Renewable and Sustainable Energy Reviews 134 (2020) 110091

Contents lists available at ScienceDirect

Renewable and Sustainable Energy Reviews

journal homepage: http://www.elsevier.com/locate/rser

Thermo-hydraulic analysis of compact heat exchanger for a simple

recuperated sCO2 Brayton cycle
V. Pandey, P. Kumar, P. Dutta *
Department of Mechanical Engineering, Indian Institute of Science, 560012, Karnataka, India

A R T I C L E I N F O A B S T R A C T

Keywords: Supercritical carbon dioxide (sCO2) Brayton cycle-based power plants are being extensively explored as viable
Supercritical carbon dioxide alternatives to traditional Rankine based steam power plants. Higher temperatures at the turbine exit in a sCO2
Recuperator cycle provide better opportunities for heat recuperation, thus improving overall cycle efficiency. Among the
Printed circuit heat exchanger
various heat exchanger configurations available, compact microchannel heat exchangers commonly known as
Thermal resistance network
Pressure drop
Printed Circuit Heat Exchangers (PCHE’s) are typically used in sCO2 power plants. In the current work, a hybrid
Zig-zag approach comprising of a Thermal Resistance Network (TRN) model coupled with a CFD model for estimating
Brayton power plant local heat transfer and pressure drops is presented for a PCHE core with straight and zigzag channel configu­
rations. Full-scale TRN model is used to optimize the overall stack dimensions based on minimum rate of heat
loss from the external surfaces of the PCHE core. The TRN model accounts for the thermo-physical property
variations of sCO2 along the channel length to effectively capture the channel pressure drop and heat transfer.
This is achieved by discretizing the heat transfer domain comprising of alternatively stacked hot and cold streams
into sub-heat exchangers to account for variations in thermophysical properties while calculating the nodal
friction factors and local heat transfer coefficients. CFD simulations are performed for a full length of a single
stack of hot and cold fluid streams to arrive at corrected heat transfer and pressure drop correlations. Thermo-
hydraulic analysis is performed for a range of channel hydraulic diameters and channel mass flow rates for both
straight and zig-zag configurations to deduce optimum stack dimensions. The efficacy of the hybrid model is
demonstrated with a case study of a counterflow recuperator used in a simple recuperated 1 MW sCO2 Brayton
power plant.

recuperated sCO2 cycle with the following conditions (Tturbine, inlet =

527 ◦ C, pturbine, inlet = 20,900 kPa, Tcompressor, inlet = 35 ◦ C, pcompressor, inlet =
1. Introduction
7500 kPa, ηisen,turbine = 75%, ηisen,compressor = 80%) [5], shows that if the
recuperator effectiveness improves from 0.65 to 0.95, the overall cycle
Supercritical carbon dioxide (sCO2) Brayton cycle-based power
efficiency shoots by nearly 50% from 18.5% to 27.58%. This simple
generation systems, have been extensively evaluated in recent years for
first-hand estimate highlights the importance of a recuperator design in
several potential applications, particularly for nuclear, concentrated
a sCO2 cycle. Present literature suggests use of diffusion bonded Printed
solar power (CSP) and waste heat recovery (WHR) units [1–3]. Several
Circuit Heat Exchangers (PCHEs) for sCO2 recuperator. PCHEs are re­
studies have revealed that sCO2 Brayton power plants offer higher cycle
ported to withstand pressures upwards of 4 × 104 kPa and temperature
efficiencies compared to conventional steam Rankine and gas Brayton
up to 1000 ◦ C with compactness densities of nearly 5000 m2/m3 [6]. The
power cycles for similar turbine inlet temperatures [4]. Cycle operation
channels in a PCHE are photo-chemically etched on independent hot and
entirely in the supercritical state coupled with lower operating pressure
cold plates. Subsequently, these are alternately stacked, and diffusion
ratios and relatively high turbine exit temperatures provides significant
bonded to form the core of the PCHE. The thickness of plates varies
opportunity for recuperation, leading to a highly efficient and compact
between 1.5 mm and 3 mm and are usually made of either stainless steel
power block. Extensive research over the past decade suggest that heat
or Inconel depending on the operating conditions [7].
exchangers, mainly the recuperator performance, is the key factor
Extensive research has gone into the optimization of the PCHE flow
contributing to successful realization of the sCO2 Brayton technology. A
paths. The flow paths are usually categorized as continuous [8–14] and
back of the envelope calculation for a recuperator operating in a simple

* Corresponding author.
E-mail address: [email protected] (P. Dutta).

https://doi.org/10.1016/j.rser.2020.110091
Received 29 January 2020; Received in revised form 9 June 2020; Accepted 12 July 2020
Available online 12 September 2020
1364-0321/© 2020 Elsevier Ltd. All rights reserved.
V. Pandey et al. Renewable and Sustainable Energy Reviews 134 (2020) 110091

Nomenclature ṁeqn Equivalent channel mass flow rate [kg/s]

