Beruflich Dokumente
Kultur Dokumente
a r t i c l e i n f o a b s t r a c t
Article history: A sensitivity analysis of various numerical parameters on the prediction of thermal stratification in a
Received 18 December 2011 2D axi-symmetric domain has been carried out by using commercial code FLUENT 6.3. The analysis is
Received in revised form 5 April 2012 carried out for both steady and unsteady flow conditions. The Reynolds number in all the cases has been
Accepted 11 April 2012
considered as Re = 19,200. The effect of under relaxation parameters, numerical schemes on convergence
while solving mass, momentum and turbulence equations was first studied for steady state conditions.
An optimal range for under relaxation parameters was found which ensures the stability of the numerical
solutions and ensures faster convergence. Further, 1st order upwind scheme for convective terms of tur-
bulent equations (k and ε) is found to give a better convergence in comparison with higher order schemes.
The study was then extended to transient flow conditions and the stratification characteristics like rising
of the stratification interface was studied. The turbulence model parameters of the standard k–ε model
were changed and predictions were observed. It is observed that the values used for turbulence model
parameters (C , C1 , C2 ) while solving the standard k–ε model for studying stratification characteristics are
different than the model values used in k–ε equation. In the lower part of reactor vessel, k–ε model with
standard value of model parameters have a very good agreement with the experimental observations of
Moriya et al. (1987) while in the upper part the model parameters take different values than the standard
values.
© 2012 Elsevier B.V. All rights reserved.
0029-5493/$ – see front matter © 2012 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.nucengdes.2012.04.008
418 S.S. Das et al. / Nuclear Engineering and Design 250 (2012) 417–435
Nomenclature
k turbulent Prandtl number for turbulent kinetic
C1 constant for k–ε model energy
C2 constant for k–ε model ε turbulent Prandtl number for turbulent energy dis-
C constant for k–ε model sipation rate
CP specific heat at constant pressure (J/kg K) T turbulent Prandtl number for energy
D diameter of the inlet of the cylindrical hot plenum
(m)
g acceleration due to gravity (m/s2 ) cold coolant from the reactor core enters the lower region of the hot
G production term for turbulent kinetic energy upper plenum slowly due to density difference, and a large portion
H total water height (m) of the coolant in the upper plenum remains hot. This phenomenon
I turbulent intensity yields thermal stratified condition and produces stiff temperature
k turbulent kinetic energy (m2 /s2 ) gradient in the axial direction. This temperature gradient at the
K coefficient of thermal conductivity (W/mK) stratification interface is very stiff which causes severe thermal
L characteristic length (m) stresses to the inner vessel and other immersed components. More-
Niter number of iterations for convergence over, the sloshing of thermal interface can lead to thermal fatigue. It
p pressure (N/m2) causes thermal stresses and thermal fatigue to the immersed com-
Pe Peclet number LU ˛ ponents in the LMFBR’s. The understanding of the hydrodynamics
Pe* non dimensionalPeclet number and heat transfer during thermal stratification is extremely impor-
Cp
Pr Prandtl number K tant for better design of LMFBR’s, due to difficulties involved to
carry out the experiments in LMFBR. Alternatively, CFD is a pow-
r
radial coordinate (m)
Re Reynolds number UD erful tool for analyzing flow pattern and temperature distribution
c −h inside the fast breeder reactor’s (FBR). Further, the experimentation
gˇD
Ri Richardson number c of characteristics like the rising speed of the stratification interface
U2
S˚ source term for general scalar variable (˚) as well as the stratification height in actual conditions in full scaled
Su source term for radial velocity FBR’s is experimentally very challenging. With a good CFD model it
Sv source term for axial velocity is possible to accurately predict these characteristics in a full scale
ST source term for temperature version of FBR’s.
Sk source term for turbulent kinetic energy Fig. 2 depicts the schematic diagram of the test section of
Sε source term for turbulent energy dissipation rate experimental set-up (Moriya et al., 1987). This set-up is a simple
t time (s) representation of LMFBR. Initially, the hot water is injected to the
t* non-dimensional time reactor vessel through the core outlet (1 in Fig. 2). Similar to the
T temperature (K) description above, to obtain a thermally stratified condition, the
Tc temperature of the cold fluid (K) cold water is injected into the domain. Since, the density of cold
Th temperature of the hot fluid (K) fluid is larger than the hot fluid it tries to settle down at the lower
T0 reference temperature (K) part of the system. This yields a steep temperature gradient in axial
T* non-dimensional temperature direction in the reactor vessel. As a result a hot-cold interface was
u radial velocity (m/s) formed.
U characteristic velocity (m/s)
v axial velocity (m/s)
V resultant velocity (m/s)
z axial coordinate (m)
1
z* non-dimensional axial coordinate 7
9
Greek letters
eff effective diffusivity (m2 /s)
˛ thermal diffusivity (m2 /s)
˛k under relaxation parameter for turbulent kinetic 2
8
energy
˛p under relaxation parameter for pressure 3
˛u under relaxation parameter for velocities 6
˛ε under relaxation parameter for turbulent energy
dissipation rate 5
ˇ thermal expansion coefficient (K−1 )
ε turbulent energy dissipation rate (m2 /s3 )
generalized notation for transport variable
molecular viscosity (kg/ms)
t eddy viscosity (kg/ms)
t kinematic eddy viscosity (m2 /s)
density of the fluid (kg/m3 )
0 reference density of the fluid (kg/m3 ) 4
c density of the cold fluid (kg/m3 ) Fig. 1. Schematic diagram of LMFBR. (1) Control plug, (2) hot pool, (3) pump, (4)
h density of the hot fluid (kg/m3 ) active core, (5) IHX window at the bottom part of the reactor vessel, (6) cold pool,
(7) IHX window at the upper part of the reactor vessel, (8) the flow into the reactor
vessel, (9) the flow into the IHX window.
S.S. Das et al. / Nuclear Engineering and Design 250 (2012) 417–435 419
Table 1A
Details of the various range and proposed correlations available in the literatures.
Authors Fluid T, temperature Non-dimensional parameters and their range Proposed correlation
difference in ◦ C of v* with Ri
Re Ri Pe Pr
Table 1B
Measurement procedure of temperatures for thermal stratification as studied in literature.
