Computation of Nusselt numbers in plate heat

exchangers using CFD with Reynolds-averaged

Navier–Stokes equations
Mads Kyed
Fachbereich Maschinenbau
Flensburg University of Applied Sciences
Kanzleistrasse 91-93, 24943 Flensburg
Email: mads.kyed@hs-flensburg.de

Nico Šubić∗
Fachbereich Maschinenbau
Flensburg University of Applied Sciences
Kanzleistrasse 91-93, 24943 Flensburg
Email: nico.subic@hs-flensburg.de
April 14, 2023

A CFD simulation based on a Reynolds-averaged Navier–Stokes model is employed to

compute Nusselt numbers in corrugated plate heat exchangers. The approach is verified
via the Dittus-Boelter formula in a simplified tube geometry. A corrugated plate is
analyzed and the effect of a chevron angle on the Nusselt number is investigated.

Keywords: Nusselt number, corrugated plate, heat exchanger, RANS models.

1 Introduction achieve higher levels of efficiency compared

to regular shell and tube heat exchangers.
A plate heat exchanger (PHE) is a de- An accurate prediction of the heat transfer
vice in which two fluids separated by solid is critical in many applications, for exam-
plates exchange heat. Corrugation pat- ple in optimal sizing of PHEs. The clas-
terns in the plates induce turbulence in sical LMTD method for PHE sizing relies
the fluids, thereby increasing the convec- on the Nusselt number to incorporate the
tive heat transfer, which enables PHEs to effect of turbulence on the convective heat

1
transfer. Computation of Nusselt numbers Reynolds numbers in ANSYS for a special
in PHEs to date typically employ heuristic type of PHEs, not including, however, geo-
correlations like the Dittus-Boelter formula metrical parameters.
and derivatives hereof, in which the coeffi- We use the OpenFoam framework to
cients are empirically adapted from mea- carry out CFD simulations, and GMSH to
surements. In this paper we propose to create meshes.
use turbulence models and CFD to com-
pute the Nusselt number in a PHE. We 2 Model
first verify our computation method on a
simple tube geometry, on which the Dittus- We use the k −ω −SST model, a two equa-
Boelter formula is known to yield a suf- tion turbulence model within the RANS
ficiently precise prediction, and then em- models, coupled with the standard energy
ploy the method to demonstrate the effect equation to compute the heat transfer per-
of a chevron angle on the performance of a formance. The turbulent length scale is
PHE. given in terms of the hydraulic diameter.
Our results show that a CFD based nu- Initial experiments showed that the k −
merical scheme can be established to ro- ω − SST model yields more accurate Nus-
bustly compute the Nusselt number of a selt number computations in our setting
PHE from the Reynolds number, Prandtl than both the k − ε and the standard k − ω
number and geometric parameters that model. The k−ε model in particular proved
describe the corrugation patterns in the to be inadequate with very large deviations.
plates. Algorithms based on such schemes From the inferior behavior of the k − ε
may be used to generate large data sets model we conclude that the more detailed
from which improved analytic formulas for treatment of wall effects in the k−ω models
the Nussselt number of a PHE can be con- are critical to Nusselt number computation
structed. We consider this paper a step on in heat exchangers.
the way towards such a framework.
The influence of geometrical parameters
3 Simulation
on PHE performance has been investigated
in previous works. Dović and Švaić [2] have Initial simulations were carried out for a
analyzed the influence the length-width re- flow in a smooth pipe. In this simplified
lation of a single plate and the corruga- geometry, the Dittus Boelter formula
tion angle have on the heat transfer coef-
ficient in a PHE. Skočilas and Palaziuk [7] N u = 0.023Re0.8 P rn (3.1)
have investigated the influence of the cor-
rugation angle on the heat transfer coeffi- (n = 0.3 for cooling and n = 0.4 for heat-
cient and on the pressure drop for a single ing) is known to be valid when the relation
channel. Dović, Palm and Švaić [3] devel- between the diameter D and the length L is
oped a generalized approach to computa- at least 10, and the flow is fully developed
tion of the Nusselt number from empirical with Reynolds number greater than 10000
measurements and CFD simulations. That and Prandtl number 0.6 ≤ P r ≤ 160.
CFD can be a viable source of data was The geometry for the pipe flow was gen-
also shown by Luan et al. [5] who created erated with GMSH. We employed a pure
a correlation between Nusselt, Prandtl and hexaeder mesh based on a 2D hex mesh of

