/- 4.8/- 4.1/- 4.8/- 4.1/-
3 mm 5.5/- 4.6/- 4.7/- 3.6/- 5.8/- 5.1/- 6/1.1 6/0.6 5.9/1.4 6/1.2
5 mm 5.9/- 4.9/- 4.7/- 3.7/- 6/0.1 5.6/- 6/>5 6/4.5 6/>5 6/>5
Fig. 2 depicts the hourly condensation along the height of the exterior air barrier for MW20 with air
channels of 5mm at both sides. This graph demonstrates that condensation clearly occurs during
winter periods and is concentrated at the upper height of this layer, indicating the importance of
natural convection within the component.

FIG. 2: Hourly condensation along the exterior air barrier for mineral wool of 20kg/m³ and air
channels of 5mm on both vertical sides of the insulation layer.
3.2 Different climates

Finally, the hygrothermal response of wall elements with an exterior air barrier is studied for various
European climate conditions. The selected locations cover a wide range of different climates from
South to North Europe. Main focus is on the latter and particularly on Scandinavian regions. Recent
building practice in Norway tends to focus on increasing the airtightness of the exterior layer, as stated
by e.g. Myhre & Aurlien (2005), Holøs & Relander (2010). Therefore, special attention was given to
this country by including Stavanger, Oslo, Bergen and Trondheim in the analysis. Figure 3 shows an
overview of all locations studied. In addition, this figure compares the selected climates by means of
averaged climate indicators.
FIG. 3: Overview of the European locations for which simulations are performed.
First, the averaged outdoor winter temperatures (θwinter), provides a direct indication of the potential for
buoyancy driven air flow within the component, and thus, the increased moisture loads. Second, the
Πy -factor expresses the severity of the exterior climate. This factor (kg/m³), proposed by Hagentoft &
Harderup (1993) for the selection of reference years, calculates the yearly averaged difference between
the vapour concentration in the outdoor air and the maximum moisture concentration on the building
envelope surface on hourly data:
𝑣 𝜃 𝑣 | year (2)
in which vout,sat (θs) represents the saturation vapour concentration at the buildings surface for a North
oriented wall and vout corresponds to the moisture concentration of the outdoor air. As this value is
calculated on a yearly basis, it represents the drying potential of the climate, and is thus, a relevant

indicator for the severity of the climate for the purpose of this study. Both the averaged outdoor winter
temperatures (θwinter) and the Πy -factor of the climates investigated are depicted in Figure 3. This
graph illustrates how the Southern European climate, Lisbon, combines a high averaged winter
temperature with a high Πy -factor. Furthermore, this figure clearly shows higher Πy -factors for
Eastern European climates (Berlin, Vienna) compared to the Western European locations (Brussels,
Nantes, Dublin). In contrast the Western European climates tested correspond to milder winter
temperatures. Most critical locations expected are Helsinki, Stockholm and Trondheim, in that they
combine low winter temperatures with low Πy -factors. Furthermore, this graph illustrates great
differences between the four Norwegian climates included.
The simulations are performed for the wall configuration discussed in section 2.1 for which either
MW30 with two vertical channels of 3mm on both sides or CL60 is applied. For the exterior air barrier
material, on the other hand, two variants are included: (a) BIFB and (b) FOIL. Figure 4 depicts the
simulation results for the wall elements insulated with mineral wool. Here, the maximum mould
growth index on the walls using BIFB exterior air barriers is indicated in black and the maximum
accumulated condensation amounts for the elements with a FOIL are given in red. These two graphs
contain the same information, however, the left hand side is given as a function of θwinter and the right
hand side is presented as a function of Πy . Out of the 12 climates tested only the Eastern European
(Berlin, Vienna), the Southern European (Lisbon) and two of the Northern European locations (Oslo,
Bergen) result in mould growth levels lower than 3. From this figure it follows that Πy is the most
dominating climate parameter in the prediction of mould growth. For these limited number of climates
it appears that no problematic mould growth is observed for Πy -factors higher than 2kg/m³. When it
comes to condensation amounts on the non-hygroscopic foil, however, it appears that the temperature
is the dominating climate parameter. Here only the coldest climates result in excessive condensation
levels (Stockholm, Helsinki, Oslo and Trondheim).
FIG. 4: Top: maximum mould index on BIFB and accumulated condensation on foil as a function of
θwinter (left) and Πy (right) for MW30 insulated elements, bottom: maximum mould index on BIFB and
maximum moisture content of cellulose insulation as a function of θwinter (left) and Πy (right) for CL60
insulated elements.

In contrast to the mineral wool insulated walls, no moisture problems were found for the wall elements
simulated with loose-fill cellulose insulation. Because no condensation was found for these
configurations, the highest moisture contents of the cellulose insulation are given here (in red).
Highest moisture content of the cellulose insulation are found for Stockholm and Helsinki.
In summary, this section explored the hygrothermal response of highly insulated wall elements with an
exterior air barrier in various European climates. In total 12 climates were verified, covering a range
from North to South Europe. From this data it follows that Northern and Western Europe are most
vulnerable to moisture problems when such wall systems are insulated with mineral wool. When these
walls are insulated with loose-fill cellulose insulation, however, no moisture problems were obtained
for the European climates tested.
4. Conclusions
This paper investigated the hygric response of a timber frame walls with an exterior air barrier in a
Belgian climate. The simulation results identified a large impact of the applied insulation material on
the hygrothermal performance of this component. No moisture problems were found for the walls
insulated with loose-fill cellulose insulation. For mineral wool (glass wool) insulated walls, however,
excessive mould growth levels were obtained on the upper parts of the exterior air barrier as a result of
natural convection within the insulation layer. Furthermore, the simulations emphasised the
importance of small air channels along the mineral wool and the adjacent layers. Such limited
deficiencies increased the risk for mould growth and harmful condensation amounts against the
exterior air barrier. The second part of the paper verified the performance of highly insulated walls
with an exterior air barrier in 12 European climates. From the limited amount of locations tested, it
followed that Northern and Western Europe regions were most vulnerable to mould problems. No
mould problems were found for the (dryer) continental climates, such as Berlin and Vienna. For
interstitial condensation on spunbonded foils, however, the decisive climate parameter appeared to be
the winter temperatures. Excessive amounts of interstitial condensation were found for the coldest
climates, such as Helsinki, Oslo, Trondheim and Stockholm.
5. Acknowledgements
Research funded by a Ph.D. grant (grant number 81153) of the Institute for the Promotion of
Innovation through Science and Technology in Flanders (IWT-Vlaanderen).
References
Brown, W. C., Bomberg, M. T., Ullett, J. M., & Rasmussen, J. (1993). Measured thermal resistance of
frame walls with defects in the installation of mineral fibre insulation. Journal of Building Physics,
16(4), 318–339.
Hukka, A & Viitanen, H (1999), ‘A mathematical model of mould growth on wooden material’, Wood
Science and Technology 33(6),475–485.
Holøs, S. B., & Relander, T.-O. (2010). Airtightness Measurements of Wood Frame Low Energy Row
Houses. In BEST conference (pp. 1–11). Portland.
Langmans, J., Klein, R., De Paepe, M. & Roels, S. (2010), ‘Potential of wind barriers to assure
airtightness of wood-frame low energy constructions’, Energy and Buildings 42(12), 2376–2385.
Langmans, J., Andreas Nicolai, Ralf Klein, and Staf Roels. (2012). “A quasi-steady state
implementation of air convection in a transient heat and moisture building component model.”
Building and Environment 58 (12) 208-218.

Langmans, J. (2013), Feasibility of exterior air barriers in timber frame construction, PhD thesis,
Department of Civil Engineering, Section Building Physics, KU Leuven.
(http://bwk.kuleuven.be/bwf/PhDs/phdLangmans)
Myhre, L., & Aurlien, T. (2005). Measured airtightness in low-energy houses. In 7th Nordic Building
Physics Symposium. Reykjavik, Iceland.
Nicolai, A., (2007). Modelling and numerical simulation of salt transport and phase transitions in
unsaturated porous building materials, PhD thesis, Syracuse University, USA.
Nore, K. (2009), Hygrothermal performance of ventilated wooden cladding, PhD thesis, Norwegian
University of science and technology, Norway.
Zillig, W. (2009), Moisture transport in wood using a multiscale approach, PhD thesis, Catholic
Universtity of Leuven, Belgium.
Swami, M. & Chandra, S. (1987), Procedures for calculating natural ventilation airflowbrates in
buildings, Technical report, FSEC-CR-163-86, Florida Solar Energy Center, Cape Canaveral,
Florida, USA.

Full paper no: 28
Computational simulation of hygrothermal performance of hollow

brick-based building envelopes
Václav Kočí, Ph.D. 1
Jiří Maděra, Ph.D. 1
Miloš Jerman, Ph.D. 1
Robert Černý, Professor 1
1
Department of Materials Engineering and Chemistry, Faculty of Civil Engineering, Czech Technical
University in Prague, Czech Republic
KEYWORDS: building envelope, hollow brick, computational simulation, heat and moisture
transport, hygrothermal performance
SUMMARY:
Computational analysis of hygrothermal performance of several types of hollow-brick based building
envelopes is presented, aimed at the optimal hygric properties of hollow brick and exterior plaster. In
the simulations, Künzel’s mathematical model of coupled heat and moisture transport is used.
Dynamic climatic data in a form of Test Reference Year for Prague are utilized as boundary
conditions. The computational results show that for a suitable hygrothermal performance of the
analyzed envelopes the key factor is the hygric properties of the exterior plaster, whose value of
moisture diffusivity has to be reduced to at least 1 · 10-10 m2/s, e.g. using hydrophobic agents. In that
case, moisture content can be held under the hygroscopic threshold for almost the whole reference
year, which ensures the elimination of the presence of liquid moisture inside the envelope.
1. Introduction
Hollow brick blocks have found a widespread use in building industry during the last decades. The
increasing requirements to the thermal insulation properties of building envelopes (Uygunoglu and
Kecebas 2011) given by national standards in Europe led the brick producers to reduce drastically the
production of common solid bricks. The brick blocks with complex systems of internal cavities (da
Silva Almeida et al. 2013, Alhazmy 2010, Arendt et al. 2011) replaced the traditional bricks and
became dominant on the building ceramics market. However, the development of the new types of
bricks was very fast and their properties are often not known with a sufficient accuracy. Consequently,
in the current durability assessment of hollow brick-based building envelope systems a complex view
is often missing. Precise and serious analyses of hygrothermal performance based on sound scientific
knowledge are not performed very frequently. However, hollow brick-based building envelopes should
be designed as a system consisting of hollow bricks, internal and external finishes and possibly also
thermal insulation layer, and in this design process, the details have to be solved using suitable
methods. In this paper, influence of the effective hygric parameters of hollow brick and exterior plaster
on the hygrothermal performance of building envelope is analyzed, in order to identify possible
problems related to the coupled heat and moisture transport. Main objective of this paper is to find
such combination of hygric parameters of exterior plaster and hollow brick, which will reduce the
moisture content in the envelope as much as possible.

2. Computational simulation
2.1 Description of building envelope
As the load bearing material, hollow brick in a thickness of 500 mm, with the cavities filled by
expanded polystyrene, was assumed. This brick was provided with 10 mm thick lime-metakaolin
plaster in the interior and 30 mm thick thermal insulating plaster in the exterior. Scheme of analyzed
building envelope is shown in Figure 1.
Thermal Lime-metakaolin
insulating plaster plaster
Hollow brick
30 500 10
FIG 1. Description of analyzed building envelope
2.2 Material parameters

All material parameters were measured in the laboratories of transport processes of the Department of
Materials Engineering and Chemistry, Faculty of Civil Engineering, Czech Technical University in
Prague using the methods common in building physics (Černý, 2010). The properties of lime-
metakaolin plaster were taken from Vejmelková et al. (2012). The properties of the thermal insulating
plaster have not been published so far. The hollow brick was assumed in the form of a continuum
comprising brick body and expanded polystyrene in the cavities and its properties were obtained using
homogenization principles according to the procedure described in Pavlík et al. (2013), Korecký et al.
(2013). The data are summarized in Table 1. The particular symbols denote: ρ – bulk density (kg/m3),
ψ − porosity (%), c – specific heat capacity (J/kg K), μ – water vapor diffusion resistance factor (-),
λdry – thermal conductivity in dry state (W/mK), λsat – thermal conductivity in water saturated state
(W/mK),κapp – apparent value of moisture diffusivity (m2/s), whyg - maximum hygroscopic moisture
content by volume (m3/m3). Determination of optimal moisture diffusivity of both hollow brick and
thermal insulating plaster was an objective of the computational analysis.
TAB 1. Material parameters
Lime-metakaolin Hollow brick filled with expanded Thermal insulating
plaster polystyrene plaster
ρ (kg/m )
3
1637 793 434
ψ (%) 33.9 wsat = 27 % 40.7
c (J/kgK) 922-1404 1052 - 1995 646-1703
μ (-) 8.5-23.6 10.6-12.8 3.1-13.9
λdry
0.664 0.084 0.097
(W/mK)
λsat
1.847 0.187 0.198
(W/mK)
objective of analysis
κapp (m2/s) 1.85e-7 objective of analysis
(1·10-8; 1·10-9; 1·10-10)
whyg
0.0299 0.005 0.02
(m3/m3)

Moisture diffusivity of porous building materials is highly dependent on moisture content and the
shape of the moisture diffusivity curve is often exponential (Carmeliet et al. 2004). It can be written
that
κ (w) = a ⋅ ebw , (1)

where κ(w) moisture diffusivity as a function of moisture content (m2/s)
a, b constants determining the shape of exponential curve
w moisture content by volume (m3/m3)
The constant b determines the range of functional values of κ(w) function and was set to 2, 10 and 20
according to the characteristic shapes of κ(w) function (Carmeliet et al. 2004). The apparent values of
moisture diffusivity usually correspond to its values at about 2/3 of water saturation (Drchalová and
Černý, 2003), therefore the constant a can be estimated using the equation
2
b⋅ wsat
κ app = a⋅e 3 , (2)
where κapp apparent moisture diffusivity (m2/s) (see Table 1)

wsat saturated moisture content (m3/m3)
The apparent moisture diffusivity of hollow brick was estimated using homogenization principles in a
similar way as in Pavlík et al. (2013), with the result 1 · 10-10 m2/s. The corresponding κ(w) functions
are summarized in Fig. 2.
5.0E-10
Moisture diffusivity (m2/s)
4.5E-10
4.0E-10
3.5E-10
3.0E-10
2.5E-10
2.0E-10
1.5E-10
1.0E-10
5.0E-11
0.0E+00
0.00 0.05 0.10 0.15 0.20 0.25
Moisture content (m3/m3)
κ(w) = 7.359e-11 · exp (2w) κ(w) = 2.158e-11 · exp (10w)

κ(w) = 4.657e-12 · exp (20w)
FIG 2. Moisture diffusivity vs. moisture content functions
2.3 Boundary condition

On the exterior side, dynamic climatic data in the form of Test Reference Year for Prague were
applied, in order to describe the conditions as realistic as possible. Test Reference Year contains
average hourly values of long-term climatic data, such as temperature, relative humidity, wind
velocity, wind direction, different types of sun radiation, or rainfalls. On the interior side, constant
values of temperature (21 °C) and relative humidity (55 %) according to the CSN 73 0540-2 (2011)
were used.

2.4 Mathematical model and computer code
The computer simulation tool HEMOT (HEat and MOisture Transport) (Černý et al. 2010), which is
based on the general finite element package SIFEL (SImple FInite Elements) (Kruis et al. 2010), was
used for the assessment of hygrothermal performance of the analyzed building envelope.
Mathematical model of coupled heat and moisture transport formulated by Künzel (1995) was
implemented in HEMOT for the calculations throughout this paper. The balance equations of moisture
(3) and heat (4) were formulated as follows:
dρ v ∂ϕ
dϕ ∂t
[
= div Dϕ gradϕ + δ p grad (ϕp s ) ] (3)
dH ∂T
dT ∂t
[ ]
= div(λgradT ) + Lv div δ p grad (ϕp s ) , (4)
where ρv partial density of moisture (kg/m3)

φ relative humidity (-)
δp water vapor permeability (s)
ps partial pressure of saturated water vapor (Pa)
H enthalpy per unit volume (J)
Lv heat of evaporation (J/kg)
λ thermal conductivity (W/(m K))
T temperature (K)
This model has been verified (Künzel and Kiessel 1996) and successfully applied in many
hygrothermal simulations before, dealing with hygrothermal performance of building envelopes
(Maděra et al. 2010), optimization of building envelope composition (Kočí et al. 2012), or service life
estimates (Kočí et al. 2009). The simulations were performed as one dimensional. All the material
parameters were determined using one dimensional experimental setup.

The investigated variations of building envelopes were marked according to Table 2.
TAB 2. Investigated variations of building envelopes
Moisture diffusivity of hollow brick (m2/s)
7.359e-11 · e2w 2.158e-11 · e10w 4.657e-12 · e20w
1 · 10-8 A1 A2 A3
Moisture
diffusivity of
1 · 10-9 B1 B2 B3
thermal insulating
plaster (m2/s)
1 · 10-10 C1 C2 C3
The results of hygric and thermal behaviour of studied building envelopes are presented in the form of
temperature and relative humidity profiles for November 22 of the reference year. This day appeared
statistically as one of the most critical from the point of view of hygric performance. According to the
reference climatic data for Prague, a 3-day period with highly increased relative humidity of the air (>
97 %) due to the effect of rainfalls ends in this day. Such high value of the relative humidity
corresponds to the overhygroscopic (liquid) moisture content, thus the presence of liquid moisture in
the pore space of exposed building materials is to be expected. Although in the subsequent days the
relative humidity of the air is still very high, it remains in the hygroscopic range and the liquid

moisture in the materials of surface layers can already begin to evaporate. Therefore, the moisture
content in the external layer of the studied envelope at the end of the mentioned 3-day period can be
considered to belong to the highest; this was confirmed in the test calculations. Taking into account the
subsequent weather development, this moisture could get frozen within the next weeks.
According to the results, hygric parameters of hollow brick did not significantly affect hygric
behaviour of the building envelope. Because of the presence of exterior plaster, moisture content of
hollow brick remained very low and almost invariable, therefore it hardly reached high values, when
the differences between moisture diffusivities of particular variations were considerable. Small
differences are visible in the part of hollow brick which is closer to the exterior. Comparison of
relative humidity profiles of variations A1 and A3 are captured in Figure 3.
100
A1
A3
95
90
Relative humidity [%]
85
80
75
70
65
60
55
0 0.1 0.2 0.3 0.4 0.5
Distance [m]
FIG. 3 Relative humidity profiles, A1 vs. A3
100
A3
C3
95
90
Relative humidity [%]
85
80
75
70
65
60
55
0 0.1 0.2 0.3 0.4 0.5
Distance [m]
FIG. 4 Relative humidity profiles, A3 vs. C3

While the hygric properties of hollow brick had only small effect on the hygrothermal behaviour of the
whole building envelope, the same could not be stated about the hygric properties of the exterior
plaster. Because it is exposed directly to the effects of weather conditions its hygric properties are
decisive, as for the amount of moisture transported inside the envelope. For this reason it is suitable to
apply external plaster with low value of moisture diffusivity; otherwise the moisture content can reach
very high values, as it is shown in Figure 4, where building envelope marked as A3 easily approaches
overhygroscopic values of moisture content.
Comparing Figs. 5 and 6 which show long-term relative humidity distribution of variations A1 and C3
during three reference years, the differences are apparent.
FIG 5. Long-term relative humidity distribution, variation A3
FIG 6. Long-term relative humidity distribution, variation C3

Unlike hygric conditions, temperature profiles of all the investigated variations were very similar.
Therefore, only one example (variation C3) is presented (Fig. 7). Because of the exterior surface layer,
the presence of liquid moisture was limited only to the external plasters and its subsequent worsening
of thermal properties did not significantly influence the thermal behaviour of the building envelope.
FIG 7. Long-term temperature distribution, variation C3
4. Conclusions
The effect of hygric properties of hollow brick and external plaster on the hygrothermal performance
of hollow-brick based envelopes was analyzed in the paper. According to the results, hygric
parameters of hollow brick did not play as significant role as the hygric parameters of the external
plaster. This confirmed that the properties of surface layers of building envelopes are very important
from the point of view of hygrothermal performance of whole building envelopes. In order to reduce
moisture intake, it is necessary to decrease the value of moisture diffusivity of external plaster. Among
all investigated variations, the apparent moisture diffusivity of external plaster equal to 1 · 10-10 m2/s
provided the best results. This value can be easily reached using external plasters with hydrophobic
admixtures.
5. Acknowledgements
This research has been supported by the Ministry of Industry and Trade of the Czech Republic, under
project No FR-TI3/085.
References
Alhazmy M.M. 2010. Numerical investigation on using inclined partitions to reduce natural
convection inside the cavities of hollow bricks, International Journal of Thermal Sciences 49 2201-
2210.

Arendt K., Kraczek M. & Florczuk J. 2011. Numerical analysis by FEM and analytical study of the
dynamic thermal behavior of hollow bricks with different cavity concentration, International
Journal of Thermal Sciences 50 1543-1553.
Carmeliet J., Hens H., Roels S., Adan O., Brocken H., Černý R., Pavlík Z., Hall C., Kumaran K. & Pel
L. 2004. Determination of the Liquid Water Diffusivity from Transient Moisture Transfer
Experiments. Journal of Thermal Envelope and Building Science 27 277- 305.
Černý R. (ed.). 2010. Complex System of Methods for Directed Design and Assessment of Functional
Properties of Building Materials: Assessment and Synthesis of Analytical Data and Construction of
the System, CTU, Prague, Czech Republic.
ČSN 73 0540-2. 2011. Thermal protection of buildings – Part 2: Requirements. Prague: Czech Office
for Standards, Metrology and Testing.
da Silva Almeida G., da Silva J.B., e Silva C.J., Swarnakar R., de Arújo Neves G. & de Lima A. G. B.
2013. Heat and mass transport in an industrial tunnel dryer: Modeling and simulation applied to
hollow bricks, Applied Thermal Engineering 55 78-86.
Drchalová J. & Černý R. 2003. A Simple Gravimetric Method for Determining the Moisture
Diffusivity of Building Materials, Construction and Building Materials 17 223-228.
Kočí V., Maděra J., Černý R. & Rovnaníková P. 2009. Application of a combined computational-
experimental approach for service life estimate of exterior plasters of historical buildings, In:
Structural Studies, Repairs and Maintenance of Heritage Architecture XI. WIT Press, Southampton,
UK, 303-314.
Kočí V., Maděra J. & Černý R. 2012. Exterior thermal insulation systems for AAC building envelopes:
Computational analysis aimed at increasing service life, Energy and Buildings 47 84-90.
Korecký T., Jerman M., Vejmelková E. & Černý R. 2013. Homogenization of physical parameters of
filled hollow bricks, Stavební obzor 22 44-47 (in Czech).
Kruis J., Koudelka T. & Krejčí T. 2010. Efficient computer implementation of coupled hydro-thermo-
mechanical analysis, Mathematics and Computers in Simulation 80 1578-1588.
Künzel H. M. 1995. Simultaneous Heat and Moisture Transport in Building Components, Ph. D.
Thesis. IRB Verlag, Stuttgart, Germany.
Künzel H.M. &, Kiessl K. 1996. Calculation of heat and moisture transfer in exposed building
components. International Journal of Heat and Mass Transfer 40 159-167.
Maděra J., Kočí J., Kočí V., Výborný J. & Černý R. 2010. Computational prediction of hygrothermal
conditions in innovated AAC-based building envelopes. In: Advanced Computational Methods and
Experiments in Heat Transfer XI. WIT Press, Southampton, UK, 291-301.
Pavlík Z., Fiala L., Vejmelková E., Černý R. 2013. Application of Effective Media Theory for
Determination of Thermal Properties of Hollow Bricks as a Function of Moisture Content,
International Journal of Thermophysics 34 894-908.
Uygunoglu T. & Kecebas A. 2011. LCC analysis for energy-saving in residential buildings with
different types of construction masonry blocks, Energy and Buildings 43 2077-2085.
Vejmelková E., Keppert M., Keršner Z., Rovnaníková P., Černý R. 2012. Mechanical, fracture-
mechanical, hydric, thermal, and durability properties of lime–metakaolin plasters for renovation of
historical buildings, Constructions and Building Materials 31 22-28.

Full paper no: 29
Comparison of Direct and Iterative Linear Equation System

Solvers for Building Component Simulation
Anne Paepcke 1
Andreas Nicolai 1
1
Institute for Building Climatology, Faculty of Architecture, Dresden University of Technology,
Germany
KEYWORDS: building envelope, numerical methods, modelling, 2D, hygrothermal transport,

building simulation, linear equation solvers
SUMMARY:
Transient hygrothermal simulations of two-dimensional detailed wall constructions set high demand
on computational capacity. When using implicit time integration combined with spatial Finite-Volume
discretization the resulting system of equations contains a large number of unknowns and the
Jacobian matrices are sparse. This motivates the application of iterative solvers, particularly Krylov
Subspace methods, to the resulting linear equation system. We compare two representative
algorithms for direct and iterative linear equation system solvers: the Gaussian elimination with LU-
decomposition versus the ILU-preconditioned GMRES. The numerical study covers three classes of
problem types that may occur in the hygrothermal simulation: diffusion-dominated transport
problems with highly nonlinear coupling of heat and moisture, diffusion-dominated transport inside
constructions with extremely varying material properties, mixed convection-diffusion hygrothermal
transport problems. The aim of the study is to evaluate the applicability of Krylov Subspace methods
to different problem types and to provide advice for the choice of the linear equation system solver.
1. Introduction
Building component simulation considers the dynamic thermal and hygric transport inside wall
constructions. We use the transport equations based on porous media theory by (Grunewald 1997) that
include both liquid water convection and vapor diffusion. The transport equations are highly
nonlinear and stiff. Therefore, only implicit time integration methods combined with established
nonlinear solution methods may solve these equations sufficiently stable and efficient.
The numerical solution requires a time and spatial discretization. When using implicit time integration
methods, many linear equation systems have to be solved. With increasing complexity of the
construction the solution of the linear equation system dominates the simulation performance
(Vogelsang 2013, Nicolai 2014). Therefore, the demand for an acceptable simulation time limits the
problem dimension. Consequently, efficient numerical solution methods are of high interest.
The discretization of partial differential equations (PDE) results in matrices with characteristic
pattern. This fact offers a high potential for optimizing the linear equation solver. The classical
approach focuses on banded matrices. Alternatively the matrix can be considered as sparse with most
elements of the matrix equal to zero. These matrix properties motivate the application of sparse matrix
linear solvers.

2. Implementation of the Numerical Solution Method
In order to guarantee unconditional stability we apply an Implicit Euler method as time integration
method to the dynamic equations. Additionally, we ensure accuracy of the numerical integration by an
automatic time step adjustment scheme based on local truncation error control (Hindmarsh 2005),
(Nicolai 2008, ch. 4.3).
For each of these integration steps a nonlinear problem is solved using a Newton-Raphson method.
The method requires solving a sequence of linear equation systems consisting of Jacobian matrix and
right hand side vector.
2.1 Direct linear equation solver

We use the Gaussian elimination which is a well-known representative of direct linear equation
system solvers. The method is implemented via decomposition of the discretization matrix into the
product of a lower triangular and an upper triangular matrix, the LU-decomposition (Meister 2008,
ch. 3.1). To reduce the computational effort a special variant is used, taking into account the banded
structure of the Jacobian matrix. The decomposition algorithm is the computational expensive part of
the equation system solver.
Further, the LU-factorization strategy allows the re-use of already calculated decomposition matrices
for the solution of several equation systems, i.e. several Newton iterations, a procedure described as
modified Newton-Raphson method (Hindmarsh 2005). If the Jacobian matrices do not change
significantly this may reduce simulation time substantially. In our implementation the Jacobian is
generally only updated once at the beginning of a Newton iteration. It is then kept in factorized form
until either a convergence failure occurs, or the time step size changes significantly.
2.2 Iterative linear equation solver

Iterative methods only approximately solve the linear equation system. Krylov Subspace methods are
established algorithms for the solution of sparse linear equation systems (Saad 2003, ch. 6). As the
discretized transport equations result in nonsymmetric matrices we choose the Generalized Minimal
Residual (GMRES) method as a suitable candidate.
The GMRES solver is particularly efficient if it converges within a few iterations. For larger number
of iterations error propagation from rounding errors and memory requirements may be problematic.
Fast convergence of the GMRES solver may only be reached by using preconditioning techniques.
Preconditioners usually approximate the inverse of the original matrix. They modify the linear
equation system with the aim to achieve better convergence of Krylov subspace methods. For the
current study we choose a standard Incomplete LU (ILU) – preconditioner (Saad 2003, ch. 10.3) and
an error bound from CVODE (Hindmarsh 2005) to check for convergence of the GMRES method.
3. Simulation examples
Typical for practical applications are two-dimensional simulation problems with a high geometrical
complexity. Sufficiently refined spatial discretization is necessary and leads to critical demand on
computational capacity. Further, for hygrothermal transport problems the intensity of nonlinear
interactions varies highly for each simulation case. Consequently, we take into consideration different
levels of physical model complexity for the current study.
We start with a thermal simulation, continue with a moderately nonlinear hygrothermal simulation,
and finish with a discussion of a highly nonlinear hygrothermal simulation of a steel detail in the
presence of real climate conditions.

