Case Studies in Fire Safety 1 (2014) 17

Contents lists available at ScienceDirect

Case Studies in Fire Safety

Impact of jet fan ventilation systems on sprinkler activation

a r t i c l e

i n f o

Article history:
Received 28 October 2013
Received in revised form 26 November 2013
Accepted 29 November 2013
Available online 12 December 2013
Keywords:
Car park re
Jet fan
Impulse ventilation
CFD modelling

a b s t r a c t
Jet fans are increasingly preferred over traditional ducted systems as a means of ventilating
pollutants from large spaces such as car parking buildings.
Outside of tunnel applications jet fans are considered to have limitations as means of
smoke control. This is because their effectiveness is limited without side walls and they
de-stratify the smoke layer. Jet fans can however, aid post-event in smoke clearance under
the control of the attending re brigade.
The prudent re mode strategy in those jurisdictions where jet fans are not used for
smoke control is for the ventilation system to shut down on a re alarm signal.
The problem lies in the jet fans receiving that signal. A re has three basic signatures;
heat, smoke, and light.
Large car parking buildings may be sprinkler-protected. Sprinklers operate on the
heat signature. A jet fan is assumed to disrupt this signature by forcing the plume
of heat downstream and diluting that plume with cool air. This may delay sprinkler
activation.
In terms of the smoke signature, the presence of pollutants such as carbon monoxide that
could be detected by a specialised detector, are the same signals that would cause a jet fan
to increase its ow in normal mode. The response to a smoke signature is contradictory
between the desired normal-mode reaction to speed up and the re-mode reaction to shut
down.
This leaves light. Fires are very rich across the non-visible light spectra and ame detectors are an effective way to provide a re signal independent of the normal-mode operation
of the jet fans. However, the cost-benet of installing ame detectors in addition to sprinklers is questionable.
A series of detailed computational analyses are therefore undertaken to quantify the
impact of the jet fans on delaying sprinkler activation on a typical car park sprinkler
arrangement. This article assumes the perspective of a jurisdiction where the re mode
strategy is for the fans to shut down. However, the sprinkler delay will also be of interest
where the re mode strategy is to continue operating.
Once sprinkler activation had occurred a re signal is assumed to have been generated to
shut down the jet fans. Separate evaluations can then be made as to whether the delay was
acceptable or not, dependent upon to the specic building geometry.
The results indicate that for a fast-growing design re, sprinkler activation occurred at
180 s with the jet fans off and at 210 s with them on: an increase of 630 s. A separate
analysis was undertaken for the case-specic egress safety margin. Comparing the two, it
was concluded, that the impact of jet fans upon sprinkler activation was not signicant. It

Tel.: +61 458 111 022.

P.A. (Tony) Enright / Case Studies in Fire Safety 1 (2014) 17

was further established that the smoke disturbance due to the jet fan ow did not
adversely affect the visibility for those escaping.
2013 The Author. Published by Elsevier Ltd. Open access under CC BY-NC-SA license.

