Professional Documents
Culture Documents
Abstract
At present, Computational-Fluid-Dynamics (CFD) with the standard k-8 model is a popular method for numerical simulation of room
airflow. The k-c model needs a lot of computing time and large a computer. This paper proposes a new zero-equation model to simulate three-
dimensional distributions of air velocity, temperature, and contaminant concentrations in rooms. The method assumes turbulent viscosity to
be a function of length-scale and local mean velocity. The new model has been used to predict natural convection, forced convection, mixed
convection, and displacement ventilation in a room. The results agree reasonably with experimental data and the results obtained by the
standard k-c model. The zero-equation model uses much less computer memory and the computing speed is at least 10 times faster, compared
with the k-c model. The grid size can often be reduced so that the computing time needed for a three-dimensional case can be a few minutes
0naK. 0 1998 Elsevier Science S.A. All rights reserved.
1. Introduction usedto obtain empirical formulas for heat transfer. The heat
transfer on room enclosuresurfaceis muchmore complicated
Proper design of indoor environment requires detailed becauseit can be a combination of forced and natural con-
information of indoor air distribution, suchasairflow pattern, vection and jet flows. The empirical formulas often fail to
velocity, temperature, and contaminant concentrations.The calculate accurately heat transfer on the surfaces.Therefore,
information can be obtained by experimental measurements the temperaturedistributions of room air andheat transfer on
and computational simulations.Experimental measurements room enclosuresurface are important input parametersfor a
are reliable but need large labor-effort and time. Therefore, building energy program. A CFD program can calculate the
the experimental approachis not feasible asa generaldesign heat transfer.
tool. A popular approachof computational simulationsis the The CFD methods solve the Navier-Stokes equationsfor
computational-fluid-dynamics (CFD) method. However, flows. For laminar flows the computed results are accurate
popular modelsusedin the CFD method to calculate turbu- and reliable. However, it is difficult to predict turbulent flows.
lenceneed a fast computer with a large memory. Very fine numerical resolution is required to capture all the
On the other hand, mostbuilding energy analysisprograms details of the indoor turbulent flow. This type of simulation
assumeuniform distributions of room air temperature and is direct numerical simulation. The direct numerical simula-
calculate convective heat exchange coefficient by empirical tion for a practical flow needsa huge computer systemthat
formulas for simpleflows. The energy programscannot accu- is not available [ 11.
rately predict energy usedby HVAC systemsif there is tem-
Indoor airflow simulations use turbulence models in the
perature stratification in a space,suchasa room with radiant
CFD approach to compute the mean values. This type of
heating/cooling systems,with convective andradiative heat-
simulationcan be donewith the capacity and speedof present
ing systems,and with displacementventilation systems.In
computers. Eddy-viscosity modelsare the most popular tur-
addition, airflow on room enclosuresurfacesis not the same
bulence models.A typical example of the modelsis the stan-
as that in an infinite long heated or cooled surface asoften
dard k-.s model [ 21. Most of the modelsrequire solving one
* Corresponding author. Tel.: + 1 617 2537714; fax: + 1 617 2536152; or more additional differential equations.The computing cost
e-mail: qchen@mit.edu is still large at present.Most HVAC designersand architects
may not have a larger computer. Therefore, in predicting turning to more sophisticated turbulence models, such as two-
indoor air distribution, designing a comfortable indoor envi- equation models and Reynolds-Stress models, zero-equation
ronment, and analyzing building energy, application of the models still deserves some attentions since they are simple,
turbulence models is limited. cost-effective and once calibrated, can predict the mean-flow
Some other turbulence models, such as zero-equation tur- quantities fairly well.
bulence models, do not solve additional differential equa- However, none of the above zero-equation models are
tions. They can be used to compute turbulent flows without developed for room airflows, especially buoyancy-driven
extensive computing power. This paper presents a new zero- flows. It is therefore necessary to develop a zero-equation
equation model for room airflow prediction. model for room airllows.
