See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/259705306

Computational Fluid Dynamics Study of Air Flow Characteristics in the

Columbus Module

Article  in  SAE (Society of Automotive Engineers) Transactions · July 2004

DOI: 10.4271/2004-01-2500

CITATIONS READS
16 489

5 authors, including:

Evgueni Smirnov Nikolay G Ivanov

Saint Petersburg State University Peter the Great St.Petersburg Polytechnic University
183 PUBLICATIONS   781 CITATIONS    114 PUBLICATIONS   451 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Numerical modeling of multiphase polydisperse flows with account of bubble nucleation View project

mass-forces effects on internal flows and heat transfer View project

All content following this page was uploaded by Nikolay G Ivanov on 13 October 2016.

The user has requested enhancement of the downloaded file.

 2004-01-2500

Computational Fluid Dynamics Study of Air Flow

Characteristics in the Columbus Module
Evgueni M. Smirnov, Nikolay G. Ivanov, and Denis S. Telnov
St. Petersburg State Polytechnic University, Russia

Chang H. Son and Valery K. Aksamentov

The Boeing Company
Copyright © 2004 SAE International

ABSTRACT predictable CFD ventilation model for the Columbus

module.
Ventilation characteristics of the Columbus module are
numerically predicted on the basis of the Reynolds- The goal of this study is to develop an advanced CFD
Averaged Navier-Stokes (RANS) and Large Eddy model for Columbus ventilation analysis. The present
Simulation (LES) approaches. Effects of air supply study has been performed with the FLUENT 6.1 CFD
diffuser modeling on computed flow are analyzed. An software. Data computed is compared with the
“effective diffuser” model that considerably reduces the Columbus experimental data [1]. Special attention is
number of computational cells for Columbus CFD paid to correlation of CFD data and the diffuser
ventilation analysis is proposed and tested. The simulation technique.
computational models are verified by a comparison with
the experimental data available. Special attention is paid COMPUTATIONAL MODELS
to distinctions in fields of the time-averaged absolute
velocity magnitude and the whole mean velocity that are The computational domain for Columbus ventilation CFD
due to the contribution of large-scale fluctuations. A analysis is illustrated in Figure 1. The domain is a
technique to evaluate spatial distribution of the time- rectangular-shaped volume with eight inlets (cabin air
averaged absolute velocity magnitude using data of supply diffusers) on the top and an outlet on the port face
RANS steady-state predictions is suggested. 3 –5
wall. The air (ρ = 1.225 kg/m , µ = 1.8×10 kg/m s) is
assumed to be an incompressible fluid.
INTRODUCTION
The present contribution covers results obtained for four
The Columbus module is a pressurized laboratory, computational cases. The main features of the problem
permanently attached to International Space Station formulations for the cases considered are summarized in
(ISS), that allows astronauts to work in a comfortable Table 1. Three cases, Cases 1 – 3, were computed with
environment. The module has a pressurized crew steady-state formulations, while a time-dependent
accessible area, which is approximately 6.1 meter long, solution was obtained in Case 4.
attached to the Node 2 port side. Revitalized air from
Node 2 is supplied into the crew cabin for ventilation by For Case 1 the uniform slit distribution with the maximum
the Columbus cabin fan, and then the circulated air aperture width, shown in Figure 2a, was adopted. It
returns to Node 2 for carbon-dioxide removal. corresponded to the uniform velocity distribution for each
slit (V = 1.47 m/s) with velocities directed normal to the
The Columbus intra-module ventilation system includes inlet boundary.
eight air supply diffusers that provide the total supply flow
3
rate of 408 m /h, and one large return grille located near An accurate resolution of jets issued from each slit and
the hatch. Four aft diffusers and four forward diffusers their interaction in the vicinity of the inlet sections
are placed in staggered rows in the overhead standoff requires a large number of grid cells that is unacceptable
faces stretched in the longitudinal direction. Each for an integrated ISS ventilation model due to its size.
diffuser consists of 12 slits, and each slit’s aperture width Consequently, the second computational case, Case 2,
and direction of flow can be adjusted. The return grille is was modeled in order to test an “effective diffuser” model
placed in the lower part of the face wall of the module (to with the rectangular jet formulation for each inlet section
the port direction). Presence of numerous inlet/outlet that substitutes multiple jets issued from 12 slits (see
boundaries produces certain difficulties for creation of a Figure 2). The width of the equivalent rectangular inlet
was set to 76 mm, with a corresponding decrease in the
inlet velocity magnitude. This value of the width of the the model constant does not significantly influence the
effective rectangular jet was chosen on the basis of a averaged flow field.
special CFD study concerning development of multiple
jets. For Case 4 the Large Eddy Simulation (LES) technique
was applied. In contrast to the RANS approach with all
For Cases 3 and 4 the jets issued from the 12 slits per the scales of the turbulence being modeled, in LES,
diffuser were resolved directly as in Case 1, but the slit dealing with the filtered Navier-Stokes equations, large
aperture and flow direction adopted in the Daimler- eddies are resolved directly in time-dependent
Chrysler Aerospace experiments and described in [1] computations, while small eddies are modeled using a
were used. The values of aperture width were within the subgrid-scale (SGS) model. The Smagorinsky-Lilly SGS
range of 2.2 to 8.0 mm. The velocity magnitude of 2.28 model [3] was used to define the subgrid-scale turbulent
m/s was set as the common boundary value, but the flow viscosity, νSGS,
angle was varied from slit to slit and from diffuser to
diffuser.
ν SGS = (L S ) 2SijSij
2
( ) 1
2 , (2)
For all cases the velocity magnitude of -0.91 m/s was set
for the return grille that corresponds to the flow rate of where Sij is the rate-of-strain tensor for the resolved
3
399.5 m /h leaving the Columbus model flow domain motion, and Ls is the subgrid-scale mixing length given
through the return grille. The remaining flow rate of by
3
8.5 m /h left the Columbus interior through the
Columbus/Node 2 hatch surface, where the pressure {
L S = min k ⋅ d; CS ⋅ V 1/ 3 , } (3)
outlet boundary condition was assumed. The no-slip
boundary condition was imposed on all of the solid
Columbus surfaces.