Lorenz et al. (1977) Temperature Thermocouples Study the influence of scale size and fluid thermal properties
Aoki and Okada (1982) Temperature Thermocouples Calculate the velocity of thermal stratification interface
Tanaka et al. (1984) Visualization of the Dye injection Study the effect of elevation of upper international structure (UIS) and
flow field Thermocouples intermediate heat exchanger intake (IHX) on thermal stratification
Temperature
Moriya et al. (1987) Visualization of the Aluminium powder as a tracer Study the effect of Reynolds number and Richardson number on
flow field Thermocouples thermal stratification
Temperature
Vidil et al. (1988) Temperature Thermocouples Studied the basic mechanism which create and maintain thermal
stratification
Tanaka et al. (1990) Temperature Thermocouples Studied the applicability of small scale water test to predict thermal
stratification
Ieda et al. (1990) Temperature Thermocouples of 1 mm diameter. Examination of threshold of thermal stratification and rising rate of
stratification interface
Ravichandran et al. (1995) Temperature Thermocouples of 0.5 mm diameter Studied the effect of Reynolds number and Richardson number on
thermal stratification in scaled down model of fast breeder reactor
Khimura et al. (1999) Temperature Thermocouples of 1 mm diameter Examined the effect of Re, Ri and Gr on thermal stratification
Table 1C
Details of various geometry for thermal stratification studies in previous literatures.
Authors Experimental set up Diameter of Diameter of the Diameter of upper Elevation of UIS Elevation of IHX
cylindrical plenum core outlet internal structure from core outlet intake (m)/aspect
(m)/width of cavity (m)/height of (UIS) (m)/Depth of (m)/height of ratio (AR)
cavity cavity channel (m)
Lorenz et al. (1977) 1/15th scale model of liquid metal 0.4 0.08 – – –
fast breeder reactor (LMFBR) 0.4 0.2
Aoki and Okada 1/10th scale of full scale model
(1982)
Tanaka et al. (1984) Simplified model of the reactor 0.8 0.2 0.2 Three different Three different
position have been position have been
taken to study its taken to study its
effect on thermal effect on thermal
stratification stratification
Moriya et al. (1987) Simplified model of the reactor 0.8 0.2 0.2 0.05 0.2
Tanaka et al. (1990) Simplified model of the reactor 0.6 0.069 0.069 0.05 0.2
1/8th model of full scale reactor 2.2 0.434 0.5 0.1 0.56
Ieda et al. (1990) 1/10th scale model of full scale 0.65 0.1652 0.1652 – 0.682
reactor
Ravichandran et al. 1/24th scale down model of fast
(1995) breeder reactor
Khimura et al. Simplified model of the reactor 1.8 0.4 0.4 0.3 1.1
(1999)
Vidil et al. (1988) Rectangular cavity with a 1.6 3.2 0.8 0.03 2
rectangular channel at the bottom
of the cavity
˛k = 1, ˛ε = 0.25. One of the earlier numerical studies on thermal the oscillatory behaviour of stratification interface and temper-
stratification was carried out by Tanaka et al. (1984). He carried ature fluctuations inside the reactor. The algebraic stress model
out CFD simulation by developing a 2D in-house code. The pre- (ASM) (Rodi, 1980; Deumaran and Rodi, 1980; Noat and Rodi, 1982.)
dicted result agreed well with the experimental observation. But, model has been used with higher order schemes for thermal strati-
in case of higher discharge of upper internal structure (UIS), the fication analysis by some researchers (Ieda et al., 1990; Muramatsu
predicted result disagreed with the experimental observation. He and Ninokata, 1994; Khimura et al., 1999). Ieda et al. (1990) stud-
observed that the movement of the stratification interface was
slower in CFD simulation. Mausi et al. (1987) numerically studied
Table 1D
thermal stratification in cylindrical geometry with an in house code.
Details of conclusions and limitations of the research work carried out by the
The predicted result agreed well with the experimental results for researchers.
lower value of Richardson number (Ri). Investigations for higher
Authors Conclusions Limitations
values of Ri were not performed. Maekawa (1990) studied the
effect of higher order numerical schemes on thermal stratifica- Lorenz et al. (1977) 1 1, 2
tion. Standard k–ε model (Launder and Spalding, 1974) was used Aoki and Okada (1982) 3, 4 1, 2, 4
Tanaka et al. (1984) 5 1–6
to model turbulent flows. Maekawa (1990) observed that the flow Moriya et al. (1987) 15, 16 2–4, 6
pattern inside the reactor significantly depends on the turbulent Tanaka et al. (1990) 10–12 1–6
eddy viscosity. Transient simulations of Maekawa (1990) showed Ieda et al. (1990) 2 2, 4, 5
that standard k–ε model with the combination of QUICK (Leonard, Ravichandran et al. (1995) 15 2, 4–6
Khimura et al. (1999) 14 2–5
1979), QUICK-FRAM (Chapman, 1984) schemes properly predicts
S.S. Das et al. / Nuclear Engineering and Design 250 (2012) 417–435 421
Limitations
grid size on thermal stratification with sodium as working fluid.
They observed that thermal boundary conditions have a signifi-
7–12
7–12
7–12
7–12
7–12
1–6
1–6
cant effect on transient temperature values. They found that below
a specific grid size one cannot resolve the stratification interface
Conclusions and its characteristics. An important finding by the authors was
that, a turbulence model with higher order numerical schemes is
18–21
17, 18
21, 13
unable to predict oscillatory behaviour of stratification interface in
6, 7
8, 9
22
sodium. This contradicts the finding by Maekawa (1990) who found
5
1–4
1–4
1–4
1–4
1–5
1–4
models (i) The k–ε model with constant turbulent Prandtl num-
ber (ii) The k–ε model with turbulent Prandtl number depend on
Time-step
local Richardson number Ri and (iii) The ASM model. The most
0.005 s
important aspect found by the authors was that the interaction
between transport equations for Reynolds stress and turbulent heat
–
HPLA Bounded scheme flux needs the adjustment of values of generic constants. Further,
a suitable combination of higher order schemes with turbulence
order accuracy in time
QUICK-FRAM, Second
QUICK,QUICK-FRAM
QUICK-FRAM
QUICK-FRAM
the ASM model gave better predictions than the other models.
The number of model constants are however large (13 constants)
method
method
Scheme
QUICK,
QUICK,
142 × 93
17 × 26
44 × 82
seen from temperature profiles that for transient analysis the pre-
Grid
important study has been made by Choi and Kim (2006). They also
Finite Difference
Finite difference
Finite difference
Finite difference
Finite difference
method
method
method
method
method
Hanjalic, 2002), V2-f model (Medic and Durbin, 2002), Two layer
Sodium
Sodium
Water
Water
Water
Water
model (Chen and Patel, 1988) and Shear stress transport model
(Mentor, 1994) which are better than the ASM model. The com-
parison of the four models has provided excellent insight into the
stratification phenomena. The authors reported that out of the four
Numerical technique
Governing equations
Navier–Stokes
Navier–Stokes
Navier–Stokes
Navier–Stokes
Navier–Stokes
Navier–Stokes
equation
equation
equation
equation
equation
equation
1.2. Conclusions
Table 2B
Details of range and proposed correlation for thermal stratification in the literatures.
Re Ri Pe Pr
Table 3
Governing equations for 2D model.