3.1 Simulation of a thermal bridge
We choose a thermal bridge example with constant boundary conditions defined in the norm
(EN ISO 10211:2007 2007) for the linear problem case. A cross section of a concrete wall with
insulation and integrated aluminium/wood detail (see FIG. 1) is simulated. On the upper side of the
geometry a constant outside temperature of 0°C is assumed, at the lower side a constant inside
temperature of 20°C.
The norm provides reference results with a tolerance band. In order to achieve this requested accuracy
we apply a non-equidistant discretization clustered near boundaries and material interfaces, resulting
in 13674 elements and a Jacobian matrix dimension of 13674x13674. The requested steady-state
temperature distribution is shown in FIG. 2.
Contrarily to the norm we are interested in the dynamic development of the temperature profile during
the first 5 days, when starting with a constant initial temperature of 10°C.
TABLE 1. Thermal bridge: simulation performance
Numerical method Number of Average time/ Overall simulation
integrator steps integrator step [s] time
Modified Newton, Band Solver 746 0.08 59 s
Modified Newton, GMRES + 748 0.5 6.168 min
ILU
This simulation case is a good candidate for the Modified Newton method. Because the problem is
linear only a few matrix assemblies and few expensive LU-decompositions for the Band Solver are
required. The actual time needed for the solution (backsolving-step) after the decomposition of the
matrix is small, explaining the short overall simulation time (see TABLE 1).
The choice of the linear equation system solver should not affect the convergence of Newton-Raphson
method and consequently lead to the same time step regime and the same number of integrator steps.
Despite this fact, the GMRES variant takes a significantly longer overall simulation time.
Generally, Krylov Subspace methods demand higher computational effort for solving the linear
equation system than needed for a backsolving-step of a Band-Solver with already decomposed
matrix. Contrarily, the effort for the preparation of the equation solving is small compared to the
Band-Solver that performs an LU-decomposition. Effectively, using GMRES shifts computational
time from setup of the equation system to solving.
Note, for a few Newton iterations the embedded GMRES solver did not converge, resulting in
additional Newton iterations. In two cases convergence failures occurred explaining the larger number
of integration steps.
3.2 Hygrothermal simulation of a floor construction

The hygrothermal study of a floor construction in contact with ground is a typical application case.
We choose real climate data of Potsdam (Germany) including rain and perform an annual simulation,
see (Paepcke 2014).
The construction is shown in FIG. 3. It consists of a brick wall that is protected by a bitumen layer
which is considered as watertight. Thus, the rain only infiltrates into the ground on the left side of the
geometry. Rising damp, water vapour diffusion and capillary condensation cause moisture
accumulation in the brick wall, see FIG. 4. The XPS layer inside the floor construction also is
enclosed by a bitumen layer and therefore it is protected from water transport.

We rate this simulation case as a moderate nonlinear hygrothermal problem. The Finite Volume
method discretizes the simulation domain into 7395 elements, a moderate problem size, with a total of
14790 unknowns.
Simulation runs for Band Solver and GMRES variants with modified Newton showed, that simulation
times for GMRES are now lower (see first two lines in TABLE 2). In contrast to the first test case, the
nonlinearity of the problem requires many more expensive LU factorizations for the Band Solver,
which dominates the overall simulation time.
FIG 1. Thermal bridge: geometry.
FIG 2. Thermal bridge: temperature distribution at steady-state.

FIG 3. Floor construction: geometry.
FIG 4. Floor construction: relative humidity at November 28 th.

FIG 5.Balcony bearing: geometry.
FIG 6. Balcony bearing: temperature distribution at June 11 th.

Another consequence of the now very nonlinear equations may be an increasing number of Newton
iterations required, and also more Modified Newton convergence failures because of outdated
Jacobians. This motivates analysis of both Modified and Standard Newton-Raphson methods. The
Standard Newton algorithm updates the Jacobian matrix data for each nonlinear iteration and
performs an LU-decomposition for the Band Solver. It improves convergence of the algorithm, hereby
also reducing the number of Newton convergence errors. Consequently, the number of integrator steps
is reduced (compare first and last two lines in TABLE 2).
TABLE 2. Floor construction: simulation performance
Modified Newton, Band Solver 20522 1.2 6.6 h
Modified Newton, GMRES + 20632 0.3 1.7 h
ILU
Standard Newton, Band Solver 17554 3.0 14.4 h
Standard Newton, GMRES + 17552 0.25 1.1 h
ILU
As expected, the Standard Newton further increases computational effort for the Band Solver. Clearly,
for this case the Modified Newton approach is best when using a direct Band Solver.
For the GMRES solver the time needed per integration step is approximately the same for Modified
and Standard Newton. Consequently, the overall simulation time decreases similarly as the number of
integrator steps. Already in this case a Standard Newton method in combination to a Krylov Subspace
linear solver can decrease simulation time by a factor of 6 if compared to the best direct variant.
3.3 Hygrothermal simulation of a balcony bearing

The last test case considers a balcony bearing consisting of a steel and concrete detail inside a brick
wall (see FIG. 5). The wall is insulated from the outside by mineral wool and the steel beam leads to a
thermal bridge. This fact motivates the application of simulation software in order to get information
about hygrothermal conditions inside the wall. We apply climate data of Potsdam including solar
radiation and rain. The simulation starts at May 15th and continues for one month, see (Paepcke 2014).
In order to guarantee a sufficient accuracy of the results a high-resolution discretization is necessary.
We use 17204 elements and the number of unknowns is increased to 34408 for the hygrothermal
simulation.
The problem is of highly nonlinear type. The steel beam conducts heat very well and therefore
dynamic changes of the outside climate immediately affect a large part of the construction detail.
Additionally, high temperature gradients occur at the contact between the steel and the external mortar
layer (FIG. 6, left side of the construction) that are also rapidly changing in time. This effect results in
very frequent variations in moisture content and temperature in the mortar close to the beam.
This simulation example reveals the limits of the classical Band Solver. The simulation time for this
problem size and the Modified Newton variant is hardly in an acceptable range (see TABLE 3). For
this highly nonlinear problem the Modified Newton method with our current update strategy is
obviously an unsuitable choice.
As before, the GMRES method drastically improves the numerical efficiency.
Note, for this case the application of the Standard Newton method drastically improves convergence
and the number of integrator steps is much lower compared to the Modified Newton variant. This
even leads to a reduction of simulation time in the case of the direct Band Solver. Obviously, the
current Modified Newton update strategy is unsuitable for this problem type.

TABLE 3. Balcony bearing: simulation performance
Modified Newton, Band Solver 151080 12.8 22.4 d
Modified Newton, GMRES + 151083 1.3 2.2 d
ILU
Standard Newton, Band Solver 2840 74.6 2.5 d
Standard Newton, GMRES + 2841 5.2 4.1 h
ILU
Since GMRES does not share the high performance penalty of direct solvers during Jacobian setup it
benefits directly from the reduced integrator steps.
4. Conclusions
The study considers three simulation cases that appear in practical applications. Requirements on
sufficient accuracy lead to high grid detail and critical simulation times. Choosing suitable numerical
methods may drastically reduce the computational effort and therefore give access to simulation
problems with an increased complexity.
In the case of a linear problem (case 1) the Band Solver combined with a Modified Newton method
proves to be a high-performance method. However, hygrothermal simulations are of nonlinear type.
We have shown that Krylov Subspace methods prove to be a very efficient alternative to direct banded
linear equation system solvers when applied to nonlinear transport problems. As the equation system
setup does not demand significant simulation time the Modified Newton strategy is no longer
necessary. We suggest the combination of the GMRES method with a Standard Newton procedure for
large nonlinear problems.
However, the GMRES method is only applicable when problem-specific preconditioned. In our tests
the standard ILU preconditioner provided a base-level for convergence acceleration. The development
of suitable preconditioning strategies will be subject of the future work.
References
Thermal bridges in building construction - Heat flows and surface temperatures - Detailed
calculations (ISO 10211:2007). German version EN ISO 10211:2007
Grunewald J. 1997. Diffusiver und konvektiver Stoff- und Energietransport. Ph.D thesis. Dresden
Hindmarsh A. C. et al. 2005. SUNDIALS: Suite of Nonlinear and Differential/Algebraic Equation
Solvers. ACM Transactions on Mathematical Software. Vol. 31(3). pp. 363-396.
Meister A. 2008. Numerik linearer Gleichungssysteme. Vieweg & Sohn Verlag, Wiesbaden.
Nicolai A. 2008. Modeling and Numerical Simulation of Salt Transport and Phase Transitions in
Unsaturated Porous Building Materials, Ph.D thesis, Dresden University of Technology.
Nicolai A. 2014. Performance improvement of HAM simulations through an optimized grid
numbering technique, 10th Nordic Symposium on Building Physics. Lund.
Paepcke A. 2014. Building constructions for the performance test of dynamic hygrothermal transport
solvers arising from practical application. In prep.
Saad Y. 2003. Iterative Methods for Sparse Linear Systems, SIAM Society for Industrial and Applied
Mathematics. Philadelphia.
Vogelsang S. 2013. Parallel Hardware Architectures. EnTool 2013 – Symposiuum Dresden.

Full paper no: 30
The effect of weather data on glazing U-value in building

performance simulation
Steffen Petersen, Assistant Professor 1
1
Aarhus University Department of Engineering
KEYWORDS: Energy use, Glazing, U-value, Building Simulation, Performance gap
SUMMARY:
The certification of building energy performance according to the European Building Performance
Directive is in many EU countries based on monthly quasi-steady-state calculation of the expected
energy use. But does this calculated energy use correspond to the actual energy use? The aim of this
paper is to contribute to the minimisation of the performance gap between calculated and actual
energy use by improving the precision of the calculated energy use. It is common practice to use the
declared U-value of a glazing in calculations of the expected energy use despite the fact that this U-
value is varying with outdoor temperature, wind speed and direction. Data presented in this paper
suggests that the dynamic changes in U-value due to weather conditions should be taken into account
to obtain a more accurate calculation of the expected energy use of a building in the Nordic
countries. Consequently, three alternative approaches to minimise the performance gap is suggested:
1) abandon the monthly quasi-steady-state method for certification in favour of a dynamic method, 2)
use the approach suggested in this paper to reduce the performance gap and then visualise the
remaining performance gap as an uncertainty of the overall calculation result, or 3) visualise the full
uncertainty of the overall calculation result due to the performance gap.
1. Introduction
The European Building Performance Directive (EPBD 2010) has since 2006 been the framework for
certification of building energy performance in European Union (EU) member states. The certification
process varies from country to country but is very often based on a calculation of the expected energy
use (Lausten et al. 2010). Various stakeholders in the building industry use this calculated energy use
for estimating operating budgets, investment security, policy making etc. Consequently, an increasing
number of stakeholders are raising the question: does the calculated energy use correspond to the
actual energy use?
EPBD prescribes the use of a standardised energy calculation method for calculating the expected
energy use. ISO 13790 (2008) describes two basic types of methods for this purpose where the
monthly quasi-steady-state method is the most widespread method in the EU member states. This
method has its benefits but it is also in many aspects a simplified representation of reality. In relation
to the above raised question, it is therefore relevant to investigate the significance of these simplified
representations. This paper focuses on the simplified representation of heat loss through glazings.
The monthly quasi-steady-state energy calculation relies on the so-called declared U-values of
glazings in its calculation of the annual heat loss. The purpose of the declared value is product
comparison and it is obtained by calculating the U-value according to EN 673 (1997) using a set of
fixed standardised boundary conditions. But EN 673 (1997) also makes it clear that the U-value of a
glazing is not a static value as it varies with the environmental temperatures on each side of the
window, outdoor wind speed and direction, and indoor air flows at the window. This paper therefore
investigate the theoretical effect on building energy calculations when taking into account the

variations in U-value due to weather conditions compared to the use of the static declared U-value.
This investigation is used to suggest how the effect can be represented when calculating the expected
energy use according to the monthly quasi-steady-state method in ISO 13790 (2008). The
investigation is limited to encompass the U-value of the central area of the glazing, i.e. ignoring any
two-dimensional effects when getting near to the window frame.
2. Method
This section describes a procedure for calculating the dynamic U-value of a glazing due to changes in
hourly weather conditions. The procedure is used in two types of energy calculations:
1. Simplified dynamic calculation.
The procedure is integrated in a simplified hourly-based tool called iDbuild (Petersen &
Svendsen 2010). The purpose is to investigate the theoretical effect of a dynamic U-value on
energy performance calculations compared to the use of the static declared U-value.
2. Monthly quasi-steady-state calculation.
The procedure is used to calculate the mean monthly U-values. These values are used for
calculating the energy use of an office space in the program Be10 (SBi 2011). Be10 is the
Danish implementation of the monthly quasi-steady-state energy calculations method
described in ISO 13790 (2008). The purpose is to investigate whether the use of mean
monthly U-values in monthly quasi-steady-state energy calculations is a reasonable approach
to take into account any effect of weather data on glazing U-value identified in the simplified
dynamic calculations.
The results from the calculations and comparisons can be found in section 3.
2.1 Procedure for calculating monthly mean U-values of a glazing

The procedure is based on the approach given in EN 673 (1997). Here the U-value is given as:
1 1 1 1
   (1)
U he ht hi
Where he external heat transfer coefficient (W/(m2·K))
ht total heat transfer coefficient of the glazing (W/(m2·K))
hi internal heat transfer coefficient (W/(m2·K))
he is the sum of a convective conduction and a radiation conduction. The convection part is the
dominating and can be 3 to 4 times larger than the radiation conduction (Cooper 1981, Palyvos 2008).
The convection conduction dependent on many factors such as the geometry of the building and its
surroundings, the position at the building envelope, the building surface roughness, wind speed, wind
direction, local airflow patterns and surface to air temperature differences. The research made in this
area is extensive and on-going and there are many different calculation methods available. It is
difficult to choose a model as there is large uncertainty associated with the use of them as they vary in
scope and detail (Mirsadeghi 2013). The model for representing he in this procedure is the NBS
polynomial model (Rowley 1930). This model correlates he to surface roughness and local surface
wind velocity. This correlation is chosen because it includes radiation to sky, ground, and air, and we
thus avoid the rather detailed calculation of the radiation conduction. he as a function of air velocity
for ‘very smooth’ materials like glass is:
he  11.56  0.036Vs
2
(2)
Where Vs air velocity at the surface (m/s)

Vs can be calculated using the procedure described by Ito et al. (1972). hi is also the sum of a
convective conduction (hc) and a radiation conduction (hr). Both conductions are dynamic as they
change with the temperature of the internal surface of the glazing, the room surfaces and the indoor
air. A simplified expression is therefore preferred because the monthly quasi-steady-state method do
not involve calculations of these temperatures. hi can be expressed in a simplified way according to
EN 673 (1997):
4.4
hi  hc  hr  3.6  (3)
0.837
Where ε corrected emissivity of the internal surface of the inner glazing pane (-)
According to EN 673 (1997) ht can be expressed as the sum of the gas space conduction (hg) and the
conduction of the individual glazing panes (hp). In this procedure, it is assumed that hp is independent
of weather conditions. hg is the sum of the radiation, conduction and convection conduction which all
are dependent on the environmental temperatures on each side of the glazing:
 The radiation conduction depends on the temperature difference between the glass surfaces
surrounding the gas space (ΔT).
 The gas property (density, dynamic viscosity and conductivity) depends on the mean absolute
temperature of the gas space (Tm) which affects the conduction as well as convective
conduction. The gas property as a function of temperature can be found ISO 15099 annex B
(2003).
 The convection conduction expressed by the Nusselt number also depends on ΔT.
The calculation of the U-value of the glazing, it is assumed that a state of equibrilium is reached in
each weather data time step. This requires an iterative process because ΔT and Tm are not known
beforehand. The iteration in each weather data time step starts with a guess of the temperatures ΔT
and Tm:
ti  t o
Tguess  0.75 (4)
n
Where n factor depending on number of panes, n=1 for two-layer glazings and n=2 for three-
layer glazings
Tm, guess  273  k t i to (5)
Where k factor depending on number of panes, k=0.5 for two-layer glazings, k=0.25 for the
cavity closest to outside in a three-layer glazing and k=0.75 for the cavity closest to
inside in a three-layer glazing
These guesses are used to calculate the initial U-value of the glazing. The initial U-value is used to
calculate the temperatures of the glazing pane surfaces facing the gas space (t s,x). ts,x of a certain
surface x can be expressed as:
x
h z
1
t s , x  ti  ti  to  z  i 1 (6)
U
Where ti indoor air temperature (°C)
to outdoor air temperature (°C)
z index referring to the heat transfer coefficients in equation 1 starting with hi
A proportion of the total solar radiation is absorbed in the panes and thus increases the pane
temperature. For example, an incident solar radiation of 500 W/m2 increases the temperature of an

outer pane with an absorbance of 0.07 by approximately 1.5 ºC. Taking the relationship between
incident solar radiation and pane temperature into account complicates the calculation of the total heat
loss coefficient as it then becomes dependent on the orientation and tilt of the glazing. The
relationship is ignored in the analysis presented in this paper but it could be added in future more
detailed analysis.
The surface temperatures ts,x are used for recalculating ΔT and Tm which then is used for recalculating
the U-value. This process is repeated in each time step until the U-value converges on the third
decimal.
The above described procedure can be used for calculating the glazing U-value for every hour of the
year. As an example of this, FIG 1 depicts the variation in glazing U-value of a two-layer and a three-
layer glazing, respectively, for a west-facing glazing in Kiruna, Sweden. Results will always differ
with orientation because the convective part of the exterior heat transfer coefficient depends on
orientation. The deviations from the declared U-value are considerable for the two-layer glazing. This
is mainly due to the radiation conduction which is the conduction most sensitive to the difference
between indoor and outdoor temperature. The U-value of the three-layer glazing is more stable as the
difference between indoor and outdoor temperature is less governing. The U-values are, however, in
general somewhat lower than the declared value which is due to weather-induced fluctuations in the
external heat transfer coefficient which often is lower than the standardised boundary conditions in
EN 673 (1997).
FIG 1. Hourly U-value of two-layer west-facing glazing in Kiruna, Sweden. The vertical dotted line is
the declared U-value using the boundary conditions in EN 673 (1997).
2.2 Dynamic energy calculations

The procedure for dynamic calculations is in general as described in section 2.1. However, the radiant
part of the internal heat transfer coefficient hi can now be calculated as the mean radiant temperature
is known:
hi 

 t s , x 4  t rm 4  (4)
t s , x  t rm
Where σ Stefan-Boltzmann’s constant, 5.67·10-8 (-)

trm mean radiant temperature of the room (K)
3. Results
The following case was used to investigate the impact of dynamic glazing U-value on energy need for
heating and cooling using the formulas specified in section 2 compared to the use of the static
declared U-value. Assume a single-sided, single-zone two-person office with a west-facing window.

The occupied period is from 8 a.m. to 5 p.m. every weekday. The lower limit for thermal comfort is
set to 20ºC and the upper limit is set to 26ºC. Further data assumptions are shown in Table 1.
TABLE 1. Data assumptions for case.
Parameter Description
Room dimensions Height × width × depth 2.8 m × 3 m × 6 m
Window Height × width 1.8 m × 2 m
Offset Symmetrical, 0.8 m from floor
Glazing (U/g/LT) Two-layer (1.09/0.31/0.65) or
Three-layer (0.73/0.34/0.58)
Frame Standard wooden frame, U=1.6 W/(m2·K),
width=0.08 m, ψ=0.05 W/(m·K)
Constructions Façade U=0.15 W/(m2·K)
Thermal capacity 120 Wh/(K·m2)
Systems Infiltration 0.10 l/s m2, always active
Mechanical ventilation Min. ventilation rate 1.48 l/s m2 corresponding to
in occupied hours class II in EN 15251:2007. Max. ventilation rate is
2.96 l/s m2. Average specific fan power of 1.0
kJ/m3 air. No mechanical cooling available
Mechanical ventilation Min. ventilation rate is 0 l/s m2, max. ventilation
in unoccupied hours rate 2.96 l/s m2
Heat exchanger Efficiency of 85 %
General lighting Dimming control, set point 200 lux, min. power 0.5
W/m2, max. power 6 W/m2, 3 W/m2/100 lux. Only
active in occupied hours
Task lighting Always on in occupied hours, set point 500 lux, 1
W/m2.
Internal load 300 W in occupied hours, 0 W in unoccupied hours
External conditions Shadows from None
surroundings
Two glazing solutions are investigated: 1) a two-layer and 2) a three-layer glazing. Both solutions are
simulated for four locations in the Nordic countries: Copenhagen (Denmark), Oslo (Norway) and
Kiruna (Sweden), and Helsinki (Finland). IWEC data files normally used for EnergyPlus simulations
(U.S. Department of Energy 2013) is used for representing hourly weather conditions.
3.1 Dynamic energy calculations

FIG 2 and FIG 3 depicts the annual sum of the hourly simulated energy need for heating and cooling
using the static declared U-value and the dynamic U-value for the two-layer and three-layer glazing,
respectively. For the two-layer glazing, simulations with dynamic U-value results in higher heating
need and a slightly lower cooling need compared to simulations with declared U-value. For the three-
layer glazing, simulations with dynamic U-value results in a somewhat lower heating need and
practically the same cooling need compared to simulations with declared U-value.

FIG 2. Energy use for heating and cooling, respectively, for the two-layer glazing at the four
locations. Dynamic energy calculations.
FIG 3. Primary energy use for heating and cooling, respectively, for the three-layer glazing at the
four locations. Dynamic energy calculations.
3.2 Monthly mean quasi-steady-state calculations

It is investigated whether the use of monthly mean temperature-weighted U-values in monthly quasi-
steady-state energy calculations is a reasonable approach to represent the dynamic effect of the U-
value on energy need as documented in section 3.1. The monthly mean temperature-weighted U-value,
Umonth, is calculated as:
n
U T
is
i i
U month  n
(6)
T
is
ref  Tout,i
Where s the annual hour number in the start of a certain month (-)
n the annual hour number in the end of a certain month (-)
ref a reference indoor temperature (°C), e.g. 20 °C
Results are depicted in FIG 4 and FIG 5. The heating demand for both glazing solutions shows the
same tendency in deviations as the dynamic calculation in FIG 1 but not the same relative or absolute
magnitude. The tendency for the cooling demand in the dynamic calculation is not repeated in the
quasi-steady-state calculation.

FIG 4. Primary energy use for heating and cooling, respectively, for the two-layer glazing at the four
locations. Monthly mean quasi-steady-state calculations.
FIG 5. Primary energy use for heating and cooling, respectively, for the three-layer glazing at the
four locations. Monthly mean quasi-steady-state calculations.
4. Conclusions
The aim of this paper is to minimise the performance gap between calculated and actual energy use by
improving the precision of the calculated energy use. Focus is on the gap caused by the use of the
declared U-value of a glazing as a static all-year input in energy calculations instead of the U-value as
a function of hourly changes in weather conditions.
Data from dynamic hourly building simulations suggest that the dynamic changes in U-value due to
weather conditions should be taken into account to obtain a more accurate calculation of the expected
energy use of a building in the Nordic countries. The use of declared U-value in performance
simulation of two-layer glazing solutions leads to an underestimation of the heating demand for all
investigated locations and an overestimation of the cooling demand. For a three-layer glazing solution
the use of dynamic U-value results in less heating demand whereas the cooling demand is practically
the same. The use of declared U-values instead of dynamic U-values in energy performance
calculations may therefore be a reason for performance gaps between calculated and actual energy
use. Based on this data, the paper investigates whether the use of monthly mean temperature-weighted
U-values is a reasonable approach to represent this effect when calculating the expected energy use
according to the monthly quasi-steady-state method in ISO 13790. The conclusion is, assuming that
the dynamic hourly calculation represents actual conditions, that the suggested approach is not
sufficient to represent the identified effect even though the approach does compensate somewhat for
the performance gap.
Three alternative approaches to minimise the performance gap is hereby suggested: 1) abandon the
monthly quasi-steady-state method for certification in favour of a dynamic method, 2) use the
approach suggested in this paper to reduce the performance gap and then visualise the remaining
performance gap as an uncertainty of the overall calculation result, or 3) visualise the full uncertainty
of the overall calculation result due to the performance gap.

5. Acknowledgements
The author gratefully acknowledges the support of this work through the project “Energy sinners in
low energy buildings” financed by the Danish energy research and development program ELFORSK.
The author would also like to thank Michael Dahl Knudsen at Aarhus University for discussions and
generation of weather data files.
References
Cooper P.I. 1981. The effect of inclination on the heat loss from flat-plate solar collectors. Solar
Energy 27 (5) 413-420.
EN 673. 1997. Glass in building – Determination of thermal transmittance (U value) – Calculation
method. European Committee for Standarization. Brussels Belgium
EPBD. 2010. Directive 2010/31/EU of the European Parliament and of the Council of 19 December
2010 on the energy consumption of buildings (recast).
ISO 13790. 2008. Energy performance of buildings – calculation of energy use for space heating and
cooling. International Organization for Standardization. Geneva Switzerland
ISO 15099. 2003. Thermal performance of windows, doors and shading devices – detailed
calculations. International Organization for Standardization. Geneva Switzerland
Mirsadeghi M., Cóstola D., Blocken B. & Hensen J.L.M. 2013. Review of external convective heat
transfer coefficient models in building energy simulation programs: Implementation and
uncertainty. Applied Thermal Engineering 56 (1–2) 134-151.
Ito N., Kimura K., & Oka J. 1972. A field experimental study on the convective heat transfer
coefficient on exterior surface of a building. ASHRAE Transactions 78 (1) 184-191
Lausten J., Brophy V. & Ryan L. 2010. Energy performance certification of buildings. IEA Energy
Conservation in Buildings and Community Systems Annex 53 – Total energy use in buildings:
analysis and evaluation methods. Paris France
Palyvos J.A. 2008. A survey of wind convection coefficient correlations for building envelope energy
systems' modelling. Applied Thermal Engineering 28 (8-9) 801-808.
Petersen S. & Svendsen S. 2010. Method and simulation program informed decisions in the early
stages of building design. Energy and Buildings 42 (7), 1113–1119.
Rowley F.B., Algren A.B. & Blackshaw J.L. 1930. Surface conductance as affected by air velocity,
temperature and character of surface, ASHRAE Transactions 36 p. 429.
SBi 213. 2011. Bygningers energibehov. 5th ed. Copenhagen, Statens Byggeforskningsinstitut. 120 p.
U.S. Department of Energy. 2013. EnergyPlus Energy Simulations Software website,
www.eere.energy.gov/buildings/energyplus/

Full paper no: 31
Hygro-thermal and Energy Related Performance of Vertical

Greening on Exterior Walls – A Field Measurement Study
Johnny Kronvall, Professor 1
Hans Rosenlund, Ph.D. 2
1
Malmö University and Green Building Science, Sweden
2
Malmö University and CEC Design AB, Sweden
KEYWORDS: vertical greening, green walls, living walls, moisture conditions, evaporative cooling,
energy balance, mould growth.
SUMMARY:
Vertical greening on external walls has lately been the subject of increasing interest, where much
hope is spent upon the benefits of enhanced energy performance of buildings. However, there is a lack
of knowledge and experiences of the technical performance of such walls, especially their hygro-
thermal and energy related behaviour under different climatic conditions.
This paper presents a long-term, full-scale field experiment study on a masonry wall in the southern
part of Sweden. Different kinds of wall greening have been included in the experiment, mainly
solutions based on commercially available cassettes, simple home-built pocket-like plant pots and
different climbers.
Monitored parameters in the technical part of the program include outdoor and indoor climate, air
gap and surface temperature and relative humidity, driving rain and thermal flux.
The wall greening systems studied quite strongly influence the hygro-thermal behaviour of the original
walls; greening may even contribute to dryer conditions by protecting from driving rain. Caution must
be paid to protect from water leakage, which may cause mould growth. The energy balance of poorly
insulated walls is slightly improved in winter by raised air gap temperatures, and in summer by
shading. No evaporative cooling effect has been found in this study.
1. Introduction
Vertical greening on external walls has during the last years been the subject of increased interest from
city planners, architects and real estate developers (Köhler 2008). Much hope is spent upon the
benefits of green walls, not only from an esthetical point of view, but also technically: e.g. enhanced
energy performance with vertical greening, cooler street canyon climate during hot days and traffic
noise reduction. Vertical greening concepts have been developed in a number of countries during the
last decades, especially in Germany, Austria, the Netherlands and the US (Perini et al. 2013, Enzi
2011, Mir 2011, Tilley et al. 2012). In the northern parts of Europe and North America there exists
however much less practice and thus experience of exterior wall greening and so little is known today
about different greening systems and adequacy of different plants as well as the technical behaviour of
wall greening in relation to the building under influence of e.g. more severe winter conditions.
In a collaborative research project with the Swedish University of Agricultural Sciences (SLU),
Malmö University and the building developer and contractor PEAB as partners, research on the hygro-
thermal and energy related performance forms one part which is performed by Malmö University and
CEC Design AB. The results from this part of the project form the basis for this paper. The most
extensive part of the project, however, is related to horticultural issues, such as planting systems, plant
choice and adequacy, irrigation solutions etc. This research is mainly performed by SLU.