The development of zero-equation turbulence models Many HVAC design engineers and architects do not have
dates back to 1925, when Prandtl [ 31 postulated the mixed- the access to a large computer. It is important to develop a
length hypothesis. In analogy to the molecular transport with simple model to simulate indoor airflow on a personal
fluid lump motions, the mixing-length model assumes a computer. The model can calculate airflow pattern and the
relationship: distributions of air velocity, temperature, contaminant con-
centrations, and pressure. The flow program should then be
coupled with an energy analysis program to simulate simul-
taneously airflow, thermal comfort, and energy consumption
where lmix is the mixing length, U is the main stream velocity of HVAC systems. The program will also allow the temper-
and y is the direction normal to the main flow. Although the ature of interior walls to be predicted. The program would
mixing-length model is not theoretically sound, it has yielded serve as a tool to accurately provide design information and
quite plausible results for some flows. For example, the mix- to properly size HVAC systems and assure comfort condi-
ing length model gives excellent agreement with measure- tions at all important locations within the space.
ment in free shear flows with proper calibration. The goal of the present investigation is to develop a simple
Many efforts had been made to extend the applicability of but reliable turbulence model that can provide design infor-
the mixing length model. van Driest [ 41 devised a damping mation to establish acceptable comfort conditions through
function to include the wall-damping effects. The van Driests the interior space. With the model, one can simulate indoor
damping function was formulated as: airflow with a minimum of training and modest desktop per-
sonal computers. Section 4 describes the new model.
lmi.=Ky[ l-e%] (2)
0 Momentum. 0 Speciesconcentrations.
For determination of pollutant or water vapor concentra-
apvi apvivj tion distribution the equation for conservation of massmust
at' axj be combined with the equation for the species.
apt apvjc- a
at+ axj
z Geff g +s, (10)
J( J 1
where p = air density; Vj = velocity component in xi-direc- where C = speciesconcentration; rc,eff = effective turbulent
tion; p = pressure; perf= effective viscosity; p = thermal diffusion coefficient for C; SC= sourceterm of C.
expansioncoefficient of air; To= temperatureof a reference Similar method to the energy equationis usedto determine
point; T= temperature; gi = gravity acceleration in i-direc- the effective diffusive coefficient for speciesconcentrationin
tion. Eq. ( 10) 9rc,eff:
The last term on the right sideof the equation is the buoy-
ancy term. Geff= 2 (11)
The turbulent influencesare lumped into the effective vis- eff
cosity as the sum of the turbulent viscosity, Pi, and laminar where effective Schmidt number, Sceff,is 1.O.
viscosity, p: Solving Eqs. (3)-( 11) except Eq. (6)) one could obtain
flow information.
CLeff=k+EL (5)
apT ap?T a aT
-=
y- + - - &ff E ++ (8) 0
ax, ax, ( ,1 P axi
where rrefi - effective turbulent diffusion coefficient for T, ac
z=O (13)
q = thermal source; C, = specific heat.
In our work we have estimatedthe effective diffusive coef-
ficient for temperaturein Eq. (8)) rTeff, by: where prerum is the pressureat a return and xi the coordinate
normal to the surface.
0 Symmetry surface.
G,,eff= pf If the xi coordinate is normal to the symmetry surface,the
eff
following equationsdescribethe boundary conditions of the
where the effective Prandtl number, Preff, is 0.9. surface:
140 Q. Chen, W. Xu / Energy and Buildings 28 (1998) 137-144
q=h(T,-T)
&=CSOUTCe (15)
where r=shear stress;h=convective heat transfer coeffi-
cient; C,,,,, = speciesconcentration source. Adiabaticfloor
7.9 m
The convective heat transfercoefficient is determinedfrom I I
the following equation, which is similar to the Reynolds Fig. 1. Sketch and boundary conditions of the natural convection case.
analogy:
h=&& (16)
Preff Ax,
-.+--+--+-,---a-.-+
,._-_ - _ - - - -v-e--
\....
5. Application examples
k-E model has been used extensively in the past and detailed _ x - h
information of the model is widely available, this paper will
not repeat all the information. The computed velocity profiles
Fig. 4. Sketch of the forced convection case.
are compared in Fig. 6 with experimental data in two vertical
sections x/H = 1 and x/H = 2 respectively and two horizontal
sections, y/H = 0.972 (through the air supply outlet) and yl
H = 0.028 (through the air exhaust inlet). The results of the
zero-equation model show a jet decay that is too strong.
Hence, the primary flow near the ceiling and the return flow ,._......4----/;
: 1
near the floor are smaller than the data. In this case, the k-g . .
.
.
.
.4---.-c&-/,
. -cc--t-cccc r ,
,
_ _ _ - . - - cccccccc /
model predicts a satisfactory result. Nevertheless, the zero-
equation model could predict the secondary recirculation on (a)
x/H=2
1.0 I
0-q mode4
_____...__--.. L-E model r::*
0.5 - . data .,.,
.I 0,
(a) (3)
ynGo.972
T 0.50 / , YM=!.o?8
O-eq. model
____
___________-._ &tie,
1 o-e.q.model f
--. .