2
 z
Property value

Mol weight 18 mol x

kg px
Density 997 m3
J
Specific heat capacity 4180 kgK

Prandtl number 6.09 β

Table 1: Fluid properties water

Figure 1: Geometric properties of the corruga-

a circle. We extruded this surface twice tion pattern
to generate an entrance region and a fully
developed region, and used the extrusion
length as entrance length for the flow. We
used the Reynolds based approach to en- b
sure an entrance length in correspondence
with the flow development:

lentrance = 1.359DRe0.25 . (3.2)

pc

We employed the material properties for

the fluid given in table 1. As boundary Figure 2: Geometric properties of the corruga-
conditions we assumed a constant heat flux tion pattern
W
of 1000 m 2 on the outer cylindrical bound-
ary, a fixed temperature of 298.15K on the coefficient with
inlet, and an adiabatic outlet (zero tem- q
perature gradient). The zero gradient con- h= . (3.3)
twall − tf luid
dition was also employed at the outlet for
the velocity field. The inlet velocity was The heat transfer coefficient was used to
set to be constant and directed in direc- obtain the Nusselt number
tion of the pipe. Simulations with Reynolds hdhyd
numbers in the range 20000 − 40000 with a Nu = (3.4)
κ
step size of 5000 were carried out. High
with
Reynolds number were chosen since the ηcp
κ= . (3.5)
Dittus Boelter equation is only valid for Pr
Re ≥ 10000. The Nusselt number was cal- For the plate simulations we employed
culated in post processing from the area av- a sinosodial profile extruded with corruga-
erage temperature of the heated wall in the tion angle β. The same scheme as for the
developed region and the bulk temperature pipe was used to compute the Nusselt num-
of the fluid. With the prescribed heat flux ber. Usually the hydraulic diameter of ar-
at the walls we computed the heat transfer bitrary geometries can be estimated with

3
 Re N uCF D N uDB Deviation
4Achannel
dhyd = , (3.6)
Acontact 20000 131.16 130.74 0.32 %
where Achannel represents the channel flow 25000 157.19 156.29 0.58 %
area and Acontact the wetted surface. How-
ever [6] presents an approach based on the 30000 183.05 180.83 1.23 %
surface enlargement factor ϕ and the chan-
35000 208.01 204.56 1.68 %
nel width b between two plates, which we
employed with the additional relation for 40000 232.69 227.63 2.22 %
the corrugation pitch according to [4]
Table 2: Results of the simulation for the pipe
2b
dhyd = , (3.7)
ϕ
4 Results
where ϕ is given by
4.1 Pipe
!0.5
1  π 2
The results for the simulations are given
ϕ= 1+ 1+ γ2 +
6 2cos(β) in Table 2. We notice that the deviation
!0.5 ! from the Dittus-Boelter formula (3.1) is
π
 2
4 1+ √ γ2 smaller for lower Reynolds numbers, which
2 2cos(β) matches the results from [1]. Further we
(3.8) notice that the deviation is positive for ev-
and γ is the aspect ratio of the corrugation ery simulation. We deduce a slight over-
defined as prediction of the turbulent effects, which
2b is known to happen in the framework of
γ= (3.9)
px k − ω models. In general the results show
good agreement with the Dittus-Boelter
with px the corrugation pitch in main flow
formula (3.1), and we thus conclude that
direction. To extract the Nusselt num-
our scheme is viable to extract Nusselt
ber from the simulation we employed again
numbers.
(3.3) and (3.4), with dhyd given by (3.7).
We simulated Reynolds numbers in the
range 2500 − 4000 with a step size of 500. 4.2 Corrugated plate
Constant mass flows were used as bound-
ary conditions on both the inlet and outlet. Concerning PHEs, we recall the following
W expected behavior:
A constant heat flux of 500 m 2 was chosen
as boundary condition on the fluid-plate in-
terfaces, which corresponds to a total heat • A steeper chevron angle reduces the
W pressure drop.
flux of 1000 m 2 similar to the pipe. The
outer plate walls were assumed to be adi-
abatic. A no-slip boundary condition was • A steeper chevron angle reduces the
used for the velocity field on the fluid-plate heat transfer coefficient and therefore
interfaces. the Nusselt number.