ṁt Total mass flow rate [kg/s]
sCO2 Supercritical carbon dioxide H Overall stack height of recuperator core block [m]
PCHE/s Printed Circuit Heat Exchanger/s W Overall stack width of recuperator core block [m]
TRN Thermal Resistance Network Δx Discretized length [m]
TIT Turbine Inlet Temperature [◦ C] Pr Prandtl number [− ]
HEx/s Heat Exchanger/s v Bulk average velocity at a cross section [m/s]
d EEdge of square cross-scetiond Edge of Edge of square Nt Total number of channels in the recuperator
cross-section [m] N Total number of channels for hot/cold streams in the
dh Hydraulic diameter of square cross-section [m] recuperator
w Width between two channels in the same plate [m] cp Specific heat [J/kgK− 1]
t Thickness between two channels of different plate [m]
m Number of columns in the PCHE core block [− ] Subscripts
n Number of rows in the PCHE core block [− ] hot Hot stream
L Channel length [m] cold Cold stream
Lc Characteristic length of external surface of PCHE core c Convection
block [m] tc Transverse conduction
Ac Cross sectional area of square channel [m2] a Axial
f Fanning factor [− ] eqn Equivalent
Nu Nusselt Number [− ] in Inlet
Re Reynolds number [− ] semi Semi-circular
Ra Rayleigh number] h Hydraulic
p Static pressure [Pa] TH Thermal
k Number of elements in the fluid stream Greek symbols
Q̇ Heat transfer rate [W] ρ Density [kg/m3]
R Resistance [K/W] μ Dynamic viscosity [Pa-s]
U Overall heat transfer coefficient [W/m2 K− 1] ε Effectiveness [− ]
Aht Heat transfer area [m2] λ Pitch of the zigzag channel [m]
ks Thermal conductivity of solid [W/mK− 1] θ Angle of turn for the zigzag channel [◦ ]
T Temperature [◦ C]
ṁc Channel mass flow rate [kg/s]

discontinuous [15–19] based on channel geometry. Straight, zigzag, and were reported by other researchers as well [26–28]. Ngo et al. [28]
wavy channels fall under the category of continuous flow channels, tested PCHE with S-shaped fined plates. This PCHE showed 6–7 times
whereas, fin type channels comprising of S-shaped and air-foil fins come lower pressure drop while maintaining almost identical heat transfer
under the discontinuous category. Use of zig-zag channels is widely performance as compared to a conventional zig-zag channel PCHE.
reported in the literature on account of high compactness density Saeed et al. [29] reported pressure drop associated with a sinusoidal
(80–200 kg/MW) and ability to handle higher heat fluxes [20]. Yang fin-based geometry to be ~2.5% lower than a zigzag channel. This work
et al. [21] optimized the size of PCHE using non-dominated sorting suggests that overall performance can be improved using sinusoidal
genetic algorithm II (NSGA-II) coupled with surrogate modelling, fin-based geometries in comparison to zig-zag channels for both low and
wherein, temperature rise and pressure drop were used as objective high Reynolds number cases.
functions. The results indicated that the height and width of the PCHE Majority of studies related to PCHE modelling use commercial CFD
core should be designed as small as possible to obtain the highest tem­ solvers with periodic and symmetric boundary conditions at the outer
perature rise. However, to reduce the pressure drop the height should be surfaces. For overall stack modelling and optimization, conventional
increased as compared to the width to achieve more flow area. Baik et al. full-scale CFD based approach is tedious and time-consuming. In addi­
[22] numerically investigated the effect of waviness factor, amplitude tion, full scale CFD models require large number of fine grids to resolve
and the periodicity of a zig-zag channel on the thermal performance of channel level flows and heat transfer over the entire heat exchanger
the PCHE. The paper reported a monotonically increasing thermal per­ domain, thus making it computationally expensive and impractical [9,
formance of wavy-channel as either of the amplitude or period was 25,30–32]. Despite these hurdles, few researchers have attempted full
increased. Lee and Kim [23] investigated use of arc-shaped ribs for sCO2 scale simulations. Figley et al. [33,34] performed a numerical simula­
cooling in zig-zag channels to enhance heat transfer rate and reduce tion to understand the thermo-hydraulic performance of full-scale PCHE
pressure drop. Jing et al. [24] numerically investigated eight multi-flow using ANSYS Fluent® with helium as the working fluid. Simulations
paths (MFPs) PCHEs with different channel configurations. Arc ribs such were performed for a column of 20 straight channels comprising of 10
as 1/3 ribs, 2/3 ribs, and semi-circular ribs were introduced into the hot and 10 cold channels arranged in a single banking configuration. In
MFP channels. The semi-circular ribs provided the best heat transfer another notable work, Daouk et al. [35] simulated a core sub-stack
with a 7.3% improvement in the Nusselt number at the cost of additional comprising of 20 channels using COMSOL Multiphysics® model
hydraulic loss compared to smooth case. In the category of discontin­ instead of the complete PCHE stack with 180 channels. The percentage
uous flow channels with sCO2 as the working fluid, Kim et al. [25] pressure drop reported was 0.15% compared to 0.13% reported by
performed a numerical study with staggered air-foil shape fins etched on Figley et al. [33] using identical number of channels operating under
plates. Owing to the increased heat transfer area and streamlined shape similar conditions.
of the air-foil fins, the footprint of the PCHE reduced significantly by Quasi one-dimensional models based on Logarithmic Mean Tem­
nearly 20 times compared to continuous zig-zag channels having iden­ perature Difference (LMTD) and effectiveness-NTU methods have also
tical pressure drop and heat transfer performance. Similar observations been reported to model the PCHE core. These bulk methods are unable