Two approaches for turbulence modeling were used. For D7 4 Aft Diffusers
Cases 1 – 3 the Reynolds-Averaged Navier-Stokes
D5
(RANS) approach was employed. The standard k-ε
turbulence model with the standard wall functions [2] was D3
used for computations. In this model the turbulent D1
viscosity, νtur, is defined as

k2
ν tur = C µ , (1)
ε
Z
where k is the turbulence kinetic energy, ε is its rate of
X
dissipation, and Cµ is the model constant. Since for
Columbus it was impossible to obtain a fully-converged Y

steady-state solution with the default formulation of the 4 Forward Diffusers

standard k-ε model suggesting Cµ = 0.09, in the present (D2, D4, D6, D8)
Return grille
computations the value of Cµ was increased up to 0.12.
Some test computations in ventilation configurations for
Figure 1. View of the Columbus computational grid of
which steady-state solutions could be obtained with the 1,732,904 cells
standard Cµ-value as well, showed that such increase in

Table 1. Problem formulation features of the cases computed

Grid, Turbulence Model Formulation Diffuser Inlet Velocity Inlet

number of cells modeling TVR
Case 1 988,566 RANS, standard k-ε model Steady 12 slits 1.47 m/s 2
Case 2 433,134 RANS, standard k-ε model Steady Effective 0.78 m/s 40
Case 3 1,732,904 RANS, standard k-ε model Steady 12 slits Experimental 2
Case 4 1,732,904 LES, Smagorinsky-Lilly SGS model Unsteady 12 slits Experimental –
10 mm 80 mm 40 mm (a) RESULTS AND DISCUSSION
(c)
20 mm
Diffuser VELOCITY FIELD ANALYSIS – Figure 3 illustrates air
10 mm velocity distributions over a vertical section cutting the
10 mm return grille, Y = 0.56 m (the point of origin for the
Columbus model in the current computations is placed in
Slit
76 mm 240 mm
the central point of the X-section passing through the
Z return grille). Results for Cases 1 – 3 computed with the
RANS approach are presented. A comparison between
Y X Figures 3a and 3b shows that the application of the
effective rectangular diffuser model does not change the
(b) airflow significantly. The high-velocity regions
corresponding to the jets issued from the inlet diffusers
Figure 2. Columbus cabin air supply diffuser modeling: (a) are located at the same place and have similar areas.
uniform distribution of slits with the maximum aperture width On the contrary, switching to the experimental boundary
used for Case 1, (b) effective rectangular diffuser used for conditions resulted in considerable changes in the
Case 2, and (c) computational grid in the vicinity of the diffuser
used for Cases 3 and 4
velocity field (Figure 3c).

(a)

where k is the von Karman constant, d is the distance to

the closest wall, V is the cell volume, and the
Smagorinsky constant, CS, is taken as 0.1 (the default ft/min
value in FLUENT 6.1). The no-slip condition for wall
surface was satisfied using the wall-function approach.
(b)
The inlet turbulence intensity was taken as 10% for all
inlet boundaries while the inlet ratio of the turbulent to
molecular viscosity (TVR), νtur/ν, was set as 2 for Cases
1 and 3 and as 40 for Case 2. These values ensure that
the inlet-jet effective Reynolds numbers,
Reeff = VinH/(ν+νtur), are within the range from 200 to 300.
Here Vin is the inlet velocity value, and H is the inlet (c)
length scale (the diffuser width).

The present computations were performed using three