.a
cj*; ~~~~~~
% . :----
w (4
Fig. 6. Comparison of velocity profiles in different sections of the room with forced convection: (a) at x/H = 1, (b) at x/H= 2, (c) at y/H =0.972, and (d)
at yl H = 0.028.
__---____--.
cFDModel Simplified Model
(b)
Fig. 7. Comparison of the airflow patterns for the mixed convection: (a) the
zero-equation model, (b) the standard k-e model.
Fig. 8. Comparison of the penetration length vs. Archimedes number for the
A grid number of 16 X 14 X 12 is minimum in order to rep- mom with mixed convection.
resent the room geometry, such as the inlet, outlets, window,
and table. Fig. 9 shows similar airflow patterns and the dis- the computed and measured results is reasonably good. The
tributions of air temperature and helium concentration com- results are nearly identical between the two models if the grid
puted by the zero-equation model with 16 X 14 X 12 grids and number is the same. This implies that the new zero-equation
the k-8 model with 3 1 X 28 X 26 grids. model is as good as the k-8 model for displacement ventila-
Fig. 10 further compares the computed results with the tion. However, the model performance depends on flow type.
experimental data. The velocity and temperature profiles are In a separated study to examine the performance of five dif-
at the center of the room and helium concentration profile at ferent k-E models, we found that a model may work better in
a line near the center of the room. The agreement between one type of flow but poorer in another type [ 231. Therefore,
Q. Chen, W. Xu /Energy and Buildings 28 (1998) 137-144 143
(c) 0-l
Fig. 9. Comparison of the airflow patterns and distribution of air temperature
(C) and helium concentration (%): (a), (b), and (c) the zero-equation
model and (d) , (e) , and ( f) the k-e model. Fig. 10. Comparison of the profiles of air velocity, air temperature, and
helium concentration in a vertical line of the room.
we cannot say that the zero-equation model is better than the
k-E model. The main advantage of the zero-equation model Table 1
is its simplicity and less computing time required, compared Comparison of computing performance of the zero-equation and k-e models
with the k-e model.
Case Model Grid number Core CPU
It is possible to use a minimum grid number of 6 X 7 X 6
memory time
with which the table in the room cannot be represented. The (bytes) (s)
accuracy of the results is relaxed but it does predict the main
features of displacement ventilation, such as temperature gra- Natural convection Zero-equation 20x 10 15,006 18
dient, non-uniform distribution of contaminant concentra- k-e 96X60 158,OcHl 3238
Forced convection Zero-equation 20x 18 25,000 9
tion, and higher risk of draft near the inlets at the floor level. k-e 50x45 177,000 593
The minimal grid number is less than that used in zonal Mixed convection Zero-equation 25X18 31,000 33
models. Therefore, the zero-equation model has a great poten- k-e 70x45 263,000 1438
Displacement ventilation Zero-equation 31 x28X26 555.000 5400
tial to be used in an hour-by-hour energy simulation program
16X14X12 75,000 311
to take into account the impact of non-uniform temperature 10x10x10 27,CQO 119
distribution on energy consumption. 6~7x6 9ooo 33
Note in all of the cases, the zero-equation model, Eq. (7), k-s 31X28X26 770,000 58,163
is exactly the same. No adjustable constants were used in the
computations. The zero-equation model is universal for room ing because the basis of the k-E model is more solid. The
airflow simulation. reason to use a zero-equation model is to reduce the comput-
ing time used by the k-E model.
6. Discussion Table 1 shows the total grid number used in the four cases
by the zero-equation model and the k-s models. It also shows
The results show that the k-E model may predict better the memory needed and CPU time used. The convergence
results than the new zero-equation model. This is not surpris- residuals, R, are the same for the zero-equation model and
144 Q. Chen, W. Xu /Energy and Buildings 28 (1998) 137-144
the k-E models. The present investigation uses R < 0.001 of References
the mass inflow, for mass continuity and R <O.Ol of energy
exchange for temperature. [ 1] F.T.M. Nieuwstadt, J.G.M. Eggles, R.J.A. Hanssen, M.B.J.M.
