You are on page 1of 15

Progress in Nuclear Energy 97 (2017) 11e25

Contents lists available at ScienceDirect

Progress in Nuclear Energy


journal homepage: www.elsevier.com/locate/pnucene

Numerical investigation of supercritical water flow in a vertical pipe


under axially non-uniform heat flux
Marcin Karol Rowinski a, b, Jiyun Zhao c, *, Timothy John White a, Yeng Chai Soh b, **
a
Energy Research Institute at NTU, Interdisciplinary Graduate School, Nanyang Technological University, Singapore 639798, Singapore
b
School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore 639798, Singapore
c
Department of Mechanical and Biomedical Engineering, City University of Hong Kong, Kowloon, Hong Kong

a r t i c l e i n f o a b s t r a c t

Article history: The authors conducted a numerical investigation of water flow in a vertical pipe during the transition
Received 6 May 2016 from sub-to supercritical conditions when non-uniform power is applied axially at the wall. The
Received in revised form Computational Fluid Dynamics calculations were first assessed with experimental results found in
22 November 2016
literature, where uniform heat flux was applied. Next, the model was used to perform calculations of four
Accepted 29 December 2016
various cases with non-uniform heat flux distribution along the axis, however, the average value is equal
to the value of the reference axially uniform-heated case. Results showed high dependency of the wall
temperature distributions on the shape of the power distribution curve applied at the wall. It was found
Keywords:
Computational fluid dynamics
that if the peak value of heat flux is close or behind occurrence of heat transfer deterioration (HTD)
Non-uniform heating phenomenon, the wall temperature increases rapidly to higher than in the reference case. However, if the
Supercritical flow peak is moved to the front part of the pipe, it is possible to achieve lower temperatures than when
Heat transfer uniform heat flux was applied. Calculations of convective heat transfer i.e. Nusselt number show that
Mass transfer currently, only two correlations were able to follow the trend of the results i.e. the Bishop's and
Heat transfer deterioration Ornatsky's formulas. From the sensitivity calculations, it was found that the wall temperature is much
more dependent on the mass flux than the operating pressure since the temperature variation is much
higher and it is followed by the change of the position where HTD occurs.
© 2016 Elsevier Ltd. All rights reserved.

1. Introduction around 1000MWe (Liu and Oka, 2013), and small, around 300MWe
(Rowinski et al., 2015), rectors. From thermal-hydraulics point of
Super-Critical Water-cooled Reactors (SCWRs) are one of the view, one of the challenges is to fully understand the phenomenon
most promising technology among Generation IV reactors accepted during the transition from sub-to supercritical conditions. Duffey
by Generation IV International Forum (GIF) (GIF, 2002; GIF, 2009). and Pioro (Pioro and Duffey, 2005) reviewed all of the most sig-
They are being under development all over the world, thanks to the nificant experiments with supercritical water flow inside pipes,
possibility of increasing the efficiency of light water reactors annuli and fuel rod bundles. They have found only two experiments
(LWRs) from current ca. 35% to ca. 48%. It is possible due to the fact that were performed with use of rod bundles, since that moment
that the steam is under much higher temperature and pressure there was one more experiment done by Wang et al. (2014).
than the contemporary reactors, i.e. the parameters are above Therefore, the amount of experimental data is very limited. How-
pseudo-critical point, similar to current fossil thermal power ever, use of Computational Fluid Dynamics (CFD) software is
plants. Oka and other researchers around the world are working on possible and it has been used to simulate supercritical flow suc-
reactor designs that will bring this technology to operation (Oka cessfully. In recent years, numerical methods were validated
et al., 1995; Koshizuka et al., 1995; Schulenberg et al., 2014; against data found in the literature e.g. by Jaromin and Anglart or
Squarer et al., 2003). This technology is suitable for both large, Gu et al. (Jaromin and Anglart, 2013; Gu et al., 2008).
Nevertheless, aforementioned experiments or simulations were
performed only when uniform heat flux was applied at the wall. In
a real situation, it is different, and the heat flux in the fuel rod is
* Corresponding author.
** Corresponding author. never uniform and therefore, it is required to conduct such evalu-
E-mail addresses: jiyuzhao@cityu.edu.hk (J. Zhao), eycsoh@ntu.edu.sg (Y.C. Soh). ation. Only Wang et al. conducted an investigation when

http://dx.doi.org/10.1016/j.pnucene.2016.12.009
0149-1970/© 2016 Elsevier Ltd. All rights reserved.
12 M.K. Rowinski et al. / Progress in Nuclear Energy 97 (2017) 11e25