grids. The first grid was generated for Case 1 and
consisted of 988,566 cells. The “effective diffuser”
model implementation allowed a reduction in the number
of cells for Columbus CFD ventilation analysis, and the
grid for Case 2 consisted of 433,134 cells. The refined Figure 3. Air velocity magnitude contours at cross-section
grid of more than 1.7 million cells was generated for Y = 0.56 m: (a) – (c) Cases 1 – 3 respectively
Cases 3 and 4 by means of three grid adaptations in the
vicinity of the inlet sections. The global view of the last
grid is given in Figure 1, and a grid cross-section, which
cuts the diffuser in the mid-line, is shown in Figure 2c.
All meshes were created using the GAMBIT 2.0.4 For the same Y-section, computational results obtained
generator. for Case 4 are presented in Figure 4. It is visible that the
LES technique allows more comprehensive airflow data
A steady or unsteady segregated solver with the second- containing both actual and time-averaged fields to be
order upwind spatial discretization and SIMPLEC obtained. The typical actual air velocity field presented in
pressure-velocity coupling was used for computations Figure 4a demonstrates a rich pattern of eddy motion
[4]. The second order time discretization was used for that noticeably differs from the RANS results (Figure 3c)
the unsteady computations in Case 4 starting from the and from the time-averaged LES data (Figure 4b,c). The
steady-state velocity field obtained for Case 3. A time- difference between the two approaches to time-average
step of 0.05 seconds was chosen. The length of the used in the present paper requires a detailed
sample attributed to a statistically developed regime was explanation.
about 500 seconds, and statistics of the mean flow
quantities was computed for over 350 seconds. Firstly, the absolute magnitude of the local mean
velocity, abs<V> ≡ Vm, was computed, and the results
are presented in Figure 4b. From here on and after, < > experimental inlet velocity distribution: for Case 3 the air
denotes time averaging. Secondly, the time-averaged from the top slit is directed upward, while the jets issued
value of the local velocity magnitude, <abs V> ≡ Va, was from two bottom slits move in downward direction. The
computed, and the contour field of this quantity is given time-averaged LES Vm-distribution (not shown) does not
in Figure 4c. As seen in Figure 4, the difference differ from the RANS results computed with the same
between Va and Vm is very significant. CFD solutions velocity boundary conditions (Case 3, Figure 5c) while
based on the LES approach directly produce data LES results for actual velocity shown in Figure 5d
sufficient for statistical evaluation of both Va and Vm. On demonstrate a complex flow structure with a number of
the contrary, CFD solutions based on the RANS vortices of various scales.
approach produce a distribution of Vm, but not Va. Note
that the contour plots given in Figures 3c and 4b for Vm-
distributions produced by two approaches look similar. It
should be stressed that estimations of LES time- (a) (b)
averaged data sample dependency allows conclusion
that the sample computed is enough for statistical
analysis. Therefore, some distinctions in the RANS and
LES Vm-fields should be attributed to the RANS ft/min
turbulence model features and to the “under-resolved”
nature of LES.