2D Equations
1 ∂
Continuity (ru) + ∂∂zv = 0
r ∂r
∂(u) 1 ∂ ∂
∂u ∂ ∂u u
1 ∂
Momentum (in radial direction) + (ruu) + (vu) = reff + eff +S
∂∂rv ∂∂z ∂∂zv v
∂t r ∂r ∂z r ∂r
∂(v) 1 ∂ ∂ 1 ∂
Momentum (in axial direction) + r ∂r (ruv) + ∂z (vv) = r ∂r reff ∂r + ∂z eff ∂z + S
∂t
∂(Cp T ) ∂ ∂ ∂
∂T ∂ ∂T T
Energy (T) + 1r ∂r (Cp ruT ) + ∂z (Cp vT ) = 1r ∂r reff ∂r + ∂z eff ∂z + S
∂t
∂(k) 1 ∂ ∂ 1 ∂
∂k
∂ ∂k k
Turbulent kinetic energy (k) + r ∂r (ruk) + ∂z (vk) = r ∂r reff ∂r + ∂z eff +S
∂t
∂(ε) 1 ∂ ∂ 1 ∂
∂ε ∂ ∂ε∂z ε
Energy dissipation rate (ε) ∂t
+ r ∂r
(ruε) + ∂z
(vε) = r ∂r
reff ∂r
+ ∂z
eff ∂z
+S
S.S. Das et al. / Nuclear Engineering and Design 250 (2012) 417–435 423
Table 4
Source terms for the transport equations.
eff S
1 ∂ ∂u
∂
∂p u
u + t C r ∂r
reff ∂r
+ ∂z
eff ∂∂rv − ∂r
− 2 ∂k
3 ∂r
−
r2
1 ∂ ∂u
∂
∂v
∂p 2 ∂k
v + t C r ∂r
reff ∂z
+ ∂z
eff ∂z
− ∂z
− 3 ∂z
+ 0 gˇ(T − T0 )
t
T K+ T
0
Table 5
Source terms for the standard k–ε model.
Variable eff S
t
k + k
G − ε
t ε
ε + (C1 G − C2 ε)
ε
k
2
2 2 2
where G is production of k=2t (∂u/∂r) +(∂v/∂z) +(u/r) + (∂u/∂z)+(∂v/∂r) .
1.3. Assumptions
1. Wide range of Re and Ri were not considered. Literature review shows that during the stratification phenom-
2. The effect of Peclet number on thermal stratification char- ena the complexity lies in selection of numerical scheme and
acteristics such as rising speed of stratification interface and turbulence model that can accurately predict the oscillatory motion
temperature gradient at stratification interface was not studied. in a thermally stratified behaviour. Different turbulence models
3. Type of the reactor under consider was simple. have been used by different researchers for studying stratification
4. Velocity distribution at different radial and axial positions in characteristics. A consensus is not achieved as to which turbulence
the reactor were not considered in both steady and transient model predicts the stratification characteristics reasonably well.
state. The real motivation of the present work is to find out whether
5. Temperature distribution at different axial positions in the selecting an optimum combination of numerical schemes for solv-
reactor were not considered in unsteady state. ing the governing equations and turbulence model parameters
6. Effects of the scale size on thermal stratification were not con-
sidered.
0.12
7. Parametric studies of turbulent models have not been studied.
8. Effect of numerical schemes on turbulent transport quantities
like k and have not been verified.
0.1
9. Numerical issues such as effect of initial guess values, under
relaxation parameters, grid sensitivity on prediction of thermal
stratification have not been addressed. 0.08
10. Various algorithms for pressure-velocity coupling such as SIM-
V (m/s)
0.04
Table 6
Boundary conditions.
800
600
Niter
400
200
0
0.5 0.6 0.7 0.8 0.9 1
αε
Fig. 6. Effect of number of iterations on convergence (Niter ) on under relaxation Fig. 8. Effect of number of iterations on convergence (Niter ) on under relaxation
parameter for pressure ˛p and for three different values of ˛u . ( ) 0.8; (- - -) parameter for turbulent energy dissipation rate ˛ε and for three different values of
0.7; (—) 0.5. ˛u . ( ) 0.8; (- - -) 0.7; (—) 0.5.
affect the prediction of stratification characteristics using a stan- 2. Boussinesq’s approximation was made. i.e. Density difference
dard k–ε model. Hence, a stepwise procedure has been devised. It is are only important in producing buoyancy. The buoyancy term
thought that first, the effect of various parameters like under relax- is assumed to vary with the temperature according to the
ation parameters and numerical schemes on standard k–ε model relation,
for at steady state flow conditions be studied and the code be stan- = 0 [1 − ˇ(T − T0 )] (1)
dardised for accuracy and boundedness. After this the turbulent
model parameters values be varied for steady state conditions and 1 ∂
ˇ=− (2)
their effect on flow field was analyzed. Then, the transient studies ∂T
be taken up and an effort would be made to see if changing the
3. Flow is axi-symmetric.
turbulent model parameters of the standard k–ε model makes an
4. Physical properties except density are assumed to be constant.
effect on the flow pattern and thermal stratification characteristics
5. Temperature range is sufficiently low to include effect of radia-
like the rising of the stratification interface.
tion. Hence radiation effect is neglected.
Fig. 3 shows the computational model with the boundary con- In order to model unsteady mixed convection turbulent flows,
ditions. the governing equations (continuity, momentum and energy) with
appropriate Reynolds stress closure have been solved. The basic
2.1. Model assumptions
Table 7
Effect of under relaxation parameters on the rate of convergence.
governing equations consist of time averaged equation of conti- ulation was carried out with a time step size 0.0001 s. All the
nuity, Reynolds equation of motion and energy. The standard k–ε simulations were carried out on i7 Intel 64 bit processor with a
model has been used. Table 3 summarizes the details of the govern- clock speed of 3.06 GHz and 8 GB memory. The various parameters
ing equations. The expression for diffusive coefficient and source like velocity, turbulent quantities, temperature, etc. were estimated
term was given in Tables 4 and 5. Table 6 depicts the boundary for every configuration by using the post-processing features of the
conditions at the different boundaries of the domain. software.
Fig. 10. Comparison between experimental and numerical predictions for velocity field. (A) Vector plot from experiment (Moriya et al., 1987), (B) vector plot from CFD
simulation.
426 S.S. Das et al. / Nuclear Engineering and Design 250 (2012) 417–435
0.08
0.02
(B)
(A)
0.01
0.04
0
u (m/s)
u (m/s)
-0.01
0
-0.02
-0.03 -0.04
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
z (m) z (m)
Fig. 11. Comparison between simulated and experimental profiles for radial velocity at various radial locations. (A) r = 0.2 m: (—) CFD; () experimental, (B) r = 0.3 m: (—)
CFD; () experimental.