circumferentially non-uniformly heat flux was applied at the pipe taken into account. Previously, all estimations related to safety in
wall (Wang et al., 2012), however in reactor pressure vessel (RPV) NPP were conservative and therefore, were evaluated with safety
the power generated in fuel assembly is non-uniform axially. Water margins and always at the worst possible conditions that may occur
properties before and after the pseudo-critical point are signifi- in the investigated system. Nowadays, the nuclear industry and
cantly different. From the power generation point of view, the research tend to use best estimate methods such as CFD compu-
density is the most influential. During the transition from sub-to tations with uncertainty evaluation.
supercritical conditions, the fluid density decreases significantly The study was conducted using available ANSYS Fluent code.
(almost one order of magnitude) and therefore, at the inlet of the Geometry consists of fluid volume with the following boundary
fuel assembly moderation is much greater than at the outlet. conditions: mass flow inlet, pressure outlet, axis, and wall. The
Tashakor et al. (2012, 2013). performed an evaluation of power mesh was validated against available data with uniform heat flux.
distribution along fuel rods and found that the power distribution The non-uniform heat flux was programmed in User Defined
in the axial direction has sinusoidal shape with its peak moved to Function (UDF) file based on the position in the axial direction. The
the upstream section of the fuel assembly, the location is at around turbulence model used in the numerical investigation was low-
15% of total fuel rod length. Reynolds k-u Shear Stress Transport (SST), and this model has
The main thermal properties of the water at 25.5MPa are shown proven its ability to accurately simulate supercritical flow (Palko
in Fig. 1. Above the supercritical pressure boiling crisis does not and Anglart, 2008; Wen and Gu, 2010). Numerical investigation
occur with an increase of the temperature, and the properties are answers the question how the wall temperature distribution along
changing continuously. However, the change of values is signifi- fuel rod behaves under non-uniform heat flux at the wall. The PCT
cant, especially the specific heat has a large peak near the pseudo- cannot exceed safety limits during operation and therefore, it has to
critical point that is around 15 times higher than at any other point. be proven that under non-uniform heat flux it is possible to achieve
The other parameters are ca. 5e8 times lower at supercritical required outlet temperature without sacrificing any of the safety
conditions than at subcritical temperature. Thermal conductivity criteria.
can be characterized by a small peak just before the pseudo-critical
point, whereas viscosity has the lowest value right after pseudo- 2. Methodology e numerical approach
critical point and slightly increases with a rise in temperature.
Nevertheless, the value of this property is much lower than at Since the experimental investigation is expensive due to the fact
subcritical conditions. The only parameter that continuously de- that it requires building a facility that will be able to perform either
creases with temperature is density. scaled or non-scaled experiment, it is convenient to use numerical
Despite the lack of boiling crisis, it is possible to observe heat methods. The numerical methods may also save significant re-
transfer deterioration (HTD) phenomenon in specific conditions sources in terms of expensive instrumentation required to collect
that may lead to undefined, nevertheless much smaller than in data during the tests. Use of computer analysis may be much
comparison to dry out, the rise in peak cladding temperature (PCT). cheaper and more efficient since it is much easier to increase
HTD is not completely defined until this day since it behaves computational power. Therefore, use of numerical approach at the
differently than boiling crisis. Therefore, most of the current re- early design stage is appropriable.
searches are focused on the determination of the mechanism of One of the most common way to simulate real experiment is to
HTD and how to avoid it. On the other hand, sometimes enhance- use Computational Fluid Dynamics (CFD). It is a powerful tool that
ment of heat transfer (EHT) occurs, however, to be conservative in allows one to predict with satisfactory accuracy the phenomena
the safety analysis, the heat transfer enhancement is not being ongoing during an experiment with the only use of a computer.

Fig. 1. Thermal properties of water at 25.5 MPa.


M.K. Rowinski et al. / Progress in Nuclear Energy 97 (2017) 11e25 13

Hence, many researchers have tried to perform calculations using


commercial or self-developed CFD software.
The water properties during the transition from sub-to super-
critical conditions varies significantly and therefore, it may affect
the accuracy of calculations. The “NIST Reference Fluid Thermo-
dynamic and Transport Properties Database (REFPROP)” contains
properties data tables of many different fluids at sub- and super-
critical conditions including water and carbon dioxide that are
going to be used in Generation IV nuclear reactors. The water
properties are based on work conducted by J. Kestin et al (Kestin
et al., 1984; Kestin and Sengers, 1986). and J. V. Sengers et al.
(Sengers et al., 1984; Huber et al., 2009), they are based on Helm-
holtz free energy. Thus, it is possible to either create an UDF file
with water properties or use incorporated in commercial software
NIST real gas model. Since ANSYS Fluent has incorporated REFPROP
database, this software was used to perform calculations.
Solver used to perform pressure based calculations uses SIMPLE Fig. 2. Normalized heat flux distribution along the axis with different N1 and N2
scheme. Second order discretization was used for pressure, density, values.
momentum, and energy to solve the main governing equation,
while first order scheme was used for turbulent kinetic energy and
specific dissipation rate. In order to ensure sufficient accuracy Table 1
Values of N1 and N2 for different investigated case.
convergence criteria were set to residuals of 106 for continuity and
energy. Case No. Variables

Case 1 N/A e uniform heat flux


Case 2 N1 ¼ 1, N2 ¼ 1
Case 3 N1 ¼ 4, N2 ¼ 1
2.1. Axial power distribution Case 4 N1 ¼ 4, N2 ¼ 0.6
Case 5 N1 ¼ 4, N2 ¼ 0.2
Heat flux distribution is a reflection of power generation inside
the fuel element, it depends on burn-up, enrichment variation,
moderation, and a burnable neutron absorber. As it is commonly 2.2. Geometry
known the fission reaction in nuclear fuel occurs at the much
higher rate when sufficient moderation is provided. In typical Computational domain was simplified from full 3D model to 2D
PWRs, the power curve can be assumed to be a sinusoid with a peak axisymmetrical geometry in order to optimize time needed for
in the middle of the fuel channel. However, in SCWRs during the computations. It is possible due to the shape of the model and the
transition from sub-to supercritical conditions the density de- direction of the applied forces i.e. gravity and mass flux. The model
creases significantly and therefore the moderation decreases to- used is based on Ornatsky experiment (Ornatsky et al., 1970) i.e. the
wards the outlet. This phenomenon causes the peak of power vertical pipe of 1.05 m length and 3mm diameter (Fig. 3). This
distribution to move from the middle position towards the channel particular example was used since the data collected is within the
inlet (as it was shown by Tashakor et al. (2012).). On the other hand, operating parameters of typical SCWR. To ensure that the results
in LWRs during different stages of the fuel cycle, the power distri- would not be influenced by flow development, two additional
bution is changing and the peak moves towards the inlet, similar to domains of the length of 50 mm at the inlet and the outlet were
the predicted distribution of SCWR. Blakeman and March-Leuba added. Therefore, the flow is developed at the inlet to the heated
(1989) used the following formula to describe power distribution region and bulk flow at the outlet can stabilize.
shape at various stages of the cycle: The geometry was created and meshed in GAMBIT software.
h  iN2 Then the mesh was validated both in radial and axial directions. In
JðzÞ ¼ sin pðx=HÞ1=N1 (1) every mesh the main criteria was to obtain y þ value below unity, as
it was found in the literature review, only then the numerical
investigation can be accurate. The main reason is that the Low-
where:
Reynolds k e u SST turbulent model does not use any wall func-
tions. Since y þ can be defined as the non-dimensional distance
x e distance from the inlet, H e total height of the element, N1 e
between the wall and the first computational node, the first node
parameter that controls the height of heat flux peak, N2 e
parameter that controls the slope of heat flux.