(a) (c) (d)

ft/min

(b)

Figure 5. Air velocity magnitude contours at cross-section

X = 3.4 m: (a) – (d) Cases 1 – 4 respectively

(c)

LOCAL VELOCITY STATISTICS – Figure 6 shows the

computed velocity magnitude and velocity components
fluctuations at a monitoring point placed in the Columbus
interior (point P1 with the coordinates x = 1.27 m,
y = -0.36 m, z = 0 is located in the section crossing
Figure 4. Results of computations for Case 4 at cross-section diffuser D3 over its middle line). The time behavior
Y = 0.56 m: (a) snapshot of instantaneous velocity magnitude
related to the statistically developed self-oscillating
field, (b) distribution of the mean velocity magnitude, Vm, and
(c) time-averaged velocity magnitude, Va regime is presented (only the last 200 seconds are given
at velocity components plots). The form of fluctuations
clearly indicates the chaotic behavior of airflow. As seen
from Figure 6, the amplitude of the fluctuations is
comparable or even higher than the time-averaged value
of the local velocity magnitude equal to Va = 0.152 m/s
Figure 5 gives additional information on the air motion in
(Figure 6a). The absolute magnitude of the local mean
the Columbus interior. The RANS results for Cases 1–3
velocity,
in a vertical X-section which intersects aft diffuser D7 in
the middle are shown in plots 5a, b, and c, while for Case
4 a snapshot of the actual velocity field at the same Vm = < u > 2 + < v > 2 + < w > 2 , (4)
section is presented. As before, the RANS results
computed with an accurate slit resolution (Case 1, Figure is equal to 0.083 m/s; that is almost twice less than Va.
5a) and the results from the effective area diffuser model This is in accordance with the above established
(Case 2, Figure 5b) are similar. The visible difference in difference in the averaged velocity fields shown in Figure
air velocity magnitude contours computed for Cases 2 4.
and 3 is due to peculiarities of the current diffuser
Velocity magnitude, m/s 0.4 (a) (b) (c)

(a) 11 12 13 14 15

0.3 21 22 23 24 25

31 32 33 34 35
0.2
41 42 43 44 45

0.1 51 52 53 54 55

0.0
200 300 400 500 600
0.2 Figure 7. Results of computations for Case 4 at cross-section
X = 2.7 m: (a) snapshot of instantaneous velocity magnitude
X Velocity, m/s

field, (b) distribution of the mean velocity magnitude, Vm, and

(c) time-averaged velocity magnitude, Va
0.0

(b) <Vx> = - 0.006 m/s

-0.2
400 440 480 520 560 600
0.3
Table 2. Results of computations for Case 4: Vm-values, m/s,
(c)
Y Velocity, m/s

0.2 at 25 points ordered as shown in Figure 7

0.1 0.047 0.029 0.055 0.039 0.041
0.0
0.046 0.037 0.018 0.034 0.068
-0.1
<Vy> = 0.027 m/s 0.065 0.046 0.076 0.053 0.071
-0.2
400 440 480 520 560 600 0.044 0.038 0.078 0.055 0.061
0.2
0.025 0.058 0.039 0.030 0.060
Z Velocity, m/s