Moriya et al. (1987). Both, steady and transient simulations were number of iterations for convergence (Niter ) was recorded in each
carried out. Initially, the optimization for various numerical param- case (steady state). All these studies were carried out at three dif-
eters such as effect of under relaxation parameters, numerical ferent values (0.5, 0.7 and 0.8) of ˛u (under relaxation parameters
schemes and turbulent model parameters on turbulent equations for velocities). Fig. 6 depicts the effect of ˛u and ˛p on Niter . For
were carried out. Then, transient simulations were studied in order ˛u = 0.8, a very good convergence is obtained for various values
to study the sensitivity of turbulence model parameters on thermal of ˛p in the range of 0.1–0.4. However, the solution diverges for
stratification. Furthermore, the predicted results were validated ˛p > 0.4. It was further observed that with decrease of ˛u it requires
with the experimental data of Moriya et al. (1987). more number of iterations for convergence and the range of ˛p
is increased. If we reduce the value of ˛u to 0.7, it was observed
3.1. CFD simulations in steady state that the range for ˛p was increased from 0.2 to 0.7. When the
value of ˛p is further increased, divergence in solution is observed.
3.1.1. Effect of the under relaxation parameters on the rate of It was interesting to note that, for ˛u = 0.5 the convergence rate
convergence (The speed at which the iterative system converges. Specifically, if
The effect of under relaxation parameters for flow variables the convergence rate is faster, then the convergence will be faster
plays a significant role in the convergence of the numerical solu- otherwise it will be slower) is independent of ˛p in the range of
tions of the fluid flow problems (Sahu and Joshi, 1995; Barron 0.3–1 (Fig. 6). Although, the range for ˛p was extended but, it
and Neyshabouri, 2003). It is a known fact that in a CFD code, affects significantly on the convergence of the numerical solution.
the governing equations for mass and momentum along with the In other words, the convergence is slower for smaller values of ˛u
equations of turbulence are discretized and solved with numerical (Fig. 6).
schemes. Although changing the relaxation parameters for better Fig. 7 shows the effect of ˛u and ˛k on Niter . It is interesting to
convergence is a common practice we think that it is important to note that, for a given value of ˛u the required number of iterations
report the sensitivity of these parameters. First, the under relax- for convergence sharply decreases as ˛k increases. It was observed
ation parameters for pressure (˛p ), turbulent kinetic energy (˛k ), that for ˛u = 0.8, the convergence is slower for smaller values of
turbulent energy dissipation rate (˛ε ) were systematically changed ␣k . It took around 1838 iterations to get a converge solution for
between the limits 0.1 and 0.9 for steady state case. Also the ˛u = 0.8 and ˛k = 0.1. As the value of ˛k (0.2–0.9) is increased the
0.03
0.025 (B)
(A)
0.02
0.01
v (m/s)
v (m/s)
0.01
-0.005
-0.02 0
-0.035 -0.01
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
z (m) z (m)
Fig. 12. Comparison between simulated and experimental profiles for axial velocity at various radial locations (A) r = 0.2 m: (—) CFD; () experimental, (B) r = 0.3 m: (—) CFD;
() experimental.
S.S. Das et al. / Nuclear Engineering and Design 250 (2012) 417–435 427
Table 8
Effect of numerical schemes on flow pattern.
Cases Turbulent energy equation, k (m2 /s2 ) Turbulent dissipation rate, (m2 /s3 ) Flow pattern
0.04
(A) 0.08
(B)
0.02
0.04
u (m/s)
u (m/s)
0
0
-0.02 -0.04
-0.04 -0.08
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
z (m) z (m)
Fig. 13. Effect of numerical schemes on radial velocity at various radial locations (A) r = 0.2 m: (—) k (1st order upwind scheme) and ε (1st order upwind scheme); (- - -) k
(2nd order upwind scheme) and ε (1st order upwind scheme); ( ) k (QUICK) and ε (1st order upwind scheme); () experimental. (B) r = 0.3 m: (—) k (1st order upwind
scheme) and ε (1st order upwind scheme); (- - -) k (2nd order upwind scheme) and ε (1st order upwind scheme); ( ) k (QUICK) and ε (1st order upwind scheme); ()
experimental.
number of iterations needed for convergence decreases. Further, Fig. 8 depicts the effect of ˛u and ˛ε on Niter . It is interesting to
when the value of ˛u is decreased from 0.8 to 0.5 (corresponding note that for a given value of ˛u , the solution diverges for ˛ε < 0.5. It
values of ˛k are 0.1–0.4), the time taken for getting a converged was also observed that the convergence is slower for small values
solution is more. For example, for ˛u = 0.5 and ˛k = 0.1 the iterative of ˛u . Several combinations of under relaxation parameters were
system converges after 2155 iterations. It was observed that for a considered. Table 7 shows the effect of relaxation parameters on the
given value of ˛u , the convergence is faster for larger values of ˛k . convergence of the flow variables. Patankar (1980) suggested that
The faster convergence was observed for the ˛u = 0.8 and ˛k = 0.9 by using ˛u = 0.5 and ˛p = 0.8 the faster convergence is obtained in
(Fig. 7). SIMPLE algorithm. However, in the present system it was observed
0.02 0.04
(A) (B)
0.03
0
0.02
v (m/s)
v (m/s)
-0.02
0.01
-0.04
0
-0.06 -0.01
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
z (m) z (m)
Fig. 14. Effect of numerical schemes on axial velocity at various radial locations (A) r = 0.2 m: (—) k (1st order upwind scheme) and ε (1st order upwind scheme); (- - -) k
(2nd order upwind scheme) and ε (1st order upwind scheme); ( ) k (QUICK) and ε (1st order upwind scheme); () experimental. (B) r = 0.3 m: (—) k (1st order upwind
scheme) and ε (1st order upwind scheme); k (2nd order upwind scheme) and ε (1st order upwind scheme); ( ) k (QUICK) and ε (1st order upwind scheme); ()
experimental.
428 S.S. Das et al. / Nuclear Engineering and Design 250 (2012) 417–435
0.15
0.15
(A) (B)
0.12
0.12
0.09 0.09
μt (kg/ms)
μt ( kg/ms)
0.06 0.06
0.03 0.03
0 0
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
z (m) z (m)
Fig. 15. Effect of C on the eddy viscosity t (A) r = 0.2 m: ( ) 0.05; (- - -) 0.07; (—) 0.09. (B) r = 0.3 m: ( ) 0.05; (- - -) 0.07; (—) 0.09.
that the following values: ˛u = 0.8, ˛p = 0.4, ˛k = 0.8, ˛ε = 0.8 gives shows that it is necessary to improve the CFD predictions inside the
the faster convergence. cylindrical plenum.
Fig. 9 shows the predicted velocity contour for the steady state.