Introduction
This paper discusses a case study undertaken on an underground car park building of approximately 30,000 m2. This paper focuses on the sprinkler delay which is typical of many geometric arrangements. The safety margin is building specic
and is not presented here for the sake of brevity and to keep the case study general in application.
Contemporary research into car park res
In 2010 BRE published a comprehensive research report titled Fire Spread in Car Parks [1]. The overall aim of the project
was to gather information on the nature of res involving modern cars and to use this new knowledge as a basis, where necessary, for updating current guidance in Approved Document B [2].
The key ndings from the study are:
 The number of res in car parks each year is about 260 per year. This represents a very small percentage of all res in the
UK and the trend is downward.
 Of these res in car parks, about 50% did not start in a car.
 Many car res are deliberately lit.
 Most res in car parks do not spread (to a car or another car).
 About seven people are injured in car park res each year and there are very few fatalities; on average less than one per
year.
 Fires in car parks for which the building is classied as car park show an injury rate which is low compared with other
occupancy types.
 Fires which spread to involve more than one car can sometimes result in signicant structural damage.
 Despite concerns, there is very little evidence (and no substantiated evidence) to show that cars fuelled by LPG are a particular danger in re.
 The experimental programme demonstrated that where a re starts in a car which is well-ventilated (i.e. open windows),
very fast growing and severe res have resulted, leading readily (when sprinklers are not present) to re spread to all
nearby cars and potentially, all cars in the car parks.
 The experimental programme also demonstrated that where a re starts in a car which is poorly-ventilated (i.e. closed
windows), the re may go out, and has demonstrated that res in engine compartments will grow slowly.
 These results largely account for the very few car park res which cause substantial or signicant structural damage.
 The effectiveness of sprinklers in limiting a re to a single car has been demonstrated.
 There are no cases to date in the UK of structural collapse of a car park due to re.
Description of the system (normal-mode)
Ventilation of the case study car park is achieved through mechanical exhaust at each level with make-up air provided
from natural ventilation openings in the perimeter and transported via impulse ventilation from jet fans.
The jet fans provide advantages through elimination of air distribution ductwork within the car parks and reduction of
supply and exhaust air ow resistance, thereby reducing the associated fans energy consumption.
The normal-mode criteria is to ensure minimum airow velocities are maintained throughout the car park areas and that
a maximum carbon monoxide concentration is not exceeded. Carbon monoxide monitoring control systems are provided to
vary the fan speeds.
Description of the system (re-mode)
As described above, this article assumes shut down in re mode. Under normal operation the jet fans increase air ows
upon detection of carbon monoxide. Unfortunately, carbon monoxide is a common product of re and so there are contradictory system objectives for the fan to both speed up and shut down.
Alternative methods of generating a re signal to shut down the jet fans were considered.
 Heat sensing devices suffer from the same delay as the sprinklers since they operate on the same principle.
 Smoke sensing devices are not effective as in normal mode operation the jet fans activate at a threshold many times lower
than would be expected in a re and dilute concentrations.
 Light sensing devices along the fan line of sight will detect res at an early stage. The UV ranges are not notably effective
in smoky environments but the IR are effective.

P.A. (Tony) Enright / Case Studies in Fire Safety 1 (2014) 17

Analyses
Computational Fluid Dynamics or CFD is used to solve fundamental ow equations, and has been applied to modelling of
re and smoke behaviour for over 20 years. For a comprehensive discussion of CFD as applied to re refer to Chapter 38 of
the SFPE Handbook of Fire Protection Engineering [3].
The suite of CFD computer models used in these re-mode calculations are FDS [4] using the PyroSim [5] pre-processor
and Smokeview [6] visualisation.
FDS is not specically validated for jet vent air ows. Other proprietary software could model that aspect more accurately. However, the analysis is not solely about air ows, it is three-part; re plume, sprinkler activation and air ow.
FDS is considered to be superior in those other two aspects. It should also be noted that CFD models in general are better
predictors than empirical models. They are more fundamentally based. This does not mean they are perfect predictors however they are good predictors. These analyses are comparative; fan off and fan on and this reduces uncertainty.
For expedience of calculation, the model domain is deliberately small to limit the number of calculations, yet large enough to be representative. The space modelled represents an area of the car park of 20 m by 20 m by 4 m high and contains
the jet fan, the re source and an array of sprinklers. The boundary conditions are concrete oor and slab above with open
sides to represent an innite space. A plan view of the model domain is shown in Fig. 1.
Two cases are modelled, with the jet fans off or on:
 Case A: Sprinklers are in-plane with the jet fans centreline. This is the worst case from a disruption point of view as it
exposes the sprinklers to the maximum centreline velocity, and
 Case B: Sprinkler array is rearranged to be as far as possible from the jet fan centreline. This is the proposed conguration
as it places the sprinklers furthest from the centreline velocity.
A ow rate of 1.35 m3/s was modelled through a nozzle 900 mm wide by 100 mm high. This is the maximum ow rate
and its use is conservative as it leads to the greatest sprinkler delay. This centreline velocity at the outlet is approximately
15 m/s maximum. The nozzle directs air ow downward by 7 however the initial cases were simplied and a jet stream at
0 was assumed. The inclined air ow is picked up in the sensitivity cases.
Where possible it is recommended that fans are installed over the roadways, as this is where the motor vehicle pollutants
are most prevalent. This is a positive feature for the re-mode case, since car res are assumed to predominantly occur in
stationary vehicles. This means that most re locations will occur more remotely from jet fan locations than the cases modelled, meaning less potential for interference.
The choice of design re is derived from the British Standard BS 7346-7 as described in BD2552 [1]. This is a 5 m  2 m
source of perimeter 14 m and with a 4 MW HRR. The purpose of the design res in BS 7346-7 is to size smoke control systems and hence the steady state re. In this case study a fast t2 growth rate is assumed. This is more conservative as a steady
state design re would lead to quicker sprinkler activation and system shut-down. The decay phase is ignored. This is modelled in FDS using the heat release rate per unit area (HRRPUA) or burner.
Petrol was selected as a fuel rather than say a plastic or rubber-based material. Because the HRR_PUA function is used and
because the visibility is directly analogous to other tenability criteria the parameters of importance are the Soot Yield and the
Visibility Factor.