2
V. Pandey et al. Renewable and Sustainable Energy Reviews 134 (2020) 110091

to capture the correct temperature profile along the channel length due It is observed that one-dimensional discretized approaches reported
to non-linear variation in the thermo-physical properties [36,37]. Ke in the literature for two channel/unit cell models do not effectively
et al. [38] performed a comparison between LMTD, constant property capture the heat transfer rate and pressure drop of a full-scale PCHE. The
segmented calculation (CPSC) and variable property segmented calcu­ present study aims to fill in this gap by introducing a hybrid model using
lation (VPSC) methods for a PCHE with straight channels having a the concept of Thermal Resistance Network (TRN) and CFD. Local scale
semi-circular cross section. VPSC was reported to be more effective in heat transfer and pressure drop correlations derived from CFD simula­
capturing the variations in thermo-physical properties of sCO2 tions serve as inputs in the TRN model for thermo-hydraulic perfor­
compared to CPSC and LMTD methods. Utamura et al. [39] proposed use mance analysis of the complete PCHE stack. Subsequently, the hybrid
of Generalized Mean Temperature Difference (GMTD) as an improve­ model is used to analyse a sCO2 recuperator employing straight and
ment over the LMTD method, while others suggested the of use of dis­ zigzag channels with a square cross-section. The overall stack di­
cretized approaches based on effectiveness or conductance [40,41] to mensions of straight and zigzag channel based PCHEs are optimized
account for the variable thermo-physical properties along the flow. based on minimum heat loss from the outer exposed surfaces. The
However, there are limitations of effectiveness and conductance models. modelling procedure is presented for a recuperator used in a simple sCO2
These methods do not consider the effects of pressure drop and pressure Brayton power cycle producing a net power output of 1 MW.
differentials along channel length. Guo [42] employed a segmented
design method for straight semi-circular PCHE to accommodate the 2. Recuperator’s role in sCO2 cycle
variable thermo-physical properties of sCO2 in PCHE. For a fixed heat
load, extremums for local heat conductance, local effectiveness, local The thermodynamic cycle for a simple recuperated sCO2 Brayton
LMTD and local entransy dissipation number corresponding to the power plant is shown in Fig. 1(a) on a T-s diagram. The process 1–2′ is
intersection of heat capacity rates of both fluids were observed. Using isentropic compression, 2–3 is the cold stream of the recuperator which
CFD based approach, Bennett et al. [43] examined the design and is heated by hot stream following the path 5′ -6, path 4–5′ is the isen­
operating factors influencing the size of the data sampling interval tropic expansion in the turbine, and 6–1 is the heat rejection in the gas
which were used in the segmented design method of straight and zigzag cooler. The corresponding schematic of the system is presented in Fig. 1
channel-based PCHEs. The results showed that the required data sam­ (b).
pling interval was primarily driven by bend angle and mass flow rate for The recuperator alone can contribute 60–70% of the total heat
zigzag and straight channel, respectively. transfer in sCO2 power cycle [47]. An efficient design of recuperator can
Other discretized approaches include calibration between the reduce the cost and complexity associated with gas cooler design. Hence,
experimental/manufacturer’s data and predicted data using 1-D it is now well established that the overall cycle efficiency of sCO2 power
modelling tool by introducing some form of multiplier coefficients cycle is highly influenced by the recuperator performance. PCHEs are
into the conventional correlations. Bone et al. [44] presented a method found to be most effective for this purpose. For sCO2 using solar heat, the
to develop accurate on- and off - design models of heat exchangers cycle efficiency has a direct influence on the cost of power generation,
wherein, a 1-D discretized heat exchanger is fitted to experimental data and the CSP community is eager to see this technology succeed as a
using non-linear least squares optimization. Marchionni et al. [45] re­ viable renewable energy option. In this context, Dutta et al. [48] dis­
ported a reduced order 1-D model using a commercial tool GT-SUITE™ cussed the influence of performance and size of recuperator on overall
for predicting the thermo-hydraulic performance of a complete PCHE cycle efficiency considering various turbine and compressor isentropic
stack. The reduced order model was validated against manufacturer’s efficiencies. Based on recuperator heat transfer ratio (RHTR), it was
data for a 630 kW PCHE recuperator. Baik et al. [46] developed a proposed that this ratio must not exceed 1 considering the critical size of
modelling tool KAIST_HXD based on 1-D finite difference method (FDM) the recuperator. To attain this condition, the minimum turbine isen­
to account for the variable thermo-physical properties by solving mo­ tropic efficiency (ηisen,turbine) should be 92% corresponding to a
mentum and energy equations. The difference between the heat transfer compressor isentropic efficiency (ηisen,compressor) of 70%. At a given ηisen,
performances obtained from experimental data and code results was less compressor of 80%, if ηisen,turbine was increased from 85% to 95%, the cycle
than 7%. Following this study, Kwon et al. [37] developed a modified efficiency improved by 11% with a reduction in RHTR from 1.05 to 1.
LMTD approach based on the non-dimensionalized form of correction Noriega [49] reported that for a fixed set of cycle conditions (Tcompressor,
factors to capture the variation in the thermo-physical properties with inlet = 31.25 C, pcompressor, inlet = 74,00 kPa, ηadia,turbine = 93.4%, ηadia,
◦

reduced computational time. compressor = 85%, ηheater = 90%, ΔPLS = 2%) and cycle thermal efficiency

Fig. 1. (a) Simple recuperated sCO2 closed loop Brayton power cycle on a T-s diagram with high side pressure of pH = 21,225 kPa and low side pressure of pL = 75,00
kPa, (b) Schematic of a simple recuperated sCO2 Brayton cycle.

3
V. Pandey et al. Renewable and Sustainable Energy Reviews 134 (2020) 110091

1 MW net output,
Trecup,inlet = 10 C
(%)

cycle

qrecup
cycle

precup
Fig. 3. Variation of cycle efficiency and recuperator heat load with pressure
drop across the hot and cold streams of the recuperator.