Modifications of N1 and N2 make it possible to change the heat


flux distribution in two ways, either flatter the curve or shift the
peak to the front (Fig. 2). Therefore, it is possible to obtain distri-
bution at different stages of fuel cycle or different degrees of
moderation along the axis.
Five different cases were evaluated in order to better under-
stand how the power distributions can influence wall temperature
distribution along the axis. The reference case 1 has applied uni-
form heat flux, whereas the other 4 cases used eq. (1) to simulate
different power distributions. Table 1 shows a summary of all Fig. 3. Diagram of the geometry and boundary conditions (dimensions in mm, model
investigated cases. is unscaled).
14 M.K. Rowinski et al. / Progress in Nuclear Energy 97 (2017) 11e25

Fig. 4. Mesh comparison in radial direction. Different interval count.

was placed 0.0005mm from the wall. Therefore, the y þ value varies 2.3. Boundary and operating conditions
in the range of 0.3e0.7.
First, the validation was performed in the radial direction Boundary and operating conditions were taken from Ornatsky
(Fig. 4). The mesh of 125 interval counts in the radial direction was experiment, and a simplified diagram is shown in Fig. 3. The
chosen for further investigation since the result was placed be- following boundary conditions were set: mass flow inlet, pressure
tween meshes of 150 and 175 interval counts and it is sufficiently outlet, axis and wall (where heat flux is applied). In all investigated
accurate to the experimental solution. Only 150 interval counts cases, the mass flux (G) of 1500 kg m2 s1, inlet temperature (Tin)
were closer to the data, however, the difference between 125 and of 368.91 K and gravity acceleration of 9.81 m s2 is constant, only
150 meshes are negligible. In axial dimension three meshes were heat flux (q00 ) varies. The reference case 1 is characterized by uni-
investigated, each of them has different interval size i.e. 0.5mm, form q00 of 1810 kW m2, while the remaining cases are non-
1mm and 2mm (Fig. 5). The mesh of 2mm interval size has the best uniform as was explained in section 2.1. The heat flux was
accuracy among the others. Therefore, for later computations, mesh normalized to achieve an average q00 as in the reference case.
with 125 intervals counts in radial direction and interval size of Operating pressure is 25.5 MPa, it is above the critical point.
2mm in the axial direction was used. The heat transfer coefficient (HTC) was calculated according to

Fig. 5. Mesh comparison in axial direction. Different interval size.


M.K. Rowinski et al. / Progress in Nuclear Energy 97 (2017) 11e25 15

the following equation:  


00 v eI þ 12 v r  
q 1
h¼ (2) þ V$vr eI þ v2
Tw  Tb vt 2
00 000
¼ V$q þ g$vr  V$vp þ V$ðt$vÞ þ q (7)
where: q00 e local heat flux, Tw e local wall temperature, and Tb e
local bulk temperature. where: eI e internal specific energy, v e velocity, q00 e heat flux.
Whereas the Nusselt number is calculated as follow:

hL 2.3.2. Turbulence models


Nu ¼ (3) Until this day, many turbulent models have been developed.
k
They can be divided into many categories but the most important is
where: h e the HTC calculated from equation (2), L e the charac- the trade-off between computation time and accuracy. In the best
teristic length (hydraulic diameter in our calculations), and k e bulk case, the algorithm should be accurate and efficient/fast. Unfortu-
thermal conductivity. nately, those two requirements are in contradiction due to the fact
that turbulent flow is a complex problem and the accuracy depends
Numerical Approach on how many parameters are taken into account during
computation.
The computational fluid dynamics is based on Navier-Stokes Different turbulent models are commonly used in most of the
equations, unfortunately, it is impossible to solve it analytically commercial and non-commercial CFD codes. Moreover, every code
for every kind of flow pattern. Therefore, numerical methods that has a couple of turbulent models built-in in order to choose the one
simplify it are used i.e. a model that approximate fluid structure. that is the most convenient in investigated case. The following
subsections are based on ANSYS Theory Guide (ANSYS®, 2012a;
ANSYS®, 2012b) and they represent a couple of commonly used
2.3.1. Governing equations turbulent models. They describe k e ε, k e u, and Shear-Stress
The fluid mechanics is based on three fundamental principles, Transport (SST) k e u turbulent models that belong to a group of
namely on the laws of conservation of mass, momentum, and en- RANS (Reynolds-averaged NaviereStokes) turbulent models. Those
ergy, as discussed below. models are commonly used in research and industrial application
due to the best trade-off between computation time and accuracy.