Fig. 1. Plan view, Case A (L) sprinkler array in-plane with jet fan centreline, and Case B (R) sprinkler array out of plane from jet fan centreline.

P.A. (Tony) Enright / Case Studies in Fire Safety 1 (2014) 17

The soot yield for petrol of 0.058 kg/kg is also representative of polyethylene 0.060 kg/kg and polypropylene 0.060 kg/kg
[7]. The visibility factor of 8 is used since exit signage is light emitting rather than light reecting which is the default factor
of 3 [4].
Having selected the model domain and design re the grid resolution and source Froude number were checked. For brevity these studies are not included here other than to say;
 Cell sizes between 100 and 400 mm are within the range to adequately resolve plume dynamics with 100 mm selected to
provide the nest grid.
 The source Froude number was within the appropriate range to be representative of a buoyant re.

Initial results
FDS was used to calculate the predicted sprinkler link temperature. The results of the Case A and Case B arrangements are
shown in Fig. 2. For clarity, only the rst sprinkler in the array to activate in included.

Fig. 2. Effect of jet fan on sprinkler activation, (L) Case A and (R) Case B.

Fig. 3. Temperature vectors Case B1 with jet fan off at 165 s (activation).

P.A. (Tony) Enright / Case Studies in Fire Safety 1 (2014) 17

Fig. 4. Temperature vectors Case B2 with jet fan on at 165 s (activation).

 Case A: Sprinkler activation is delayed by 165 92 = 73 s. It is not the sprinkler immediately above the re that activates
rst, but rather the one downstream.
 Case B: The sprinkler temperature lags that of the non-operating case as in the early stages of re growth, but as the HRR
increases that lag decreases to the point that there is a negligible activation time delay.
Animations may also be viewed on YouTube at http://youtu.be/NRd_R2W0OkQ, and http://youtu.be/zhvdVUFP2iw. In
these, it can be observed that in the early stages of re growth, where the HRR is low, the re plume leans sharply towards
the downstream in response to the effect of the jet fan ow. This straightens as the HRR increase and at about 1000 kW the
re plume has sufcient buoyancy and momentum to penetrate the jet stream. The Case B results are also shown as temperature vectors in the following Figs. 3 and 4 at sprinkler activation.

Sensitivity cases
Additional analyses were undertaken that tested the sensitivity of sprinkler activation to varying factors. In each Case a
pair of analyses are undertaken, i.e.; with (2) and without (1) the jet fans running. The cases are summarised in Table 1.

Table 1
Summary of scenarios.
Scenario

Jet fan

Description

Case A

1 = Off
2 = On
1 = Off
2 = On
1 = Off
2 = On
1 = Off
2 = On
1 = Off
2 = On
1 = Off
2 = On
1 = Off
2 = On
1 = Off
2 = On
1 = Off
2 = On

Sprinkler array in-plane with jet fan centreline (the so called bad arrangement)

Case B
Case BA
Case C
Case CA
Case D
Case E
Case FA
Case G

Sprinkler array out of plane from jet fan centreline (the so called good arrangement)
Case BA: As per Case B above with the affect of an angled (inclined) air ow modelled
As per Case B but with a shielded re source
As per Case C with the affect of an angled (inclined) air ow modelled
As per Case B but with a varied sprinkler height
As per Case B but with a varied growth rate
As per Case B but the sprinkler array offset forward by 1 m relative to the nozzle and with inclined air ow modelled
A nal sensitivity case not to determine sprinkler delay but to determine the disruption to visibility of the jet fans up

P.A. (Tony) Enright / Case Studies in Fire Safety 1 (2014) 17

Fig. 5. Matrix of results.