2. Hypotheses and methods
The following hypotheses were formulated for the behaviour and hygro-thermal and energy related
performance of vertical greening on exterior walls:
• The vertical greening modifies the microclimate for the exterior wall behind the greening in
the following respects:
o Cooler summer conditions
o Warmer winter conditions
o Relative humidity is affected by the air temperature in the gap between the wall
greening and the wall behind
o Vapour content of air nearly equals that of ambient outdoor air
o The wall behind the wall greening is effectively protected against driving rain
• The risk for mould growth behind the wall greening must be paid attention to when designing
wall greening systems
• Vertical greening on exterior walls influence the energy balance of poorly insulated buildings:
o in wintertime by creating an increased thermal insulation
o in summertime by decreasing the cooling load of the building, while
o for modern well-insulated buildings these effects are negligible
The hygro-thermal and energetic behaviour of outer walls with and without vertical greening was
studied by means of field measurements described below.
3. Experimental setup
The study is using two experimental sites, both located in an industrial area in the city of Malmö in
southern Sweden (N 55.6108, E 12.9896), see FIG 1.
FIG 1. Experimental sites overview. View from South (Source: Google).

3.1 Site 1 – Driving Rain
Site 1 is a south-west-facing corrugated steel façade of an industrial warehouse building. The wall was
prepared for the propagation of climbers by mounting a standard reinforcement mesh #100 mm at a
distance of 150 mm outside the wall surface, thus creating a trellis. A number of different climbers
were planted and established in the ground and consequently climbing up in the trellis. The objective
of the study at this site was to monitor the ability of wall vegetation to protect the wall behind against
driving rain. In order to quantify this protection capacity, driving rain was monitored behind the
vegetation layer and compared to the amount hitting the free, unprotected wall. A special meter
device, the type of which has been used earlier and validated in a number of extensive studies of
driving rain on buildings (e.g. Sandin 1984) was used, see FIG 2. The driving rain collected by the
meter is further led to a standard tipping bucket rain gauge producing electrical pulses to the
monitoring system depending on the intensity of the driving rain hitting the façade.
FIG 2. Site 1 with driving rain meters. FIG 3. The wall-greening systems (Site 2).
Left: Humulus lupulus (Hops) Left: Felt pocket module
Right: Fallopia baldschuanica (Bukhara fleeceflower) Right: Modular panel
3.2 Site 2 – Temperature, Moisture and Thermal Flow

Site 2 is an old unoccupied, yet heated, office/workshop building with a south-facing ≈ 40 cm thick
massive yellow masonry wall, rendered on the inside. Two different wall greening systems were
mounted on the outside of the wall: 1) Modular panels (Vertigreen™ ) from Zinco Gmbh: a) placed
with an air gap of 100 mm between the back of the panel and the wall and b) with mineral wool
insulation bat (70 mm thick) close to the back of the panel leaving an air gap of 30 mm between the
back of the bat and the wall and 2) On-site-made felt pocket modules placed with an air gap of 30 mm
between the back of the panel and the wall, FIG 3. Also the microclimate close to the wall created by
an already present and well-established plant of Virginia Creeper (Parthenocissus quinquefolia)
growing on the wall was the subject of field measurements.
On locations on the inside of the exterior wall, where there was a greening system on the outside or
where there was an undisturbed reference area without any greening, heat flow meters were glued to
the wall. Further, indoor air temperature was monitored in the seasonally heated office room.
3.3 Monitoring System

Ambient weather conditions were monitored by means of a weather station built up with commercially
available sensors for air temperature (radiation protected), relative humidity, total solar radiation
hitting a horizontal plan, wind velocity and direction, precipitation and driving rain (site 1).
Weather data, as well as data regarding temperatures, relative humidity levels and heat fluxes were
continuously collected for further analyses by means of two data loggers, see FIG 4.

FIG 4. Monitoring setup. Left: Weather station and Site 1 with driving rain tests. Right: Site 2 with
hygro-thermal and energy performance measurements.
4. Results and Discussion

This section presents measured data, either as monthly averages during a year or as details from two
winter and two summer days. The days chosen; 17–18 January are cold with easterly wind, and 17–18
July are averagely warm and sunny with winds around west.
4.1 Thermal Performance

The thermal performance is evaluated through the temperature difference between the outdoor air and
the air in the gaps or close to the wall surface (FIG 5).
The winter case shows considerable temperature rise, especially in the thinner gaps behind the
insulated modules and pockets. The latter also has a steel sheet cover on top, decreasing the ventilation
rate.
In the summer, the air space temperatures are closer to outdoor, except for the climber and the
reference area, where the solar radiation makes temperatures increase during daytime. This effect is
also seen in the January case, where the first day is sunny, but the second is not.
Also the wind affects the temperatures; higher wind speeds increase the ventilation rate and make the
temperatures approach the outdoor ones.

K Climber Pockets RefArea Module (NoIns) Module (Insul) WindSpeed m/s
16 5
14 4
12 3
10 2
8 1
6 0
4 -‐1
2 -‐2
0 -‐3
-‐2 -‐4
00:00 06:00 12:00 18:00 00:00 06:00 12:00 18:00
K Climber Pockets RefArea Module (NoIns) Module (Insul) WindSpeed m/s
16 5
14 4
12 3
10 2
8 1
6 0
4 -‐1
2 -‐2
0 -‐3
-‐2 -‐4
00:00 06:00 12:00 18:00 00:00 06:00 12:00 18:00
FIG 5. Air temperature differences in the air spaces or close to the wall compared to outdoor air
during; 17–18 January (top); and 17–18 July (bottom).
It is often argued that vegetated walls and roofs generate a considerable evaporative cooling effect that
reduces the need for active cooling of the building. The backside temperatures of the systems were not
measured, but in FIG 6 we see the differences between wall surface and air gap temperatures of the
test areas. A negative balance could thus indicate a radiative cooling effect from the green systems.
However, this effect seems to be very marginal. Probably most of the evaporative heat exchange will
take place between the plants and the surrounding air outside the plants, thus cooling the air in e.g. a
street canyon. Later in the project this will be studied by means of simulations.
K Climber Pocket RefArea Module Module-‐Insul
6
4
Difference: Tsurf — Tair
3
0 Missing
data
-‐1
okt-‐12 nov-‐12 dec-‐12 jan-‐13 feb-‐13 mar-‐13 apr-‐13 maj-‐13 jun-‐13 jul-‐13 aug-‐13 sep-‐13 okt-‐13
FIG 6. Monthly average temperature differences between wall surfaces and air gaps.

4.2 Hygric Performance
In both winter and summer, the vapour concentration levels are quite similar in all locations, except
behind the pockets, where they are considerably higher (FIG 7). Winter conditions are stable, while in
the summer moisture levels vary more due to evaporation processes from plants and surroundings.
g/m³ Climber Pockets RefArea Module (NoIns) Module (Insul)
5
4,6 5,3 4,6 4,8 4,8
Above: Monthly average (g/m³)
4
1
00:00 06:00 12:00 18:00 00:00 06:00 12:00 18:00
g/m³ Climber Pockets RefArea Module (NoIns) Module (Insul)
15
12,1 13,4 11,8 12,2 12,0
Above: Monthly average (g/m³)
14
13
12
11
00:00 06:00 12:00 18:00 00:00 06:00 12:00 18:00
FIG 7. Vapour concentrations in the air spaces or close to the wall during;
17–18 January (top); and 17–18 July (bottom). Monthly averages also included.
There are two reasons for the higher moisture level behind the pockets; 1) the above mentioned steel
sheet cover on top of the space reduces ventilation; 2) at the same time water transport through the felt
structure into the space was discovered. Conditions suitable for mould growth occur from time to time
and a preliminary investigation performed with an endoscope seems to confirm the existence of cob
web and signs of mould growth. High moisture levels found in the wall surface could also indicate
water leakage from the irrigation system. These issues will be further investigated.
4.3 Energy Performance

The thermal transmittance of the poorly insulated south-facing heavy brick wall varies considerably
with outdoor weather conditions and seasons, see Fel! Hittar inte referenskälla.Fel! Hittar inte
referenskälla.FIG 8. During the winter, when solar radiation is low, all test areas have similar
transmittance values. However, increasing solar radiation accumulation in the summer counteracts
energy losses from the building, clearly seen at the unprotected reference area. We also notice the
effects of shading; by the modules, but also by the leaves of the climber from May to September.

Thermal Transmittance Solar radiation (kWh/m²)
W/m²K Climber Pocket RefArea Module Module-‐Insul I-‐GL T-‐ave Temperature (°C)
1,5 180
Annual
0,55 0,88 0,45 0,82 0,67 160
average:
140
1,0
120
100
0,5 80
60
40
0,0
20
0
-‐0,5 -‐20
nov-‐12 dec-‐12 jan-‐13 feb-‐13 mar-‐13 apr-‐13 maj-‐13 jun-‐13 jul-‐13 aug-‐13 sep-‐13 okt-‐13
FIG 8. Annual and monthly average variations in thermal transmittance of the different test areas and
the bare brick wall. Monthly global solar radiation and average outdoor temperature included.
The monthly values presented are based on diurnal averages, since such a representation of the
thermal transmittance should fairly well reflect the thermal inertia effects of the approximately 40 cm
massive masonry wall. The variations of the modules’ performance in the warmer period, when the
flux is not unidirectional, could thus be explained by the effects of time lag versus outdoor
temperature stability. For instance, in May the range of average diurnal outdoor temperatures was
13 K, while in June only 6 K. Average monthly wind speeds during the summer were 2,5–2,9 m/s,
which range should not affect the transmittance to a noticeable degree.
The average annual energy loss from this poorly insulated wall is lowest without any vegetation
systems. However, the important winter transmissions are very similar, except for the insulated
modules, which is slightly lower. Further, it should be kept in mind that in summer the heat gains
through windows are usually dominating those through walls.
4.4 Driving rain

Even if the resolution of the driving rain gauges are rather coarse for detailed analyses, when
summarized over the year we get remarkable differences between the reference area and the collectors
behind the two climbers. The vegetation reduces the driving rain penetration of the wall by more than
half, see FIG 9. Also during the winter, when only branches and some dead leafs remain, there is a
clear protection effect of the vegetation.
l/m² Fallopia (left) RefArea (mid) Hops (right)
10
Total: 13,6 30,9 12,7
8
44% 100% 41%
6
Corrupted data
0
okt-‐12 nov-‐12 dec-‐12 jan-‐13 feb-‐13 mar-‐13 apr-‐13 maj-‐13 jun-‐13 jul-‐13 aug-‐13 sep-‐13 okt-‐13
FIG 9. Driving rain. Two climbers compared to bare reference area on a wall facing prevailing wind.

5. Conclusions
The hygro-thermal and energy related performance of a number of vertical vegetation systems on
outer walls in Malmö, southern Sweden, has been investigated in this year-round field measurement
study. In summary, the result of the study can be concluded as follows:
The thermal conditions in air gaps behind the greening systems are slightly warmer in both winter and
summer. However, solar radiation raises temperatures close to the bare wall and also in climbers.
Vapour concentrations are marginally elevated behind the greening, but since temperatures vary
diurnally, relative humidity does not reach high and stable levels that indicate risk for mould growth.
Nevertheless, mould growth was found in one case, where water leaked into the air gap, probably
through the felt material and/or from the irrigation system. These risks require close attention at design
and construction stages.
Wall greening effectively protects against driving rain. Climbers reduced the annual amount of
driving rain on the facade by more than half.
The energy balance of poorly insulated buildings benefits from the slightly raised air gap
temperatures. In summer the shading of the greening reduces solar heat load and thus risk for
overheating. Climbers, dropping their leaves in the winter, combine this beneficial shading with
allowing for solar access in the winter, thus reducing heat loss. These effects decrease with increasing
thermal insulation, and is not estimated to play any significant role for modern standard buildings.
This study has not been able to identify any evaporative cooling effect of the green module systems –
an effect that is often argued for in the debate.
6. Acknowledgements
The Swedish Research Council FORMAS and the building developer and contractor PEAB have
financially supported the research study. The support is gratefully acknowledged.
References
Enzi V. et al. 2011. Leitfaden Fassadenbegrünung. (Guide book for vertical greening.) ÖkoKauf Wien,
Arbeitsgruppe 25 Grün- und Freiräume. www.oekokauf.wien.at
Köhler M. 2008. Green facades – a view back and some visions. Urban Ecosystems, 2008, Vol.
11:423-436. Springer.
Mir M.A. 2011. Green facades and building structures. Master thesis CIE5060, TU Delft.
Nevander L.E. & Elmarsson B. 1994. Fukthandbok, Praktik och teori. (Moisture handbook, Practise
and theory) 2nd ed. Stockholm, AB Svensk Byggtjänst and the authors. 538 p.
Perini K. et al. 2012. Vertical greening systems, a process tree for green facades. Urban Ecosystems,
2012, Vol. 16:265-277. Springer.
Sandin K. 1987. Fukttillstånd i autoklaverade lättbetongväggar. (The moisture condition in aerated
lightweight concrete walls.) Report TVBM-3026. Div. of Building Materials. Lund Institute of
Technology, Lund, Sweden
Tilley D. et al. 2012. Vegetated walls: Thermal and growth properties of structured green facades.
Final report to Green Roofs for Healthy Cities – Green Walls Group. UM-09040836. Toronto.
www.greenroofs.org

Full paper no: 32
Characterising the actual thermal performance of buildings:

round robin experiment within IEA EBC Annex 58
Staf Roels, Professor 1
Geert Bauwens, M.Sc. 1
Gilles Flamant, M.Sc. 2
Peder Bacher, PhD 4
Maria José Jiménéz, PhD 3
Henrik Madsen, Professor 4
1
Building Physics Section KU Leuven, Belgium
2
Belgium Building Research Institute, Belgium
3
Energy Efficiency in Buildings R&D Unit, CIEMAT, Spain
4
Technical University of Denmark, Denmark
KEYWORDS: actual energy performance, in situ characterisation, dynamic data analysis, grey box
models
SUMMARY:
Several studies have shown that actual thermal performance of buildings after construction may
deviate significantly from that anticipated at design stage. As a result, there is growing interest in full
scale testing of components and whole buildings. The IEA EBC Annex 58-project ‘Reliable Building
Energy Performance Characterisation Based on Full Scale Dynamic Measurements’ is developing the
necessary knowledge, tools and networks to achieve reliable in-situ dynamic testing and data analysis
methods that can be used to characterise the actual thermal performance and energy efficiency of
building components and whole buildings. The research within this project is driven by case studies.
As a first simple case, an experiment on testing and data analysis is performed on a round robin test
box. This test box can be seen as a scale model of a building, built by one of the participants, with
fabric properties unknown to all other participants. Full scale measurements have been performed on
the test box in different countries under real climatic conditions. The obtained dynamic data are
distributed to all participants who have to try to characterise the thermal performance of the test
box’s fabric based on the provided data.
This paper describes the rationale, aim and objectives of IEA EBC Annex 58 and presents more in
depth the first results obtained on the round robin experiment. It is shown how different techniques
can be used to characterise the thermal performance of the test box, ranging from a simple stationary
analysis to advanced dynamic data analysis methods.
1. Introduction
The rise of living standards, the scarcity of natural resources and the awareness of climate change
resulted in an international pressure to significantly reduce the energy consumption of buildings and
communities. In several countries more stringent requirements are imposed by energy performance
legislation and also an increased awareness for environmental issues in building codes can be noticed.
Mostly, requirements and labelling of the energy performances of buildings is done in the design
phase by calculating the theoretical energy use. Several studies showed however that the actual
performance after realisation of the building may deviate significantly from this theoretically designed
performance. Part of the deviations can be explained by user behaviour, but the other part has to be
attributed to the physical features of the building and its systems. For the latter, building performance
characterisation based on full scale testing – testing of building components or whole buildings under

realistic dynamic conditions – could help to bridge the gap between theoretically predicted and real
life performance of buildings. Full scale dynamic measurements are e.g. helpful to investigate the
performances of building components and whole buildings as built in reality, including the influence
of workmanship. This is illustrated in Figure 1. The left image shows the impact of air looping due to
poor workmanship on the U-value of a cavity wall (Hens et al. 2007). While the designed U-value
corresponds to a high insulation level (U=0.2 W/m²K), the value based on full scale testing measures
0.8 W/m²K, which corresponds to an increase of more than 300%. The graph at the right in Figure 1
compares the designed and realised overall heat loss (W/K) of 18 dwellings in the UK. The overall
heat losses are obtained with in situ co-heating tests (Wingfield et al. 2011). As can be seen none of
the houses realises the intended performance and the measured heat losses of the houses may be up to
200% of the designed value.
U=0.86 W/m²K U=0.21 W/m²K
FIG 1. Left: infrared pictures of the outer leaf of two full scale highly insulated cavity walls in the
VLIET-test building at KU Leuven. The impact of workmanship on the thermal performances of the
walls is clearly visible. Right: Measured versus predicted whole house heat losses (W/K) for 18 new
build dwellings in the UK. None of the houses is able to reach the designed values.
Examples as those mentioned above, explain why at present several in situ testing activities are going
on. A recent international workshop showed the interest for full scale testing from all over the world
(Janssens et al. 2011). A growing activity is observed in both full scale testing on building components
(as e.g. in Paslink-cells or in situ on components of real buildings) and on whole buildings (to
characterise thermal performance and energy efficiency of either test buildings or real buildings). So it
is clear that, contrary to what was expected, the numerical building component and building energy
simulation models did not make full scale testing of building (components) redundant. On the
contrary, together with an increased application of numerical simulations, a renewed interest in full
scale testing can be observed. This is not so strange, because dynamic full scale testing showed not
only to be of interest to study building (component) performances under different real conditions – and
as illustrated, quite often a huge difference is observed between predicted and realised performances –,
it is also a valuable and necessary tool to integrate advanced components and systems into simulation
models. Jiménez et al. (2008a) among others used full scale tests on BIPV-façade (BIPV: Building
Integrated Photovoltaic Cells) to develop a so-called grey box-model. A grey box model is based on a
combination of prior physical knowledge and statistics by identifying the unknown parameters of the
system with dynamic data analysis. Once identified, the grey box model is able to predict the thermal
dynamic response of ventilated photovoltaic double skin facades under different climatic conditions.
This way it can be ensured that the behaviour of new advanced building components is integrated in a
correct way in building energy simulation (BES) models. A similar approach of parameter
identification based on dynamic measurements can be used to identify suitable models to describe the
thermal dynamics of whole buildings including building systems (Bacher and Madsen 2011).
Characterising the dynamic behaviour of buildings is an essential and very valuable input e.g. when
optimising energy grids for building communities.

Despite the renewed interest in full scale testing, practice shows that the outcome of many on site
activities can be questioned in terms of accuracy and reliability. The focus of nearly all full scale
testing activities is on the assessment of the components and buildings, often neglecting the necessity
of reliable assessment methods and quality assurance issues. Full scale testing however requires
quality on all topics of the process chain (Strachan and Baker 2008), starting with a good test
environment: test cells or real buildings, accuracy of sensors and correct installation, data acquisition
software,… Only when this is present a good experimental set-up (test lay-out, imposed boundary
conditions,…) can be designed producing reliable data that can be used for dynamic data analysis
based on advanced statistical methods. This in order to come to a characterisation with reliable
accuracy and use of the results. As soon as the required quality fails on one of the topics, the results
become inconclusive or might even be wrong. To this extent an international collaboration in the
framework of the ‘Energy in Buildings and Communities’-programme (former ECBCS) of the
International Energy Agency has been set up. Launched in September 2011, the IEA EBC Annex 58-
project will work four years with international experts from all over the world on the topic of ‘Reliable
building energy performance characterisation based on full scale dynamic measurements’. The present
paper shortly describes the Annex 58-project, but mainly focusses on a round robin experiment
performed within the project. In this experiment a test box – a scale model of a building – has been
built by one of the participants, with unknown fabric properties for all other participants. The test box
is shipped to different institutes (different climatic conditions) with the aim to perform a full scale
measurement of the box under real climatic conditions. The obtained dynamic data are distributed
amongst all partners who have to try to characterise the thermal performance of the test box based on
the provided dynamic data. A description of the round robin experiment is given in section 3, followed
by a presentation and discussion of the first results of the experiment.
2. The IEA EBC Annex 58-project

The global objective of Annex 58 is to develop the necessary knowledge, tools and networks to
achieve reliable in situ dynamic testing and data analysis methods that can be used to characterise the
actual energy performance of building components and whole buildings. As the focus of the project is
on the development of the testing and analysis methodology, no limitations are set to the type of
components, nor to the type of buildings. The research project is organised around the full scale
process chain as illustrated in Figure 2 and five subtasks are defined.
Collection and evaluation of in situ activities
Subtask 1
Experimental set-‐up
Test infrastructure
ST1. State of the art on full scale testing and dynamic data
Use of results
Data analysis
ST 1. State of the art on full scale testing and dynamic data analysis
analysis
ST 2. Optimising full scale dynamic testing
ST2. Optimising full scale dynamic testing

ST 3. Dynamic data analysis and performance characterisation
ST 4. Application of the developed framework

ST3. Dynamic data analysis and performance characterisation
ST 5. Setting up a network of excellence on in situ testing
Subtask 2 Subtask 3
ST4. Application of the developed framework
Network of Excellence ST5. Setting up a network of excellence
Subtask 5
Application of developed concepts

Subtask 4
FIG 2. Schematic overview of Annex 58 and organisation of different subtasks
Subtask 1 is a short introductory subtask. Based on a literature review an overview and evaluation is
made of previous and ongoing in situ test activities. An inventory is made both on full scale test
facilities available all over the world as well as on common methods to analyse dynamic data.
Subtask 2 establishes the procedure how to realise a good test environment and test set-up. A decision

tree will be developed to help research institutes and industry in setting up full scale dynamic testing,
with emphasis on quality of test environment, measuring sensors and monitoring systems,.. in the light
of the envisaged accuracy. Subtask 3 focuses on the dynamic data analysis. The accuracy and
reliability of different dynamic analysis methods will be compared by testing them on different data
sets. This way quality procedures and guidelines for dynamic data analysis can be developed. Subtask
4 will apply the developed concepts and show the applicability and importance of full scale dynamic
testing for different issues with respect to energy conservation in buildings and communities. Finally
in Subtask 5 a network of excellence on ‘in situ testing and dynamic data analysis’ will be installed.
To guarantee continuity when the Annex-project is finalised, a collaboration with DYNASTEE
(http://www.just-pm.eu/dynastee) is established.
3. Round robin experiment

To determine the state of the art on full scale measurements and dynamic data analysis a round robin
experiment has been set up in the framework of Subtask 3 of Annex 58. The global objective of the
round robin experiment is to perform a well-controlled comparative experiment on testing and data
analysis. To this extent, a test box (a scale model of a simplified building) has been built by KU
Leuven. KU Leuven is the only partner within the Annex 58-project aware of the exact composition of
the test box. After construction the box has been shipped to different partners (different climatic
conditions and different acquisition equipment) with the aim to perform a full scale measurement of
the test box under real climatic conditions. The obtained dynamic data is distributed to different
institutes who have to try to characterize the test box based on the provided experimental data.
3.1 Description of the experiment

The investigated test box has a cubic form, with exterior dimensions of 120x120x120 cm³. Figure 3
gives an overall schematic view of the round robin test box. The floor, roof and wall components of
the box are all identical and have a thickness of 12cm, resulting in an inner volume of 96x96x96cm³.
One wall contains an operable wooden window with overall dimensions of 71x71 cm² and a glazed
part of 52x52 cm². A structure is provided around the box, so that the box remains free from the
thermal influence of the ground. Hence, the box can be considered as floating in free air.
FIG 3. Overall schematic view of the round robin test box.
Winter 2012-2013 the test box has been tested at the premises of the Belgian Building Research
Institute in Limelette, Belgium (50°41’ N, 4°31’ E). Afterwards the box has been shipped to Spain,
where it was measured under summer conditions in Almeria (37.1° N, 2.4° W). In general, the weather
conditions in Belgium are temperate, with a mild, but rainy, humid and cloudy winter. The weather at
Almeria on the other hand is dry and extremely hot in summer, with large temperature amplitudes

between day and night. During day time, solar radiation is very high on horizontal surfaces and the sky
is usually very clear. Figure 4 shows the test box at both sites.
FIG 4.Test box during winter at the measuring site at BBRI. Belgium (left) and during summer at the
Plataforma Solar de Almeria, Spain (right).
At both sites, different experiments have been performed, ranging from co-heating tests with constant
indoor temperature, over free floating temperature runs, to imposed dynamic heating sequences
(ROLBS-signals). During the experiments, heat fluxes on all internal surfaces, together with internal
and external surface temperatures, indoor temperature and delivered heating energy within the box
have been measured. In addition, both test sites are equipped with an outdoor weather station,
measuring all relevant boundary conditions (temperature, relative humidity, wind direction and speed,
diffuse and direct solar radiation, long wave radiation,…). The measured data has been provided to all
participants in the Annex 58-project. They are requested to characterise the thermal performance of
the round robin test box as good as possible based on the provided dynamic data. Both stationary
properties, e.g. the overall heat loss coefficient, and dynamic properties of the test box are aimed for.
3.2 Data analysis methods

Based on the provided dynamic data, different analysis methods have been used by the participants of
Annex 58 to characterise the thermal performance of the test box. The techniques vary from simple
stationary methods to advanced dynamic data analysis methods. In the next paragraphs a short
description of the most important characterisation methods is given together with their main
possibilities and limitations.
3.2.1 Averaging method

Averaging methods are typically used in winter conditions to estimate the thermal resistance of
building elements from in situ surface temperature and heat flux measurements (ISO 9869, 1994). The
method assumes that the (average) heat flow rate and temperatures over a sufficient long period of
time give a good estimate of the values in stationary conditions. By averaging the (dynamic) measured
data the steady state values are calculated. This way, making use of the measured heat input and
indoor/outdoor temperature difference, the overall (stationary) heat loss coefficient of the box can be
determined. The method is only valid if the thermal properties and heat transfer coefficients can be
treated constant over the test period and if the effect of heat storage is negligible. As a result, it is clear
that the method can be of use for the parts of the data measured during winter conditions in Belgium
(when also the indoor temperature is kept constant and solar gains are negligible), but that the method

loses his applicability for the Almeria data. Furthermore, only the stationary thermal properties of the
box can be determined.
3.2.2 Single and multiple linear regression

Apart from the averaging method, linear regression techniques are typically used to determine the
stationary thermal properties. By fitting the linear correlation between the heat input and
indoor/outdoor temperature difference, the overall heat loss coefficient can be determined. But where
in the averaging method detailed (short interval data) can be used and the stationary values follow
from the averaging technique, the linear regression typically makes use of daily averaged values, to
cancel out short-term effects of thermal mass (Bauwens et al., 2012). Applying multiple linear
regression, allows to determine not only the overall heat loss coefficient, but to gain also some
information on the solar transmittance. Major drawback is again that only the stationary properties can
be determined and no characterisation of the dynamic thermal behaviour of the box can be made.
3.2.3 ARX-models and ARMAX-models

Compared to the previous methods, ARX and ARMAX -models allow to include the dynamics of the
system. In the abbreviation AR stands for AutoRegressive: the current output is related to the
previous values of the output; MA (Moving Average) refers to the noise model used and X for the fact
that eXternal inputs are used: the system relies not only on the current input value, but also on the
history of the input. For identifying generic systems AR(MA)X-models are the standard methodology.
The most used ARX model structure is the simple linear difference equation which relates the current
output at time t to a finite number of past outputs and inputs. ARX and ARMAX models have among
others been applied by Jimenez and Heras (2005) and Jimenez et al. (2008b) for modelling the heat
dynamics of buildings and building components. Main problem when applying AR(MA)X-models on
the data of the round robin box is first of all the selection and validation of the model, but then also
how to interpret the model to get information on the thermal characteristics of the test box. Bacher
and Delff (2013) show that by stepwise increasing the model order until most significant
autocorrelation and crosscorrelation is removed, a reliable modelling of both stationary and dynamic
properties of the box is feasible.
3.2.4 State space or so-called grey box models

A final methodology to characterise the round robin box is making use of state space models. State
space models make use of simple resistance/capacitance schemes to simulate the dynamic behaviour
of the box. Mostly a forward selection approach is used. In this approach the analysis starts with fitting
a very simple model, which is then stepwise extended until the loglikelihood no longer increases
significantly compared to the previous model and the model validation shows that the residuals (the
difference between the measured and predicted output) correspond to white noise. As both the initial
model as well as all possible extensions are expected to represent a simplified version of the round
robin test box, this requires – in contrast to the ARMAX-model – some prior physical knowledge.
That is why state space models are often also referred to as grey box models. Figure 5 shows as an
example a two-state grey box model for the round robin test box, taking into account heat input by
heater and solar radiation, capacity of the interior and walls of the box and (conductive) heat flow
through the walls of the box. To identify all relevant dynamic characteristics of the box, preferably a
predetermined heating power signal (e.g. ROLBS- or PRBS-signal) is imposed to excite the box
around its expected time constants, whilst remaining uncorrelated with outdoor weather conditions.