2.3.1.1. Mass conservation. The mass conservation states that the 2.3.2.1. k e ε turbulent model. One of the most commonly used
mass in investigated system cannot be created nor destroyed, hence turbulent models in the free turbulent stream is k e ε model. It is a
it must remain constant. In other words, the mass of all inlets has to semi-empirical model that is based on transport equation model of
be equal to the mass at all outlets. This can be written as: turbulence kinetic energy (k) and its dissipation rate (ε) and
therefore, it can be said that it is a two-equation model. The tur-
V$v ¼ 0 (4) bulence kinetic energy k and dissipation rate ε are obtained from
transport equations:
where: v e velocity vector.
"  #
v v v m vk
ðrkÞ þ ðrkui Þ ¼ mþ t þ Gk þ Gb þ rε  YM þ Sk
2.3.1.2. Momentum conservation. The momentum conservation is
vt vxi vxj sk vxj
based on Newton's second law, which states that the acceleration of (8)
an object is dependent upon two variables - the net force acting
upon the object and the mass of the object. Therefore, the mo- and
mentum conservation equation has form of following differential "  #
equation: v v v m vε ε
ðrεÞ þ ðrεui Þ ¼ mþ t þ C1ε ðGk þ C3ε Gb Þ
vt vxi vxj sk vxj k
V$rvv ¼ Vp þ V$t þ rg (5)
ε2
 C2ε r þ Sε
where: r e density, p e pressure, t e shear stress vector, g e the k
gravity acceleration. (9)

where:
2.3.1.3. Energy conservation. The energy conservation is based on
the first law of thermodynamics, which states that the total energy Gk e generation of turbulence kinetic energy due to mean ve-
of the isolated system is constant. Since the energy cannot be locity gradients, Gb e generation of turbulence kinetic energy
created or destroyed it can only be transformed from one form to due to buoyancy, YM e contribution to fluctuation dilatation in
another. In case of fluid volume, it might be represented by: compressible turbulence to the overall dissipation rate, C1ε, C2ε,
  C3ε e are constants determined from experiments with air and
V$ rcp vT ¼ kV2 T (6) water, sk and sε e the turbulent Prandtl numbers for k and ε,
respectively, and Sk and Sε e user-defined source terms.
where: ET e the total energy of the fluid including potential and
kinetic energy, q e is the heat added to the control volume per unit This model is only valid for fully turbulent flows. It is possible to
time, P e is the work done by the system by the unit time. distinguish three versions of k e ε model and the differences are in
The corresponding energy conservation equation in differential accuracy for different types of flow e.g. low and high Reynolds
form can be expressed as: numbers.
16 M.K. Rowinski et al. / Progress in Nuclear Energy 97 (2017) 11e25

2.3.2.2. k e u turbulent model. The standard k e u model was During the years researchers tried to find a formula valid in both
developed by Wilcox (Wilcox and I. DCW Industries, 1998) that sub- and supercritical region. It is possible to distinguish two ap-
predicts free shear flow spreading rates. It is an empirical model proaches i.e. modification of Dittus-Boelter's equation or derivation
based on transport equations for turbulence kinetic energy (k) and of new correlations.
the specific dissipation rate (u), which is basically a ratio of ε and k. In 1957, Shitsman (Pioro et al., 2004) proposed a new formula
! based on the Dittus-Boelter relation:
v v v vk
ðrkÞ þ ðrkui Þ ¼ Gk þ Gk  Yk þ Sk (10)
vt vxi vxj vxj Nu ¼ 0:0243Re0:8 0:8
b Prmin (13)

and where Reb e states for Reynolds number calculated at a bulk fluid
! temperature and Prmin e states Prandtl number calculated at either
v v v vu bulk fluid temperature or the wall surface temperature, depending
ðruÞ þ ðruui Þ ¼ Gu þ Gu  Yu þ Su (11)
vt vxi vxj vxj on which is smaller.
Krasnoshchekov and Protopopov in 1959 (Pioro et al., 2004)
where: proposed a new formula for forced convective heat transfer in
water and carbon dioxide (CO2):
Gk e generation of turbulence kinetic energy due to mean ve-
locity gradients, Gu e generation of u, YM and Yu e dissipation of  0:11  0:33  0:35
mb kb cb
k and u, GM and Gu e effective diffusivity of k, and u, Sk and Sε e Nu ¼ Nu0 (14)
mw kw cpw
user-defined source terms.
where:
The k e u predicts with sufficient accuracy all effects at near wall
since it does not have to use any wall functions, while the flow in
the free stream is not so accurate. ðx=8ÞReb Pr
Nu0 ¼ pffiffiffiffiffiffiffiffiffiffiffi 2=3  (15)
12:7 ðx=8Þ Pr  1 1:07
2.3.2.3. Shear-Stress Transport (SST) k e u model. The SST model
was developed in order to combine accurate prediction of k e u at and:
near wall region with the performance of k e ε in the free stream in
the far field of the flow. It is done by converting k e ε model into k e 1
u with the use of multiplications results from both models by x¼ (16)
blending function and adding them together. Hence, the blending ð1:82 log10 Reb  1:64Þ2
function is unity in near-wall region in order to use k e u model and
where subscript b e states for evaluation at bulk temperature, w e
zero away from the surface and therefore, k e ε model is used. It can
for evaluation at wall temperature. The Prandtl number and specific
be concluded that the SST k e u combines k e u and k e ε in order to
heat should be averaged over the ranges of thermophysical varia-
provide accurate results both at near wall and in the free stream.
tions. The above correlation is valid in the range of pressure be-
Hence, it can be used in a wider spectrum of applications. More-
tween 22.3 and 32 MPa for water and 8.3 MPa for CO2.
over, researchers such as Wen and Gu and Kao et al (Wen and Gu,
Another common correlation for Nusselt number is the one
2011; Kao et al., 2010). have validated few turbulent models
developed by Bishop et al. (1964). proposed in 1964. The correlation
against supercritical flows and concluded that SST k e u is the most
was based on experiments done under following conditions:
accurate.
pressure in the range of 22.8e27.6 MPa, bulk temperature
282e527  C, mass flux of 651e3662 kg m2 and heat flux of
2.4. Convective heat transfer 0.31e3.46 MW m2.