Fig. 2. Variation of cycle efficiency with net specific work output for a turbine
inlet temperature of 527 ◦ C and compressor inlet conditions of 75,00 kPa and 2.1. Effect of recuperator pressure drop on cycle efficiency
35 ◦ C [5].
Variation of cycle efficiency and heat load with pressure drop across
both streams of the recuperator are presented in Fig. 3. The cycle state
points (except recuperator inlet pressure conditions) including the
of 50%, higher values of regenerator effectiveness improves the thermal compressor inlet conditions presented in Table 1 are fixed throughout
energy recovery and hence decreases the TIT. the analysis. As pressure drop across the recuperator increases (on ac­
In this section of the current work, a preliminary analysis pertaining count of variation in the inlet recuperator pressure), the cycle efficiency
to the impact of recuperator pressure drop and pinch temperature on decreases whereas, heat load across it increases monotonically. The net
overall cycle efficiency is presented. The recuperator boundary condi­ specific work output decreases on account of increase in pressure drop
tions are selected based on thermodynamic analysis performed for a across the recuperator. Hence, to obtain 1 MW net output from the cycle,
simple recuperated sCO2 Brayton plant producing a power output of 1 the total mass flow in the cycle increases. Increase in the total mass flow
MW [5]. For a base line condition considered (Tturbine, inlet = 527 ◦ C, rate, correspondingly increases the heat load across the recuperator.
Tcompressor, inlet = 35 ◦ C, Pcompressor, inlet = 75,00 kPa, ηisen,turbine = 75%, ηisen,
compressor = 80%) [5], the variation of cycle efficiency with net specific
2.2. Effect of recuperator pinch on cycle efficiency
work output is obtained with increasing high side pressure presented in
Fig. 2. Initially, as high side pressure increases, the cycle efficiency in­
Fig. 4 show the variation of cycle efficiency and heat load with
creases with increase in net specific work output. After attaining a
temperature pinch (minimum temperature difference needed for heat
maximum, the efficiency decreases with increase in high side pressure.
transfer) across the recuperator. As the pinch temperature across the
The high side pressure corresponding to the region of maximum
recuperator increases, the enthalpy gain in the recuperator decreases.
cycle efficiency is selected to arrive at recuperator inlet conditions. For a
Consequently, the required enthalpy from the heater to achieve the
pressure drop of 50 kPa across the hot and cold streams of the recu­
desired TIT increases, thereby reducing the cycle efficiency. Even
perator with 275 kPa and 50 kPa pressure drop across heater and gas
though gross cycle mass flow rate is constant, increase in pinch tem­
cooler, respectively, the inlet temperatures of hot and cold streams
perature decreases the heat load across the recuperator due to reduced
entering the recuperator are estimated to be 427 ◦ C and 110 ◦ C,
respectively. The pressures at the inlet of hot and cold streams are 76,00
kPa and 21,225 kPa, respectively. For the selected recuperator condi­
tions, the state points are tabulated in Table 1. 1 MW net output,
recup= 50 kPa

Table 1
cycle

Cycle information, state points, and recuperator conditions.

qrecup

ηcycle = 29.11%, ṁt = 16.11 kg/s, q̇recup = 5.66 MW, ΔT = 10 ◦ C

State Location T [◦ C] p [kPa]

Point

1 Compressor inlet/Gas cooler exit/compressor 35 7500

inlet
2 Compressor outlet/recuperator cold inlet 110 21,225
3 Recuperator cold outlet/heater inlet 354.34 21,175
4 Heater outlet/turbine inlet 527 20,900
5 Turbine exit/recuperator hot inlet 427 7600 Trecup,pinch
6 Recuperator hot outlet/gas cooler inlet 120 7550
Fig. 4. Variation of cycle efficiency and recuperator heat load with tempera­
ture pinch across the recuperator.

4
V. Pandey et al. Renewable and Sustainable Energy Reviews 134 (2020) 110091

enthalpy gain.
From above discussion, it is evident that the pressure drop and
effectiveness of the recuperator significantly affects the cycle efficiency.
In the subsequent case study presented later in Section 4.2, the mass
flow rate of 16.11 kg/s is selected to obtain optimum stack dimensions
based on minimum rate of heat loss.

3. Computational domain and thermo-hydraulic model

3.1. 1-D model

A 1-D model based on Thermal Resistance Network (TRN), is

developed for modelling a full PCHE core. The model is described for a
recuperator core shown in Fig. 5(a). The recuperator geometry has a
counter-flow arrangement of hot and cold plates stacked alternatively on
top of each other to form a single bank configuration. Each plate in the
recuperator comprises of straight channels having a square cross-
section. Fig. 5(b) represents a unit cell of the PCHE core highlighting
the geometrical features. Fig. 5(c) shows a typical column of channels.

3.1.1. Description of the thermal resistance model

The sectional view of a column of channels (Fig. 5(c)) is used to
represent the TRN model which is based on an electrical circuit pre­
sented in Fig. 6(a). In this model, the flow paths of the fluid streams are
Fig. 5. (a) Schematic of recuperator core block with hot and cold plates stacked discretized along the channel length to capture the variation in the
together, (b) a two-channel configuration showing geometrical dimensions, (c) thermo-physical properties.
a column of channels. In the analysis, following are the simplifying assumptions made in
the proposed TRN model:

a) Outer surfaces are adiabatic, i.e.

Fig. 6. (a) Representation of a Thermal resistance network with the help of a half sectional view of a column of channels,(b) a typical node in the fluid streams, (c) a
typical node in the top solid wall, (d) a typical node in the bottom solid wall, and (e) a typical node in the middle solid wall.