The predicted flow patterns were shown with the experimen-
3.1.2. Effect of different numerical schemes on turbulent
tal data (Fig. 10A and B). Two circulation cells were observed
equations
both in experiments and CFD simulations. Fig. 11A and B shows
The effect of different numerical schemes (Table 8) were stud-
the comparison between the experiment and CFD predictions for
ied on the turbulent equations (k and ε). It was observed that the
radial velocity for various radial locations (r = 0.2 and 0.3 m). It was
numerical schemes on the turbulent equations have significant
observed that the predicted result agreed well with the experimen-
effect on the flow field (Table 8). It was found that Power Law
tal data for z > 0.2 m (Fig. 11A). But, for z < = 0.2m it over predicts the
scheme for k equation is not a suitable scheme to model the con-
radial velocity (Fig. 11A). A good observed prediction was observed
vective terms in k equation. Similarly, for ε equation higher order
between CFD and experiment (Fig. 11 B) for radial velocity for z < 0.1
schemes such as Power Law (Patankar, 1980), 2nd order upwind
and z > 0.2 m. Over predictions were observed for radial velocity in
schemes (Ferziger and Peric, 2002) and QUICK schemes (Leonard,
range of axial coordinate for 0.1 m < z < 0.2 m (Fig. 11 B). The max-
1979) were not robust schemes. It was observed that by using
imum value of radial velocity at r = 0.3 m from the experiment is
higher order schemes for ε equation produces one circulation cell
0.01 m/s, where as the maximum predicted value is 6 times the
in the domain (Table 8). Further, it was observed that the predicted
experimental value. Fig. 12A and B shows the comparison between
flow pattern from case 1 (solving both k and ε equation with 1st
the experiment and CFD predictions for axial velocity for vari-
order upwind scheme), case 2 (solving k equation with 2nd order
ous radial locations (r = 0.2 and 0.3 m). For z < 0.2 m, a very good
upwind and ε equation with 1st order upwind) and case 3 (solving k
agreement between experiment and CFD predictions is observed
equation with QUICK scheme and ε equation with 1st order upwind
(Fig. 12A). But, for z > 0.2 m the predicted values under predicts the
scheme) agreed well with the experimental flow pattern. Therefore,
axial velocity (Fig. 12A). A close agreement was observed between
these three different cases were considered and further analyses
CFD predictions and experimental data for z < 0.1 m (Fig. 12B). How-
were carried out to study the effects of the numerical schemes on
ever, it was observed that CFD over predicts the axial velocity for
turbulent equations.
0.1 m < z < 0.2 m (Fig. 12B). For z > 0.2 m, the predicted axial velocity
Fig. 13A and B shows the effect of numerical schemes for radial
under predicts the experimental data (Fig. 12B). The above study
velocity at various radial locations (r = 0.2 and 0.3 m). The predicted
0.04 0.08
(A) (B)
0.06
0.02
0.04
u (m/s)
u (m/s)
0 0.02
0
-0.02
-0.02
-0.04 -0.04
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
z (m)
z (m)
Fig. 16. Effect of C on radial velocity at various radial locations. (A) r = 0.2 m: (—) 0.09; (- - -) 0.07; ( ) 0.05; experimental. (B) r = 0.3 m: (—) 0.09; (- - -) 0.07; ( )
0.05; () experimental.
S.S. Das et al. / Nuclear Engineering and Design 250 (2012) 417–435 429
0.02 0.04
(A) (B)
0.03
0
0.02
v (m/s)
v (m/s)
-0.02
0.01
-0.04
0
-0.06 -0.01
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
z (m) z (m)
Fig. 17. Effect of C on axial velocity at various radial locations. (A) r = 0.2 m: (—) 0.09; (- - -) 0.07; ( ) 0.05; experimental. (B) r = 0.3 m: (—) 0.09; (- - -) 0.07; ( )
0.05; () experimental.
flow pattern obtained from case 1 agreed well with the experimen- for all the three cases (Fig. 14B). An under prediction was observed
tal data for z > 0.2 m (Fig. 13A). For z ≤ 0.2 m, case 1 over predicts in the region for z > 0.2 m for all the three cases (Fig. 14B).
the radial velocity, whereas, the profiles for radial velocity obtained
from cases 2 and 3 are similar (Fig. 13A). The predicted radial veloc- 3.1.3. Effect of the turbulent model parameters on the flow field
ity from cases 2 and 3 over predicts the experimental data for In standard k–ε model, the values of eddy viscosity are calcu-
z ≤ 0.2 m. It was observed that the predicted radial velocity under lated by the solving of k and ε equations in which
predicts the experimental data in the range of z > 0.3 m for all the
k2
three cases. Fig. 13B depicts that CFD prediction with three different t = C (3)
combinations of numerical schemes for radial velocity at r = 0.3 m. ε
It was observed that flow pattern from all the three cases were The equilibrium between the production and dissipation of tur-
almost same except in the region for 0.1 m < z < 0.2 m (Fig. 13B). A bulent kinetic energy in a shear layer indicate that the value of C
over prediction was observed in the range 0.1 m < z < 0.2 m for all is 0.09. Modelling of ‘ε’ equation gives rise to two more important
three cases (Fig. 13B). Fig. 14A and B shows the effect of numer- constants namely C1 and C2 .The measurements of rate of decay of
ical schemes for axial velocity at various radial locations (r = 0.2 turbulence behind a grid indicate that the value of C2 is in the range
and 0.3 m). The predicted flow pattern obtained from all the cases of 1.8–2.0. These values were obtained by carrying out experiments
agreed well with the experimental data for z ≤ 0.15 m (Fig. 14A). For in relatively simply geometries, such as shear layers, decay of tur-
z > 0.15 m, all the cases under predicted radial velocity (Fig. 14A). bulence behind grid, etc. These relatively simple flows are mostly
The flow pattern obtained from both the cases (cases 2 and 3) are one dimensional. The flow occurring in the present system is mainly
almost same (Fig. 14A). The comparison among three cases shows two dimensional, axi symmetric, and highly nonlinear. Assuming
that for z > 0.15 m from case 1 is closer to the experimental data the same values of these parameters was thus questionable.
(Fig. 14A). It was observed that the predicted result for axial veloc-
ity for all the three cases show the same behaviour (Fig. 14B). But, 3.1.3.1. Effect of C on the flow field. In the present section, we
there is a slight variation in the axial velocity obtained from case 1. carried out sensitivity analysis of C on the flow field. For these
An over prediction was observed in the region for 0.1 m < z < 0.2 m purpose, several values of C were taken. It was observed that for
0.04
0.08
(A) (B)
0.02
0.04
u (m/s)
u (m/s)
0 0
-0.02 -0.04
-0.04 -0.08
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
z (m) z (m)
Fig. 18. Effect of C1 on radial velocity at various radial locations. (A) r = 0.2 m: (—) 1.44; (- - -) 1.5; ( ) 1.64; experimental. (B) r = 0.3 m: (—) 1.44; (- - -) 1.5; ( )
1.64; () experimental.