During the years many empirical and semi-empirical correlation  0:43  


were formulated in order to predict the heat transfer coefficient in rw D
Nux ¼ 0:0069Re0:9 0:66
x Prx 1 þ 2:4 (17)
the supercritical flow. Therefore, the numerical investigation is rb x x
followed by a comparison of experimental and simulation results
with six correlation to find if any of available correlation is able to where D is the diameter of the pipe and x is an axial position.
correctly predict the obtained values of Nusselt number, hence the Swenson et al., in 1965 (Pioro et al., 2004) had found that
HTC. standard correlation for HTC is not valid in a rapid change of
The most common relation for Nusselt number is the one pro- properties during transition from sub-to supercritical environment
posed by Dittus-Boelter in 1930 (Zang et al., 2014): and therefore the following expression is recommended:

Nu ¼ 0:023Re0:8 Prn  0:923    


(12) hD DG ðHw  Hb Þmw 0:613 rw 0:231
¼ 0:00459 (18)
kw mw ðTw  Tb Þkw rb
it is valid for ratio of pipe length to diameter L/DH > 60, Reynolds
number Re > 10,000 and Prandtl number 0.7 < Pr < 100. The The following ranges of experimental results were used to
exponent n is equal to 0.4 for heating and 0.3 for cooling. The develop the correlation: pressure 22.8e41.4 MPa, mass flux
properties should be estimated at mean temperature i.e. bulk 542e2150 kg m2s1, wall temperature from 93 to 649  C and bulk
temperature. In our calculations, the n exponent has a value of 0.4. fluid temperature within 75e576  C. Hence, it covers the greatest
Albeit the fact that this correlation was not developed specifically spectrum of the results. Moreover, it also yields good accuracy with
for supercritical flow estimation, it is used here for comparison CO2.
since it is the most commonly used formula for computation of HTC Ornatsky et al., in 1970 (Ornatsky et al., 1970) proposed a vari-
in the case of forced convection in turbulent flows. ation of Shitsman's relation as follow:
M.K. Rowinski et al. / Progress in Nuclear Energy 97 (2017) 11e25 17

Fig. 6. HTC and temperature distributions against distance from the inlet. Reference case 1.

 0:3 are compared with the prediction of convective heat transfer co-
rw efficient calculated as described in section 2.4.
Nu ¼ 0:0243Re0:8 0:8
b Prmin (19)
rb

where subscript b e states for evaluation at bulk temperature, w e 3.1. Uniform heat flux
for evaluation at wall temperature. The above correlation was based
on his team's experimental data. 3.1.1. Temperature and HTC
The uniform heat flux computations is a benchmark case for the
results with non-uniform heat flux. Case 1 shows in Fig. 6 the
3. Results and discussion distribution of heat transfer coefficient (HTC) and temperatures
distribution along the pipe in reference case i.e. the heat flux is
Our results provide information on how the wall temperature constant. The rapid changes of HTC and wall temperature at the
responses to uniform and non-uniform heat flux. The main inlet and the outlet are related to the fact that the heated wall is
emphasis is on the correlation between the wall temperature and connected to two faces that are not heated and therefore, they are
heat transfer coefficient in particular cases. The numerical results not taken into further consideration. Wall temperature distribution

Fig. 7. Comparison of numerical investigation results with analytical prediction and with Ornatsky results.
18 M.K. Rowinski et al. / Progress in Nuclear Energy 97 (2017) 11e25

Fig. 8. HTC and temperature distributions against distance from the inlet. Case 2, N1 ¼ 1, N2 ¼ 1.

at the inlet of the pipe is linear, at around 0.6 m from the inlet the Protopopov's formula over predicts the results while the physics of
temperature rises quickly as HTC decreases i.e. the HTD phenom- HTD is preserved. The Bishop's and Ornatsky's formulas are the
enon occurs. However, once the HTC stabilizes and starts to in- most accurate in the case of analytical analysis. At the subcritical
creases again, the temperature starts to decrease before the outlet. region both shows very good agreement with experimental data,
The peak wall temperature is 802.85 K. The bulk temperature in- while after HTD they tend to over predict the convective heat
creases nearly linearly with the distance from the inlet. transfer. However, the Bishop's correlation shows the result to be
closer to the experiment in a higher range of coolant enthalpy.
3.1.2. HTC investigation
The results of Nusselt number comparison between Ornatsky
experimental results, numerical and the analytical investigation is 3.2. Non-uniform heat flux
shown in Fig. 7. Numerical results tend to slightly under predict the
convection effect, however, it is the closest to the results obtained The results of non-uniform heat flux i.e. cases 2e5 are presented
by Ornatsky in his experiments (uniform heat flux) in the whole in this section. The results show that it is possible to achieve lower
range of coolant enthalpy. While most of the analytical solutions maximum wall temperature. Nevertheless, the outlet temperature
are able to predict the place where HTD occurs (except Dittus- in every case is the same i.e. 660 K. As it was shown in Fig. 2 the heat
Boelter's and Shitsman's correlations), the results are usually over flux at the inlet and outlet are zero, hence the values do not change
predicted. Swenson's correlation is the best example of over pre- rapidly at those points when non-uniform heat flux is applied. The
diction, while most of the correlations can be characterized by investigation emphasis is put on the wall temperature distribution.
small peak right before the occurrence of HTD, this shows 3e4.5 The conduction heat transfer from the heated region to the un-
times higher peak than any other result. The Krasnoshchekov and heated regions is not included.