5
V. Pandey et al. Renewable and Sustainable Energy Reviews 134 (2020) 110091

Table 2
Nodal energy balances for various typical nodes.
Description Algebraic equation

Nodal energy balance for the nodes present in the hot and cold fluid Ti+1,j − Ti,j Ti− 1,j − Ti,j ⃒ ⃒
+ + ṁc ⃒hi,j+1 − hi,j− 1 ⃒ = 0
streams Rtc(i+1,j/2) + Rc(i,j/2) Rc(i− 1,j/2) + Rtc(i,j/2)
⃒ ⃒
For hot stream ṁc ⃒hi,j+1 − hi,j− 1 ⃒ = ṁc cp (i,j) (Ti,j+1 − Ti,j− 1 )
⃒ ⃒
For cold stream ṁc ⃒hi,j+1 − hi,j− 1 ⃒ = − ṁc cp (i,j) (Ti,j+1 − Ti,j− 1 )where i = 2,4,6 … …2n and j = 2,4,6 …. .2k.
at x = L: Ti = 2,4,6 …. 2n, j=2k+1 = Thot,in
at x = 0: Ti = 2,4,6 …. .2n, j=1 = Tcold,in
Nodal energy balance for the nodes present in the middle solid walls Ti+1,j − Ti,j Ti− 1,j − Ti,j Ti,j− 1 − Ti,j Ti,j+1 − Ti,j
+ + + =0
Rtc(i+1,j− 1) + Rc(i,j− 1) Rtc(i,j− 1) + Rc(i− 1,j− 1) Ra((i+1)/2,j− 1) Ra((i+1)/2,j)
where i = 3,5,7 … …2n-1 and j = 2,3,4 …. .k+1.
Ti,j=1 − Ti,j=2
at x = 0, =0
Ra((i+1)/2,1)
Ti,j=k+2 − Ti,j=k+1
at x = L, =0
Ra((i+1)/2,k+1)
Nodal energy balance for the nodes present in the top solid wall Tts,j − T2n+1,j T2n,j − T2n+1,j T2n+1,j+1 − T2n+1,j T2n+1,j− 1 − T2n+1,j
+ + + = 0where j = 2,3,4 …. .k+1.
Rtc(2n+2,j− 1) Rtc(2n+1,j− 1) + Rc(2n,j− 1) Ra(n+1,j) Ra(n+1,j− 1)
Tts,j − T2n+1,j
at y = H, =0
Rtc(2n+2,j− 1)
T2n+1,1 − T2n+1,2
at x = 0, =0
Ra(n+1,1)
T2n+1,k+2 − T2n+1,k+1
at x = L, =0
Ra(n+1,k+1)
Nodal energy balance for the nodes present in the bottom solid wall T2,j − T1,j Tbs,j − T1,j T1,j+1 − T1,j T1,j− 1 − T1,j
+ + + =0
Rtc(2,j− 1) + Rc(1,j− 1) Rtc(1,j− 1) Ra(1,j) Ra(1,j− 1)
where j = 2,3,4 …. .k+1.
Tbs,j − T1,j
at y = 0, =0
Rtc(1,j− 1)
T1,1 − T1,2
at x = 0, =0
Ra(1,1)
T1,k+2 − T1,k+1
at x = L, =0
Ra(1,k+1)

⃒ ⃒ ⃒ ⃒
∂T ⃒⃒ ∂T ⃒⃒ ∂T ⃒⃒ ∂T ⃒
= 0, = 0, = 0, ⃒⃒ =0 (1) Table 3
∂Y ⃒Y=0 ∂Y ⃒Y=H ∂X ⃒X=0 ∂X X=L
Expression for resistances involved in the TRN model.
Description Resistances
b) The model considers the steady-state operation.
Wall conduction along Y Δy
c) The flow is uniformly distributed across all channels of hot and cold Rtc(i,j) = , Δy = t, i = 1,2,3 …. .2n+2, j
direction 2ks (Δx)d
streams, i.e. = 1,2,3 …. .k.
Wall conduction along X Δx
(ṁhot )i=1,2...n/2 = (ṁcold )i=1,2....n/2 = ṁc (2) direction
Ra (i,j) =
ks Δyd
, i = 1,2,3 …. .n+1, j = 2,3 …. .k.
Δx
at x = 0 and x = L, Ra (i,j=1,k+1) =
2ks Δyd
d) At the inlet of all hot/cold streams, the temperature and pressure Resistance for transverse
Rc(i,j) =
1
, i = 1,2,3 …. .2n, j = 1,2,3 …. .
convection between wall and h(i,j) (Δxd)
conditions are identical which are mass-weighted average at the
fluid k.
cross-section.