430 S.S. Das et al. / Nuclear Engineering and Design 250 (2012) 417–435
0.02 0.05
(A) (B)
0.04
0
0.03
v (m/s)
v (m/s)
-0.02 0.02
0.01
-0.04
0
-0.06 -0.01
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
z (m) z (m)
Fig. 19. Effect of C1 on axial velocity at various radial locations. (A) r = 0.2 m: (—) 1.44; (- - -) 1.5; ( ) 1.64; experimental. (B) r = 0.3 m: (—) 1.44; (- - -) 1.5; ( )
1.64; () experimental.
C > 0.09, the solution becomes oscillatory. Experimental data was predicted axial velocity under predicts the experimental data for all
compared with the predicted values for C (0.05, 0.07, 0.09) keep- values of C1 (Fig. 19A). Fig. 19B depicts that the comparison between
ing all the other parameters as constant. Fig. 15A and B shows the experimental data and CFD prediction for axial velocity at r = 0.3 m.
effect of C on eddy viscosity at different radial locations (r = 0.2 It was also observed that for z < 0.15 m, there is agreement between
and 0.3 m). It is interesting to note that the eddy viscosity increases the experimental data and CFD predictions for C1 = 1.5 (Fig. 19B). It
with an increase in C (Fig. 15A and B). was observed that for 0.15 m < z < 0.2 m, the predicted result over
Fig. 16A depicts that the comparison between experimental data predicts the experimental data for all values of C1 (Fig. 19B). For
and CFD prediction for radial velocity at r = 0.2 m. All three values z > 0.2 m, the predicted result from CFD simulation under predicts
of C (0.05, 0.07, 0.09) over predicts the experimental data in the for all values of C1 (Fig. 19B).
range of z < 0.2 m. It means that the predicted radial velocity over
predicts below the IHX intake. For z > 0.2 m, the result obtained from 3.1.3.2. Effect of C1 on the flow field. In this section, the effects of C1
all the three values of C under predicts the experimental data were studied on the flow field at various radial location for several
(Fig. 16A). Fig. 16B shows the comparison between experimental values of C1 . And also a range for C1 was observed (1.44 ≤ C1 ≤ 1.64).
data and CFD prediction for axial velocity at r = 0.3 m. A good agree- Experimental result shows that there are two circulation cells, how-
ment between the experiment and CFD for all values of C was ever when C1 < 1.44 only one circulation cell was observed in upper
observed in the range of z < 0.1 m (Fig. 16B). For 0.1 m < z < 0.25 m, part of the reactor. Further, three different values of C1 were con-
there is an over predication for radial velocity for all values of C sidered (1.44, 1.5, 1.64).Two circulation cells were observed for all
(Fig. 16B). For 0.25 m < z < 0.4 m, a good agreement between experi- values of C1 . But, for C1 > 1.64, the solution becomes oscillatory.
mental data and CFD predictions were observed for all values of C Fig. 18A and B illustrates the comparison between the experi-
(Fig. 16B). Near the free surface (for z > 0.4 m), the predicted radial mental and CFD predictions for radial velocities at various radial
velocity under predicts the experimental data for all values of C locations (r = 0.2 and 0.3 m). Fig. 18A depicts the comparison
(Fig. 16B). Fig. 17A illustrates that the comparison between exper- between experimental and CFD prediction for radial velocity at
imental data and CFD prediction for axial velocity at r = 0.2 m. A r = 0.2 m. It was observed that for z < 0.2 m, the predicted result
close agreement between the experiment and CFD for all values of obtained from all values of C1 over-predict the experimental data
C in the range of z < 0.2 m was observed (Fig. 17A). For z > 0.2 m, the (Fig. 18A). For z > 0.2 m, the result obtained from C1 = 1.44 agrees
0.04 0.08
(A) (B)
0.02 0.04
u (m/s)
u (m/s)
0 0
-0.02 -0.04
-0.04 -0.08
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
z (m) z (m)
Fig. 20. Effect of C2 on radial velocity at various radial locations. (A) r = 0.2 m: (—) 1.92; (- - -) 1.82; ( ) 2.0; experimental. (B) r = 0.3 m: (—) 1.92; (- - -) 1.82; ( )
2.0; experimental.
S.S. Das et al. / Nuclear Engineering and Design 250 (2012) 417–435 431
0.02 0.04
(A) (B)
0.03
0
0.02
v (m/s)
v (m/s)
-0.02
0.01
-0.04
0
-0.06 -0.01
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
z (m) z (m)
Fig. 21. Effect of C2 on axial velocity at various radial locations. (A) r = 0.2 m: (—) 1.92; (- - -) 1.82; ( ) 2.0; () experimental. (B) r = 0.3 m: (—) 1.92; (- - -) 1.82; ( )
2.0; () experimental.
with the experimental data (Fig. 18A). The predicted radial velocity (z > 0.4 m) for C2 = 2 (Fig. 20A). An underprediction was observed
from C1 = 1.5, 1.64 under predicts the experimental data (Fig. 18A). between CFD and experimental data for C2 = 1.82 and 2. Fig. 20B
Fig. 18B depicts that the comparison between experimental data depicts that the comparison between experimental data and CFD
and CFD prediction for radial velocity at r = 0.3 m. It was observed prediction for radial velocity at r = 0.3 m. A under prediction was
that the predicted radial velocity under predicts the experimen- observed for all values of C2 (Fig. 20B). For 0.1 m < z < 0.2 m, there is
tal data in the range of z < 0. 06 m (Fig. 18B). A good agreement a over predication for radial velocity for all values of C2 (Fig. 20B).