4
 Nu simulation 7000 Delta P 70 degree Angle
Nu Dittus Boelter Delta P 40 degree Angle
220 6000

200 5000

Delta P [pa]
4000
Nu [1]

180
3000
160
2000

140
1000

20000 22500 25000 27500 30000 32500 35000 37500 40000 2000 2250 2500 2750 3000 3250 3500 3750 4000
Re [1] Re [1]

Figure 3: Nusselt numbers from simulation Figure 5: Pressure drop from simulations with
and from the Dittus Boelter formula a 70 and 40 degree chevron angle

Nu 70 degree Angle
160 Nu 40 degree Angle
is observed for the pressure drop. The ob-
150
served effect of changing chevron angles on
140
the Nusselt number is also in accordance
130
the expected behavior.
Nu [1]

120

110
We note that a validation based on com-
100
parative measurements was not carried out.
90
We intend to do so in future works.
2000 2250 2500 2750 3000 3250 3500 3750 4000
Re [1]

5 Conclusion
Figure 4: Nusselt numbers from simulations
with a 70 and 40 degree chevron an-
We computed Nusselt numbers for PHEs
gle
with CFD simulations based on turbulence
modelling. We first validated our scheme
• With increasing Reynolds number, the by computing the Nusselt number for a flow
pressure drop and Nusselt number in- in a pipe and comparing the results with
crease. the Dittus Boelter formula. We then em-
ployed the algorithm on corrugated plates
• With increasing Reynolds number, the and investigated the effect of the corruga-
increment in the Nusselt number for tion angle on the Nusselt number, which we
steeper angles is lower than that of observed to be in accordance with expected
broader angles. physical behavior. We conclude that our
scheme is valid.
• With increasing Reynolds number, the
increment in pressure drop for steeper
angles is lower than that of broader an- References
gles.
[1] X. Chi and T. I.-P. Shih. Bulk Temper-
In our simulations, we find that the ature, Heat-Transfer Coefficient, and
Nusselt number increases with increasing Nusselt Number-Revisited. Wiley, 2012.
Reynolds number as expected. The same 4

5
[2] D. Dovic and S. Svaic. Influence
of chevron plates geometry on per-
formances of plate heat exchangers.
Tehnicki Vjesnik, 14:37–45, 01 2007. 2

[3] D. Dović, B. Palm, and S. Švaić. Gen-

eralized correlations for predicting heat
transfer and pressure drop in plate heat
exchanger channels of arbitrary geome-
try. International Journal of Heat and
Mass Transfer, 52(19):4553–4563, 2009.
2

[4] C. S. Fernandes, R. P. Dias, J. M.

Nóbrega, and J. M. Maia. Laminar flow
in chevron-type plate heat exchangers:
Cfd analysis of tortuosity, shape fac-
tor and friction factor. Chemical En-
gineering and Processing: Process In-
tensification, 46(9):825–833, 2007. Se-
lected Papers from the European Pro-
cess Intensification Conference (EPIC),
Copenhagen, Denmark, September 19-
20, 2007. 4

[5] Z. jian LUAN, G. min ZHANG,

M. cheng TIAN, and M. xiu FAN.
Flow resistance and heat transfer char-
acteristics of a new-type plate heat ex-
changer. Journal of Hydrodynamics,
Ser. B, 20(4):524–529, 2008. 2

[6] H. Martin. A theoretical approach

to predict the performance of chevron-
type plate heat exchangers. Chemical
Engineering and Processing: Process
Intensification, 35(4):301–310, 1996. 4

[7] J. Skocilas and I. Palaziuk. Cfd simu-

lation of the heat transfer process in a
chevron plate heat exchanger using the
sst turbulence model. Acta Polytech-
nica, 55(4):267–274, Aug 2015. 2