FIG 5. Example of a two-state grey box model applied by one of the participants (Bacher and Delff,
2013)
3.3 Characterisation of the test box – discussion of the results

As some of the methods are only able to determine the stationary properties of the box, Table 1
compares the obtained overall heat loss coefficient as determined by different participants. Comparing
the results, it can be seen that most methods result in an overall heat loss coefficient around 4 W/K.
Only when using the linear regression technique on short term data a significant lower value is found.
At the moment a round robin task force has been initiated within Annex 58 to go in more depth on the
pro’s and cons of the different methodologies.
TABLE 1. Determined overall heat loss coefficient (W/K) of the round robin test box by different
modelling teams and making use of different data analysis methods
Team Winter data Belgium Summer data Spain
1 Averaging method 3.77-3.92
State space model (RC using LORD) 3.07-3.42
Linear regression (5’-data) 2.84-4.11
Linear regression (daily averaged data) 3.68-4.12 4.32-4.48
AR(MA)X-models 4.06 4.07-4.32
State space models (RC using LORD) 3.98-4.04 4.23
3 Multiple linear regression (hourly data) 4.77-5.24
Multiple linear regression (daily data) 3.73-4.39
4 State space models 4.27-4.56
5 Linear regression (daily averaged data) 3.99-4.08
State space models (RC using CTSM-R) 3.99
6 State space models (RC using Matlab) 3.97
7 ARX-models 3.95 4.05-4.10
State space models (RC using CTSM-R) 3.84 3.96
Linear regression (5’-data) 2.98-3.94
AR(MA)X-models 4.01-4.08
State space models (RC using CTSM-R) 4.48
4. Conclusions
The Annex 58-project of the IEA EBC-programme shows that there is currently a larger international
interest in full scale testing and dynamic data analysis. This can be explained by the fact that full scale
testing remains necessary for several reasons. It is for instance the only way to verify our numerical
BES-models. Furthermore, full scale testing allows evaluation and characterisation of the thermal

performance of building components and whole buildings in reality. To illustrate this, as a first step a
round robin test box experiment has been performed within the framework of Annex 58. The global
objective of the round robin experiment was to perform a well-controlled comparative experiment on
testing and data analysis. It is shown how different techniques can be applied to characterise the
thermal performance of the test box ranging from (quasi)stationary techniques towards dynamic
system identification. Where the first ones are only able to estimate the steady state properties of the
box (e.g. overall heat loss coefficient), the latter can give additional information on the dynamic
behaviour of the box and can be used to simulate the dynamic response of the box in a simplified way.
In a next step the investigated methods will be applied to characterise real buildings.
5. Acknowledgements
The construction of the IEA EBC Annex 58 round robin test box was financially supported by Knauf
Insulation. This as well as the input from the different Annex 58 members who participated in the
common exercises is greatly acknowledged.
References
Bacher P and Madsen H, 2011. Identifying suitable models for the heat dynamics of buildings. Energy
and Buildings, 2011, 43(7): 1511-1522.
Bacher P. and Delff P. 2013. IEA EBC Annex 58 Common Exercise 4: ARX, ARMAX and grey-box
models for thermal performance characterisation of the test box. Internal report.
Bauwens G., Standaert P., Delcuve F. & Roels S. 2012. Reliability of co-heating measurements.
Building Simulation and Optimization Conference 2012, Loughborough, UK. 348-355 p.
Hens H., Janssens A, Depraetere W, Carmeliet J. and Lecompte L. 2007. Brick cavity walls: a
performance analysis based on measurements and simulations. Journal of Building Physics, 31(2):
95-124.
ISO 9869. 1994. Thermal insulation. Building elements. In-situ measurements of thermal resistance
and thermal transmittance (ISO 9869:1994(E)). International Organization for Standardization ISO.
Janssens A., Roels S., Vandaele L. 2011. Full scale test facilities for evaluation of energy and
hygrothermal performances. Proceedings of the International Workshop, Brussels, Belgium, 30-31
March 2011.
M.J. Jimenez and M.R. Heras. 2005. Application of multi-output ARX models for estimation of the u
and g values of building components in outdoor testing. Solar Energy, 79(3):302–310.
M.J. Jiménez, H. Madsen, H. Bloem, B. Dammann. 2008a. Estimation of Non-linear Continuous Time
Models for the Heat Exchange Dynamics of Building Integrated Photovoltaic modules. Energy and
Buildings. 40, pp. 157-167.
M.J. Jimenez, H. Madsen, K. Andersen. 2008b. Identiﬁcation of the main thermal characteristics of
building components using matlab. Building and Environment, 43(2): 170–180.
Strachan P.A., Baker P.H. 2008. Editorial. Outdoor testing, analysis and modelling of building
components. Building and Environment 43: 127-128.
Wingfield J., Miles-Shenton D., Bell M. 2011. Comparison of Measured versus Predicted Heat Loss
for New Build UK Dwellings – Unpublished Data, Leeds Metropolitan University, Leeds, UK

Full paper no: 33
The application of a modified co-heating test and grey-box

modelling for predicting building thermal dynamics
Paul Steskens, Ph.D.1
Guillaume Lethé, M.Sc.1
Gilles Flamant, M.Sc.1
1
Belgian Building Research Institute, Belgium.
KEYWORDS: Grey-box modelling, co-heating, building thermal dynamics
SUMMARY:
The applicability of the co-heating test for the estimation of a building’s thermal dynamic
characteristics, i.e. the heat loss coefficient, time constants, and effective thermal capacity, is
investigated. Different grey-box models are identified based on a simulated co-heating test while the
heating power supplied to the building is controlled using: (1) a constant indoor air temperature set-
point, (2) multi-sinusoidal signals, and (3) a pseudo random binary sequence signals.
The study shows that measurement data obtained from a co-heating test with the indoor air
temperature maintained constant is not suitable for estimation of the building’s thermal dynamics,
while multi-sine and PRBS based data provide a better basis for model identification. Multi-sine or
PRBS signals covering a larger range of the frequency spectrum provided the best data set for
identification heat dynamics, since the influence of the outdoor environmental conditions is reduced.
While the advantage of the PRBS signal compared to a multi-sine signal is that its frequency spectrum
is continuous, a PRBS is more stringent with respect to the selection of the sampling frequency for
data collection. However, a multi-sine signal demands a well-designed frequency spectrum in order to
excite the building properly.
1. Introduction
The consideration of the heat losses through the building envelope is crucial in order to develop
energy efficient buildings and to reduce carbon emissions. During the last decades, the thermal
insulation of building components has increased a lot, resulting in lower U values. However,
experimental investigations and field studies (Hens et al. 2007) (Wingfield et al. 2009) indicated that
the U values, and hence, the heat loss to the environment, in practice are often much higher compared
to the theoretical U values resulting from calculations (EN ISO 6946). Moreover, the U value in
practice depends strongly on the design and the installation of insulation layers.
A valuable test method to measure the heat loss of a building (both the building fabric and background
ventilation) to the environment is a co-heating test. During the test, the building (often a dwelling) is
heated electrically, using electric resistance point heaters, to an elevated mean internal temperature
(typically 25oC) over a specific period of time, typically 3 winter weeks. The electrical energy
consumption which is required for maintaining the building at the elevated temperature is measured,
resulting in the heat loss coefficient (HLC) by plotting the daily heat input against the daily
temperature difference between the inside and outside air. The development of the co-heating method
started in the late 1970s and early 1980s and has been further developed, resulting in the current
experimental guidelines available in the UK (Wingfield et al. 2011). Moreover, improvements to the
co-heating test for characterization of the dynamic thermal properties of a building are currently
studied in the European project IEA ECBCS Annex 58.

Nowadays, the co-heating method is mainly used for the estimation of the heat loss coefficient of a
building assuming steady-state conditions and daily averaged measurement data, in an attempt to
smooth the dynamic effects throughout the course of the day. However, dynamic effects and thermal
inertia, represented by a building’s time constants and effective thermal capacity, are important as
well, for example when considering thermal comfort, overheating, and indoor climate control. In the
standard EN ISO 13790 buildings are classified into inertia classes according to the value of the time
constant, which is a function of the average active thermal capacitance Cm, defined as the heat
accumulated inside the building, when the internal air temperature oscillates according to the sinusoid
with the period t-24h and the amplitude 1K. However, the value of the internal thermal capacitance of
the building depends on the amplitude of temperature oscillation in the premises. And, for indoor
climate control the effective thermal capacity at other frequencies than the frequency corresponding to
periods of 24h are important as well.
In this paper, the applicability of the co-heating test for the analysis of the dynamic thermal response
of a building is investigated. A single zone building is selected for analysis. The building is modelled
using the TRNSYS software for building energy performance simulation. A co-heating test is
simulated while different input signals for the heating power supplied to the building are used: (1) a
standard co-heating test, (2) multi-sinusoidal heating power, and (3) pseudo random binary sequence
heating power. Based on the simulation results, grey-box models of the building’s thermodynamics are
developed. The grey-box models are used to estimate the dynamic thermal characteristics of the
buildings, i.e. the heat loss coefficient, effective thermal capacity, and time constants. The influence of
the different heating power signals on the accuracy of the estimated thermal properties is investigated.
2. Analysis
A case study building is selected and modelled using the TRNSYS software for building energy
performance simulation. The software is used to simulate different co-heating experiments.
2.1 Case study

The geometry of a building which is defined along the lines of Common Exercise 1 of the IEA-ECBS
Annex 41, is selected for analysis. The basic geometry of the test case building is presented in Figure
1. Table 1 shows the properties of the building, including the theoretically determined heat loss
coefficients (UA values) for respectively the walls, floor, roof, and windows. Two building
configurations have been used: a lightweight and a heavyweight building envelope, corresponding to a
theoretical heat loss coefficient of 105.6 W/K. The walls of respectively the lightweight building and
heavyweight building consist (from inside to outside) of a plasterboard (12mm), fibreglass quilt
insulation (66mm), a wood siding (9mm), and of concrete blocks (100mm), foam insulation (62mm),
wood siding (9mm). The floor of the lightweight building is an insulated timber floor, and an insulated
concrete floor for the heavyweight building. The building is naturally ventilated with an air change
rate of 0.5 h-1, corresponding to a heat loss of 21.6 W/K.
TABLE 1: Properties of the building
UA
[W/K]
Exterior walls [ m2] 63.6 32.4
Floor [ m2] 48 1.9
Roof [ m2] 48 15.2
Windows [ m2] 12 34.5
Constant air change rate [h-1] 0.5 21.6
FIG 1: Geometry of the building

2.2 Heating power sequence
The co-heating tests have been simulated during two periods: a first period comprising a winter week
from January 5 to January 19 and a second period consisting of a week in spring between April 1 and
April 15. The building has been heated applying different sequences for the heating power supplied to
the building. External boundary conditions were applied using the typical meteorological year for
Belgian (Brussels) outdoor climatic conditions.
2.2.1 ‘Standard’ co-heating: constant indoor air temperature

Figure 2 presents the outdoor and indoor air temperature, the solar radiation, and the predicted heating
power during a co-heating test with a controlled indoor air temperature and a set-point of 25oC in
January and April (only the results for the heavyweight building are presented in Figure 2).
FIG 2. Indoor and outdoor air temperature (Ti and Te [oC]), solar radiation (Q [kJ/(hr.m2)]), and
heating power (Q [W]) during the co-heating experiment with constant indoor air temperature in the
heavyweight building during two weeks in January and April
2.2.2 Sinusoidal heating power sequence

Alternatively, the building is excited applying a signal consisting of multiple sines:
= ∑ sin + (1)
with a user defined set of excitation frequencies {ωk}k=1,…,r and associated amplitudes {α}k=1,..,r. The
phases {φk} k=1,…,r are chosen in such a way that a minimal amplitude in the time domain, i.e.
Schroeder-phased sines, is obtained. Obviously, when a system is excited with such a multi-sine,
information is only obtained in a limited number of frequencies. The frequencies at which the system
is excited are calculated based on the estimated system’s lower and higher time constants, respectively
τL and τH. The first identification attempts, which seem to be straightforward, are based on excitation
of the system by generated sinusoidal signals with τL and τH corresponding to 60 and 240 minutes
respectively, which may be sufficient in sense of building dynamics.

2.2.3 Pseudo random binary sequence PRBS heating power
As an alternative to a multi-sine signal, which only excites the system at a limited set of predefined
frequencies, a pseudo random binary sequence is used for controlling the heater of the building. The
heating power is a deterministic periodic signal fulfilling the requirement of continuous excitation of
all modes in the system. The signal changes value (0 to 1) at a time point that is a multiple of the pulse
period T, while the smallest and largest time constants are defined as τL = 2·T and τH = N·T/5. The
shortest pulse is T time units and the longest n·T time units.
Table 2 presents the different signals and their properties which have been applied in the study.
TABLE 2. Properties of the different signals

Signal Description τL [h] τH [h]
Const. Controlled indoor air temperature, set point of 25oC - -
S1 Multi-sine function (10 sines) 1 24
P1 PRBS 5 120
P2 PRBS 1 24
3. System identification
The following iterative methodology has been applied for selecting, identification and validation of a
suitable model of the heat dynamics of the building:
1. The first seven days of each data set, comprising respectively data from January 5 to January
12, and data from April 1 to April 8, are selected for model identification.
2. A sampling period containing sufficient information regarding the heat dynamics, is chosen.
Average values of the signals are calculated for the sampling period.
3. The model’s dependent variable(s) are selected, being the outdoor air temperature, total solar
radiation on the horizontal plane and the heating power as input variables, while the indoor air
temperature is selected as the output variable.
4. An ARMAX model is identified and validated with time series plots of the residuals and inputs,
based on the auto-correlation function (ACF) of the residuals and cross-correlation functions
(CCFs) from the residuals to the inputs. The data periods used for validation comprise January
12 to January 20 and April 8 to April 14.
5. If necessary, the mathematical order of the model is increased until the model’s output does not
contain any significant autocorrelation and cross-correlation.
6. The heat loss coefficient, time constants and effective thermal capacity of the building are
determined.
3.1 ARMAX model

The heat dynamics of the building is modelled applying autoregressive integrated moving average
models with exogenous input (ARMAX). While applying these models, it is assumed that the general
properties of the building materials and the heat transfer equations are linear within the relatively
narrow range of temperatures involved. And, causality exists between the different processes in the
sense that the present value of the interior temperature is a unique function of the past values of the
interior temperature, exterior temperature, auxiliary heat input, and solar heat gains.
The ARMAX model is written in the general form represented by:
= + (2)

where A(q), B(q) and C(q) are the p’th, s’th, and q’th order polynomials, respectively, in the backshift
operator q, Y, U and e are respectively the model’s output, input and the process noise. Considering
the analysis of the thermal dynamics of the building, the most straightforward option is to choose the
indoor air temperature Ti as dependent variable.
= + + ! + (3)
o
where Q is the heating power [W], Ti and Te the indoor and outdoor air temperature [ C], and Gv the
total horizontal radiation on the horizontal plane [W/m2].
3.2 Interpretation of physical parameters

The physical parameters describing the heat dynamics of the building can be directly obtained from
the ARMAX model’s coefficients. The heat loss coefficient (HLC [W/K]) of the building is calculated
based on the steady-state relationship:
= "# − − % ! + (5)
o
where Q is the heating power [W], Ti and Te the indoor and outdoor air temperature [ C], gA the solar
aperture [-], and Gv the total horizontal radiation on the horizontal plane [W/m2]. The equation is
required to coincide with the steady-state equation representing the ARMAX model, where q=1, and
the thermal characteristics are directly obtained from the ARMAX model’s coefficients.
The transient behaviour can be characterized by the system’s time constants, defined by the response
to a sudden change of Ti, while all driving forces are held constant, i.e. the step response. As far as the
change is concerned, the driving forces are set equal to zero. The system’s time constants are derived
from the ARMAX model’s roots applying the methodology presented by (Jiminez et al. 2008) In order
to calculate the building’s effective thermal capacity, the methodology presented in (Rabl. 1988) is
followed. Starting from a general differential equation describing the heat dynamics of a building, the
effective thermal capacity is derived based on the system’s time constants:
+
&& = "# '( + ( + ⋯ + (* − +,-./,1 3 (6)
,-./,2
where τ is a time constant and the term αheat,1/ αheat,0 could be interpreted as the sum of time
constants for Q. With respect to the identified ARMAX models and the corresponding coefficients A,
B1, B2, and B3 obtained. The coefficients αi are replaced by the corresponding ARMAX parameters.
It should be noticed that the thermal capacity Ceff is only as exact as the coefficients of the model,
which in practice are determined by means of the data set obtained. The normal diurnal cycling does
not give enough time for a heat pulse to penetrate all parts of a massive building, which means that Ceff
is an effective thermal capacity. It is a measure of the heat storable under normal cycling. For slower
cycling Ceff would increase, and so would the time constants of the data fit. This illustrates to what
extent the parameters of the model are dependent on the frequencies represented in the data.
4. Results
This section presents the simulation results and the physical parameters estimated by the different
ARMAX models. Figure 3 presents a comparison of the measured indoor air temperature and the
predicted indoor air temperature by a second order ARMAX model, while using a multi-sinusoidal
heating power signal (S4). In addition, the figure presents plots of the ACF of the output’s residuals
and the CCFs between the residuals and the inputs, respectively the outdoor air temperature (Te), the
heating power (Q) and the solar radiation (Gv). No significant auto-correlation and cross-correlation is
observed while the residuals lie near the confidence band of 95%. With respect to the indoor air
temperature and the solar radiation, it could be argued that significant cross-correlation is observed.
Additional investigations, which are not presented in this paper, showed that using a separate input for
the direct solar radiation on each vertical facade instead of the total horizontal radiation on the

horizontal plane improves the model significantly. However, this increased the number of inputs with
4, i.e. one input for each facade. In order to reduce the complexity and the size of the models, it has
been decided to use the total solar radiation as an input.
FIG 3. Comparison of the measured indoor air temperature and the indoor air temperature predicted
by a second order ARMAX model using a multi-sinusoidal (S4) heating power signal and the
corresponding autocorrelation and cross-correlation functions for the lightweight building
Figure 4 presents a comparison of the estimated HLC by the models based on the heating power
signals (Table 2) both for the lightweight and the heavyweight building and for January and April. The
ability of the models to predict the steady-state thermal properties of the building, i.e. HLC and the
solar aperture are analysed. The models are compared using a relatively small sample time of 5
minutes (including both fast and slow system dynamics) and larger sample times of 15, 60, and 120
minutes focusing on the slow dynamics and the determination of the steady-state thermal parameters.
First of all, figure 4 shows that the models identified based on data obtained from the standard co-
heating tests showed to be incapable of always giving a reliable prediction of the building’s HLC.
Since the indoor air temperature is constant during the measurement period, a reliable model
describing the building’s thermal dynamics is difficult to identify. Information regarding heat
dynamics is only obtained from over-heating periods. If overheating does not occur or seldom occurs,
which is the case in the heavyweight building, the capturing the building’s thermal dynamics is
impossible.
Second, the figure shows that the predicted HLCs for the lightweight building lie close to the
theoretical value of 105.6 W/K, while the predicted standard deviation is relatively small, both for
January and April based estimates. Moreover, model identification and prediction of the HLC was
possible for every signal except for signal S3 in April. This signal contains too little information to
excite the system sufficiently, comprising only 5 sines in a relatively large range of frequencies.

FIG 4. Estimated heat loss coefficient (HLC [W/K]) and solar aperture (gA [m2]) for the lightweight
building based on the data sets for January and April for sample times of 5, 15, 60 and 120 minutes.
Considering the predicted HLCs for the heavy weight building, a larger difference between the models
is observed. The accuracy of the predicted HLC is largely influenced by the outdoor environmental
conditions. The figure shows that the conditions in spring require a ‘richer’ signal in order to identify a
reliable model. Model identification appeared only to be possible for those signals which cover a
larger bandwidth of the frequency spectrum, i.e. a multi-sine signal consisting of 10 sines (S4) and
PRBS signals. These signals showed to be capable of exciting the building dynamics in a wider range,
clearly enhancing model identification.
Comparing multi-sine signal S4 and PRBS signals P1 and P2, the observations from this study showed
that models based on PRBS heating power sequences are more susceptible to the sampling time of the
data. Compared to the multi-sine signal based model, PRBS based models’ predictions showed to be
more influenced by redundance and reluctance of the data. If the sampling time Ts is too short the data
become highly correlated (singular) and problems due to ill-conditioning arise, and redundance is
observed. On the other hand, if the sampling time Ts is too long adjacent data points tend to become
uncorrelated, i.e. reluctance.
TABLE 3. Estimated time constants for the lightweight building
Lightweight Heavyweight
Const. 13.3 1.18 - 149.8; 0.018
S1 1.9 1.20; 0.010 7.7 10.0; 0.010
S2 2.1 1.15; 0.012 - 7.4; 0.010
S3 1.1 1.10; 0.012 7.3; 0.019 -
S4 1.1 1.20; 0.013 7.5; 0.018 7.3; 0.016
S5 1.2; 0.015 1.10; 0.012 8.1; 0.018 -
P1 1.4; 0.015 1.20; 0.013 11.8; 0.011 9.3; 0.011
P2 0.4; 0.04 1.15; 0.012 10.8; 0.012 12.0; 0.011
Table 3 presents the estimated time constants and effective thermal capacity for the lightweight and
heavyweight building for the models based on different heating power sequences. For all identified

models, a sample time of 5 minutes has been applied. With respect to the lightweight building, two
well-separated time constants of 0.012 and 1.1 hours and a corresponding effective thermal capacity of
0.12 kWh/K are predicted by the models based on multi-sine signals SINE3, SINE4, and SINE5.
Similarly, the estimated time constants and effective thermal capacity of the heavyweight building are
respectively 0.018 and 7.5 hours, and 0.8 kWh/K. With respect to the PRBS signal, slightly larger time
constants are predicted.
In addition, the table shows that the models identified based on data obtained from the standard co-
heating test showed not to be capable of giving a reliable prediction of the building’s time constants.
Since the indoor air temperature is constant during the measurement period, a reliable model
describing the building’s thermal dynamics and time constants could not be identified, resulting in a
poor prediction of the building’s thermal dynamics.
5. Conclusions and Recommendations

Based on the results presented in this study it is concluded the measurement data obtained from a co-
heating test with the indoor air temperature maintained at a constant temperature is not applicable for
the prediction of building thermal dynamics, i.e. the building’s time constants and effective thermal
capacity. Heating the building using a multi-sine or PRBS signal based heating power showed to
provide a better data set for identification of the heat dynamics.
When applying multi-sine or PRBS signals covering a larger bandwidth of the frequency spectrum, the
influence of the outdoor environmental conditions on model identification is reduced, and, hence, the
thermal characteristics of the building can be predicted with a smaller uncertainty. The investigated
PRBS signals showed always to be capable of identifying a reliable model, though these models
appeared to be more influence by redundance and reluctance of the data compared to a multi-sine
signal. The findings also showed that a multi-sine signal requires a well-thought signal design and
selection of suitable frequencies to excite the system.
The observations and research results presented in this paper serve as a basis for additional study and
analysis of the dynamic thermal characteristics of multi-zone buildings. In addition, they are applied
for the design of the co-heating tests carried out at the Belgian Building Research Institute.
References
EN ISO 6946. 2007. Building components and building elements. Thermal resistance and thermal
transmittance. Calculation method. (EN ISO 6946:2007). International Standards Organisation.
EN ISO 13790. 2008. Energy performance of building – Calculation of energy use for space heating
and cooling. (ISO 13790:2008). International Standards Organisation
Hens H, Janssens A, Depraetere W, Carmeliet J, Lecompte J. 2007. Brick Cavity walls: performance
analysis based on measurements and simulations. Journal of building physics 31(2), pp. 95-124.
Jiminez MJ, Madsen H, Andersen KK. 2008. Identification of the main thermal characteristics of
building components using MATLAB. Building and Environment, vol. 43, pp. 170-180.
Madsen H. 2008. Time Series Analysis. Chapman and Hall.
Rabl. 1988. Parameter Estimation in Buildings: Methods for Dynamic Analysis of Measured Energy
Use. Journal of Solar Energy Engineering, vol. 110, pp. 52-66.
Wingfield, J. Bell M, Miles-Shenton D, Seavers J. 2011. Elm Tree Mews Field Trial – Evaluation and
Monitoring of Dwellings Performance Final Technical, Leeds Metropolitan University

Full paper no: 34
Sensitivity analysis of the energy demand of existing buildings

based on the Danish Building and Dwelling Register
Anker Nielsen, professor 1
Kim B. Wittchen, senior researcher 1
Niels H. Bertelsen, senior researcher1
1
Danish Building Research Institute, Aalborg University, Denmark
KEYWORDS: Energy demand; Danish Building Regulations; Calculation method; Uncertainty;

Energy ratings
SUMMARY:
The EU Directive on the Energy Performance of Buildings requires that energy certification of
buildings should be implemented in Denmark so that houses that are sold or let should have an energy
performance certificate. The result is that only a small part of existing houses has an energy
performance certificate. The Danish Building Research Institute has described a method that can be
applied for estimating the energy demand of dwellings. This is based on the information in the Danish
Building and Dwelling Register and requirements in the Danish Building Regulations from the year of
construction of the house. The result is an estimate of the energy demand of each building with a
variation. This makes it possible to make an automatic classification of all buildings. The paper
discusses the uncertainties and makes a sensitivity analysis to find the important parameters. The
variations are compared with measured energy demand. The method can be applied in other countries
with modifications for local building requirements, climate and building registers.
1. Introduction
In Denmark, energy performance certification has been in use since 1997 with energy performance
rating and for new houses (since 2006), including sold or let buildings. The energy rating classifies the
building on an efficiency scale ranging from A (high energy efficiency) to G (poor efficiency) (EPBD
2011). The rating is important when buying and selling houses. The problem is that many houses still
do not have an energy rating, and that it will take many years to cover the entire building stock. Until
ultimo 2013, only 22% of single-family houses had an energy rating. Energy ratings are given in
connection with a visit by an energy expert and the result is a rating and a description of possible
energy-saving measures. For houses that have not been energy rated, we do not know the energy
demand and potential energy savings. An automatic calculation of the energy rating would thus be
very useful. It is possible that some owners would make energy saving measures, if they were
informed about the energy rating of their property. It would also make it possible for authorities to
give an economic subsidy or tax deduction for the buildings with the poorest energy efficiency. The
result should be more energy saved for the same money.
It is possible to apply a normal certification method based on monthly values and details on heating
system and constructions to make energy demand calculations of existing houses. The problem is that
we have to visit the houses and we will not always be able to obtain information about the
constructions at the site or in drawings. Another important point is that this would be a very time-
consuming and expensive method. The Danish Energy Performance Certification Scheme for existing
buildings includes energy demand calculations based on a monthly calculation method (Aggerholm
and Grau 2005). This can be used at the visit as you can find some information and have additional
information from a guideline with typical constructions. The cost of this is paid by the owner before
selling the building. As it will take many years before all buildings have been officially rated, the
Danish Building Research Institute (SBi) has described a method (Bertelsen et al. 2011) that can be
used without physically visiting the house but only based on available information.