Fig. 6. Collated results, HRR versus time to sprinkler activation.

P.A. (Tony) Enright / Case Studies in Fire Safety 1 (2014) 17

Collation of results
The results in terms of time and time-differential to a link temperature of 68 C are given in Fig. 5 below. The important
result is the time-differential recorded in the right-most column. These results are also expressed graphically in Fig. 6.
Problems associated with the delay in Case A can be eliminated by specifying and installing the sprinkler array offset from
the centreline velocity. Case E with the medium growth rate is included for illustrative purposes. In the companion egress
calculations the ASET assumed a fast t2 re. The delay in sprinkler activation is longer for a medium t2 re but this also results in a longer ASET.
Cases B, BA, C, CA, D and F are of the most interest. All adopt the optimum sprinkler array arrangement. These all result in
a <30 s delay. Note also, that the HRR is less than the 4 MW steady-state re assumed.
An additional Case G was run to determine the impact on visibility downstream of the fan. This is not related to sprinkler
activation but included for completeness. The model was run to 240 s because sprinkler activation occurs at 221 s in Scenario
CA2 it is assumed the fans will shut down at about this time. Contrary to worsening the situation, the jet fans improve visibility by diluting the smoke density.
Conclusions
Fires in car parks are rare events. According to UK research of >3000 car park res only about half originate in cars. Following an experimental program it was found that:
 Fires in engine compartments will grow slowly.
 Where a re starts in a car which is poorly-ventilated (i.e. closed windows) the re will most probably go out.
 Where a re starts in a car which is well-ventilated (i.e. open windows), very fast growing and severe res have resulted,
leading readily (when sprinklers are not present) to re spread to all nearby cars.
 The effectiveness of sprinklers in limiting a re to a single car has been demonstrated.
Ventilation of car parks using jet fans to reduce pollutants is becoming an increasingly popular alternative to fully-ducted
systems.
 Jet fans are considered to have limitations as means of smoke control. This is because their effectiveness is limited without side walls.
 Jet fans can however, aid post-event in smoke clearance under the control of the attending re brigade.
The convention is for jet fans to stop in re mode. The problems are how and when this is achieved. The generation of
carbon monoxide in the early stages of re will cause the jet fans to increase their speed.
 This causes disruption of the smoke signature by dilution of re products and makes shut down based on a high concentration of carbon dioxide unreliable.
 Similarly, the increased air ow can disrupt the heat signature by disrupting the re plume and ceiling jet. This could
delay sprinkler activation.
In order to quantify the potential delay to sprinkler activation caused by jet fans continuing to operate sixteen CFD analyses are undertaken to determine whether (i) it is signicant and (ii) it detrimentally impacts upon the means of escape in
the car park undertaken in a separate companion analysis (not presented here for brevity). The results of the analyses indicate delays of 630 s to sprinkler activation where the sprinkler and jet fans layout was coordinated so the sprinklers are inplane with the jet fan nozzle. This does not signicantly increase the hazard to occupants with the safety margin in the
egress design being many times greater than this. Furthermore, the jet fans do not adversely affect the visibility in means
of escape.
References
[1] BRE. BD2552 re spread in car parks; 2010.
[2] HMSO. Approved document B (re safety) volume 2 buildings other than dwelling houses. 2006 ed. Her Majestys Stationery Ofce; 2010.
[3] McGrattan K, Miles S. Modelling enclosure res using computational uid dynamics (CFD). SFPE handbook of re protection engineering. fourth ed.
National Fire Protection Association; 2008.
[4] McGrattan K. et al. Fire dynamics simulator (version 5) users guide NIST special publication 1019-5. s.l. Gaithersburg, MD: National Institute of
Standards and Technology; 2009.
[5] Thunderhead. Pyrosim: a model construction tool for re dynamics simulator (FDS) Pyrosim users manual. s.l.: Thunderhead Engineering; 2010.
www.thunderheadeng.com.
[6] Forney G. Users guide for smokeview version 5 A tool for visualising re dynamics simulation data. NIST special publication 1017-1. s.l.
Gaithersburg, MD: National Institute of Standards and Technology; 2009.
[7] Tewarson A. Generation of heat and chemical compounds in re. SFPE handbook of re protection engineering. fourth ed. National Fire Protection
Association; 2008.