0.0
Table 3. Same as Table 2 but with Va-values, m/s
-0.2 0.078 0.072 0.088 0.080 0.084
(d) <Vz> = - 0.078 m/s
0.090 0.077 0.070 0.077 0.112
-0.4
400 440 480 520 560 600 0.118 0.096 0.106 0.114 0.136
Flow Time, seconds
0.118 0.117 0.118 0.127 0.128
Figure 6. Velocity evolution at monitoring point P1 computed
for Case 4: (a) velocity magnitude and (b) – (d) velocity
0.106 0.134 0.110 0.123 0.105
components
Table 4. Experimental air velocities, m/s, measured
at 25 points ordered as shown in Figure 7 [1]

0.067 0.087 0.101 0.100 0.107

The Vm and Va plots for the section that crosses the
Columbus interior between diffusers D5 and D6, 0.075 0.061 0.058 0.069 0.127
X = 2.7 m, are given in Figure 7. Statistics of the mean 0.090 0.066 0.071 0.086 0.124
flow quantities gathered during the LES computations at
25 points located in this section are presented in Tables 0.087 0.100 0.129 0.155 0.116
2 and 3, the points are shown in Figure 7. The point 0.081 0.099 0.119 0.146 0.117
numbers (see Figure 7c) correspond to the table rows
and columns order. The experimental data [1] is
represented in Table 4 for the same points. The Vm-
values are significantly lower than the experimental CABIN AIR VELOCITY DISTRIBUTION – To examine
velocities, while the Va-values are comparable with the the ISS cabin velocity requirements, the air velocity
experimental data. The reason is that the anemometer magnitudes in the fluid zone were separated into 5
used in the Columbus velocity measurements produced groups: below 0.036 m/s, 0.036 – 0.076 m/s, 0.076 –
data for the time-averaged value of the local velocity 0.203 m/s, 0.203 – 1.016 m/s and above 1.016 m/s. The
magnitude, Va, but not for the absolute magnitude of the flow fields for all the cases computed were processed in
local mean velocity, Vm. Overall the difference between accordance with these velocity ranges and the results
the computed local Va-values and experimental data are summarized in Table 5 in comparison with the
does not exceed 15% for most of the points. experiments (there were no air velocities in the
computational and experimental data exceeding 1.016 could be attributed to the fact that the relatively low-
m/s, and the corresponding column is excluded from the velocity region in the vicinity of Node2/Columbus hatch
table). For Cases 3 and 4, results produced by two was taken into account. However, even in this case
different post-processing procedures are presented in more than 67% of the air velocity values are within the
Table 5. The first post-processing was performed for the range of 0.076 to 0.203 m/s.
whole Columbus interior, just as for Cases 1 and 2. The
second procedure was applied to the Columbus sub- The time-averaged value of the local velocity magnitude,
volume containing the experimental points while the Va, is the quantity that should be compared with the data
regions in the vicinity of Node2/Columbus hatch and the measured by probes that are not sensitive to the flow
Columbus back wall were excluded. It should be noted direction. The air ventilation requirements fulfillment
that, in accordance with the ISS ventilation requirements, analysis should be performed on the basis of Va-data as
only fluid zones 0.15 m away from the Columbus walls well. The problem is how to transform the RANS-
were under consideration. computed distribution of Vm to the Va-field. This problem
can be approximately solved if one uses a turbulence
First, consider the RANS Vm-results for Cases 1 – 3 that model based on the transport equation for the turbulent
were directly extracted from the computed steady-state kinetic energy, k. It is just the case when the k-ε model
velocity field distributions. In order to meet the is used. Some theoretical analysis given in the Appendix
ventilation requirements at least 67% of the air velocity A1 introduces a correction to the RANS-computed
values computed in the habitable area have to be absolute magnitude of the local mean velocity to account
between 0.076 and 0.203 m/s. This is not the case for for a large-scale fluctuations contribution. Taking into
Vm-fields computed by RANS: most of the air in the account these contributions of intensive large-scale
Columbus interior has lower velocities, and only about fluctuations, the time-averaged absolute velocity
30% of the air velocity values are within the range magnitude, Va, can be evaluated as
mentioned above. A further requirement states the
velocity shall not be below 0.036 m/s. However, RANS
5
results for Vm-distribution give us great portions of Va = Vm2 + k. (5)
stagnant air moving with velocities less than 0.036 m/s. 3
For Case 3 the disagreement between the computed Vm-
distribution and the experimental data is very large. where Vm is the velocity given by a RANS solution, and k
is the computed turbulent kinetic energy. Results of
The processing of Vm-data for Case 4 resulting from applying formula (5) to the present RANS-computations
statistical evaluation of LES-computed actual velocity of Columbus ventilation are given in Table 5. The RANS
fields gives the distribution with the same shift to lower Va-distributions computed for Cases 1 and 2 are shifted
velocities. On the contrary, the processing of Va-data towards the higher velocity magnitude bands as
give us remarkably better results. The LES-generated compared with the Vm-data. These distributions are
Va-distribution for the Columbus sub volume containing close to each other and almost match the air ventilation
experimental points is in excellent agreement with the requirements. The Va-data computed for Case 3
experimental data. The Va-distribution computed for completely satisfies the air ventilation requirements and,
Case 4 for the whole Columbus interior gives a slight moreover, agree well with the LES computational results
shift towards the lower velocity magnitude bands that and the experimental data.