2. BBR data
The Danish Building and Dwelling Register (BBR) has information that can form the basis for the
calculation. This has to be supplemented with information from the Danish Building Regulations from
the year of construction and some estimates from previous research and information collection
(Wittchen 2009 and 2012) by SBi. The following data from BBR are used: building (building type),
year of construction, useful floor area in m2, number of storeys, building footprint in m2,useful attic
area in m2, basement area in m2 and external wall material.
The described method can be used for dwellings like single-family houses and blocks of flats. In the
following, we only describe the use for single-family houses. The construction year is used to find the
actual energy requirements in the Danish Building Regulations. In Denmark, we can expect that these
requirements for thermal insulation are respected. We know that many buildings have better insulation
or improved windows and we take that into account in the method.
The useful floor area is used for calculating the amount of heat from persons, light, equipment and
solar radiation. The number of storeys, the building footprint, useful attic area, and basement area are
used in the calculation of the geometry of the building. The external wall material defines the U-value
level in the Danish Buildings Regulations.
3. Geometric model
BBR contains no information on building height, length or width. So we have to make an estimate of
the geometry. The simplest form is a box-shaped building. We know the roof and floor area as it
equals the building footprint. For buildings without attic and basement, we use a box-shaped building.
We need the building width. Based on typical Danish buildings, we estimate the building width to be 9
m for single-family houses. The building length is the building footprint divided by the width.
Calculations with other length/width proportions show that it does not influence the surface area very
much. The storey height is assumed to be 2.8 m as a typical value. The volume and the external wall
area can then be calculated.
If the building has a useful attic area, we get an extra wall area based on the attic which is also a box.
The box is placed on top of the building box and does not change the floor or roof area. The length of
the attic is estimated to be the length of the building. The width and volume of the attic and wall area
can now be calculated.
For houses with a basement, this is calculated as a box placed under the building. As it is placed
underneath the building, there is no change in floor or roof area. The walls of the box give the
basement wall area. Note: We calculate as if the whole basement area is heated, as most people use it
like that. The length of the basement is estimated to be equal to the length of the building. The width
and volume of the basement and the basement wall area can now be calculated.
3.1 Areas used in the energy calculations

The roof area and the floor area are both equal to the building footprint. The external wall area is the
sum of the wall areas of the building box and the attic box. The volume of the building is the sum of
the volume of the three boxes. This is a simplification as some part of the volume might be unheated,
e.g. staircases.
Another important information is the area of the windows of the building, as windows have a greater
heat loss than the rest of the wall area. There is no information on window area or number of glass
panes in BBR, so we need to estimate the values. The window area also includes door areas. It is
estimated that before 1999 most buildings had a window area covering 15% of the façade area. After
1999, the windows covered 22% based on Wittchen 2009 and 2012. The basement wall has no
windows. If we have an attic, then we estimate a window area covering 10% of the attic wall area.

Now we can calculate the window area for the building based on the construction year and possible
presence of an attic. The wall area excluding windows is then easy to find.
4. U-value model
U-values for the different building parts are taken from the Danish Building Regulations in force when
the building was constructed. Energy requirements were changed in 1961, 1967, 1972, 1977, 1982,
1985, 1995, 1998 and 2008. The five U-values used in the calculation are outer wall, basement wall,
floor, roof and window. We do not take into account that we can have different outer wall
constructions in the same house.
TABLE 1. Typical U-values and requirements in different periods
Heavy outer wall Roofs Windows
(> 100 kg/m2)
Before 1930 1.2 W/m2K Before 1950 0.97 W/m2K Before 1930 3.7 W/m2K
1931-1950 1.5 W/m2K 1951-1960 0.6 W/m2K 1931-1994 2.9 W/m2K
1951-1960 1.5 W/m2K 1961-1976 0.4 W/m2K 1995-2007 1.8 W/m2K
1961-1971 1.1 W/m2K 1977-1994 0.2 W/m2K From 2008 1.5 W/m2K
1972-1976 1.0 W/m2K From 1995 0.15 W/m2K Note: We assume that most old
1977-1994 0.4 W/m2K windows are upgraded.
1995-2007 0.3 W/m2K
From 2008 0.2 W/m2K
BBR information on the exterior wall material is used to decide whether it is a light or heavy wall as
the Danish Building Regulations define different U-values for these wall types. Data about older
houses are supplemented by the experience of SBi. Table 1 is an example of the U-value variation
over the years for heavy outer walls, roofs and windows. Data for other construction types and periods
are given in a Danish report (Bertelsen et al. 2011).
5. Model for energy balance

The total transmission heat loss is found by multiplying the areas with the U-values from the
construction year and making a summation of all constructions in the house. For floor and basement
walls, we reduced the heat loss by 30% because of the ground.
The volume of the building is needed for calculation of the ventilation loss. As mentioned, the volume
also includes internal construction and unheated areas such as staircases and basement. The ventilation
is estimated based on SBi’s experience gained from other projects. Here it was found that before
1961the air change rate (1/h) was 0.45, from 1961 to 1978 it was 0.4 and after 1978 it was 0.35.
Calculation of heat from sun, persons, and equipment can be complicated. It is theoretically possible
to calculate the solar radiation through windows, but there are uncertainties in both the areas and
orientation of the windows. The heat from persons, light, equipment etc. is also very uncertain.
Another problem is that only part of this heat will reduce the heating energy demand of the building.
The solution is to use a value for the useful part of the internal loads from SBi. The useful internal heat
gain is estimated to be 55 kWh/m2year, where the sun accounts for half the amount.
To calculate the energy loss, it is necessary to know the indoor and outdoor temperature. The indoor
temperature is fixed at 20 °C. We use degree days. In Denmark the degree-day number for the energy
calculation is 3600. The useful internal heat gain is calculated for the living area (given in BBR), and
does not include for instance basement areas not accepted for living. The energy demand for heating
the building in kWh/year (EX) can then be calculated as the sum of the heating and ventilation losses
minus the useful internal heat gain.

6. Uncertainty of the parameters
The calculated energy demand is the best estimate based on the information in BBR and the estimates
made by SBi based on the knowledge of the Danish building stock. Table 2 lists all the uncertainty
parameters.
TABLE 2. List of uncertain parameters

Geometry U-values Energy calculation
Width Walls Ventilation rate
Length Roof Airtightness
Storey height Floors Useful heat (solar, persons,
Volume Window type, number of glass light and equipment)
Attic area and width panes translucent area, glass Degree-hour number
Basement area and width coating, orientation Renovation
Window area
The method ignores complex geometric buildings like U or L shaped ones. The uncertainty of the U-
values is large for buildings constructed before the first nationwide Danish Building Regulation in
1961 – after that time the uncertainty is lower as buildings are normally built in compliance with the
Danish Building Regulations. There is no information in BBR as to whether extra insulation has been
added after the construction year. In practice, we can expect that energy-saving measures have been
implemented in many older buildings.
7. Variations of the result

Measurements of the energy demand of buildings show large variations for individual buildings in
spite of their being uniform in construction (Petersen 1997). This is caused by the influence from the
users and variations in workmanship etc. This calculation method is based on the variations in the
building and only partly takes into account the variation caused by the users and their use of the
building. If that variation was fully included, the variation would increase.
An estimation of the uncertainty of the calculations depends on the construction year. Old houses have
a large variation and new houses a lower variation. The method provides an upper limit and a lower
limit, as given in Table 3. The variation is not symmetric. These limits are in reality upper and lower
quartiles as we estimate that 25% of houses have an energy demand higher than the upper limit and
25% of houses have an energy demand lower than the lower limit
TABLE 3. Variations in upper and lower limits during different periods
Upper limit Lower limit
Before 1961: EX * 1.25 Before 1931: EX * 0.60
(it is 25% above the average) (it is 40% below the average)
1961-1971: EX * 1.15 1931-1961: EX * 0.70
From 1972: EX * 1.10 1962-1977: EX * 0.80
1978-1995: EX * 0.85
After 1995: EX * 0.90
The lower limit has more variations than the upper limit which is explained by the fact that many old
houses have been retrofitted with extra insulation and new windows. The result is an energy demand
that is well below the average calculated value (EX). The values in kWh/year are calculated as energy
demand per m2 as this is the basis for the Danish Energy Performance Certification Scheme.

8. Automatic energy rating
The described method to perform automatic energy demand calculations can be used for automatic
energy certification and rating of houses that do not have an official energy performance certificate.
The auto-rating method can be described as follows:
Using the average energy demand per m2 (EX) as a basis for automatic rating results in a number of
houses would get a better rating than if they had been visited by an energy expert. This problem is
solved by calculating the energy demand from the upper-limit energy demand (worst case). Use of this
value makes nearly all houses obtain a better rating when an official certification is made later.
The method has been used in a case study of a street with 56 single-family houses. Figure 1 shows
variations in energy demand, based on the construction year. It is seen that most houses are built from
1940 to 1960. Most houses with an automatic rating are rated either F or G, the two lowest ratings.
FIG 1. Calculated energy demand of 56 single-family houses versus construction year.
If we compare the official energy rating for the five houses that have an official energy certificate with
the result of the automatic rating method based on BBR it is seen that the officially calculated energy
demand by the energy expert lies between the upper and lower limits of the automatic method. As
expected, the value is closer to the lower limit, as some energy-saving measures were made. These are
typically additional insulation, tightening and retrofitting of windows and new heating system.
9. Sensitivity analysis for automatic calculation

A sensitivity analysis is done on a house from 1964 with a Monte Carlo simulation using random
values to recreate the variation of the parameters. The calculation data is described in Table 4.

TABLE 4. Case of a single-family house with average values and variations
Geometry
• Building footprint 100 m2 (variations from 95 to 105 m2)
• Storey height 2.8 m (variations from 2.4 to 3.0 m)
• Length of house (area fixed – variation in length/width)
• Areas of windows 15% (variation from 10 to 25%)
• Insulations and constructions in W/m2K. It is taken into account that there is a possible later
insulation of the constructions.
• Walls 1964 1.1 W/m2K ( variation from 0.6 to 1.3 W/m2K)
• Windows 1964 2.9 W/m2K ( variation from 1.6 to 3.0 W/m2K)
• Roof 1964 0.4 W/m2K ( variation from 0.25 to 0.5 W/m2K)
• Floor 1964 0.4 W/m2K ( variation from 0.3 to 0.5 W/m2K)
Other parameters
• Air change rate 0.4 1/h (variation from 0.2 to 0.6 1/h)
• Indoor climate – constant standard value
• Outdoor climate – constant standard value
• Heat from persons, equipment and solar radiation – constant standard value
The calculation of 10000 cases gives an average value and a variation for the building:
• Average energy demand 192 kWh/m2
• In 10% of the cases, the demand is less than 157 kWh/m2
• In 10% of the cases, the demand is higher than 226 kWh/m2
• So 80% of all values ranged between 157 and 226 kWh/m2
This variation is similar to the variation found in other research on buildings’ from around 1965. The
calculation makes it possible to evaluate the sensitivity of the parameters in the calculation. Table 5
show which parameters have the most influence on the final result.
TABLE 5. The calculated variance for the 9 parameters in the sensitivity analysis
Parameter Variance Sum of Variance
Wall U-value 36% 36%
Storey height 19% 55%
Air change rate 15% 70%
Window area 10% 80%
Roof U-value 7% 87%
Windows U-value 6% 93%
Building footprint 4% 97%
Floor U-value 2% 100%
Length/width proportion 0% 100%
The most important parameter is the U-value of the wall, and after that the storey height, air change
rate and window area and hereafter the rest of the U-values and the built area. If we do a more correct
calculation, we need more information. From this analysis, we see that if we can get better information
on the U-values of the wall, the storey height and window area, we reduce the variance in the energy
demand calculation as they account for 65% of the variation. The air change rate is very difficult to
know without making measurement and it will change during the year – so it is not realistic to get
better information on the air exchange rate. For the other three parameters, it is possible to ask the
owner about the values and whether extra insulation has been added. It is interesting to note that an
improvement of information for a number of the parameters does not have much effect on reducing
the variance.

If we use the method to calculate the total energy demand for a group of houses, the estimate will be
much better, if we assume that the houses have random variations. For instance we can estimate a
single house with a variation of 56%, but for a group of 10 houses is it 18% and for more than 50
houses is it 8%.
10. Measured energy use

Is the assumption realistic if we compare with measured energy consumption in buildings? Figure 2
shows measured energy consumption for houses in relation to construction year (Jensen 2004). The
red line is the level of the rules in the Danish Building Regulations from 1961 and onwards. It is seen
that for most houses from 1960 to 1996, the energy consumption is below the requirement level. So
our assumption that houses follow the rules is realistic and also that many has been improved. After
1996, more houses have a higher energy use than the Danish Building Regulations, probably because a
slightly higher indoor temperature will only give a small increase in the energy bill. For older houses,
this effect of the indoor temperature variation is blurred.
FIG 2. Measured energy consumption (dots) in Danish houses for different construction year. The red
line is the requirement in the Danish Building Regulations.
Figure 3 shows that old houses have a high energy demand and large variation and new houses has
lower energy demand and less absolute variation (Jensen 2004). The arrows indicate the average
values. This shows that our assumptions for variations are realistic.
11.Conclusion and use in other countries

The comparison between the official Danish energy performance certification scheme and the
automatic system shows that the automatic system can be used to identify buildings with potentially
high energy consumption. This can be done based on information in BBR and the knowledge of
typical Danish buildings. This provides a possibility of targeting incentives for better energy
performance in the buildings, where you obtain the highest energy savings for the investment. The
comparison with measurements shows that our assumptions are realistic.

This method can also be used in other countries using similar basic national building registers and
building regulations over time, climate and information of the typical building and their use.
FIG 3. Variation in energy demand of houses with different construction year
References
Aggerholm, S. and Grau, K. (2005). Bygningers energibehov - PC-program og beregningsvejledning.
(Building energy demand – PC program and user guide) SBi-Anvisning 213. Statens
Byggeforskningsinstitut (SBi), Hørsholm, Denmark
Bertelsen, N. H., Nielsen, A., Sørensen, N.L. and Wittchen, K.B. (2011) Automatisk
energirammeberegning for den eksisterende bygningsmasse (Automatic energy performace
calculation for the existing buildings), SBi-rapport 2011-20, Statens Byggeforskningsinstitut (SBi),
Hørsholm, Denmark
Danish Building Regulations (in Danish: BygningsReglementet. Erhvervs- og Byggestyrelsen,
København. Use www.ebst.dk – to find older building regulations
EPBD (2011). Implementation of the Directive on the Energy Performance of Buildings - Country
reports 2010. EPBD Buildings Platform, Brussels, European Commission ISBN: 978-92-9202-090-
3.
Jensen, O. M. (2004). Barrierer for realisering af energibesparelser i bygninger. (Barriers for
realisation of energy saving in buildings), Statens Byggeforskningsinstitut (SBi), Hørsholm.
Petersen, T.D. (1997). Uncertainty analysis of energy consumption in dwellings, NTNU, Dr.avh.
1997-122, Trondheim, Norway
Wittchen, K. B. (2009). Potentielle energibesparelser i det eksisterende byggeri. (Potential energy
savings in existing buildings) Forskningsprojekt SBi 2009:05. Statens Byggeforskningsinstitut,
Hørsholm, Danmark
Wittchen, K.B, and Kragh, J. (2012) Danish building typologies, Danish Building Research Institute,
SBi 2012:01, Hørsholm, Denmark

Full paper no: 35
Thermal performance of a window with in-between solar shading

- Hot box measurements and simulated performance comparison
Steinar Grynning M.Sc, PhD-candidate 1,2*
Berit Time, Senior Principle Research Scientist, PhD2
Egil Rognvik2, Engineer
Arild Gustavsen, Professor, PhD 1,2
1
Department of Architectural Design, History and Technology, Norwegian University of Science and
Technology (NTNU), NO-7491 Trondheim, Norway,
2
SINTEF Building and Infrastructure, Building Materials and Constructions, Høgskoleringen 7b, Trondheim,
Norway
* Corresponding author: [email protected], tel. +47 97 56 61 03, fax +47 73 593380.
Keywords: Windows, glazing, solar shading, thermal transmittance, U-value, Hot box, laboratory,
simulation
SUMMARY:
Shading systems are widely used in conjunction with glazed facades, especially in office buildings.
Solar gains in combination with high internal gains will often lead to a cooling need during
operational hours. Thus, shading devices are traditionally used in order to reduce cooling demands
during the time of day when the sun is up.
The possible effects of a solar shading device on the thermal transmittance value (U-value) of the
glazing unit has, however, not been given much attention.
In order to assess energy saving potentials by applying shading devices, detailed information about the
component level characteristics is necessary. Thermal and optical properties of these components are
important input data for simulation tools and the calculation of cooling and heating demands as well as
daylight accessibility in buildings.
Shading systems can be categorized into three main types; external, internal and in-between panes
shading devices. When an in-between pane shading device is deployed (i.e. closed shut), it can act as
an additional layer in the insulating glazing unit (IGU) and thus reduce the U-value of the IGU. In a
cold climate, such a system could be utilized in order to reduce heat losses during periods of time
when a room is not occupied and when no view to the outdoors is necessary.
In this article, the thermal properties of such a system have been explored. Hot-box measurements
have been carried out for a window with 4-panes. The glazing unit is built up as a 3-pane IGU with an
external coupled fourth pane. A venetian blind shading device is placed between the outermost,
coupled pane and the IGU. U-values have been measured for various slat-angles of the blinds. These
results have been compared with calculated values using the THERM and WINDOW simulation
software. This is software developed for calculations of thermal and optical properties of entire
windows.
1 Introduction
Some of the previous work pertaining to louvered blinds has been done for a horizontal blind sealed
between the panes of an Insulated Glazing Unit (IGU). Garnet et al. (1995) carried out an

experimental investigation using a guarded hot plate apparatus to determine the thermal performance
of an IGU with a between-the-panes venetian blind. They found that the effect of the venetian blinds
on the U-value of the glazing unit ranged from a 10 % increase for blinds in horizontal position to a
20 % decrease with the venetian blinds in a closed position. It is also concluded that, from a thermal
performance perspective, it is better to close the blinds hot side up, i.e. the slats should point upwards
on the warm side of the window. Similar blind configurations, within a sealed IGU, have also been
studied using analytical and numerical models (Rheault and Bilgen 1989).
In spite of a large amount of theoretical work carried out in order to assess thermal and optical
performance of solar shading systems, few measurement campaigns have been carried out. In 2001,
Breitenbach et al. (2001) presented thermal resistance measurements for a two-pane IGU with
integrated venetian blinds. The authors found that the thermal transmittance value (U-value) varied
almost linearly with the slat angle from 2.92 ± 0.14 W/(m²K) with blinds retracted to 1.44 ±
0.21 W/(m²K) with the blinds shut and slats in vertical, closed position. This contradicts the values
found by Tzempelikos (2005) during a measurement campaign performed on a double-pane window
with in-between venetian blinds. Here it was found that a blind-tilt angle of 120° gave the lowest
thermal resistance and that blinds closed with the slats in vertical (i.e. 180°) position gave the highest
thermal resistance.
None of the two aforementioned publications present values for a situation with the blinds in a
retracted position and only the centre-of-glass U-values are presented. Presenting only centre-of-
glazing U-values does not take into account any effects caused by the mounting system and any motor
of the venetians on the total U-value of the glazing unit. Such effects could be substantial as mounting
systems and motors may contain large amounts of continuous metal spanning from the warm to the
cold side.
In 2006 Huang et.al (2006) presented measurements performed using a Guarded Heater Plate
apparatus (GHP). Here, centre-of-glass U-values (Ucog) for double-glazing units with and without
shades, were measured using Heat Flow Meters (HFM). This was carried out by controlling the
temperature on each side of the sample using the GHP. 12 sample configurations were measured. The
measured values were used for benchmarking/validation of CFD simulations. Values are extrapolated
to also include solar heat gain coefficients (SHGC) and visual transmittance (Tvis) values.
A study performed by Shahid H. and Naylor D. (2005) presents a numerical method for a window
with internal shading. The method is validated by use of previous measurement results for tall vertical
cavities. No validation using measurements on real windows was performed.
In 2009, Laouadi (2009) presented an overview of existing studies related to modelling of glazing
units coupled with solar shading systems. A model for calculating properties of the centre-of-glazing
area were proposed and found to correspond to measurements performed by Huang within a 7 %
margin of error. The model is only verified for double pane glazing units with a limited (lower than
25 mm) cavity thickness. The U-value of IGU's with low-e coated glazings and higher cavity
thicknesses (40 mm was modelled) were underestimated with the model. Furthermore, the authors
point out that at that time, the models regarding the thermal performance of shading systems are based
on simple algorithms and that further validation work needs to be carried out.
2 Methodology and experimental design

2.1 Window sample and test series overview
U-values are measured for a 4-pane window with integrated in-between-pane venetian blinds. The
window consists of a 3-pane IGU with argon fillings. An external 4th coupled glass pane is mounted

on the exterior side. A manually operated (i.e.no motor) venetian type blind is mounted in the exterior
cavity as shown in Figure 1.
Figure 1. Graphical illustration of the 4-pane window. Shading device, in dark blue, shown in
retracted position. The shading device is integrated in the cavity behind an exterior coupled glass
pane. The frame is insulated with polyurethane foam (coloured yellow).
Table 1 shows an overview and description of the window configurations that have been measured.
Generic data as described in ISO 15099 (ISO 2003) has been used as material data for the shading
slats.
Table 1: Window sample description.

Measurement
Window and material description
description
• Window size: 1200 x 1200 mm (w x h) Blind position
• Operable window 1. Retracted blind
• 3-pane IGU + 1 exterior coupled pane 2. deployed blind
• Glazing construct: 4-24-4E-12Ar-4-12Ar-E4 horizontal slats
• Venetian blind in external (coupled) cavity 3. deployed blind
• Wood frame insulated with 17 mm polyurethane foam 45° slats
o Sill width 101 mm 4. deployed blinds
o Jamb width 94 mm vertical slats
o Head width 105 mm
o Frame depth (all parts) 105 mm
• Materials used in simulations
o Wood thermal conductivity = 0.12 W/(mK)
o Polyurethane thermal conductivity = 0.023 W/(mK)
o Window spacer equivalent conductivity = 0.158 W/(mK)
• Slat material A from ISO 15099
o Conductivity = 160 W/(mK)
o Front- and back-side emissivity = 0.9
o Front- and back-side reflectance = 0.7
Figure 2 shows the four blind positions as described in Table 1.

Figure 2. Blinds positions 1-4 as described in Table 1. Position 1 on the left and 4 on the right.
2.2 The hot box test facility – test procedure and instrumentation
Measurements have been carried out in a guarded hot box apparatus, as pictured in Figure 2, according
to procedures described in ISO 8990:1994 Determination of steady-state thermal transmission
properties Calibrated and guarded hot box (ISO 1994) and ISO 12567-1:2010 Thermal performance
of windows and doors - Determination of thermal transmittance by the hot-box method - Part 1:
Complete windows and doors (ISO 2010).
Figure 2. The guarded hot box test facility used for the measurements
The tests were performed at steady state conditions of +20 °C interior and 0 °C exterior temperatures.
Window U-values are calculated based on the measured heat flows, surrounding temperatures and
window area.
The window was mounted in a template constructed as a sandwich element consisting of a 100 mm
thick Extruded Polystyrene (XPS) core, clad with 12 mm plywood on the faces exposed to the hot and
cold sides of the hot box. The joints between the window and the surround panel are sealed with tape
on both sides to ensure an air tight seal. Minor gaps between window frame and template were filled
with Expanded polystyrene (EPS).
The metering area of the hot box is 2.45 m x 2.45 m. The window is placed in the middle of the
metering area template wall at a distance of 1.0 m from the floor to the lower edge of the frame.

24 hourly, measured averaged values are used as basis for establishing the U-values. Each of the
hourly measurement periods are again averages of values measured every 6-8 seconds.
Surface thermal resistances were adjusted close to the standardised ones prior to the tests by adjusting
air flow velocities adjacent to the template surface on the cold side. Natural convection driven air-flow
were maintained on the warm side. However, during the measurements the surface thermal resistances
differ slightly from the standardized values. Corrections have thus been made for these deviations, so
that all U-values are stated with normalised surface thermal resistance values as specified in ISO
8994:1994 (ISO 1994). The standardized conditions are; interior surface thermal resistance,
Rsi = 0.13 W/m²K and external surface thermal resistance Rse = 0.04 W/m²K.
Centre-of-glazing U-values (Ucog) were measured using two PU 43 T heat flow meters, from
Hukseflux (Hukseflux), with a declared accuracy of ± 5 % at 20°C. These were mounted mid-height
approximately 10 cm from the vertical symmetry axis of the IGUs on the warm side of the samples.
2.3 Uncertainty assessments

The uncertainties associated with the hot box measurements have been assessed in accordance with
the procedure presented in ISO 12567-1:2010 (ISO 2010) The total uncertainty propagation of the
measured U-value, ΔPUw/Uw has been derived using the root-mean-square method (RMS) as shown in
Eq.1.
2 2 2
Δ PU w ⎡ Δ P Φw ⎤ ⎡ Δ P Α w ⎤ ⎡ Δ P δθie ⎤
= ⎢ ⎥ + ⎢ ⎥ + ⎢ ⎥ (1)
Uw ⎣ Φw ⎦ ⎣ Α w ⎦ ⎣ θie ⎦
Where ΔPΦw/Φw = Uncertainty in sample heat flow (W)

ΔPAw/Aw = Uncertainty of projected area of sample (m²)
ΔPδθie/δθie = Uncertainty in temperature difference between warm and cold side (K)
The uncertainty in the sample heat flow is based on the heat balance equation for the metering
chamber. The uncertainties in test sample specimen heat flow, ΔPΦw/Φw, is expressed using Eq. 2.

2 2 2 2
Δ P Φw ⎛ Δ P ΦIN ⎞ ⎛ Δ P Φsur ⎞ ⎛ Δ P ΦEXTR ⎞ ⎛ Δ P ΦFL , w ⎞
= ⎜ ⎟ + ⎜ ⎟ + ⎜ ⎟ + ⎜⎜ ⎟⎟ (2)
Φw ⎝ Φw ⎠ ⎝ Φw ⎠ ⎝ Φw ⎠ ⎝ Φw ⎠
Where ΔPΦIN = Uncertainty in power input to metering chamber (W)

ΔPΦsur = Uncertainty in surrounding template heat flow (W)
ΔPΦEXTR = Uncertainty in metering chamber walls heat flows (W)
ΔPΦFL,w = Uncertainty in test sample flanking heat loss (W)
A calibration experiment for the thermocouples was carried out prior to the hot box measurements
using a reference temperature bath. The relative scattering in measured temperatures between the
thermocouples were found to be lower than 0.02 °C. Thus it can be concluded that the influence from
the factor ΔPδθie as described in Eq.1 is negligible.
The uncertainties stated in this work, are given with a coverage factor of two standard deviations and
the corresponding 95 % confidence interval.