between the experimental data and CFD predictions for all values A close agreement between experiment and CFD was observed for
of C1 were observed in the range of 0.06 m < z < 0.12 m (Fig. 18B). It all values of C2 in the range of 0.2 m < z < 0.3 m (Fig. 20B). Near the
was observed that for 0.12 m < z < 0.2 m, there is a over predication free surface (for z > 0.4 m), the predicted radial velocity under pre-
for radial velocity for all values of C1 (Fig. 18B). A close agreement dicts the experimental data for all values of C2 (Fig. 20B). Fig. 21A
between experimental data and CFD predictions for all values of C1 depicts the comparison between experimental data and CFD pre-
were observed in the range of 0.2 m < z < 0.4 m (Fig. 18B). Near the diction for axial velocity at r = 0.2 m. An excellent prediction was
free surface (for z > 0.4 m), the predicted radial velocity under pre- observed between CFD and experimental data for all values of C2
dicts the experimental data for all values of C1 (Fig. 18B). Fig. 19A in the range of z < 0.2 m (Fig. 21A). An improvement in the pre-
and B shows the comparison between the experimental and CFD dicted axial velocity was observed for C2 = 2 in the range of z > 0.2 m
predictions for axial velocities at various radial locations (r = 0.2 and (Fig. 21A). The predicted axial velocities under predicts the experi-
0.3 m). Fig. 19A depicts that the comparison between experimental mental data for all other values of C2 (Fig. 21A). Fig. 21B shows that
data and CFD prediction for axial velocity at r = 0.2 m. A good agree- the comparison between experimental data and CFD prediction for
ment between experimental and CFD predictions were observed for axial velocity at r = 0.3 m. A good agreement between experimental
all values of C1 in the range of z < 0.2 m (Fig. 19A). It was observed data and CFD predictions was observed for C2 = 1.82 in the region
that for 0.15 m < z < 0.2 m, the predicted result over predicts the z < 0.1 m (Fig. 21B). It was also observed that for 0.1 m < z < 0.2 m,
experimental data for all values of C1 (Fig. 19A). For z > 0.2 m, the the predicted results over predict the experimental data for all val-
predicted axial velocity under predicts the experimental data for ues of C2 (Fig. 21B). An under prediction was observed for all value
all values of C1 (Fig. 19A). Fig. 19B depicts that the comparison of C2 in the region z > 0.2 m (Fig. 21B).
between experimental data and CFD prediction for axial velocity at
r = 0.3 m. It was also observed that for z < 1.5 m, there is agreement 3.2. CFD simulation in transient case
between the experimental data and CFD predictions for C1 = 21.5
(Fig. 19B). It was observed that for 0.15 m < z < 0.2 m, the predicted In the previous section, we have studied the effect of model
result over predicts the experimental data for all values of (Fig. 19B). parameters on the flow field. In this section, the effect of turbulent
For z > 0.2 m, the predicted result from CFD simulation under pre- model parameter was studied on thermal stratification character-
dicts for all values of (Fig. 19B). istics such as rising speed of stratification.
The height of stratification interface (z) is defined by Moriya
et al. (1987) which states that, “height in hot plenum where the
3.1.3.3. Effect of C2 on the flow field. In the present section, we have
non-dimensional temperature (T* ) was 50%”, non-dimensional
studied the effect of C2 on the flow field at various radial locations.
time (t* ) and non-dimensional height of interface (z* ) are defined
Several values of C2 were considered. A range was observed for
as follows,
C2 (1.7 < C2 ≤ 2). For C2 < 1.7, the numerical solution becomes oscil-
latory. One circulation cell was observed for C2 > 2. Further, three T − Tc
T ∗ = 100 × (%) (4)
different values of C2 (1.82, 1.92, 2) were considered. The experi- Th − Tc
mental data was compared with the predicted values of C2 keeping
Ut
all other parameters constant. t∗ = (5)
D
Fig. 20A and B shows the comparison between experimental and
z
CFD prediction for radial velocity at various radial locations (r = 0.2 z∗ = (6)
and 0.3 m). An over prediction was observed for all values of C2 in H
the range of z < 0.1 m (Fig. 20A) It was observed that in the range Table 9. depicts the experimental test conditions for the simu-
of 0.2 m < z <0.4 m, the predicted radial velocity from all values C2 lation.
under predicts the experimental data (Fig. 20A). A good agreement Fig. 22A–D illustrates the time sequential development of
was observed between CFD and experiment near the free surface stream function in the reactor. It was observed that initially two
432 S.S. Das et al. / Nuclear Engineering and Design 250 (2012) 417–435
Fig. 22. Time sequential contour plots for stream function for Re = 19,200. (A) t = 30 s (B) t = 60 s (C) t = 120 s (D) t = 240 s.
counter rotating circulation cells were developed approximately at predictions of temporal variation of stratification interface height.
25–30 s (Fig. 22A) below the inner outlet (IHX intake) of the reac- It was observed in the experiment, that the interface movement is
tor. Also, a small circulation cell was observed near the bottom of faster below the outlet of the inner cylinder (IHX intake). Similar
the reactor (Fig. 22B). At around 60 s, these two counter rotating kind observation was made by CFD model in the lower part of the
cells merge into a single cell below the inner outlet (IHX intake). outlet of the inner cylinder (IHX intake). This faster movement of
Then, this cell moves towards the left wall. In the mean time, the stratification interface below IHX intake may be attributed due to
small circulation cell which was observed earlier below the bottom the formation of two counter rotating circulation cells (Fig. 22A
started growing and moves slowly towards the wall below the IHX and B). A good agreement was observed between experimental
intake (Fig. 22C). This circulation becomes stronger. Then, it moves data and CFD model in the lower part of the vessel (Fig. 24).
slowly and becomes slightly elongated below the IHX intake. For But, in the upper part of the vessel, CFD under predicts the ris-
the long period time it stays there. No circulation was observed in ing speed of the stratification interface. This may be attributed due
the upper part of the IHX intake. to non-formation circulation cell in the upper part of IHX intake
Fig. 23A–D shows the time sequential development of the tem- (Fig. 22D).
perature contours in the reactor vessel. In a very few seconds, The sensitivity of C was carried out in the transient case. Three
a faster heat transfer was observed between hot and cold water different values of C (0.05, 0.07, 0.09) were taken to study the
below IHX intake. A steep temperature gradient was seen below effect of model parameter’s on the thermal stratification character-
the IHX intake. In the simulation, distinct thermal stratification istics such as rising speed of the stratification interface (Fig. 25). It
was observed in the reactor at around 120 s (Fig. 23C). Fig. 24 was observed that in the lower part of the outlet (z* < 0.45), the pre-
shows the comparison between the experimental data and CFD dicted result obtained from standard k–ε model (C = 0.09) shows
better agreement with the experimental data as compared with the
Table 9 other values of C . CFD predictions obtained from 0.05, 0.07 under
Experimental test conditions.
predict the experimental observation in the lower part of outlet.
v (m/s) Tc (K) Th (K) Re Ri It is interesting to note that for all the values C , CFD predictions
0.095 283.05 334.75 19200 3.7
under predict the experimental observation in the upper part of the
outlet.
S.S. Das et al. / Nuclear Engineering and Design 250 (2012) 417–435 433
Fig. 23. Temperature contours at various times for Re = 19200; Ri = 3.7. (A) t = 30 s (B) t = 60 s (C) t = 120 s (D) t = 240 s.
0.75
0.75
0.6
0.6
0.45
0.45
z*
z*
0.3
0.3
0.15
0.15
0
0
0 50 100 150 200 250
0 50 100 150 200 250
t* t*
Fig. 24. Comparison between experimental and numerical result for stratification Fig. 25. Effect of C on the rising speed of the stratification interface. ( ) 0.07;
interface height. (—) CFD; () experimental. (- - -) 0.05; (—) 0.09; () experimental.