Table 5. Computed Columbus cabin air velocity distribution in comparison with the measurements

Velocity Magnitude (m/s)

Below 0.036 0.036 – 0.076 0.076 – 0.203 0.203 – 1.016
Vm Va Vm Va Vm Va Vm Va
Case 1, whole interior 19.3% 3.0% 49.3% 26.4% 27.6% 62.3% 3.8% 8.3%
Case 2, whole interior 21.9% 3.4% 48.2% 32.2% 26.1% 58.0% 3.8% 6.4%
Case 3, whole interior 17.6% 1.0% 45.4% 17.6% 32.6% 72.9% 4.4% 8.5%
Case 3, exp. points 13.7% 0.0% 44.3% 6.3% 37.5% 82.4% 4.5% 11.3%
Case 4, whole interior 13.5% 0.9% 46.5% 20.7% 36.0% 72.2% 4.0% 6.2%
Case 4, exp. points 7.1% 0.0% 51.0% 13.2% 35.4% 78.2% 6.5% 8.6%
Experimental Data [1] 0.0% 13.5% 79.7% 6.8%
CONCLUSION 3. J. Smagorinsky, General circulation experiments
with the primitive equations. I. The basic
Ventilation characteristics of the Columbus module were experiments, Mon. Weather Rev., vol. 91, pp. 99-
numerically predicted on the basis of the RANS and the 164, 1963.
LES approaches using the CFD software FLUENT 6.1. 4. Fluent 6.1 User’s Guide, Fluent Inc, 2003.

Special attention was paid to the analysis of the influence

of inlet diffuser specification on computed airflow. An CONTACT
“effective diffuser” model with the rectangular jet
formulation for each inlet section, substituting multiple Evgueni M. Smirnov, DSc
jets issued from 12 slits, was suggested and tested. The Professor, Department of Aerodynamics
“effective diffuser” model application allows prediction of St. Petersburg State Polytechnic University, Russia
the ventilation characteristics with almost the same Phone: 7-812-552-6621
quality as the model directly resolving 12 inlet jets per E-mail: [email protected]
diffuser, and reduces computational cost noticeably. The
rectangular jet formulation for the Columbus diffusers is Chang H. Son, Ph.D.
attractive for a large CFD model, such as an integrated Principal Engineer
ISS model. The Boeing Company
Phone: 281-226-6269
The present LES computations have reproduced the E-mail: [email protected]
available experimental data very well with respect to the
integral Columbus cabin air velocity distribution. Local
values of time-averaged distribution of the velocity APPENDIX A1
magnitude were properly predicted as well.
Formula (5) was derived on the basis of the following
Direct processing of the RANS data for the absolute theoretical considerations.
magnitude of the local mean velocity leads to a
significant decrease in velocity values as compared with Using a local Cartesian coordinate system with the x-axis
the experimental data obtained with probes non-sensitive co-directed with the Vm-vector, the time-averaged value
to the flow direction. A technique to introduce a large- of the velocity magnitude, Va, is written as follows
scale velocity fluctuation contribution into the RANS-
computed results was suggested. The corrected RANS
velocity distributions agree well with the measurements. Va = (Vm + u')2 + v'2 + w '2 , (A1)