Each discretized element of hot and cold channels is connected To evaluate the significance of axial conduction in the PCHEs, the
through convection and conduction resistances facilitating heat ex­ following equation is used [22]:
change between channels. The fluid flow is modelled as an advection
current [50]. For each discretized element in the fluid channels, there ks Asp
<< 1 (3)
are two advection currents present at the boundary. The entering and ṁc cp L
exiting current at the boundary of the elements are the source and sink The axial conduction is neglected in comparison to convection heat
enthalpy flow rates (ṁh), respectively. The thermo-physical properties transfer when Eq. (3) is satisfied. In Eq. (1), ks is the thermal conduc­
are assumed to be constant in each discretized element. In Fig. 6, con­ tivity of the solid material, Asp is the cross-sectional area of the solid
duction resistances are present along with the wall (axial conduction plate for the conduction heat transfer, ṁc is the channel mass flow rate of
resistance-Ra) and perpendicular to the wall (transverse conduction the fluid stream, cp is specific heat, and L is the channel length. The term
resistance-Rtc), whereas, convection resistances (Rc) are present between ksAsp/ṁccpL is estimated to be order of (0.0001) in the present study
the wall and the fluid stream. based on the combination of extremities of property variations. Hence,
Following are the salient features of the hybrid TRN model: the effect of the axial condition is not pronounced in the present anal­
ysis. The typical nodes present in the fluid stream and solid wall, are
1) The variation in the thermo-physical properties of fluid streams are shown in Fig. 6(b–e), which are used to obtain the nodal balances for
accounted with reasonable accuracy. heat flow rates. The indices i and j represent the temperature nodes in Y
2) TRN model coupled with CFD corrects for the heat transfer and and X directions, respectively. A typical node in the fluid stream (Fig. 6
pressure drop correlations along the flow in the channel. (b)) is surrounded by two advection currents and two neighbouring
3) Hybrid TRN-CFD models the complete PCHE stack to arrive at op­ temperature nodes. The temperature at central node for an element
timum overall stack dimensions by minimizing the rate of heat loss along the fluid stream is assumed to be the average of boundary tem­
from the external surfaces of the PCHE core. peratures for that element. A typical node in the adjacent middle wall

6
V. Pandey et al. Renewable and Sustainable Energy Reviews 134 (2020) 110091

Fig. 7. (a) Unit cell model for straight channel PCHE (b) Hexahedral mesh generated for straight channel (c) a zigzag channel (d) Hexahedral mesh for zigzag
channel generated in ANSYS ICEM CFD.

(Fig. 6 (e)) is surrounded by four neighbouring temperature nodes. Fig. 6

Table 4
(c) and (d) show a typical node in the top and bottom solid walls,
Detail of mesh and material.
respectively. At the outer surfaces (top, bottom, front, and back) of the
column of channels, the conduction resistances do not participate in the Mesh Quality Skewness Material
Stainless steel 316 L (ASTM Grade)
nodal energy balances because of assumed adiabatic boundary condi­
tions. The nodal balances of heat flow rates for various typical nodes Hexahedral >0.95 <0.38 ρ [kg/ cp [ kJ/kg. k [W/m.
cells m3] K] K]
present in the fluid stream and solid walls are tabulated in Table 2.
8360 0.417 16.27
The expressions for various thermal resistances appearing in the
nodal balances of the heat flow rates are listed in Table 3. The convec­
tion heat transfer coefficient (h) appearing in the expression of con­
where f is the friction factor calculated from Filonienko relationship for
vection thermal resistance is not the specific enthalpy used for the fluid
smooth tubes [53]:
streams presented in Fig. 6(a) and (b). As mentioned earlier that the TRN
( )2
model considered in the paper is a 2-D thermal network, hence, only the 1
f= , (5)
top and bottom heat transfer areas are considered. 1.82log(Re) − 1.64
The momentum and energy conservation equations obtained from
the nodal balances of heat flow rates in the TRN using Kirchhoff’s laws The pressure drop along the channels is computed based on fluid
are solved with the correlations for heat transfer coefficient and friction friction. The frictional pressure drop for an element along the channel
factor for a straight channel. The resulting temperature and pressure length is given by:
fields are used to update thermophysical properties from REFPROP [51] f ρv2 Δx
NIST data base. The algorithm uses correlations for smooth pipe based ΔP = (6)
2d
on hydraulic diameter of square cross-section. Subsequently, these cor­
relations are replaced with CFD based correlations to obtain corrected
performance. The heat transfer coefficient is calculated using the 3.2. 3-D CFD model
Gnielinski correlation [52]:
A unit cell PCHE model comprising of two channels (for hot and cold
Nu =
(f /8)(Re − 1000)Pr
√̅̅̅̅̅̅̅( ) (4) streams), is shown in Fig. 7(a). The model is subjected to periodic
1.07 + 12.7 f /2 Pr2/3 − 1 boundary conditions (BCs) across top and bottom surfaces, with adia­
batic BCs imposed on the left and right walls with the assumption of
for negligible heat transfer between two adjacent hot/cold channels in the
same plate.
0.5 ≤ Pr ≤ 2000, 3000 ≤ Re ≤ 5 × 106 Fig. 7(b) shows the hexahedral cells with mesh quality greater than

7
V. Pandey et al. Renewable and Sustainable Energy Reviews 134 (2020) 110091

Table 5 flow models, are solved in ANSYS Fluent® 2019 R2. The solver is
Geometrical parameters for simulations using 3-D CFD and 1-D coupled with NIST database REFPROP to update the thermo-physical
approach. properties of sCO2 at the end of each iteration.
Edge of square cross section (d) [mm] 1.5 A mesh sensitivity analysis based on Grid convergence Index (GCI) is
Straight channel length (L) [mm] 1000
performed. The details are listed in Appendix B. 8.68 million cells have
Zigzag channel length (L) [mm] 393.6 been used for mesh generation with a discretization error of less than
Pitch of zigzag channel [mm] 24.6 0.3%,. For 1-D TRN model, 100 elements are used with a discretization
Angle of turn [◦ ] 150 error of less than 0.45% for all solution variables.
Plate thickness (t) [mm] 1.5
Half side width (w) [mm] 0.75
Hydraulic diameter (dh) [mm] 1.5 4. Performance metrics
Wetted perimeter (P) [mm] 6
Cross-sectional area (Ac) [mm2] 225
4.1. Effectiveness and pressure drop

The effectiveness of the PCHE can be defined as:

Table 6
Boundary conditions for simulations using 3-D CFD and 1-D approach. (Δh)cold
ε= (7)
min[(Δhmax(cold) ), (Δhmax(hot) )]
Boundary conditions Hot stream Cold stream

Mass flow rate [kg/s] (ṁc = Recold,in μcold,indh) (ṁc = Recold,in μcold,indh) where, (Δh)cold is the overall cold stream enthalpy gain obtained when
Inlet temperature [◦ C] 427 110
‘n’ number of channels (from ‘n’ number of rows) are modelled together.
Outlet pressure [kPa] 76,00 21,225
It can be expressed as follows:
Δhcold = hcold,out − hin (8)
95% obtained using ANSYS ICEM® CFD. Fig. 7(c) shows a zig-zag
channel configuration with a pitch of 24.6 mm and turn angle of 150◦ , ṁc hcold,1 + ṁc hcold,2 ..........ṁc hcold,n/2
respectively. These values are selected from Ref. [54] to facilitate where hcold,out = (9)
(n/2)ṁc
comparison with experimental results presented later in the paper. The
corresponding hexahedral mesh is shown in Fig. 7(d). The details of The denominator in Eq. (7) is the minimum of the two maximum
mesh and solid material used in the simulation are listed in Table 4. specific enthalpies gain calculated using the inlet conditions for hot and
The geometrical dimensions and boundary conditions for 3-D CFD cold streams. Δhmax(hot) and Δhmax(cold) are the differences between the
simulation are listed in Table 5 and Table 6, respectively. From the inlet enthalpies of the hot and cold streams based on hot and cold side
perspective of developing correlations, simulations are performed for pressures, respectively.
varying channel mass flow rates corresponding to three cold inlet Rey­ The hot side pressure drop is more significant than cold side pressure
nolds numbers: Recold,in = 5 × 103, 10 × 103 and 20 × 103. For all drop, details of which are explained in section 5.1.1. This can be ob­
simulations (3-D CFD and TRN model), t and w are fixed at 1.5 mm based tained as follows:
on simplified stress analysis reported in Appendix A. ( )
Reynolds-averaged Navier-Stokes (RANS) model is selected for Δpstack = mean Δphot,1 , Δphot,1 ........Δphot,n/2 (10)
solving Navier-Stokes equations [55]. The GEKO (generalized k-ω)
two-equation turbulence model belonging to the family of k-ω models
(offering superior performance for wall-bounded boundary layer flows), 4.2. Optimum number of rows
is used to model turbulence. The GEKO model offers better flexibility
allowing the user to adjust free parameters for a specific type of appli­ A large combination of PCHE core dimensions can be obtained for a
cation without affecting the basic calibration of the model. Enhanced target thermo-hydraulic performance based on channel dimensions and
Wall Treatment (EWT) (corresponding to a wall-adjacent cell height mass flow rates. The present section describes a method to obtain op­
based on y+ ≈ 1) is the sole treatment for the category of k-ω models. timum stack dimensions by minimizing the rate of heat loss from the
The steady-state equations are solved using a pressure-based scheme external surfaces of the PCHE core. The method is illustrated for a PCHE
by selecting pseudo transient option. The spatial discretization of gov­ core employing both zigzag and straight flow paths with m channels on
erning equations is performed using a second-order upwind discretiza­ each plate as shown in Fig. 8(a) and Fig. 8(b), respectively.
tion scheme. The residual for all variables is set at 10− 10. The governing For straight channels, the top surface area of the recuperator core
equations for mass, momentum, energy, and two-equation turbulence block (Fig. 5(a)) for an arbitrary n (= Nt/m) rows is obtained as:

L L

Flow channel

Wall/fin thickness
Wzigzag
Wstraight

Fig. 8. (a) A plate with m zigzag channels and (b) a plate with m straight channels.

8
V. Pandey et al. Renewable and Sustainable Energy Reviews 134 (2020) 110091

phot
phot
n n
d = 1 mm d = 1.5 mm d = 2 mm Recold,in d = 1 mm d = 1.5 mm d = 2 mm Recold,in
d = 1 mm d = 1.5 mm d = 2 mm L d = 1 mm d = 1.5 mm d = 2 mm L

phot
n
d = 1mm d = 1.5mm d = 2mm Recold,in
d = 1mm d = 1.5mm d = 2mm L

Fig. 9. Variation of effectiveness and pressure drop with number of rows for cold inlet Reynolds number of (a) 5 × 103, (b) 10 × 103, and (c) 20 × 103.

2L(d + w)Nt
Atop,straight = (11) Q̇loss = 2hside Aside ΔTside + hbottom Abottom ΔTbottom + htop Atop ΔTtop (17)
n
The first, second and third terms in Eq. (17) represent the convection
where (Nt/n)(d + w) is the stack width (Wstraight). The top surface area is
heat loss from the two sides faces, bottom surface and top surface,
same as bottom surface area. The side surface area is obtained as:
respectively. Various temperature differences (ΔT) appearing in Eq. (17)
Aside,straight = 2n(d + t)L (12) are the differences between the corresponding surface temperatures
obtained from the TRN model and ambient temperature. The tempera­
where n is the number of rows which is obtained as Nt/m, and n(d + t) is ture for side/top/bottom surface is the representative average temper­
the stack height. The total number of channels (Nt) is estimated to be: ature obtained by averaging the temperatures along the solid walls/
channels. The convection heat transfer coefficient for air (hair = Nu(kair/
2ṁt
Nt = 2N = (13) Lc)) is obtained from Nusselt and Rayleigh number relations discussed
ṁc
below.
The channel mass flow rates for all hot and cold streams are obtained For side surfaces considering air at ambient condition (Pr = 0.72)
based on the minimum Reynolds number (Recold,in) of the flow domain [56], the relation between Nu and Ra is written as:
using the following relation: ( )2
(18)
1/6
NuLc,side = 0.825 + 0.328RaLc,side
ṁc = Recold,inlet μcold,inlet d (14)