The computations were conducted on a 486 personal com- Pourquie, Direct and large-eddy simulations of turbulence in fluids,
Future Generation Comput. Syst. 10 (1994) 189-205.
puter. The zero-equation model uses much less memory than
[ 21 B.E. Launder, D.B. Spalding, The numerical computation of turbulent
the k-e model and is at least 10 times faster than the k-E flows, Comp. Meth. Appl. Mech. Energy 3 (1974) 269-289.
model. This is because the k-E model solves two more trans- [ 31 L. Prandtl, Uber die ausgebildete Turbulenz, ZAMM 5 ( 1925) 136-
port equations and the non-linear interaction in all the equa- 139.
tions makes it difficult to converge. The results show that [4] E.R. van Driest, On turbulent flow near a wall, J. Aeronaut. Sci. 23
(1956) 1007.
most room airflow simulation can be done with a personal
[5] F.H. Clauser, The turbulent boundary layer, Advances in Applied
computer and the computing time for each case is on the order Mechanics: IV, Academic Press, New York, 1956, pp. l-51.
of a few seconds for a two-dimensional problem and a few [6] M.P. Escudier, The distribution of mixing-length in turbulent flows
minutes for a three-dimensional case. near walls, Heat Transfer Section Report TWF/TN/ 12, Imperial Col-
lege, 1966.
[7] S. Corrsin, A.L. Kistlter, The free-stream boundaries of turbulent
flows, NACA TN 3133 (1954)
[ 81 P.S. Klebanoff, Characteristics of turbulence in a boundary layer with
7. Conclusions zero pressure gradient, NACA TN 3178 (1956)
[9] T. Cebeci, A.M.O. Smith, Analysis of turbulent boundary Layers,
This paper proposes a new zero-equation for the prediction Series in Applied Mathematics and Methods: XV, Academic Press,
of room airflow patterns and the distribution of air tempera- 1974.
[ 101 D.C. Wilcox, Turbulence modeling for CFD. DCW Industries, 1993,
ture and contaminant concentrations. The model is derived
p. 51.
from the Navier-Stokes equations. Using the concept of [ 111 B.S. Baldwin, H. Lomax, Thin-layer approximation and algebraic
eddy-viscosity, turbulent viscosity is approximated by a models for separated turbulent flows, AIAA Paper, Huntsville, AL,
length scale and mean velocity. The main difference between 1978, pp. 78-257.
the zero-equation model and the k-8 model is that the former [ 12 ] A.A. Ameri, A. Amone, Prediction of turbine blade pass heat transfer
using a zero and a two-equation turbulence model, ASME Proceedings
does not solve transport equations for turbulent quantities.
of the International Gas Turbine and Aeroengine Congress and Expo-
The zero-equation model determines turbulent quantities by sition, Paper 94-GT- 122, June 13-16, Hugus, Neth ASME, New York,
an algebraic equation. 1994, pp. l-8.
The study demonstrates the capability of the zero-equation [ 131 D.E. Nikitopoulos, E.E. Michaelides, Phenomenological model for
dispersed bubbly flow in pipes, AIChE J. 41 (1) (1995) 12-22.
model by applying it to predict the airflow with natural
[ 141 H. Liu, M. Ikehata, Computation of free surface waves around an
convection, forced convection, mixed convection, and dis-
arbitrary body by a Navier-Stokes solver usingpseudocompressibility
placement ventilation in rooms. The predicted results are technique, Int. J. Numerical Methods Fluids 19 (5) (1994) 395413.
compared with experimental data and the results with the [ 151 S.-W. Chyou, C.A. Sheicher, Can one- and two-equation models be
standard k-8 model. The zero-equation model can predict modified to calculate turbulent heat transfer with variable properties?,
Ind. Eng. Chem. Res. 31 (3) ( 1992) 756-759.
indoor airHow patterns and the distributions of air tempera-
[ 161 PHOENICS Reference Manual, CHAM, London, 1996.
ture and contaminant concentrations with reasonable [ 171 S.V. Patankar, Numerical Heat Transfer and Fluid Flow, Hemisphere
accuracy. Publishing, New York, 1980.
Since the zero-equation model does not solve transport [ 181 D.A. Olson, L.R. Glicksman, Transient natural convection in enclo-
equations for turbulence, the computer memory needed is sures at high Rayleigh number, ASME J. Heat Transfer 113 ( 1991)
635-642.
much smaller, and the convergence speed is 10 times faster
[ 191 C.K.G. Lam, K. Bremhorst, A modified form of the k-c model for
than that of the k-E model. With the zero-equation model, predicting wall turbulence, ASME J. Fluid Eng. 103 ( 1981) 456-460.
simulation of a three-dimensional, steady-state flow in a room [20] W. Xu, Q. Chen, Numerical simulation of air flow in a room with
can be made in a personal computer. differentially heated vertical walls, ASHRAE Trans. 104 ( 1) (1998)