2.4 Numerical simulations
Simulations of the thermal properties of the window with the integrated shading device have been
carried out using the THERM 7.1 and WINDOW 7.1 software developed at Lawrence Berkeley
National Laboratories in Berkeley (LBNL 2013; LBNL 2013). THERM is a finite element method
program suitable for calculating U-values of window frames, whereas the WINDOW software
calculates thermal and optical properties of glazing units. Thus, THERM has been used for calculation
of frame U-values and linear edge transmission losses caused by the spacer used. Cavities in the
frames have been modelled according to the NFRC 100-2001 definitions in THERM. WINDOW has
been used for calculating the centre-of-glazing U-values including the shading device. The results are
compared to the values measured in the hot box.
3 Results
3.1 Measurement results
From Figure 3 and Table 2, one can see that the effect of operating the blinds has minor or no effect
on the measured mean U-value. Deploying the blinds with horizontal slats, increases the U-value
slightly, whereas closing the slats shut (vertical slats) will give a slight decrease in U-value. The
alternations of the U-value as function of blind positions were found to be negligible in terms of
statistical significance, which means that the individual measured mean U-values lies inside the
measurement error boundaries.
Figure 3. Measured U-values for the window. Upper and lower uncertainty values shown with lines.
The measured U-values for the whole window, Uwindow, and the center-of-glazing U-values, Ucog, are
shown in Table 2.
Table 2. Measured mean U-values, Uwindow, and center of glazing U-values, Ucog, for the window.
retracted deployed blinds, deployed blinds, deployed blinds,
Measured value
blinds horizontal slats 45°deg slats vertical slats
Uwindow (W/m²K) 0.84 ± 0.04 0.85 ± 0.04 0.83 ± 0.04 0.81 ± 0.04
Ucog (W/m²K) 0.61 ± 0.03 0.60 ± 0.03 0.58 ± 0.03 0.57 ± 0.03
3.2 Calculation results
The outermost cavity in the glazing unit is an air filled cavity. Tests carried out using THERM to
model this cavity treating the cavity as both a slightly ventilated and un-ventilated cavity according to
the NFRC definition used in THERM (LBNL 2013), gave U-value-results comparable to the values
calculated assuming an air-filled cavity using WINDOW. Assuming a slightly ventilated cavity gave
center-of glazing U-values approximately 4 % higher than an un-ventilated cavity. However, based on
the minor openings constricting air cross-flow in the cavity, it has been treated as an un-ventilated

cavity using the simulation results from WINDOW. Whole window, Uwindow, and center-of-glazing U-
values, Ucog, calculated using THERM and WINDOW are shown in Table 3.
Table 3. Calculated mean U-values, Uwindow, and center of glazing U-values, Ucog, for the window.
retracted deployed blinds, deployed blinds, deployed blinds,
Calculated value
blinds horizontal slats 45°deg slats vertical slats
Uwindow (W/m²K) 0.65 0.66 0.65 0.64
Ucog (W/m²K) 0.52 0.53 0.51 0.50
4 Discussion and future work

The solar shading device positioned in the outermost cavity of the window does not have a statistically
significant effect on improving the U-value of the window when in closed position compared to the
open.
If one looks at the measured center-of-glazing U-values (Ucog), the same behavior as for the Uwindow-
values can be seen. Thus the two independent measurments support each other in that the effects on
the window U-value as function of shading device slat angle and position is negligible.
The U-values found from simulations using THERM and WINDOW were consistently lower than the
measured values but also here, the effects of blind posistion and angle ion the U-value were found to
be negligible. The deviation between the measured and calculated values can likely be explained by
differences between declared (from the producer) and actual properties of the low-e coatings as well as
the amount of argon in the gas-filled cavities of the IGU.
As opposed to finding any potential gains in U-value by the introduction of such a shading device, it
was found to have a negative effect on the U-value of the window compared to one without the
shading device. The protruding aluminum components of the shading device and venetian blinds, act
as thermal bridges. Numerical simulations showed that the U-value of the window with the shading
device will increase with approximately 5 % compared to a window without an integrated shading
device. Thus, any beneficial effects expected to be achieved by using the venetians as an additional
layer in the IGU, was found to be more than counteracted by the thermal bridging.
5 Conclusions
Measurements have been carried out in order to investigate the U-value of a window with a 4-pane
glazing unit incorporating an in-between panes venetian blind shading device. The aim of the study
has been to assess the effect of operating the blinds on the window U-value. Various slat angles and
blind positions have been studied. It was found that no statistically significant effect could be found on
the U-value as function of various slat angles.
Numerical simulations showed that the U-values were underestimated compared to the measured
values. This can have several reasons, including deviations in gas-filling levels and quality of the low-
emissivity coatings in the IGU. Identifying the reasons for these deviations is the aim of further studies
that are currently being carried out within the Research Centre on Zero Emission Buildings (ZEB).
6 Future work
There are several possibilities which should be explored in order to achieve more effective shading
devices, and the following should be explored
• Reduce slat thermal conductivity in order to reduce thermal bridging effects

• Improve surface properties of slats (i.e. reduce emissivity)
• Improve airtightness of shading layer by reducing openings between slats when in closed
position
7 Acknowledgements
The authors gratefully acknowledge the support from the Research Council of Norway and several
partners through the Research Centre on Zero Emission Buildings (ZEB).
8 References
Breitenbach, J., S. Lart, et al. (2001). "Optical and thermal performance of glazing with integral
venetian blinds." Energy and Buildings 33(5): 433-442.
Garnet, J. M., R. A. Fraser, et al. (1995). Effect of internal Venetian blinds on center-glass U-values.
Window Innovations Toronto, Canada.
Huang, N. Y. T., Wright, J.L., Collins, M.R. (2006). "Thermal resistance of a window with an
enclosed Venetian blind: Guarded Heated Plate measurements." ASHRAE Transactions
112(2): 13-21.
Hukseflux. "PU Series." Retrieved 06.08.2013, from
http://www.hukseflux.com/sites/default/files/product_brochure/PU%20series%20v0717.pdf.
ISO (1994). ISO 8990:1994 - Determination of steady-state thermal transmission properties Calibrated
and guarded hot box.
ISO (2003). NS-ISO 15099 – Thermal performance of windows, doors and shading devices - Detailed
calculations, International Organization for Standardization.
ISO, N.-E. (2010). NS-EN 12567:2010 Thermal performance of windows and doors - Determination
of thermal transmittance by the hot-box method - Part 1: Complete windows and doors.
Laouadi, A. (2009). "Thermal modeling of shading devices of windows NRCC – 51121." NRCC-
51121, ASHRAE Transactions 115(2): 1-20.
LBNL. (2013). "THERM 7.1 beta." Retrieved 06.08, 2013, from
http://windows.lbl.gov/software/therm/7/index.html.
LBNL. (2013). "WINDOW 7.1 beta." Retrieved 06.08, 2013, from
http://windows.lbl.gov/software/window/window.html.
Rheault, S. and E. Bilgen (1989). "Heat transfer analysis in an automated Venetian blind window
system." Solar Energy Engineering 111(1): 89-95.
Shahid, H. and D. Naylor (2005). "Energy performance assessment of a window with a horizontal
Venetian blind." Energy and Buildings 37(8): 836-843.
Tzempelikos, A. (2005). A methodology for integrated daylighting and thermal analysis of buildings.
Faculty of Engineering and Computer Science - Building, Civil and Environmental
Engineering, Concordia University.

Full paper no: 36
Modelling and Implementing efficient Three Dimensional

Anisotropic Heat Air and Moisture Transport
Stefan Vogelsang1
Andreas Nicolai1
1
Institute of Building Climate Control, Department of Architecture, Technische Universität Dresden,
Germany
KEYWORDS: simulation, framework, simulation and modelling techniques, transient transport

simulation
SUMMARY:
The development of complex physical models in building simulation creates a great demand for
powerful and flexible transient simulation programs, and thus numerical solvers. Especially when
turning to three dimensional models computational speed and memory use strongly limits the problem
size of an application. The calculation complexity increases further more when taking anisotropic
transport processes into account. This article describes a solution for a transient anisotropic three
dimensional solver for combined heat, air, and moisture transport in porous materials of building
envelop systems, utilizing the Delphin solver framework. After describing the physical model,
boundary conditions, and certain aspects of internal solver optimisation regarding memory use and
calculation speed, three initial academic verification examples will be presented. The influence of
anisotropic material properties onto three dimensional simulation results is researched, targeting an
evaluation of simulation performance and correctness while comparing to isotropic simulation results
of the same experiments. The paper closes with an outlook onto parallelisation of the introduced
solver solution.
1. Introduction
Prediction of damage in constructions (e.g. timber beam heads, interior insulations, etc.) is recently
done by performing transient transport simulations. It is state of the art to utilise continuum physics
for describing conserved quantities coupled through systems of partial differential equations. Our
Delphin solver framework (Nicolai 2011) implements continuum physics models for porous media
(construction materials) by defining a representative elementary volume (REV) following Whitaker
(1969), Bear and Bachmat (1992) and Whitaker (1986). Grunewald (1997) and Nicolai (2007) defined
essential conservation equations for hygrothermal, pollutant as well as salt transport. Moreover a
lower bound (infimum) as a function of the pore radius spectrum was defined as a constraint for an
accurately selected REV size.
The numerical solution of resulting partial differential equations is done using a spatial discretisation
that introduces location-fixed control volumes and leads to a large set of ordinary differential
equations to be integrated. Numerical integrators developed to solve such problems implement some
form of time discretisation and solve the problem in a step-wise manner. Each integration step (time
step) involved hereby a solution of a non-linear equation system, which in turn requires solving large
sparse linear systems of equations. The actual implementation is done in several cascaded
computational loops (see Listing 1).
Increasing dimension of simulation space (e.g. 1D, 2D to 3D) introduces new degrees of freedom, and
thus complexity to all calculation procedures executed during a simulation. It basically increases

execution counts of the kernel loops. The cost of such a solution process is almost always governed
by the solution of the linear systems especially for large scale problems (≥4.5 million elements).
LISTING 1. A general numerical integrator for systems of coupled ordinary differential equations
while (Time Integration) {
while (Non Linear Iteration) {
while (Linear Systems Solution) {
lots of computing work
}
}
}
Introducing anisotropic material properties into transient transport simulation creates a mapping
problem for material characteristics defined in a materials reference system (anisotropic nature of the
material) to a simulation reference system. Much research has been devoted to anisotropic material
modelling (Zillig 2009; Segerholm 2007; Siau 1984; Krabbenhoft 2003, etc.) and anisotropic material
properties as found in timber or sand stone.
We are interested in modelling three-dimensional transport in combination with anisotropy. Currently
available hygrothermal simulation programs do not readily offer anisotropic and three-dimensional
simulations. They are either limited to single materials, or single transformation orientations. A
combination of wood-based construction elements with different longitudinal/radial orientations, or
different anisotropic materials cannot be considered simultaneously. Hence, we decided to implement
this functionality into the current Delphin 6 code. This paper presents the implemented anisotropic
material model and related data handling.
Also discussed are optimisation approaches that make 3D simulations with anisotropy efficient. The
new implementation presented here is labelled Delphin version 6.1 to distinguish it from the current
Delphin 6 simulation code.
2. Three Dimensional Anisotropic Flux Calculation

Anisotropic transport models differ from standard isotropic hygrothermal transport models
(Grunewald 1997; Nicolai 2008; Janssen 2007; Hagentoft et al. 2002) in the formulation of the flux
quantities. Transport models within porous media define a linear relation between a flux and its
driving potential, see Eq. (1).
k
j k  L X k  RT Lk X k
(1)

L  Lu , Lv , Lw 
Where jk is a flux density of a quantity k
k
X is the driving potential for quantity k
L is the transport coefficient tensor formulated with respect to the MRS
L is the transport coefficient tensor formulated with respect to the SRS
RT is a transformation matrix
To emphasis additional effort of multidimensionality a vector arrow marks all divergence symbols,
usually this vector arrow is implicitly defined in a divergence and therefore omitted. In isotropic
models the transport coefficient is a scalar. In anisotropic models the transport coefficient tensor L is

material dependent and such defined in an right-hand rectangular material reference system (MRS,
with directions: u,v,w). The MRS may not necessarily be identical with the simulation space reference
system (SRS). An anisotropic solver thus requires a transformation process from material to
simulation reference system for each calculated flux in simulation space (see FIG. 1 for an example).
FIG 1. Example for a material and simulation reference system that are not aligned
The relation between material reference system and simulation reference system may change relative
to the spatial location. In FIG. 1 the detail (right) shows the MRS for a specific location timber beam
end. The direction W marks the tangential direction, V marks radial direction with respect to the
material (MRS). Moving to a different location will change the orientation of the MRS within the
global simulation coordinate system.
The mathematical transformation process responsible for mapping these two reference systems is thus
location dependent. For completeness we briefly review the derivation of rotation matrices involved.
Detailed information can be found in standard literature on the matter.
FIG 2. Rotation on Point P

Figure 2 shows an ordinary rotation of a point P by an angle  . For a start respectively an end
position of a rotation it is possible to deduce:
uP  l cos  uP*  l cos(   )

(1) (2)
vP  l sin  v p*  l sin(   )
In this transformation the w-coordinate remains unchanged. A dependency between ( uP , vP ) and

( uP* , vP* ) becomes obvious, as soon as additional theorems of trigonometric functions are applied.
Thus the equations for uP* and vP* can be reformulated as:
uP*  l cos(   ) v p*  l sin(   )

(3) (4)
 l (cos  cos   sin  sin  )  l (sin  cos   cos  sin  )
A starting position ( uP , vP ) may now be substituted, following equations (1 - 2) into the found
equations (3 - 4) , which drops out all unknowns. The rotation equation, written in matrix notation, for
a rotation by an angle  around the w-axis reads:

cos   sin  0  u P*   uP 
Rw   sin  cos  0 with v   R v 
   P*  w  P
 0 0 1  0   0 
The rotation axis u and v can be derived equivalently:
1 0 0  cos  0  sin  

Ru  0 cos   sin   (5) Rv   0 1 0  (6)
   
0 sin  cos    sin  0 cos  
A three dimensional point in a u,v,w coordinate system is transformed by stepwise execution of single
direction rotations until it matches with the target x,y,z coordinate system.
pxyz  RT puvw
RT  Ru Rv Rw
 cos  v cos  w cos  v  sin  w  sin  v 

RT   sin  u sin  v cos  w  cos  u sin  w (  sin  u )sin  v (  sin  w )  cos  u cos  w  sin  u cos  v 
 
 cos  u sin  v cos  w  sin  u sin  w cos  u sin  v  sin  w  sin  u cos  w cos  u cos  v 
In analogy a material parameter/transport coefficient tensor should be used in SRS, this coefficient
tensor is first represented in MRS as Lk , and afterwards multiplied with the transformation matrix
k
RT .  L  RT Lk
With respect to the numerical solution we have to keep in mind that the rotation matrix is
location dependent.
3. Implementation and Numerical Solution

The Delphin 6 model framework is a modularized implementation of the governing equations for
hygrothermal transport. Calculation of state variables from conserved quantities (Balance Equation
modules), internal and boundary fluxes (Flux and BCFlux modules) and sources/sinks
(FieldCondition modules) are customizable. For the purpose of anisotropic extension the modules for
calculation of material transport properties and flux calculation modules were extended.
Since the Delphin 6 framework is built on top of a Finite Volume (FV) spatial discretisation method,
the calculation of transport coefficients, both isotropic and anisotropic is done for each volume
element and only one transformation matrix is computed for each element. Since transformation
matrices may vary with spatial coordinates we work with volume-averaged transformation matrices.
Hence, the size of the discretisation cells must thus be limited for accuracy reasons to avoid excessive
changes in rotation matrices from one element to the next.
For the calculation of anisotropic fluxes the implementation requires local definition of materials and
their respective orientations. The definition of material orientation within the construction definition
(project file) is done in either simple or complex format. In the simple format each material
assignments is a direct mapping of a material reference system to the simulation reference system. For
example, if a timber stud (placed vertically) is modelled its u-direction shall be aligned to the y-
direction. Then the mapping VUW (respective target coordinate system xyz) is specified alongside the

material placement within the calculation domain. The complex format allows arbitrary rotation
matrices stored for each Finite Volume. Those rotation matrices are stored in simplified Euler-Matrix-
1 0 0 
Format. A line [1;0,0.5,-0.5;0,0.5,0.5] corresponds to the matrix: 0 0.5 0.5 .

 
0 0.5 0.5 
To assist in generation of rotation matrices we have developed a tool that generates matrices for
materials with rotation symmetric material reference systems (i.e. wood/timber). Anisotropic
measured material data and material functions including directional properties are stored and
organised utilising the standardised IBK material format (Vogelsang et al. 2013).
For the purpose of integrating the set of ordinary differential equations resulting from the FV
discretisation the Delphin 6 model is embedded into an advanced integration framework (Nicolai
2013). This framework implements several integrators including a BDF-type numerical integrator
(backward-difference-formula) that is strictly mass and energy conserving and error controlled. The
nonlinear equation systems arising from the BDF time discretisation are solved with a modified
Newton-Raphson algorithm. In the standard configuration the integrator uses a direct solver for the
solution of the linear equation systems within each Newton iteration.
The new implementation has been validated with isotropic validation cases from HAMSTAD
benchmarks, DIN EN ISO 15026, and DIN EN ISO 10211 (Sontag et al. 2013) utilising an automated
validation framework by Vogelsang et al. (2014). Since benchmarks or test cases for anisotropic
simulations have not been published/defined yet, we are currently developing validation cases for the
correct application of rotation and transformation matrices.
4. Improving Efficiency of Anisotropic Flux Evaluation

A transformation of material properties into the reference system of simulation space can be done
automatically for each finite volume element when all rotation matrices are calculated in a
preprocessing step of a simulation. Thus cost intensive evaluation of sinus and cosines functions
during simulation time can be omitted. The additional effort of one transformation can such be
minimised to six additions and nine multiplications per super positioned flux calculation (material
parameter assignment). Moreover other de facto necessary projection methods, e.g. application of a
two dimensional material in a three dimensional simulation space, can easily modelled through
1 0 0
convenient R m . For example: Rm  1 0 0 maps a material parameter measured and rotated to
 
0 1 0
simulation dimension X onto x, and y component and the Y dimension onto the z component of
simulation space. When mapping matrix and rotation matrices are applied in a preprocessing step, the
k
whole procedure can be realised by an single rotation matrix R R : L  Rm RT Lk  RR Lk .
For large 2D and generally all 3D simulation setups efficient memory handling alongside a fast flux
calculation is essential.
5. Efficient Solution of Three-Dimensional Problems

Solving sparse linear systems is one of the most challenging cost factors while executing transient
transport simulations (see Figure 3). Two concurrent families of different solution methods exist:

Direct method solvers use a form of Gaussian elimination. Typically, an LU factorization of the linear
equation system matrix is obtained, before one or more right-hand-sides (RHS) are solved (via
backsolving). They are guaranteed to find a solution of a non-singular system in an exact count of
arithmetic operations. They are therefore called robust and general. For instance LU decomposition
with pivoting can be done for any non-singular matrix. In this sense the corresponding linear system is
always solvable, a property required when the method is embedded inside a Newton iteration. In
exceptional cases, the solution process of a linear system via LU is not backward stable, i.e. the
backsolving process fails. However, this occurs rarely in practice. It depends on the so-called growth
factor of the entries in the upper triangular matrix U, which may cause rounding errors to grow until
they reach or exceed the magnitude of the desired solution. With respect to application in a Finite
Volume solver, a pivoting version of a banded LU factorization routine will take into account the
special structure of the Jacobian matrix involved.
An inevitable problem of direct methods is that calculated matrix factors are often denser than an
original matrix. The resulting memory requirements dominate when large scale systems, as created by
discretising three dimensional partial differential equations systems (large constructions), need to be
solved. Also, banded LU factorisation routines have quadratic complexity with respect to bandwidth
and linear complexity with respect to system size. This implies that factorisation time and memory
use may grow beyond exceptional limits whenever system sizes become large. An application of
direct methods is hence restricted by calculation effort and memory use.
Iterative methods approximate the solution rather than compute the solution exactly. The most
relevant group of iterative methods are the Krylov subspace methods, projection methods where a
subspace of search directions is enlarged with each iteration. The different Krylov subspace methods
essentially differ in the choice of this construction of the subspace and the quality of the
approximation to the solution obtained and improved in each step. An iterative solution of an equation
system is finished once the solution is accurate enough, and consequently needs a suitable
convergence criterion. Within a Newton iteration the accuracy of the Krylov subspace approximation
must allow the Newton iteration to progress.
Unfortunately, Krylov subspace methods may converge very slowly or even stall without further
improvement of the approximation. The latter problem may be caused by too large subspace
dimensions and the involved rounding errors when increasing it further. Also breakdowns are
possible. The details of the various methods are beyond the scope of this article and we refer to
standard literature on the matter (Saad 2003).
With respect to hygrothermal transport problems the resulting Jacobian matrices may be very ill
conditioned. This may lead to excessive iteration counts in the linear iterative solvers or bad
approximations causing the Newton-Raphson method to fail. Therefore efficient preconditioning is
mandatory. The preconditioning operation transforms the iteration matrix into another matrix that
resembles the identity matrix more closely. Basically preconditioners (M) improve convergence
properties of a linear system Ax  f by transforming it into a linear system M 1 Ax  M 1 f . The
coefficient matrix M 1 A then posses a (hopefully) better distribution of Eigenvalues (so called
clustering) and approximates the identity matrix more closely than the original matrix A. Selecting M
is done in a way that applying M 1 onto a vector is particular efficiently and preferably optimally
scales on parallel computer hardware architectures. This procedures creates additional calculation
costs at each iteration for creating M or M 1 , but may reduce iteration count substantially. Different
classes of preconditioners exist that improve convergence speed (see Saad 2003 for a review).
One dimensional and small to medium two dimensional problems (less 10000 cells) are best solved by
direct band solvers. For large scale two dimensional hygrothermal problems iterative solvers may be
more efficient.

3D simulation models often have a grid cell count that exceeds practical limits for direct methods.
Therefore, for large 2D grids (Paepcke 2014) and any 3D simulation, we always use Krylov subspace
iterative solvers within our Newton iteration. Within the Delphin 6 framework, the ILU
preconditioner is a standard accelerator for Krylov subspace methods and used in conjunction with
single core GMRES implementation. This allows the computation of up to 4.5 million grid cells in
reasonable time.
FIG 3. Simulation Time Comparison of the Direct (left) and the Iterative GMRES Solver (right) both
utilizing the CVODE integrator.
6. Conclusions
An anisotropic three dimensional simulation model was developed and implemented in C++ within
the Delphin 6 modelling framework. The framework was tested with scientific test cases to illustrate
the functionality of the anisotropic property transformations with different mapping assignments (FIG
4).
FIG 4. Anisotropic Transport for Spruce, Oak and Beech (24h) sorption in a corner experiment
The computational speed for the 3D tests was enhanced by selecting an iterative GMRES solver with
ILU preconditioning. But large hygrothermal 3D cases take still too much time while running as serial
iterative code, future work will be devoted to a parallel implementation of the Delphin 6.1 model. The
preconditioner ILU must thus be replaced since its code is hard to parallelise. Another approach will
be applying domain decomposition techniques to split large problems onto multiple processors.
For an accurate comparison of different solution algorithms a standardised set of three dimensional
test benchmarks is required. Those test shall not only specify results, constructions, materials, and
conditions but also truncation error control parameters to ensure a valid comparison of a simulations
wall clock times or other convergence properties after valid simulation results are meet.
7. Acknowledgements
The authors gratefully acknowledge funding support by the German Federal Ministry of Economics
and Technology under the Contract No. 0329663N.

References
Bear J., & Bachmat Y. 1992. Introduction to modeling of transport phenomena in pourous media,
Kluwer Academic Publishers.
Grunewald, John. 1997 (September). Diffusiver und konvektiver Stoff- und Energietransport in
kapillarporösen Baustoffen. Ph.D. thesis, Technische Universität Dresden, 01062 Dresden. pp.
104.
Janssen, J. & Blocken, B. & Carmeliet, J. 2007. Conservative modelling of the moisture and
heat transfer in building components under atmospheric excitation, International Journal of
Heat and Mass Transfer, http://dx.doi.org/10.1016/j.ijheatmasstransfer.2006.06.048
Kalagasidis, A.S. & Hagentoft, C. 2002. Simulink modelling tool for HAM system analyses in
building physics, 6th Nordic Symposium on Building Physics
Krabbenhoft, K. 2003. Moisture Transport in Wood - A Study of Physical-Mathematical Models and
their Numerical Implementation. Ph.D. thesis, Department of Civil Engineering, Technical
University of Denmark, Denmark.
Nicolai, A. 2007 (December). Modeling and Numerical Simulation of Salt Transport and Phase
Transitions in Unsaturated Porous Building Materials. Ph.D. thesis, Syracuse University, NY,
USA. Printed at Institute of Building Climatology - Dresden University of Technology, Dresden,
2008.
Nicolai, A. 2011 (May). Towards a Semi-Generic Simulation Framework for Mass and Energy
Transport in Porous Materials. Page 559ff. of: Vinha J. & Pirronen J. & Salminen K. (ed), 9th
Nordic Symposium on Building Physics - NSB 2011, vol. 2. ISBN: 978-952-15-2573-5.
Paepcke, A. 2014. Comparison of Direct and Iterative Linear Equation System Solvers for Building
Component Simulation. 10th Nordic Symposium on Building Physics. To be published.
Saad, Y. 2003. Iterative methods for sparse linear systems. SIAM, 2. ed., ISBN: 0898715342
Segerholm, I. 2007 (December). Moisture transport processes in Scots pine – Anomalous capillary
suction. Nonisothermal diffusion. Ph.D. thesis, Chalmers University of Technology, Goteborg,
Sweden.
Siau J. F. 1984. Transport processes in wood. Springer series in wood science. Berlin, New York:
Springer-Verlag.
Sontag L. & Nicolai A. & Vogelsang, S. 2013. Validierung der Solverimplementierung des
hygrothermischen Simulationsprogramms Delphin. Tech. rept. Technische Universität Dresden.
http://nbn-resolving.de/urn:nbn:de:bsz:14-qucosa-128968.
Vogelsang S. & Fechner H. & Nicolai A. 2013. Delphin 6 Material File Specification. Tech. rept.
Technische Universität Dresden. http://nbn-resolving.de/urn:nbn:de:bsz:14-qucosa-126274.
Vogelsang S. & Nicolai A. & Sontag L. 2014. Automated Validation of Building Performance and
Building Envelope Simulation Tools. To be published.
Whitaker S. 1969 (December). Advances in Theory on Fluid Motion in Porous Media. Page 14ff. of:
Flow Through Porous Media Symposium, vol. 12/61. Industrial and Engineering Chemistry.
Whitaker S. 1986. Flow in Porous Media I: A Theoretical Derivation of Darcy’s Law. D. reidel
Publishing Company.
Zillig W. 2009 (May). Moisture Transport in wood using a multiscale approach. Ph.D. thesis,
Katholieke Universiteit Leuven, Kasteelpark Arenberg 40, B-3001 Leuven.