With the slit aperture width and inlet flow direction varied
from diffuser to diffuser as in the experiments, the LES where u′, v′ and w′ are velocity fluctuations in x, y and z
and the conversed RANS data show that approximately directions.
80% of the velocity magnitudes are in the 0.076 to 0.203
m/s range, i.e. satisfy the air ventilation requirements. It The actual kinetic energy of fluctuations is written as
is in a full accordance with the measurement data. The
alternative inlet diffuser configuration considered (the u' 2 + v '2 + w ' 2
k' = (A2)
uniform slits distribution with the maximum aperture 2
width) gives a slight shift towards the lower velocity
magnitude bands. Let’s consider the simplest model of motion with
fluctuations with the amplitude A and frequency ω
In summary, it is believed that the steady-state RANS
approach with the "effective" diffuser model and the data u' = v' = w ' = A sin ωt . (A3)
conversion suggested can serve as a reliable and robust
method for evaluation of air ventilation in ISS modules.
LES of airflow is rather time-consuming but there is a The time-averaged value of k′, denoted as k, is given by
strong motivation in using LES for deep understanding of
unsteady behavior of multiple-diffuser configurations 3A 2 T 2π
typical for air ventilation arrangement in the ISS.
k=
2 ∫0
sin 2 ωtdt , where T =
ω
, (A4)

REFERENCES and the relation between the amplitude A and the kinetic
energy k is written as
1. Columbus Cabin Ventilation Qualification Test
Report, COL-DOR-TR-3002, Daimler-Chrysler 4
Aerospace. February 14, 2000. A2 = k. (A5)
3
2. B. E. Launder and D. B. Spalding. Lectures in
Mathematical Models of Turbulence. Academic Substitution of (A3) in formula (A1) gives
Press, London, England, 1972.
 Let’s consider three cases.
Va = Vm2 + 3 A 2 sin 2 ωt + 2Vm A sin ωt . (A6)
1. A << Vm (ξ → 0), i.e. fluctuations are much lower
2
The value of sin ωt, being always positive, may be than the averaged velocity. In this case Va ≈ Vm2 .
replaced approximately by the averaged one, equal to
1/2, so that 2. A >> Vm (ξ → ∞), i.e. fluctuations are much higher
than the averaged velocity, then Va ≈ Vm2 + 2k .
3 2 3. The case of the minimum of function F(ξ).
Va ≈ Vm2 + A ⋅ F(ξ ) , (A7)
2 Accurately, this minimum, equal to 0.949 ≈ 0.95, is
observed at ξ = 0.82 (see Figure 8). For ξ = 1
here, (A = Vm) one obtains F = 0.951 ≈ 0.95. So, for the
case when the averaged velocity and fluctuations
have the same order (this case is of special interest
2ξ A 3 2
F(ξ) = 1+ sin ωt , ξ= . (A8) for ISS ventilation analysis) Va = Vm2 + A ⋅ 0.95 .
3ξ 2 Vm 2
1+ The same value of Va at A = Vm is obtained when using
2

The plot of function F(ξ) is given in Figure 8. Va = Vm2 + αA 2 , (A9)

where α = 1.256 ≈ 5/4.

1.00
Substituting (A5) in relation (A9), one obtains
F( ξ )

5
Va = Vm2 + k. (A10)
0.96 3

Formula (A10) corresponds to using the sum of the

0 2 4 6 8 ξ 10 mean flow kinetic energy and the turbulent kinetic energy
given by a RANS turbulence model and taken with a
factor of 5/6.
Figure 8. Function given by formula (A8)

View publication stats