Fig. 9. HTC and temperature distributions against distance from the inlet. Case 3, N1 ¼ 4, N2 ¼ 1.
Fig. 10. HTC and temperature distributions against distance from the inlet. Case 4, N1 ¼ 4, N2 ¼ 0.6.

Fig. 11. HTC and temperature distributions against distance from the inlet. Case 5, N1 ¼ 4, N2 ¼ 0.2.

Fig. 12. Comparison of numerical investigation results with analytical prediction in case 3.
20 M.K. Rowinski et al. / Progress in Nuclear Energy 97 (2017) 11e25

3.2.1. Temperature and HTC phenomenon can be observed (Fig. 10). The maximum wall tem-
The case 2 (Fig. 8) shows the highest maximum wall tempera- perature can be found at a 0.5 m distance from the inlet. Moreover,
ture among all considered cases i.e. 1117.56 K. It is not a surprise the maximum wall temperature is slightly higher than in reference
since the peak heat flux value is in the middle of the pipe, it is near case, where uniform heat flux was applied and the value is 818 K.
the place where HTD is most likely to occur. HTD can be easily Fig. 11 shows the correlation between temperature and HTC in
noticed at around 0.5 m from the inlet, hence with rapid decrease of case 5. This case is characterized by the most uniform heat flux
HTC, the wall temperature increases rapidly with a peak around among others and by the lowest maximum wall temperature
0.7 m from the inlet. Then the temperature decreases with the swift among the investigated cases with non-uniform heat flux. The
increase of HTC towards the outlet. maximum value is lower than any other investigated cases
The heat flux in case 3 (Fig. 9) has higher peak value than the including the reference one, the value is 776.23 K. It is placed
case 2, however, due to the fact that it is moved to the front, the around 0.6 m from the inlet and it is followed by a rather rapid
maximum wall temperature is more than 100 K lower (1000.83 K). decrease of temperature.
Nevertheless, it is still significantly higher than in the reference
case. The decrease of the HTD is smoother than in case 2, moreover, 3.2.2. HTC investigation
the bulk critical temperature is achieved closer to the inlet and the The results of Nusselt number are used to make a comparison
peak wall temperature is around 0.3 m from the inlet. between analytical solutions proposed over the years by different
Albeit the fact that the peak heat flux value in case 4 is not the researchers and the results obtained during the numerical inves-
lowest and it is placed at the inlet region similar to case 3. The HTC tigation of case 3 (Fig. 12). Again, the Dittus-Boelter's correlation is
is steady and constant until about 0.45 m from the inlet where HTD unable to predict the HTD phenomenon. This time, Shitsman's

Fig. 13. Bulk flow properties distribution, case 1 (top) and case 3 (bottom).
M.K. Rowinski et al. / Progress in Nuclear Energy 97 (2017) 11e25 21

formula is able to predict it, nevertheless, the Nusselt number is reference case. It can be noticed that in the reference case the
over predicted in the whole range of enthalpy. Swenson's correla- specific heat at the outlet does not reach a peak point, whereas in
tion shows a high peak around the place where HTD occurs which case 3 it can be observed the curve has shape sinusoid-like and is
is followed by under prediction of the results after HTD occurs. The about to turn its direction or stabilize. All other properties reach
Krasnoshchekov and Protopopov's formula over predict the their minimum value at the outlet. In case 3 the properties tend to
convective heat transfer in the whole spectrum of enthalpy, how- stabilize at the exit.
ever, it is able to show where approximately HTD occurs. Again, Case 3 is the most interesting one since the heat flux distribu-
Bishop's and Ornatsky's formulas show the best agreement with tion is the closest to the power distribution found in calculations
the simulation results. However, it is unable to choose the best done by Tashakor et al. (2012). Therefore, comparison of the
correlation since the Bishop's formula is more accurate at the lower properties distribution along the pipe wall between case 1 (top)
enthalpy range and Ornatsky's correlation follows the trend in and case 3 (bottom) is shown in Fig. 14. In each case, the peak of
higher ranges. specific heat indicates where the transition from sub-to supercrit-
ical at the wall occur and moreover where HTD starts. Other
properties i.e. thermal conductivity, viscosity and density rapidly
3.2.3. Properties distribution decrease during the transition. When non-uniform heat flux is
Similar comparison was done for bulk flow. Fig. 13 shows how applied on the wall the properties fluctuates downstream after the
the bulk flow properties vary downstream. Reference case 1 (top) occurrence of HTD. Whereas, the properties downstream are rather
and case 3 looks very alike, the main difference is the range of constant when uniform heat flux is applied. The differences are
particular properties and how fast they change. Especially, the especially visible close to the outlet of the investigated domain. In
specific heat maximum value in case 3 is nearly twice as much the

Fig. 14. Properties distribution along the pipe wall, case 1 (top) and case 3 (bottom).
22 M.K. Rowinski et al. / Progress in Nuclear Energy 97 (2017) 11e25

Fig. 15. Radial distribution of fluid properties at different axial locations in case 1 (left) and case 3 (right).
M.K. Rowinski et al. / Progress in Nuclear Energy 97 (2017) 11e25 23

Fig. 16. Wall temperature dependence on mass flux in case 3.