Full paper no: 37
Evaluating the desired accuracy for mould index calculations and

recommendations for crawlspace simulations
Anssi Laukkarinen, B.Sc. 1
Petteri Huttunen, B.Sc. 1
Juha Vinha, Professor 1
1
Tampere University of Technology, Finland
KEYWORDS: Crawlspace, temperature field, simulations
SUMMARY:
This text studies the required temperature accuracy criterion for mould growth index calculations in
crawlspace. First outdoor air conditions data and the updated VTT-TUT mould growth model are
used to evaluate the accuracy criterion for temperature field. Then heat conduction and air change
simulations on crawlspace are made and the temperature accuracy criterion is used to give guidelines
on modelling a crawlspace structure.
When the ground under the crawlspace is initially at outdoor air average temperature, it is
recommended to have a minimum of three years pre-conditioning time. Based on the literature and
simulations, it is recommended to have the cut-off planes in the ground at a distance of at least three
times the periodic penetration depth or the width of the building. The temperature accuracy limit
±0.13 °C doesn’t include many essential parameters, but gives direction on how sensitive the updated
VTT-TUT mould growth model is for the change in temperature and humidity conditions.
1. Introduction
1.1 General
According to the European directive 2010/31/EU all new buildings must be near zero energy buildings
by the end of year 2020. To design and build truly energy efficient and durable buildings it is essential
to understand the hygrothermal behaviour of building structures. Numerical simulations provide a
powerful way to contribute to this knowledge.
Calculation of temperature field is an essential part of evaluating the hygrothermal behaviour of
building components. For example the relative humidity is a key factor in material moisture content
and many degradation phenomena. Relative humidity depends essentially on temperature and because
of that the temperature conditions should be evaluated as accurately as possible. Different authors
however use different criterion for evaluating the applicability of a simulation model. This text aims to
contribute to that by suggesting a method to determine limits for temperature accuracy.
Crawlspace is a building foundation type where there is an air space between the soil and living areas
of the building. The structure allows making the building on several types of soils and for example on
top of piles. Crawlspace structure is also suitable when making the building from volume elements
and the crawlspace is also recommended in radon areas. However the major drawback of the structure
is its inadequate hygrothermal behaviour in the summertime. The temperature accuracy criterion is
used to give reference values when modelling crawlspace structure.
1.2 Literature review

International standard SFS-EN ISO 10211 (2008, chapter 5.2.4) gives guidelines on selecting the cut-
off planes in the ground. For the calculation of both heat flows and surface temperatures the building

is divided from the symmetry plane. On the outside the ground is included for the length 2.5 x b in all
directions, where the length b is the smaller length of the rectangular building foot (width). The cut-off
planes on the symmetry plane and outside the building are adiabatic boundaries. For example, if the
width of a building is 8 m, then 2.5 x 8 m = 20 m of soil is to be included in the calculation model
outside the building.
International standard SFS-EN ISO 13370 (2008) states that the default values of λ = 2 W/(mK) and
ρc = 2 x 106 J/(m3K) as thermal conductivity and volumetric heat capacity of soil can be used as
default values if no other information is available. These values for soil have been used in all
calculations.
Cheng (2013) has used a selection of the IEA BESTEST cases and the calculation results from that
project to evaluate the results of an approximate steady-state and dynamic 3D calculation model for
uninsulated slab. The selected cases have a slab dimension of 12 m x 12 m, ground depth of 15 or 30
m (one case for shallow ground at 2 m depth) and the vertical cut-off plane 8, 15 or 20 m from the
outside surface of the building wall. The deep ground boundary condition was the average outdoor air
temperature and the vertical cut-off planes were adiabatic boundaries. It is also brought up that for a
relatively shallow deep ground boundary the average temperature would be higher than the average
outdoor air temperature. Because of that the use of average outdoor temperature would cause error in
the calculations.
Pallin & Kehrer (2012) used a 1D model to evaluate the temperature conditions in 1 m depth in a free
ground in absence of a building. Adiabatic boundary was set at 20 m depth from the ground surface
and they used 10 years presimulation time. Moisture transfer was also included in the calculations.
Zoras et al. (2011) presented work on the numerical generation of response factors (NGRF). It was
stated that in dynamic situation the time required to reach equilibrium can vary depending on the
accuracy of the initial ground and structure temperature. It is also related to the meteorological
conditions and the location. Three years simulation time for thermal equilibrium is given as an
example.
Rees et al. (2007) compared 2D numerical simulations to full-scale measurements of a basement.
Building foot was 6.5 m x 6.5 m and the building was divided from the symmetry line. The distance
from outside wall surface to vertical cut-off plane was 9.7 m and the vertical distance to deep ground
was 12.1 m. They used four annual cycles as pre-conditioning before the actual evaluation cycle. For
the deep ground boundary condition a constant temperature was chosen, which was noticed to
correspond well to preliminary simulation results.
Janssen et al. (2004) used numerical simulations to study the impact of soil moisture transfer on the
building heat losses to the ground. When creating the 2D simulation model the building was first cut
from the symmetry line. In the simulation model the basement was 5 m x 2.5 m and the amount of soil
on the outside of the building was 15 m in the lateral direction and 12.5 m in the vertical direction.
Adiabatic boundary condition was used in both vertical and horizontal cut-off planes. One-
dimensional simulations were also done, which were started from average outdoor temperature and
run for 15 years. The 2D simulations were initialised from the steady-state conditions where the deep
ground conditions from 1D simulations were assigned to the outside surface. The 2D cases were
simulated for 10-15 years until steady-periodical situation was attained with good accuracy.
Hagentoft (2001, ch. 3.2.5) has shown an analytical solution to the temperature response of a half-
infinite slab to a sinusoidal temperature variation. The amplitude of the temperature variation drops
exponentially when going deeper to the slab. At a depth of e.g. three times the periodic penetration
depth the amplitude has dropped to 5 % of the amplitude at the surface. Periodic penetration depth δp
[m] is defined as δp = sqrt(atp/π), where a [m2/s] is the thermal diffusivity of the material and tp [s] is
the period length. Example of δp > 2 is given.

Adjali et al. (2000) discuss the selection of either constant temperature or adiabatic boundary
condition to the deep ground. Temperature measurements from 10 m depth under a building showed 3
°C higher average temperature than the outdoor air average temperature. As stated also by Janssen
(2004) the ground temperature depends also on the building itself and of the ground surface
temperature, not the air temperature. Adjali et al. (2000) chose to use adiabatic deep ground boundary
condition at 20 m depth. The vertical “far-field” cut-off planes were also adiabatic at 10 m distance
from the external walls. Measured temperature values were used as initial conditions.
2. Methods
2.1 Temperature accuracy criterion
The temperature accuracy criterion was determined with the following steps:
1. Hourly T/RH data for 30 years was acquired from the Finnish Meteorological Station from
four different localities (Sodankylä, Jyväskylä, Jokioinen and Vantaa).
2. For every hour of the year, new relative humidity was calculated by assuming a constant
temperature change and a constant amount of water vapour in that time step. When
temperature was decreased the relative humidity was limited to a maximum of 100 % RH.
3. The updated VTT-TUT mould growth model (Ojanen et al. 2010) was used to calculate the
maximum value of mould index for every year and as a function of temperature change.
Mould index calculation was done for the material class: “Very sensitive” (although no actual
material surface was present).
4. Cumulative distribution function of the impact of temperature change was plotted for the set
of 30 values (ΔM(ΔT)) at a time. Two maximum values (ΔM = 0.6 and ΔM = 1) were used
and the temperature change where less than 90 % of the years stay under the limit was chosen.
2.2 Simulations
The crawlspace geometry and other simulation model information are shown in the next drawing.
Figure 2.1 The geometry of studied crawlspace. Floor insulation is 200 mm and the thickness of
foundation wall layers is 100 mm. The outside ground is 500 mm lower than the floor inside surface.
For the 300 mm length of wall, adiabatic boundary condition is used. The amount of ground in the
simulations (parameter z) is given values of 1, 2, 4, 8, 16 and 32 m. Monitoring point Tairmiddle is
also shown.
As initial condition the yearly average of outdoor air temperature was used (5.63 °C). Long-wave
radiation in the y-direction in crawlspace was taken into account where it was possible, but not
between the wall and horizontal surfaces.
In Comsol Multiphysics the “Finer” Physics-controlled mesh was used for all calculations. For Wufi
calculations the default discretization was used without modifications. In Delphin calculations first the

automatic discretization was done with minimum element width 1 mm and an expansion factor 1.6.
Towards adiabatic surfaces the small elements were merged to reduce the amount of elements.
Material data and the boundary conditions that were used in the calculations are presented in the next
table.
TABLE 1. Left: Material data. Right: Boundary heat transfer coefficients.
Heat transfer
λ ρ cpa
Material Boundary coefficient,
W/(mK) kg/m3 J/(kgK)
W/(m2K)
100 or Indoor air
Air (*, **) 1.25 1000 5.88
2.353 (21 °C)
Outdoor air
Concrete 2.3 2350 800 25
(TRY Jokioinen 2012)
EPS 0.039 35 800 Crawlspace top and 0.7 + Long-wave
Mineral wool 0.035 30 800 bottom (*) radiation
Soil 2 2000 1000 Crawlspace wall (*) 7.7
(*) In Comsol and Delphin a large value for thermal conductivity of air was used and the surface heat
transfer was taken into account with surface resistances or heat transfer coefficients. Long-wave
radiation was included in the calculations. In Wufi the thermal conductivity of air was calculated from
the surface heat resistances (2 x 0.17 m2K/W) and height of the crawlspace (0.8 m).
(**) In Comsol Multiphysics the crawlspace air was handled in two ways: 1) Air was modelled as a
material with high thermal conductivity and “Thin thermally resistive layers” were used at crawlspace
surfaces or 2) Crawlspace air was analyzed as a lumped system, in which case the physics interface:
“Global ODEs and DAEs -> Global equations” was used. In the latter case the lumped crawlspace
temperature was named Tcs (u in Comsol documentation) and the equation to be solved
(f(u,ut,utt,t)=0) was written as:
·
Q
Tcs - (1)
C ×V
·
Where Tcs first time derivative of lumped crawlspace temperature (Tcst), K/s
Q heating power from materials surfaces and ventilation, W/K
C volumetric heat capacity of air, constant value of 1250 J/(m3K) was used
V volume of crawlspace air, (w x h x d) = 4 m x 0.8 m x 1 m = 3.2 m3.
Initial temperature (u_0) was set to 5.63 °C (TRY average temperature) and the initial value of
temperature time derivative (u_t0) 0 °C/s. The heating power was defined as a local variable and it
was written as:
Q = (aveop1(T ) - Tcs ) × h1 × A1 + (aveop 2(T ) - Tcs ) × h 2 × A2 + nCV (Te - Tcs ) (2)
Where aveop1 average operator (local definitions), which calculates the average
temperature of crawlspace roof and ground surface, K
aveop2 average operator (local definitions), which calculates the average temperature
of crawlspace wall, K
h, A convective heat transfer coefficients, W/(m2K) and area, m2
n air exchange rate between crawlspace and outdoor, 1 1/h
Te outdoor air temperature (global variable), which was interpolated from a text file, K.
For yearly variations the periodic penetration depth in soil is δsoil = sqrt(1·10-6 m2/s ·8760·3600 s/π) =
3.2 m

3. Results
3.1 Temperature accuracy criterion based on outdoor air conditions
The next figures show the impact of temperature change on the mould index calculated from the
outdoor air conditions.
Figure 3.1. The impact of temperature change on the calculated mould index (sensitivity class: very
sensitive). One line is associated with a one year of outdoor air T/RH conditions. The outdoor air
moisture content is kept constant while the air temperature is decreased or increased. The maximum
value of mould index is plotted against the temperature change.
Big differences exist between years. For example during 1980-2009 in Jokioinen the maximum value
of the yearly outdoor air mould index has varied from 0.6 to 5.8 (ΔT = 0 °C). In Jyväskylä, Jokioinen
and Vantaa the variability between years becomes smaller the better (warmer) or worse (colder) the
conditions become. Sodankylä is located in Northern Finland and the average temperature there is
lower than in the other three cities. It seems that the outdoor air conditions in Sodankylä are not as
critical for mould growth as they are in other cities and even with some average temperature drop, the
maximum mould index is not very high for all years.
The temperature has a big impact on the resulting mould index. A temperature change of 0.5 °C can
change the mould index value over two units (scale 0...6). This leads to the conclusion that the
determination of temperature field accurately is essential for accurate evaluation of mould growth and
the hygrothermal behaviour of building envelope structures. A second conclusion is that it is possible

to either improve or impair the hygrothermal behaviour of building structures with solutions that affect
the temperature field.
Desired accuracy can be chosen in different ways. One possibility is to allow error of 10 % (ΔM = 0,6)
in 10 % of the years. Larger temperature difference than this means that less than 90 % of the years
stay in the 10 % (ΔM=0.6) limit. This condition leads to the following values. Also limits for mould
index error of ΔM = 1.0 are given.
TABLE 2. Minimum/maximum constant temperature error with different accuracy criterion.
City ΔTallowed ΔTallowed
≥ 90 % of years ΔM ≤ 0.6 ≥ 90 % years ΔM ≤ 1
Jokioinen -0.13...0.13 -0.22...0.21
Jyväskylä -0.13...0.13 -0.22...0.21
Sodankylä -0.15...0.28 -0.27...0.70
Vantaa -0.13...0.12 -0.21...0.21
3.2 Simulation results
Crawlspace simulation results consist of hourly temperature data from certain monitoring points. From
this data the yearly averages were calculated alongside with the hourly differences in each year
compared to the last simulation year. The next figures show the development of yearly average
temperatures in the crawlspace air and the maximum difference in each year compared to the hourly
data of the last simulation year (15th year).
Figure 3.2 Results from crawlspace simulations. Left: Impact of simulation time and the choice of
simulation software on the yearly average crawlspace temperature (z = 8 m, adiabatic). Right:
Maximum difference in the hourly values compared to the last simulation year (Crawlspace air,
Delphin, q = 0.03 W/m2).
Temperature changes faster in the beginning of the simulation than after a few years. The conditions
approach asymptotically their periodical equilibrium. In the beginning of each year the conditions
differ more from the last simulation year than the end of the year. In many cases two years pre-
simulation time would have been enough but three years pre-simulation was enough in all calculated
cases.
The next figures show the impact of increasing the amount of soil in the simulation model.

Figure 3.3 Crawlspace air average temperature for the last simulation year. Left: Impact of model size
and deep ground boundary condition. Right: Difference to the largest (z = 32 m) simulation model.
When the amount of soil is small the periodic steady-state condition changes depending on the amount
of soil in the simulation model and the deep ground boundary condition. The building itself behaves as
a heat source and the amount of soil in the simulation model affects directly the amount of heat
capacity and thermal resistance under the building. The yearly average temperature in crawlspace is
few degrees higher than the average outdoor air temperature. If the average outdoor air temperature is
used as a deep ground boundary condition it holds also the crawlspace temperature on a lower level. If
adiabatic boundary condition is used the heat can’t escape quickly enough from the ground below the
building and the temperature of the thermal pillow rises. The constant heat flux 0.03 W/m2 (= 2
W/(mK)) x (1.5 °C/100 m)) differs only little from adiabatic boundary condition.
Comparison with mould indices calculated from outdoor air vapour content and crawlspace air
temperature doesn’t give very much new information because in many cases the mould index for
crawl space air (with very sensitive class) is very high. The mould index rises in the summertime and
it is almost M = 6 as long as there is colder period in the summer than the outdoor air.
Modelling crawlspace air as lumped system produced similar results than in Fig. 3.3.
4. Discussion and conclusions

The required temperature accuracy is very high. If mould index accuracy of ΔM ≤ 0.6 is pursued the
temperature accuracy should be in the range of ± 0.13 °C. With mould index accuracy ΔM ≤1.0 the
temperature accuracy should be ± 0.21 °C. The accuracy of climate data, material parameters, mould
growth modelling etc. is not explicitly evaluated at the moment. For example if more resistant mould
growth class would be used the allowed temperature range would be wider. Conditions change also
more slowly on materials surfaces, which likely also allow wider acceptable temperature range. Also
the temperature change was now given constantly to the same direction. The simulation results can be
however sometimes warmer and sometimes colder than the measured values, which is likely to even
out differences. The values are seen as a starting point for evaluating results when simulations are
compared to measurements.
The big influence of the temperature field on the calculated mould index also means that small
favourable changes in structural solutions or boundary conditions can have a big positive impact on
the hygrothermal behaviour of envelope assemblies.
When the initial ground temperature was set to the average outdoor air temperature, three years was
enough for the crawlspace temperature changes to become smaller than the desired limit. Based on the
literature and the simulations the amount of ground should be at least three times the periodic
penetration depth or the width of the building, which one is larger.

Different deep ground boundary conditions produced the same results when the size of the simulation
model was increased enough. However due to the wide use of adiabatic conditions as deep boundary
condition in the literature, it is recommended with the use of large enough of amount of ground.
Setting up maximum time step was necessary if it wasn’t already done in the calculation program. The
impact of maximum time step wasn’t varied but when hourly data was used, a maximum time step of
half an hour was selected.
Crawlspace is known as a risky solution and the summertime temperature deficit appeared also in the
current simulations. In summer the lower temperature compared to outdoor air causes relative
humidity to rise and risk for mould and moisture problems to increase. Besides minimizing moisture
loads, also the temperature behaviour of traditional crawlspace should be changed with structural
solutions.
References
Adjali, M.H., Davies, M., Ni Riain, C. & Littler, J.G. (2000) In situ measurements and numerical
simulation of heat transfer beneath a heated ground floor slab, Energy and Buildings 33 (2000) 75-
83.
Cheng, Dong (2013) Dynamic three-dimensional heat transfer calculation for uninsulated slab-on-
ground constructions. Energy and Buildings 60 (2013) 420-428.
Hagentoft, Carl-Eric (2001) Introduction to Building Physics. Studentlitteratur. Lund, Sweden. ISBN
91-44-01896-7
Janssen, Hans; Carmeliet, Jan & Hens, Hugo (2004) The influence of soil moisture transfer on
building heat loss via the ground. Building and Environment 39 (2004) 825-836.
Ojanen, Tuomo; Lähdesmäki, Kimmo; Viitanen, Hannu; Vinha, Juha; Peuhkuri, Ruut & Salminen,
Kati (2010) Mold growth modeling of building structures using sensitivity classes of materials.
Presented at the Thermal Performance of the Exterior Envelopes of Whole Buildings XI
International Conference.
Pallin, Simon & Kehrer, Manfred (2012) Hygrothermal simulations of foundations: Part I: Soil
material properties. Journal of Building Physics 37(2) 130-152.
Rees, S. W., Zhou, Z. & Thomas, H.R. (2007) Ground heat transfer: A numerical simulation of a full-
scale experiment. Building and Environment 42 (2007) 1478-1488.
SFS-EN ISO 10211. 2008. Thermal bridges in building construction – Heat flows and surface
temperatures – Detailed calculations (ISO 10211:2007). Finnish Standards Association SFS.
SFS-EN ISO 13370. 2008. Thermal performance of buildings. Heat transfer via the ground.
Calculation methods (ISO 13370:2007). Finnish Standards Association SFS.
Zoras, Stamatis; Georgakis, Chrissa; Kosmopoulos, Panagiotis & Dimoudi, Argiro (2011) Multi-year
application of the three-dimensional numerical generation of response factors (NGRF) method in
the prediction of conductive temperatures in soil and passive cooling earth-contact components.
Solar Energy 85 (2011) 2275-2282.

Full paper no: 38
Evaluation of the short-wave solar radiation transmission through

complex fenestration systems in existing office buildings in
Chilean warm climate
Sergio Vera, Assistant Professor, Ph.D. 1,3
Waldo Bustamante, Professor, Ph.D. 2,3
Alejandro Prieto, Research Associate, M.Sc. 2
Claudio Vásquez, Associate Professor, Ph.D. 2,3
1
Department of Construction Engineering and Management, School of Engineering, Pontificia
Universidad Católica de Chile, Chile
2
School of Architecture, Pontificia Universidad Católica de Chile, Chile
3
Center for Sustainable Urban Development (CEDEUS), Pontificia Universidad Católica de Chile,
Chile
KEYWORDS: Complex Fenestration Systems, Short-wave solar radiation transmission, Office

buildings, warm climate
SUMMARY:
Overheating, glare and high energy consumption for air conditioning and ventilation are recurrent
problems in office buildings in Santiago of Chile over the whole year. Solar heat gains are the major
cooling load. Large variety of complex fenestration systems (CFS) such as perforated screens,
louvers and motorized rollers are being implemented on modern office buildings. However, it is
unknown their effectiveness to control short-wave solar radiation before it is transmitted through the
glazed building façades. This paper aims to study the performance of CFS currently used in existing
office buildings in Santiago of Chile by measuring the transmission of short-wave solar radiation in
four office buildings at different façade´s orientations. The main results show that perforated and
undulated exterior screens and louvers and motorized rollers reduce peaks of solar heat transmission
up to 2% of incident solar radiation, while solar transmissions can be as low as 8%. Also, this study
reveals that CFS are more effective to reduce solar heat transmission than spectrally selective
windows with low/medium SHGC. However, non-proper architectural arrangements of CFS can
significantly diminish their performance.
1. Introduction
Overheating, glare and high energy consumption are recurrent problems in office buildings in
Santiago of Chile over the whole year. This situation is consequence of the combination of a warm
climate and modern architecture of office buildings. Figures 1a and 1c show the hourly temperature
and beam solar radiation of an average day in January, March, July and October, while Figure 1b
shows the global horizontal solar radiation during the whole-year (ASHRAE 2005). It is observed that
Santiago’s climate (S 33° 22'; W 70° 46') is characterized by high temperatures and solar irradiance
for around 8 months of the year. This climatic condition exposes office buildings to large heat gains
due to solar radiation and heat conduction.
This situation would not be a problem if office buildings would be properly designed for this weather
condition. Otherwise, modern architecture imposes buildings with completely or large glazing façades
as shown in Figure 2. Modern architecture, warm weather and internal gains (people, equipment and
lighting) cause excess of solar heat gains and daylighting transmission to the indoor environment. For
instances, it is observed in Figure 2b that internal rollers in Titanium Tower are down in the north

façade, which is either a sign of overheating or glare in perimeter offices despite of the low Solar Heat
Gain Coefficient (SHGC) of the glazing system.
(a) (b) (c)
FIG 1: Weather data of Santiago of Chile. a) Hourly temperature. b) Global horizontal solar
radiation. c) Hourly beam solar radiation.
Solar heat gains through windows are the major component of the building energy performance and
occupant’s comfort. Several authors have reported the large contribution to cooling loads of solar heat
gains through fenestration in warm climates (Reilly and Hawthorne 1998; Li and Lam 2000;
Winkelmann 2001; Kuhn 2006). Exterior shading devices are the most effective way to reduce solar
heat gains through glazed façades because they intercept direct solar radiation before it reaches the
glass. Fully shaded glazing façades can reduce solar heat gains up to 80% (ASHRAE, 2005).
CorpBanca Building Titanium Tower

FIG 2: Buildings with large glazing façades in Santiago of Chile.
2. Case studies and monitoring methodology

2.1 Buildings studies
Short-wave (SW) solar radiation transmission was monitored in four office buildings in Santiago of
Chile. Figure 3 shows pictures of the monitored buildings and their main characteristics are presented
below:

 Building A: Itaú Bank Building is a 6th story office building. Its northwest façade consists of
double glazing with unknown SHGC and visible transmittance (VT). This façade is implemented
with external motorized rollers. The rollers are set to down at certain hour each day to control
solar heat gains. According to the building manager these rollers only operate on weekdays, but it
was found that they also operate on weekends sometimes. The rest of façades does not include the
external rollers. Short-wave solar radiation transmission was monitored on the northwest façade
only.
 Building B: DHL-Intersystems Building is an office building with large glazing façades.

Northeast façade is composed of double glazing windows with SHGC of 0.39.
 Building C: San Agustín Building is a five-story university building with teaching labs, computer
labs, classrooms and offices. The east façade is composed of single-pane clear glass, while the
west façade consist of double glazed clear glass and exterior perforated and undulated screens on
the 3rd and 4rd floors of the building.
 Building D: MideUC Building is a university building mainly composed of offices and meeting
rooms. The envelope consists of double glazed clear glass with different CFS on each façade. The
north façade includes perforated and undulated louvers, while undulated and perforated screens
are placed on the east and west façades.
The CFS implemented in these buildings are motorized exterior rollers in Building A, perforated and
and ondulated louvers in Building D, and ondulated and perforrated screens in Buildings C and D.
Building B does not count with CFS but it has double glazing windows with medium SHGC. Finally,
the east façade of Building C is composed of windows with a single clear glass.
Building A Building B
Building C Building D
FIG 3: Monitored buildings

The CFS, such as the undulated and perforated screen shown in Figure 4, are complex to be modelled
in building energy simulation tools when their Bidirectional Scattering Distribution Functions
(BSDF) are unknown, which is a common situation. Thus, this study aims to evaluate the
effectivennes of CFS via monitoring how well they control solar heat gains in comparison with single
glass windows and spectrally selective windows.
FIG 4: Undulated and perforated screen installed in Buildings C and D.
2.2 Methodology of monitoring

SW solar radiation transmission was measured using piranometers SP Lite2 of Kipp&Zonen at
different locations of the fenestration system as shown in Figure 5a. The outer piranometer measures
incident solar radiation on the CFS, while the inner piranometer measures how much solar heat is
transferred through the fenestration system. When a CFS was part of the fenestration system, a
piranometer was placed between the CFS and glass as shown in Figure 5a.
(a) (b)
FIG 5: a) Mounting diagram of piranometers (dimensions are in millimeters). b) Comparison of
vertical solar radiation measurements of piranometers SP Lite 2 and CMP 11.
SP Lite2 sensor has a spectral range from 400 to 1100 nm. Since this range covers a fraction of the
whole solar spectral range, thus they measure a range much smaller than more precise piranometers, it
was necessary to evaluate if SPLite2 were suitable for this study. This was done by means of
comparing the measurements of solar radiation of three sensors SP Lite2 against the measurements of
sensor CMP11 of Kipp&Zonen, which has a spectral range from 285 to 2800 nm. Measurements were
carried out for four days during winter in a vertical plane oriented to north. Figure 5b shows that

measurements of SP Lite 2 are in very good agreements with CMP11 measurements, which evidences
that SP Lite2 sensors are suitable for measuring solar radiation on vertical planes. In consequence,
piranometers SP lite 2 could be used to indirectly evaluate the performance of fenestration systems to
reduce and control solar heat gains. The main limitation of this methodology is that long-wave solar
radiation emitted to indoors by the window is not taken into account. However, the only method that
allow to measure solar heat gains more precisely correspond to experimental calorimeters, which can
not be implemented for measurements in existing buildings.
Monitoring was continuously carried for two or more days at different seasons in different building
façades from July 2012 (winter in south hemisphere) to January 2013 (summer in south hemisphere).
New measurements are being carried out to complete data for different façade orientations and types
of fenestration systems. The objective of these measurements is to obtain experimental data for E, NE,
N, NW and W façades for clear glass only (single or double glazed), double glazed with low SHGC,
and glazing systems with CFS.
3. Results and analysis

Figure 6 summarizes the measurements (each circle) of the ratio between the daily integral of the measured SW
solar radiation by the inner and outer piranometers. This ratio reflects the effectiveness of the fenestration system
to control solar heat gains. This ratio varies between 0 to 1. Values close to 0 indicate very low SW solar
radiation transmission, while values close to 1 evidence that large portion of SW solar radiation is transmitted
indoors through the fenestration system. These results clearly show that CFS, such as screen rollers and
undulated and perforated screens/louvers can effectively reduce the solar heat gains.
FIG 6: Measured SW solar radiation transmission through different fenestration systems, buildings,
façade orientations, and season.
The following sections presents detailed results of incident solar radiation on building façades, solar radiation
between the CFS and glass, and short-wave solar radiation transmitted through the fenestration system towards
the indoor environment.
3.1 Building A
Figure 7 shows the measurements of solar radiation on the west façade in July 2012 (winter) and
January 2013 (summer). In both figures is clear the effectiveness of exterior rollers to control the
transmission of SW solar radiation. Despite of the high incident solar radiation on the northwest
façade, the peak of transmitted SW solar radiation is below 20 W/m2 in winter and summer. This

means this CFS blocks more than 85% of solar radiation when it is down and reduces the peak of
transmitted SW solar radiation by more than 95%.
Figure 7a also shows a couple of days when rollers are not down during the weekend. During these
days, the transmitted SW solar radiation is about 300 W/m2, which is significantly higher than the
transmitted SW solar radiation when the rollers are down. Figure 6 shows three days when this
situation happened, thus the ratio between the inner and outer piranometer measurements is 0.37,
which is significantly higher than the same ratio when the rollers are down.
Figure 7a also demonstrates that rollers should be down before midday in winter because the
transmitted SW solar radiation increases significantly between 9 and 12 AM reaching a peak of 200
W/m2. This could cause important cooling loads during the morning even though in winter.
(a) (b)
FIG 7: Incident and transmitted solar radiation through fenestration systems on northwest façade of
Building A: a) July 2012. b) January 2013.
3.2 Building B
Figure 8 deploys the measurements of incident solar radiation on the northeast façade in August
(winter) and October (spring). Figure 8a shows peaks of incident solar radiation around 800 W/m2 and
600 W/m2 in sunny days of winter and summer, respectively. In summer, incident solar radiation is
lower than that in winter due to higher incident sun angles. Similarly to the measurements in Building
A, it is shown that the transmitted SW solar radiation is higher in winter than that in spring and
summer. This fact, in addition to that the people occupancy ratio of buildings is usually higher in
Chile than that in North America and Europe, causes high energy consumption for cooling in office
buildings even though in winter.
Figures 8a and 8b show that the double glazed window only transmits around 25% of the incident
solar radiation. This indicates that spectrally selective windows with medium to low SHGC allow
controlling solar heat gains effectively during the whole year. However, they still allow transmitting
20 to 25% of incident solar radiation, which is significant in climates with high solar radiation even in
winter.