434 S.S. Das et al. / Nuclear Engineering and Design 250 (2012) 417–435
0.8 outlet the predicted interface height is same from the both 1.82 and
1.92, respectively. Whereas, the prediction of interface height from
2.0 is smaller as compared to other values of C2 . It was observed
that, for all values of C2 , the predicted result underpredicts the
0.6
experimental observation in the upper part of the reactor vessel.
0.4
z*
4. Conclusions
Fig. 26. Effect of C1 on the rising speed of the stratification interface. (- - -) 1.5; (—)
1.44; () experimental. 1. For larger value of ˛u , the range of ˛p is restricted. Whereas, for
smaller value of ˛u , the range of ˛p is not restricted. It takes val-
0.75 ues from 0.1 to 1. But, it has been observed that for smaller values
of ˛u , the convergence is slower. It was interesting to note that,
no restriction has been found for ˛k (0.1–0.9) for different value
0.6 of ˛u . But, for ˛k = 1.0, the solution becomes oscillatory. Also, it
has been observed that when ˛ε < 0.5, the solution become oscil-
latory. For, the present flow conditions, ˛u = 0.8, ˛p = 0.4, ˛k = 0.8,
0.45 ˛ε = 0.8 gives a faster convergence for the suitable choice of guess
values.
z*
Acknowledgements Laxman, D., Gupta, P.K., Kumar, A.S., Shivakumar, N.S., Gopal, C.S., Padmakumar, G.,
Prakash, V., Anandbabu, C., Vaidyanathan, G., 2006. Experimental investigations
of thermal stratification in a reduced scale model of PFBR. In: 18th National and
Authors are thankful to Indira Gandhi Centre for Atomic 7th ISHMT-ASME Heat and Mass Transfer Conference, IIT Guwahati, India.
Research (IGCAR) for the facility of research and funding it has Moriya, S., Tanaka, N., Katano, N., Wada, A., 1987. Effects of Reynolds number and
provided. Richardson number on thermal stratification in hot plenum. Nucl. Eng. Des. 99,
441–451.
Mausi, A., Koga, T., Moriya, S., Tanaka, N., 1987. Study of turbulent analysis code.
References Nucl. Eng. Des. 99, 109–116.
Maekawa, I., 1990. Numerical diffusion in single-phase multi-dimensional thermal
Aoki, T., Okada, K., 1982. Experimental study on thermal stratification. In: Proc. IAEA hydraulics analysis. Nucl. Eng. Des. 120, 323–339.
Specialists. Meeting Internal Working Group on Fast Reactors, Grenoble. Muramatsu, T., Ninokata, H., 1994. Investigation of turbulence modelling in thermal
Barron, R.M., Neyshabouri, A., 2003. Effects of under relaxation parameters on tubu- stratification analysis. Nucl. Eng. Des. 150, 81–93.
lent flow simulations. Int. J. Numer. Methods Fluids 42, 923–928. Mentor, F.R., 1994. Two equation eddy-viscosity turbulence models for engineering
Choi, S.K., Kim, S., 2006. The role of turbulence models for predicting a thermal applications. AIAA J. 32, 1598–1604.
stratification. J. Press. Vessel Technol. 128, 65–662. Medic, G., Durbin, P.A., 2002. Toward improved prediction of heat transfer on turbine
Chapman, M., 1984. FRAM-nonlinear damping algorithms for the continuity equa- blades. ASME J. Turbo Machine 124, 187–192.
tion. J. Comput. Phys. 44, 84. Mancaeau, R., Hanjalic, K., 2002. Elliptic blending model: a new near wall Reynolds
Chen, H.C., Patel, V.C., 1988. Near wall turbulence models for complex flows includ- stress turbulence closure. Phys. Fluids 14, 744–754.
ing separation. AIAA J. 26, 641–648. Noat, D., Rodi, W., 1982. Numerical simulation of secondary currents in channel flow.
Deumaran, A.O., Rodi, W., 1980. Calculation of turbulence driven secondary motion J. Hyd. Div. ASCE 108, 948–968.
in non circular ducts. J. Fluid Mech. 140, 189–222. Patankar, S.V., 1980. Numerical heat transfer and fluid flow. In: Hemisphere. Taylor
Ferziger, J.H., Peric, M., 2002. Computational Methods for Fluid Dynamics, third rev. and Francis Group, New York.
ed. Springer. Ravichandran, G., Padmakumar Vaidyanathan, G., Titus, G., Rajkumar, A., Chetal, S.,
Ieda, Y., Maekawa, I., Muramatsu, T., Nakanishi, S., 1990. Experimental and analyt- Kale, R., 1995. Experimental and theoretical studies of thermal stratification in
ical studies of thermal stratification phenomenon in the outlet plenum of fast hot pool of PFBR , SACAI, IGCAR, Jan 4–6, Kalpakkam, India.
breeder reactor. Nucl. Eng. Des. 120, 403–414. Rodi, W., 1980. Turbulence Models and Their Application in Hydraulics – A State of
Khimura, N., Nishimura, M., Hayashi, K., Kamide, H., 1999. An investigation of ther- the Art Review. IAHR, Delft, The Netherlands.
mal stratification phenomenon in fast breeder reactors sodium experiments and Sahu, A.K., Joshi, J.B., 1995. Simulation of flow in stirred vessels with axial flow
analysis. In: 7th International Conference on Nuclear Engineering, Tokyo, Japan, impellers: effects of various numerical schemes and turbulence model param-
pp. 19–23. eters. Ind. Eng. Chem. Res. 34, 626–639.
Launder, B.E., Spalding, D.B., 1974. The numerical computation of turbulence flows. Tanaka, N., Moriya, S., Wada, A., 1984. Thermal hydraulic characteristics in a liquid
Comput. Methods Appl. Eng. 3, 269. metal fast breeder reactor hot plenum. Nucl. Eng. Des. 88, 445–463.
Lorenz, J.J., Carlson, R.D., Howard, P.A., 1977. The influence of scale size and fluid Tanaka, N., Moriya, S., Ushijima, S., Koga, T., Egachi, Y., 1990. Prediction method for
thermal properties in simulating LMFBR outlet plenum behaviour. AIChE Symp. thermal stratification in a reactor vessel. Nucl. Eng. Des. 120, 395–402.
73, 120–126. Vidil, R., Grand, D., Leroux, F., 1988. Interaction of recirculation and stable strat-
Leonard, B.P., 1979. A stable and accurate convective modelling procedure based on ification in a rectangular cavity filled with sodium. Nucl. Eng. Des. 105,
quadratic interpolation. Comput. Methods Appl. Mech. Eng. 19, 59–98. 321–332.