case 3 it is possible to distinguish the second peak of specific heat at Moreover, the peak's width at each axial location is wider. What is
the pipe exit, it due to the decrease of heat flux and when its value worth to notice is the fact that the peaks at first two locations in
tends to zero the specific heat value gets close to the bulk flow case 1 are very similar, whereas the peak at a second location in
value. Similar with other properties, they increase downstream due case 3 is lower than at first location. The bulk specific heat increases
to lower values of heat flux, hence, tend to reach values of bulk flow downstream.
values. Viscosity behaves very similarly in both described cases. How-
Comparison of radial fluid properties at different axial positions ever, the values in case 1 are much more similar to each other and
between case 1 and case 3 is shown in Fig. 15. The radius distance is have lower values. The case 3 values have a wider spread and are
normalized and limited only to the part of interest e close to the 15e20% higher. Close to the wall, the properties values decrease
wall, where the properties vary the most. The first cross section at only to increase rapidly close to the pipe centre at all locations
distance of 600mm and 150mm (case 1 and case 3, respectively) (except the first one). The most significant difference is the fact that
from the inlet is located roughly where HTD starts to occur, the this decrease in values is more significant in case 3. The bulk vis-
following cross sections are located at the early phase of HTD, at the cosity decreases downstream.
place where highest temperature was found, when the wall tem- Density behaves similar in both investigated cases, the only
perature starts to decrease and when the wall temperature de- difference is the fact that the values have higher values in case 3.
creases. The values of specific heat are higher in case 1, especially The trend is the same in both cases. The bulk density decreases
the value located at last measured cross section is much higher. downstream.

Fig. 17. Wall temperature dependence on operational pressure in case 3.