(a) (b)
FIG 8: Incident and transmitted solar radiation through northeast glazing façade in Building B: a)
August 2012. b) October 2012.
3.3 Building C
Measurements were carried on the east and west façades in an unoccupied office and a small library,
respectively, that are located on the third floor. Main results are presented below.
3.3.1 East façade: single clear glass

Figure 9a shows 8-day measurements on the east façades by the end of November 2012 (late spring).
It can be seen that peaks of incident solar radiation on the east façade are around 900 W/m2 in sunny
days, which is a very high value. At the end of the measurement period is observed much lower values
of incident solar radiation because these days were cloudy. Figure 9a also shows that transmitted SW
solar radiation through the single clear glass is about 700 W/m2 in sunny days. This means that around
75% of incident solar radiation is transmitted through the clear glass. As expected, these results show
that this type of window do not allow controlling solar heat gains, especially at low incident angles
that occur on the east façade early in the morning. This situation might cause high cooling loads, glare
and uncomfortable indoor conditions in the work places close to the windows during the morning.
3.3.2 West façade: double glazed clear glass with CFS

Figure 9b shows the measurements on the west façades also in November 2012 (late spring). Peaks of
incident solar radiation on the west façade have similar magnitude than the peaks observed on the east
façade. On the contrary to what happened on the east façade, transmitted SW solar radiation through
the fenestration system on the west façade is very small. This result is because the CFS blocks more
than 85% of the transmitted SW solar radiation and reduces the magnitude of peaks by 90%. These
results clearly show the effectiveness of the undulated/perforated screen on reducing solar heat gains.
The CFS implemented on the west façade of this building is similar to the device shown in Figure 4,
which blocks solar radiation very effectively but allows the transmission of natural light.
3.4 Building D
Figure 10 shows that the CFS on the east and north façades significantly blocks the incident solar
radiation. On the east façade (Figure 10a), the peak of transmitted SW solar radiation is about 380
W/m2 but it occurs for very short time (< 30 min). The rest of time when sun rays strike the east
façade, the transmitted SW solar radiation is lower than 70 W/m2, which is only 10% of the incident

solar radiation. On the north façade, the results are better due to the perforated and undulated fix
louvers (Figure 10b), thus the peak of transmitted SW solar radiation is lower than 42 W/m2. This
represents less than 6% of the peak incident solar radiation. However, it should noticed that higher
values of transmitted SW solar radiation might be found due to unshaded sections of the north façade
due to architectural arrangement of the CFS as shown in Figure 11c.
(a) (b)
FIG 9: Incident and transmitted SW solar radiation in San Agustín Building in November: a) East
façade (single glazed clear glass). b) West façade (double glazed clear glass with perforated and
undulated exterior screen).
(a) (b) (c)

FIG 10: Incident and transmitted solar radiation through CFS in Building D: (a) East façade
(January). (b) North façade (December). (c) West façade (January)
Despite that the east and west façades have the same CFS, Figure 10c shows that the transmitted SW
solar radiation on the west façade is much higher than that in the east façade in summer. While the
CFS on the east façade allows transmitted SW solar radiation around 12%, the same CFS on the west

façade permits the transmission of 35-40% of the SW solar radiation (see Figure 6). Moreover, the
transmitted SW solar radiation through the west façade of this building is much higher than the value
found for the west façade of building C (11%) that has a similar CFS. The high transmission of SW
solar radiation on the west façade of building D is consequence of a non-proper architectural design
of the CFS, which allows solar transmission at high solar altitudes through the large unshaded part of
the window (see Figure 11a). This fact significantly deteriorates the performance of a good CFS.
Despite the architectural design problems of the CFS on the west façade of Building D, this CFS
performs well in winter due to lower solar incident angles. Figure 6 shows that the transmitted SW
solar radiation is about 12% of the incident solar radiation in July (winter).
(a) (b) (c)

FIG 11: a) Perforated/undulated screen on west façade with uncovered sections. b) Perforated and
undulated louvers on north façade. c) North façade with uncovered sections with louvers.
4. Conclusions
This paper presented a study of the effectiveness of typical glazing façades and CFS to control solar
heat gains in existing buildings in Santiago of Chile. The main conclusions that could be drawn from
this study are:
 CFS are more effective than double glazed windows with low/medium SHGC to reduce
transmitted SW solar radiation. Perforated and undulated exterior screen/louvers and motorized
rollers reduce peaks of SW solar radiation transmission up to 2% of incident solar radiation,
while SW solar transmissions can be as low as 8%.
 Architectural design of CFS can extremely influence the effectiveness of CFS to control solar
heat gains. When CFS are not arranged properly, their effectiveness to control solar heat gains
can be significantly diminished. Therefore, it is important that designers take care how CFS are
installed to maximize their effectiveness to control solar heat gains.
 Since high internal gains in Chilean offices buildings due to high occupancy ratios, significant
SW solar radiation transmission during wintertime might cause the need for cooling even in
winter. In consequence, designers should take into consideration the implementation of CFS to
control solar heat gains over the whole-year.

5. Acknowledgements
This work was funded by the National Commission for Science and Technological Research (Conicyt)
of Chile under the grant Fondecyt 1111001 and supported by the research grant CONICYT/FONDAP
15110020.
References
Appelfeld, D., A. McNeil & S. Svendsen. 2012. An hourly based performance comparison of integrated micro-
structural perforated shading screen with standard shading systems. Energy and Buildings, 50: 166-176.
ASHRAE. 2005. Handbook Fundamentals. Atlanta: American Society of Heating, Refrigerating and Air-
Conditioning Engineers, Inc.
Khun, T.E., S. Herkel, F. Frontini, P. Strachan & G. Kokogionnakis. 2011. Solar control. A general method for
modeling solar gains through complex façades in building simulation programs. Energy and Buildings, 43
(1): 19-27.
Khun, T.E. 2006. Solar control: A general evaluation method for façades with venetian blinds or other solar
controls systems. Energy and Buildings, 38 (6): 648-660.
Li, D.H.W. & J.C. Lam. 2000. Solar heat gain factors and the implications to building designs in subtropical
regions. Energy and Buildings, 32 (1): 47-55.
Reilly, S. & W. Hawthorne. 1998. The impact of windows on residential energy use. ASHRAE Transactions, 104
(2): 791-798.
Winkelmann, F.C. 2001. Modeling windows in EnergyPlus. Proceedings of international IBPSA Conference,
Building Simulation, Rio de Janeiro, august 13-15, pp. 457-464.

Full paper no: 39
On the applicability of quantitative infrared thermography on

window glazing
Kim Carbonez, M.Sc. 1
Nathan Van Den Bossche, Assistant Professor1
Marijke Steeman, Assistant Professor 1
Sven Van De Vijver, MEng. 1
Arnold Janssens, Full Professor1
1
Ghent University, Belgium
KEYWORDS: infrared thermography, glazing, quantitative analysis, U-value,
SUMMARY
Energy efficient buildings are an essential factor to reduce the energy consumption by 2020. New
buildings have to meet severe requirements, whereas older buildings need renovation to reduce the
heat losses through the building envelope. Infrared thermography (IRT) might be an improvement
over existing methods to assess the thermal performance of an existing wall in a non-destructive way,
or to check upon the as-built quality, specifically in the case of window glazing. The technique
instantly visualises the surface temperature of a whole building part, and in turn might allow to
deduce the thermal transmittance accordingly. However, many parameters can influence the surface
temperature and lead to distorted conclusions. This paper reports on the impact of different indoor
and outdoor boundary conditions for the assessment of the U-value of glazing, using the results from
a numerical simulation model. After an analysis of 6 types of windows, it is concluded that for
specific conditions, IRT might allow to estimate the U-value with an acceptable accuracy, based
upon the instantaneous indoor surface temperature. In future research, experiments will be
performed to validate this conclusion and the assessment methodology will be improved.
1. Introduction
Energy efficiency and sustainability are major concerns of our time. In 2011, the European
Commission decided that the energy consumption in general should decrease with 20% by 2020
(compared to 1990) (EC, 2011). To reach this goal, the building industry has to comply with stringent
regulations in respect to insulating performance, airtightness quality, renewable energy sources, etc.
However, new buildings only constitute a minority of the building stock. In Belgium, 62 % of the
buildings have been constructed before 1970 (WTCB, 2005), and typically do not comprise any
insulation in the building envelope. Consequently, deep energy renovation and refurbishment has
become essential to hit the target by 2020.
To verify whether the thermal performance of a building envelope meets the requirements, or to
determine where renovation is necessary, the as-built state has to be evaluated. To this end, it is
generally accepted to consider the thermal transmittance (U-value). This value can be calculated for
steady-state conditions from technical standards (EN673, 1997, ISO6946, 2007) or be measured on
site by means of heat flux sensors (ISO9869(E), 1994). The latter technique determines the U-value of
a building element by measuring the heat flux on one side and the surface temperature on both sides.
It is important to collect data during a considerable time span (preferably about 2 weeks) as in reality
the boundary conditions are always fluctuating. With a substantial dataset it is typically possible to
derive a precise U-value. This method is time consuming, and in principle different points on the wall
should be analysed to exclude singularities, which renders this approach rather elaborate. Infrared

thermography (IRT) could potentially improve the efficiency of on-site U-value determination. The
duration of the measurement procedure with a thermographic camera is short, and this technique
allows to analyze the surface temperature of a whole building part at once. In this way a more
complete overview of the heat flows of a building is attained, which is a clear advantage over the
point wise data of the existing methods. Evidently, the instantaneous measurement only yields a
single value in time, thereby limiting its applicability and accuracy.
Nowadays, IRT is a popular tool for qualitative building diagnostics (Lucchi, 2011, Kalamees, 2007,
Straube and Burnett, 1999, Burn and Schuyler, 1980, Taylor et al., 2013, Balaras and Argiriou, 2002),
either for supervision of the building quality during construction or to detect defects after
completion. The temperature gradients on a building surface can indicate e.g. missing or damaged
insulation, air leakages or sources of moisture. More recent studies focus on quantitative application
of IR-images: Asdrubali et al. (2012) used IRT to classify thermal bridges, other researchers tried to
derive the U-value of building components (Dall'O et al., 2013, Fokaides and Kalogirou, 2011,
Lehmann and Ghazi Wakili, 2013). However, this is a complex procedure due to the fact that many
variable parameters influence the instantaneous surface temperature of the object, such as solar
radiation, atmospheric long wave radiation, wind velocity, outside temperature fluctuation, material
characteristics, indoor heating. Lehman et al. (2013) performed a sensitivity analysis for a large
number of parameters on 6 different wall types (brick cavity wall, concrete wall, timber framed wall
and 3 types of plaster-brick walls). They concluded that the external surface temperature strongly
depends on the wall assembly and its thermal performance. Solar- and IR-radiation turned out to
impose the strongest restrictions for quantitative IRT, because these introduce a direct temperature
increment on the surface of the object. It was concluded that sunshine had to be avoided somewhere
between 1 hour and 2 days before the measurements, depending on the thermal capacity and assembly
of the wall.
In general, the time that is needed to reach a uniform temperature over the section of a wall is called
the time constant τ , which is proportional to the square root of the thermal conductivity (λ), the
density (ρ) and volumetric heat capacity (c). The larger the time constant, the more stringent the
boundary conditions for IRT become in order to reduce the impact of transient effects. Windows
typically have a low time constant (30min for a 4mm glass pane vs. 12h for a cavity brick wall), hence
the restrictions for quantitative thermography on glazing are perhaps less severe. The use of IRT on
windows could be a powerful application in the case of renovation or energy audits, because flux
measurements on glazing units are not evident. Even a visual inspection of the thermal performance
of the glass is hardly possible on site, contrary to an insulated brick wall. Next to that, IRT could have
the potential to reveal degradation in time due to gas leakage.
In this paper, the investigation to suitable boundary conditions for quantitative IRT on glazing is
presented. As explained before, an IR-image is a record of one moment under specific circumstances,
which typically does not correspond to the steady state. A robust assessment method of the U-value
should thus include a confidence interval. Six glazing types are analysed, using numerical
simulations. A sensitivity analysis to internal and external climatic variations is performed.
2. Method - numerical analysis

Every object emits radiant thermal energy from its surface as long wave radiation (heat). Hence the
energy received by an infrared sensor of a thermal camera consists of the emitted energy of the target.
But also the surroundings emit energy, which is partially reflected by the object and in this way
captured by the camera as well. Furthermore, the infrared emission of the atmosphere between the
camera and the object contributes, and in the case of transparent elements, long wave radiation from
behind the element is partially transmitted through and should be taken into consideration as an
additional term (FIG. 1). The standard equation for incident radiation of the camera becomes:

FIG. 1 Aspects of incident thermal radiation on the thermal sensor of an IR camera
I = τ a ⋅ ε o ⋅ I o° + τ a ⋅ (1 − ε o ) ⋅ I sur + (1 − τ a ) ⋅ I a + τ o ⋅ I i
(1)
Where τ transmittance
ε emissivity of the surface
I irradiance
° blackbody
a, o, sur,i atmosphere, object, surroundings, interior
The incoming energy is then converted to temperature values according to the Stefan-Boltzmann law.
In other words, when the parameters for atmosphere and surroundings are correctly set to the camera
software, thermography allows to determine the instantaneous surface temperature of a construction
(either interior (θsi) or exterior (θse)).
Next to that, the internal and external air temperature (θi and θe) can easily be measured on site as
well. With these data and an appropriate internal or external heat transfer coefficient (hi or he), the
heat flux (Q) through the wall can be calculated from the wall-air temperature difference.
Q = hi ⋅ (θ i − θ si ) or Q = he ⋅ (θ e − θ se ) (2)
The heat flux can also be expressed in function of the indoor-outdoor air temperature difference,
Q = U ⋅ (θ i − θ e ) (3)
leading to the following expression of the U-value:

θ i − θ si θ −θe
U = hi ⋅ ( ) = he ⋅ ( se ) [W/(m².K)] (4)
θi − θe θi − θe
By definition, Eq. 4 is only valid under static boundary conditions, but it will be applied to transient
conditions to assess to what extent it might provide useful information. Since the value for he is
function of the strongly varying wind velocity (Emmel et al., 2007), this study emphasises on the
determination of the U-value from the interior side. A sensitivity analysis of θsi to variations in θi is
performed by subjecting six types of windows to 3 different types of indoor temperature regimes.
Additionally, the influence of the most critical outdoor parameters is analysed. Note that specific
constraints in respect to thermography on glazing units are not addressed here. It is assumed that the
surface temperature can be determined accurately with an IR camera. Evidently, uncertainties in
emissivity and background temperature will propagate in the uncertainty interval of the estimated U-
value, as well as the accuracy of the camera itself.
2.1 Cases
FIG. 2 illustrates the different glazing assemblies that are studied. These configurations are
representative of those found in the Belgian building stock, and the broad variety of thermal
performance levels ensures the wide applicability of the results. The properties of each layer are

derived from the European and international standards (EN673, 1997, ISO10456, 2007), and listed in
TABLE 1.
FIG. 2 Overview of the 6 investigated glazing types. Diversity in glass thickness, cavity fill and
coating broadens the scope of the analysis.
TABLE 1 Properties of the 6 investigated glass types

Material Properties t [m] λ [W/m.K] ρ [kg/m³] c [J/kg.K] ε [-]
A Soda lime silica glass 0.004 1 2500 750 0.837
B Air 0.012 0.025 1,232 1008 -
C Argon 0.012 0.017 1.699 519 -
D Low ε-coating - - - - 0.04
E PVB-interlayer 0.0038 - - - -
Glass type Assembly (exterior - interior) U-valuea [W/(m².K)]
1. Single Glass (Si) A 5.75
2. Double Glazing-air (Do+Ai) A-B-A 2.85
3. Double Glazing-argon (Do+Ar) A-C-A 2.67
4. Low E Glass (LE) A-C (0.015m)-D-A 1.09
5. Acoustic Glass (Ac) A (0.006m)-D-C (0.015m)-A-E-E-A 1.08
6. Triple Glazing (Tr) A-D-C-A-C-D-A 0.7
a
with standard internal and external heat transfer coefficient hi = 7.7 W/(m².K) and he = 25 W/(m².K)
2.2 Simulation model

The numerical simulations are performed with the thermal analysis software VOLTRA . This program
calculates transient heat transfer in 3D-objects, according to European and international standards. A
solar processor takes into account of dynamic solar heat gains, based on the actual temperature and
long wave radiation (Physibel, 2011).
The main purpose of the simulations is to calculate surface temperatures θsi and θse, from which U-
values will be deduced to evaluate the use of Equation 4 under dynamic boundary conditions. Heat
transfer through conduction, convection and radiation is considered separately, according to EN 673
(1997): the radiative heat transfer is view factor based, whereas the convective heat transfer in the
cavity is proportional to the Nusselt number of the gas and its thermal conductivity. The external
convective heat transfer coefficient (hec) expressed in function of the wind speed (v) (Taki and
Loveday, 1996), the internal convective heat transfer coefficient (hic) is calculated iteratively, using
the temperature difference between the indoor air and surface (Thomas et al., 1990) :
hec = 5.85 ⋅ v 0.52 hic = 1.77 ⋅ 4 θ i − θ si

(5)
2.2.1 Outdoor boundary conditions

The simulations were carried out for a 10-days period during winter (December 17 - December 26).
The meteorological data of a reference year of Uccle, Belgium was used as input. The chosen period

contains considerable variations in wind speed (v), sunny and overcast days (Is) and variable outdoor
air (θe) and sky (θsky) temperature, as can be seen from FIG.3.
FIG. 3 Variable outdoor climate data for the period of December17-26, Uccle, Belgium, and 3
different temperature regimes for the indoor air (0= Dec 17 0h00, 10= Dec26 24h00 )
2.2.2 Indoor boundary conditions

In this study, special attention was paid to the influence of variations in indoor air temperature.
Because an IR-image shows the thermal state of a target for one specific moment, the thermal
transmittance is derived by assuming these instantaneous circumstances as a steady-state condition.
Nevertheless, in reality the indoor and outdoor temperatures are constantly fluctuating. To assess the
effect of this assumption on the estimated U-value, simulations for each glazing assembly were
carried out for 3 different indoor air temperature regimes (FIG. 3)
• θi_ct: constant temperature of 20°C
• θi_sin: fluctuating temperature 20±1°C, introduced as a sinus function with a period of 12h
• θi_ns: variable temperature with night setback (measurement data of a Belgian family
house)
3.1 Derivation of the thermal transmittance
FIG. 4 U-value assessment based on θsi for 3 θi- regimes and 6 glazing types. Note: in some cases the
estimated U-value became negative, but these values are excluded from the graph.
For each indoor temperature regime, the U-value for the 6 types of glazing is estimated every 10
minutes, based on the internal surface temperature, using Equation 4 with hi equal to 7.7 W/(m².K). In

FIG. 4 the distribution of the estimated U-value is reported. In general, it can be noticed that 50% of
the assessed values (grey boxes) approximates the theoretical (red dots) thermal transmittance (an
average deviation of 0.2W/(m².K), maximal 0.5W/(m².K)). Even though no restrictions to the
boundary conditions are made yet, this already indicates the potential of quantitative IRT on glazing.
However, strongly diverging outliers (up to 4W/(m².K) make it impossible to rely on the result of one
specific moment, neither to determine the correct type of glazing based on a single measurement
without any constraints.
It can be seen that the effect of a different indoor climates only has a minor influence for less
insulating windows (type 1-3). As their time constant is small enough, the surface temperature can
adapt at (almost) the same frequency as the ambient air temperature. For the better insulating
windows (type 4-6), the error level becomes larger, due to the higher time constant. For type 1-3, the
error interval increases with about 0.2W/(m².K), whereas for type 4-6, it augments up to 0.8
W/(m².K). This is especially true for glazing type 5, where the thickest glass pane is situated at the
interior side.
3.2 Restrictions to the boundary conditions
clear overcast
FIG. 5 Influence of the boundary conditions for the assessment of the U-value, illustrated for window
type 3. During sunshine hours and clear skies, the deviation to the theoretical value increases.
FIG. 6 Distribution of the assessed U-value(based on θsi ) for 3 θi - regimes and 6 glazing types. The
results are either only excluding sunny periods (no sun) or also moments of clear sky(cloudy) .
To increase the reliability of the estimation, events that cause extreme aberrations should be excluded
from the data. From FIG. 5 it is clear that the largest deviations of the estimated value (Utsi) come to

front in the presence of sunshine (Is) (grey circles). Due to the solar radiation, the internal surface
temperature (θsi) of the glazing augments. It gives the impression of a better insulating quality, leading
to an underestimation of the U-value. As soon as the sun disappears, this effect fades quasi instantly.
Next to that, the long wave radiation from the sky is one of the most influencing factors. This has also
been pointed out by Lehmann et al. (2013) in the case of wall assemblies. Clouds reflect the earth's
long wave radiation, leading to warmer surface temperatures of exposed objects. Consequently, a
clear sky corresponds to a low sky temperature (θsky), which reduces the surface temperature of the
glazing and therefore leads to a higher estimated U-value (FIG.5). To assess the influence of these
two outdoor parameters, the data at moments of solar radiation (Is ≠ 0 W/m²) are excluded from the U-
value assessment ('no sun', FIG. 6). This roughly corresponds with the exclusion of data between
9AM and 5PM. Additionally, the data obtained during clear skies (θsky<-5 °C ) are also excluded
('cloudy', FIG.6).
3.3 Main findings

• The range on the results decreases substantially when direct solar radiation is avoided. Apart
from single glazing, the maximal deviation from the theoretical U-value diminishes from 2.85
to 1.05 W/(m².K). It improves even more when clear sky moments are excluded, to only 0.81
W/(m².K). On average, the median deviation is only 0.24 W/(m².K), which is very accurate.
Note that these values refer to the worst case scenario of internal heating pattern: night setback.
• In absence of solar radiation, this method allows to distinguish between poor insulation value
(type 1), moderate insulation value (type 2, 3) and high insulation value (type 4, 5, 6).
• Assuming that the indoor temperature is kept constant, taking minor fluctuations of the heating
system into account (sinusoidal regime – still realistic for on-site measurements), the U-value
can be estimated with an uncertainty of at most 0.65W/(m².K) when the data is collected under
the right circumstances (cloudy), except for single glazing.
• In general, the estimated U-value is systematically higher than the theoretical value. This can be
attributed to a higher hi that is used for the estimations (7.7W/(m².K)) compared to the variable
value in the simulation model, which is typically depending on the type of glazing and
boundary conditions and in the range of 7.0 - 7.5W/(m².K). This means that for 70 to 100% of
the cases this approach yields a conservative estimation of the U-value.
4. Conclusions
This paper discusses the application of quantitative infrared thermography on window glazing. Based
on numerical simulations for 6 window assemblies with different thermal performances, a sensitivity
analysis has pointed out the most critical restrictions to the boundary conditions for a reliable
assessment of the U-value from the interior surface temperature. Assuming that this temperature can
be determined correctly, the limited simulation results show that the thermal transmittance can be
estimated within a minor confidence interval of 0.65 W/(m².K) when direct sunlight and clear skies
are avoided (except for single glazing). The regime of the indoor air temperature turned out to be of
minor importance for the quality of the results. These findings are promising for the application of IR-
thermography in assessing the insulation quality of glazing, but further research is needed to extend
them to other types of windows and other periods of the year. Of course, the confidence interval will
enlarge for on-site measurements because θsi can only be determined within a certain range, e.g. due to
the accuracy of the camera, the variability in emissivity, and the influence of the background. Future
research will extend the error estimation by means of experiments, to verify the presented conclusions
and include additional noise. Furthermore, the use of external surface temperatures for the assess-
ment of the U-value will be investigated. Up to now, this data is ignored in the analysis, but including
them might improve the methodology as information of both sides of the assembly is used. Also,
consecutive measurements in a short time span may add to the accuracy of the U-value calculation.

5. Acknowledgements
This research is founded by the agency for Innovation by Science and Technology-Belgium (IWT),
project 130210. The authors thank dr. P. Standaert for his help in the simulation modelling.
References
ASDRUBALI, F., BALDINELLI, G. & BIANCHI, F. 2012. A quantitative methodology to evaluate
thermal bridges in buildings. Applied Energy, 97, 365-373.
BALARAS, C. A. & ARGIRIOU, A. A. 2002. Infrared thermography for building diagnostics. Energy
and Buildings, 34, 171-183.
BURN, K. & SCHUYLER, G. 1980. Applications of infrared thermography in locating and
identifying building faults. journal of international institute for Conservation 4, 3-14.
DALL'O, G., SARTO, L. & PANZA, A. 2013. Infrared Screening of Residential Buildings for Energy
Audit Purposes: Results of a Field Test. Energies, 6, 3859-3878.
EC 2011. Europe 2020 targets. EU EUROPEAN COMMISSION.
EMMEL, M. G., ABADIE, M. O. & MENDES, N. 2007. New external convective heat transfer
coefficient correlations for isolated low-rise buildings. Energy and Buildings, 39, 335-342.
EN673 1997. Glass in building. Determination of thermal transmittance (U value). Calculation
Method. European Standard EN.
FOKAIDES, P. A. & KALOGIROU, S. A. 2011. Application of I thermography for the determination
of the overall heat transfer coefficient in building envelopes. Applied Energy, 88, 4358-4365.
ISO6946 2007. Building components and building elements. Thermal resistance and thermal
transmittance - calculation method. ISO International Organization for Standardization
ISO9869(E) 1994. Thermal insulation - Building elements. In-situ measurement of thermal resistance
and thermal transmittance. International Organization for Standardization ISO.
ISO10456 2007. Building materials and products - Hygrothermal properties - Tabulated design values
and procedures for determining declared and design thermal values. ISO International
Organization for Standardization.
KALAMEES, T. 2007. Air tightness and air leakages of new lightweight single-family detached
houses in Estonia. Building and Environment, 42, 2369-2377.
LEHMANN, B. & GHAZI WAKILI, K. 2013. Effects of individual climatic parameters on the
infrared thermography of buildings. Applied Energy, 110, 29-43.
LUCCHI, E. 2011. Non-invasive method for investigating energy and environmental performances in
existing buildings. PLEA Conference on Passive and Low Energy Architecture. Belgium.
PHYSIBEL 2011. Voltra 7.0 Manual. Physibel Software.
STRAUBE, J. & BURNETT, E. F. P. 1999. Rain Control and Design Strategies. Journal of Building
Physics, 23, 41-56.
TAKI, A. H. & LOVEDAY, D. L. 1996. External convection coefficients for framed rectangular
elements on building facades. Energy and Buildings, 24, 147-154.
TAYLOR, T., COUNSELL, J. & GILL, S. 2013. Energy efficiency is more than skin deep: Improving
construction quality control in new-build housing using thermography. Energy and Buildings,
66, 222-231.
THOMAS, F., GLANZMANN, J., KELLER, B., QUEISSER, A. & RAGONESI, M. 1990. Element
29: Wärmeschutz im Hochba, Zürich, Faktor Verlag.
WTCB 2005. Visietekst 'Bouwen en Innoveren'- Bouwen aan de toekomst. Belgisch Bouwplatform.

Air Tightness

Uploaded by

Copyright:

Available Formats

Air Tightness

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Air Tightness

Uploaded by

Copyright:

Available Formats

TOPIC

Full papers - NSB 2014 page 6

A quantitative evaluation of airtightness measurement

KEYWORDS: Air tightness, Buildings, Field Measurements, Survey, ZEF

Full papers - NSB 2014 page 7

2.2 Surveyed building types for the quantitative methods

Full papers - NSB 2014 page 8

Full papers - NSB 2014 page 9

4.1 Qualitative methods

Full papers - NSB 2014 page 10

Full papers - NSB 2014 page 11

Full papers - NSB 2014 page 12

Full papers - NSB 2014 page 13

Full papers - NSB 2014 page 14

Analysis and prediction of transient flow during night ventilation

KEYWORDS: measurement, mechanical ventilation, unsteady ventilation, tracer gas, night

Full papers - NSB 2014 page 15

2. Description of test cell

Full papers - NSB 2014 page 16

Full papers - NSB 2014 page 17

2.3 Measurement campaign

Full papers - NSB 2014 page 18

Full papers - NSB 2014 page 19

3.2 Global evaluation of flow type as a function of Ri

Full papers - NSB 2014 page 20

(a) Change in flow pattern (b) Richardson number

FIG 6. Schematic presentation of flow pattern changes

Full papers - NSB 2014 page 21

Full papers - NSB 2014 page 22

Determination of the ventilation rate in a full scale night

KEYWORDS: measurement, mechanical ventilation, unsteady ventilation, tracer gas, night

Full papers - NSB 2014 page 23

2. Description of test cell and measurement methodology

Full papers - NSB 2014 page 24

2.2 Measurement equipment

Full papers - NSB 2014 page 25

Where average value over period T

3.1 Tracer gas based determination of ACH

Full papers - NSB 2014 page 26

3.2 Pressure difference based determination of ACH

Full papers - NSB 2014 page 27

Full papers - NSB 2014 page 28

3.3 Corrected ACH

Full papers - NSB 2014 page 29

Full papers - NSB 2014 page 30

KEYWORDS: convection, leakage flow, discharge coefficient, air tightness, simulation

Full papers - NSB 2014 page 31

FIG 1. Circular opening FIG 2. Sharpe edged opening

Full papers - NSB 2014 page 32

According to (DIN EN ISO 5167-2), the parameter N is powered by 4. An explanation of the

Full papers - NSB 2014 page 33

Full papers - NSB 2014 page 34

Full papers - NSB 2014 page 35

Full papers - NSB 2014 page 36

Full papers - NSB 2014 page 37

Full papers - NSB 2014 page 38

Full papers - NSB 2014 page 39

2. In-situ measurement –Building Information

Full papers - NSB 2014 page 40

1.1 Sensor Location