24 M.K. Rowinski et al. / Progress in Nuclear Energy 97 (2017) 11e25

Thermal conductivity behaves similar to viscosity, case 3 is power distribution curve predicted in a real reactor. The specific
characterized by the wider spread of values and higher values in heat is higher in case 1, however, all other properties are higher in
general. The case 3 also has a small decrease in value close to the case 3. What is interesting is the fact that the viscosity and thermal
wall. What is worth to mention is the fact that the rapid increase of conductivity decrease close to the wall in case 3. Resulting in less
thermal conductivity is followed by almost instant stabilization. efficient heat removal than in case 1, where aforementioned
Again, the bulk values decrease towards downstream. properties are either stable or increases slowly close to the wall.
Therefore, the maximum wall temperature is higher in case 3.
3.3. Mass flux and pressure influence Sensitivity investigation in the case of mass flux and operating
pressure was also conducted. It was found that the temperature
As it was mentioned in the previous subsection, the case 3 is the distribution is more sensitive to changes in mass flux i.e. the Rey-
most interesting from our point of view. Hence, the aforementioned nolds number. The 5% difference in this parameter changes not only
case was further investigated to check how mass flux and operating the temperature distribution but also the place where HTD occurs,
pressure would influence the wall temperature distribution. In both whereas variation in operating pressure influences only the tem-
cases the values varies from 5% to þ5% of the reference values. perature distribution along the axis.
Variation in mass flux from 1425 to 1575 kg m2 s1 influences
temperature distribution significantly (Fig. 16). Especially, the References
smaller amount of coolant flow causes a high rise in temperature
distribution and moves the HTD phenomenon closer to the pipe ANSYS®, 2012. Academic Research, Release 12.0, Theory Guide, 4.4 Standard, RNG,
and Realizable k-epsilon Models. ANSYS, Inc.
inlet. The maximum wall temperature is 1157 K. ANSYS®, 2012. Academic Research, Release 12.0, Theory Guide, 4.5 Standard and
The operating pressure has less influence on the wall temper- SST k - Omega Models. ANSYS, Inc.
ature distribution than the variation of the mass flux. As it is shown Bishop, A.A., Sandberg, R.O., Tong, L.S., 1964. Forced Convection Heat Transfer to
Water at Near-critical Temperatures and Supercritical Pressures. P.I. Report
in Fig. 17, the maximum temperature barely achieves 1106 K at WCAP-2056, Editor. Westinghouse Electric Corp., Pittsburgh, USA.
operating pressure of 26.775MPa, the HTD phenomenon occurs at Blakeman, E.D. and J. March-Leuba, A Parametric Analysis Of Decay Ratio Calcula-
almost the same position as in the case with default operating tions In A Boiling Water Reactor Model, in Seventh Power Plant Dynamics,
Control and Testing Symposium. 1989: Knoxville, Tennessee, USA.
pressure. GIF, 2002. A Technology Road Map for Generation IV Nuclear Energy Systems.
GIF, 2009. GIF R&D Outlook for Generation IV Nuclear Energy Systems.
4. Conclusions Gu, H.Y., Cheng, X., Yang, Y.H., 2008. CFD analysis of thermalehydraulic behavior in
SCWR typical flow channels. Nucl. Eng. Des. 238 (12), 3348e3359.
Huber, M.L., et al., 2009. New international formulation for the Viscosity of H 2 O.
The numerical investigation of water flow at supercritical con- J. Phys. Chem. Ref. Data 38.
ditions in a vertical pipe at non-uniform heat flux at the wall has led Jaromin, M., Anglart, H., 2013. A numerical study of heat transfer to supercritical
to the following conclusions. Wall temperature distribution along water flowing upward in vertical tubes under normal and deteriorated condi-
tions. Nucl. Eng. Des. 264 (0), 61e70.
the axis is higher in three cases where non-uniform heat flux was Kao, M.-T., et al., 2010. Heat transfer deterioration in a supercritical water channel.
applied than the reference uniform case, whereas in only one case Nucl. Eng. Des. 240 (10), 3321e3328.
the maximum wall temperature was smaller. Therefore, it can be Kestin, J., Sengers, J.V., 1986. New international formulations for the thermody-
namic properties of light and heavy water. J. Phys. Chem. Ref. Data 15.
concluded that by adjusting the shape of heat flux distribution Kestin, J., et al., 1984. Thermophysical properties of fluid H2O. J. Phys. Chem.
curve it is possible to achieve lower temperatures. Assuring higher Ref. Data 13.
safety margins. This behavior is believed to be due to the fact that Koshizuka, S., Takano, N., Oka, Y., 1995. Numerical analysis of deterioration phe-
nomena in heat transfer to supercritical water. Int. J. Heat Mass Transf. 38 (16),
the heat flux peak is moved upstream where the coolant is 3077e3084.
subcritical and has better cooling abilities, moreover, the heat flux Liu, Q., Oka, Y., 2013. Core design for super fast reactor with all upward flow core
peak is lower. Hence, case 5 have the lowest maximum wall tem- cooling. Ann. Nucl. Energy 57 (0), 221e229.
Oka, Y., et al., 1995. Supercritical-pressure, light-water-cooled reactors for
perature among all investigated cases, including the reference one
improving economy, safety, plutonium utilization and environment. Prog. Nucl.
where uniform heat flux was applied. Energy 29 (Suppl. (0)), 431e438.
The occurrence of heat transfer deterioration (HTD) phenome- Ornatsky, A.P., L.P. Glushchenko, and E.T. Siomin. The research of temperature
conditions of small diameter parallel tubes cooled by water under supercritical
non is moved upstream the pipe even if the heat flux peak is in the
pressures. in Proceedings of the Fourth International Heat Transfer. 1970. Paris-
middle of the pipe. Heat transfer coefficient (HTC) behaves differ- Versailles, France.
ently than in constant heat flux case. The HTC tends to be uniform Palko, D., Anglart, H., 2008. Theoretical and numerical study of heat transfer
until the point where the HTD occurs, whereas when uniform heat deterioration in high performance light water reactor. Sci. Technol. Nucl. In-
stallations 2008, 5.
flux was applied the HTC was increasing. Pioro, I.L., Duffey, R.B., 2005. Experimental heat transfer in supercritical water
The comparison of numerical investigation with analytical so- flowing inside channels (survey). Nucl. Eng. Des. 235 (22), 2407e2430.
lutions and experimental data taken from Ornatsky experiment Pioro, I.L., Khartabil, H.F., Duffey, R.B., 2004. Heat transfer to supercritical fluids
flowing in channelsdempirical correlations (survey). Nucl. Eng. Des. 230 (1e3),
shown that numerical investigation has great advantages. While 69e91.
most of the analytical methods are able to predict the occurrence of Rowinski, M.K., White, T.J., Zhao, J., 2015. Small and Medium sized Reactors (SMR): a
HTD, only Bishop's and Shitsman's correlations are able to follow review of technology. Renew. Sustain. Energy Rev. 44 (0), 643e656.
Schulenberg, T., Leung, L.K.H., Oka, Y., November 2014. Review of R&D for super-
the trend. Unfortunately, in both cases, the results are over pre- critical water cooled reactors. Prog. Nucl. Energy 77, 282e299.
dicted at high values of enthalpy. Similar comparison was prepared Sengers, J.V., et al., 1984. Representative equations for the thermal conductivity of
for case 3 since no experimental data was available, the comparison water substance. J. Phys. Chem. Ref. Data 13.
Squarer, D., et al., 2003. High performance light water reactor. Nucl. Eng. Des. 221
is only between numerical and analytical solutions. This time, only (1e3), 167e180.
Dittus-Boelter correlation was unable to predict the occurrence of Tashakor, S., et al., 2012. Neutronic analysis of HPLWR fuel assembly cluster. Ann.
HTD. Again, the Bishop's and Ornatsky's correlations are the closest Nucl. Energy 50 (0), 38e43.
Tashakor, S., et al., 2013. Thermalehydraulic analysis of HPLWR fuel assembly
to the numerical solution, however, while Ornatsky's formula over
cluster. J. Supercrit. Fluids 77 (0), 91e99.
predicts the values it follows the trend for the whole range of Wang, H., et al., 2012. Heat transfer characteristic of water at near critical pressure
enthalpy. in circumferentially non-uniformly heated vertical tubes. Int. J. Therm. Sci. 54
It was found that the properties distribution along the wall are (0), 167e175.
Wang, H., et al., 2014. Experimental investigation of heat transfer from a 2x2 rod
slightly different in two compared cases i.e. when uniform heat flux bundle to supercritical pressure water. Nucl. Eng. Des. 275 (0), 205e218.
is applied and when applied non-uniform heat flux is similar to the Wen, Q.L., Gu, H.Y., 2010. Numerical simulation of heat transfer deterioration
M.K. Rowinski et al. / Progress in Nuclear Energy 97 (2017) 11e25 25

phenomenon in supercritical water through vertical tube. Ann. Nucl. Energy 37 Wilcox, D.C., 1998. In: I. DCW Industries (Ed.), Turbulence Modeling for CFD. La
(10), 1272e1280. Canada, California.
Wen, Q.L., Gu, H.Y., 2011. Numerical investigation of acceleration effect on heat Zang, J., et al., 2014. A method of extending subcritical heat transfer correlations to
transfer deterioration phenomenon in supercritical water. Prog. Nucl. Energy 53 supercritical conditions. Nucl. Eng. Des. 266 (0), 186e193.
(5), 480e486.

You might also like