Considering zigzag channels, the top surface area of the recuperator The above expression is valid for all Rayleigh numbers. For the top
core block (Fig. 5(a)) for any arbitrary m channels in each plate is ob­ surface, the convection heat transfer coefficient is obtained using the
tained using the following expression: following equation [57]:

Atop,zigzag = 2[m(d + w) + w + λ cot(θ / 2)]L (15) NuLc,top = 0.15Ra1/3

Lc,top , 107 < RaLc,top < 109 (19)

where (m(d + w)+w+λcot(θ/2)) is the stack width (Wzigzag). The side For the bottom surface, the convection heat transfer coefficient is
surface area of the recuperator core block with zig-zag channels on each obtained using the following equation [58]:
plate is similar to the case of straight channels.
(20)
1/4
NuLc,bottom = 0.27RaLc,bottom , 105 < RaLc,bottom < 1010
Aside,zigzag = Aside,straight (16)
In the above equations, the Rayleigh number is defined as:
The rate of heat loss from outer surfaces can be expressed as follow:

9
V. Pandey et al. Renewable and Sustainable Energy Reviews 134 (2020) 110091

gβΔTs L3c
RaLc = (21)
να

where Lc is the characteristic length of the corresponding surfaces. This

is defined as [59]:
Aplane
(22)

NTUeqn
Lc =
pplane

The characteristic length for each outer wall of the recuperator core
block is obtained from the stack height, stack width and channel length.
The properties of air in contact with the PCHE core walls are evaluated
at corresponding film temperature which is the mean of the surface (Ts)
and ambient temperature (T∞).
Minimization of the rate of heat loss is obtained by differentiating Eq.
(17) with respect to n, as follows;

dQ̇loss
=0 (23)
n
dn
This results in optimum number of rows (or optimum number of
columns) for a given set of channel dimensions and channel mass flow L d Recold,in
rate. The obtained nopt is mathematically complex and therefore
expressed in the general form as: Fig. 10. Variation of NTUeqn with number of rows.
(
nopt = f d, L, Recold,in, t, w, λ, θ, ΔTside , ΔTtop, ΔTbottom,
) (24) identical channel dimensions, criteria for minimum pressure drop for
air properties based on various film temperatures
the cold side is automatically satisfied if the hot side pressure drop is
curtailed within the prescribed limits during the design stage. Intui­
5. Results and discussion tively, the pressure drop is strongly related to channel Reynolds number
and channel hydraulic diameter. Fig. 9(a–c) show that the pressure drop
The hot and cold channel flow paths in the PCHE are either elec­ (50 kPa) is still within the maximum limit of 130 kPa corresponding to
trochemically etched [16,60] or micro milled on stainless steel plates. the highest Recold,in = 20 × 103 and lowest d = 1 mm considered in the
The plates are vertically stacked on top of each other in an alternate current study. From Fig. 9(a–c), it is evident that the changes in the
manner as shown in Fig. 5(a). The stack is diffusion bonded in a vacuum effectiveness and pressure drop are minimal if the number of rows (n) is
furnace under pressure to create the PCHE core. The manifolds for hot increased beyond 40 (comprising of 20 hot and 20 cold plates). As n is
and cold fluid entry and exit are welded to the diffusion bonded stack to increased, the equivalent conductance (UeqnAht,eqn)n and equivalent
complete the manufacturing process. Due to practical dimensional limit minimum heat carrying capacity ((ṁeqncp)min)n/2 among both the
of the hot zone in the vacuum furnace and the associated bonding streams, increases. Consequently, equivalent NTU (NTUeqn) increases
pressures, the maximum PCHE stack dimensions are currently restricted with n up to ~40 rows. Increase in NTUeqn beyond 40 rows is minimal as
to 1.5 mm3 (1500 × 1500 × 1500 mm3) [60]. Hence, in the current observed from Fig. 10. The equivalent conductance based on n rows is
thermo-hydraulic analysis the maximum linear dimensions of the PCHE given by:
stack (stack height, channel length or stack width) is restricted to 1500 ( )
mm. The minimum hydraulic diameter considered in the current work is ( ) Q̇cold
1 mm which is fixed on account of permissible pressure drop limit. For a Ueqn Aht,overall n = n
(25)
(LMTD)n
1500 mm long channel having a hydraulic diameter below 1 mm results
in pressure drop greater than 150 kPa. Considering the maximum ( ) ( )
operating conditions of the cycle in the present study (Tmax = 527 ◦ C and where Q̇cold = (n / 2)ṁc hcold,out − hcold,in (26)
n
Pmax = 21,225 kPa), carbon steel of grade SS 316 L has been selected as
where (LMTD)n is based on n rows which can be calculated as:
the PCHE stack material. The thickness of the channel side walls shown
in Fig. 5(b) is considered to be 1.5 mm based on thermal stress criteria (ΔT1 )n − (ΔT2 )n
(LMTD)n = (27)
described in Appendix A. The boundary conditions for thermo-hydraulic ln((ΔT1 )n /(ΔT2 )n )
analysis are listed in Table 6.
( )
where ΔT1 = Thot,in − Tcold,out n and (28)
5.1. Analysis for straight channel-based PCHE ( )
ΔT2 = Thot,out − Tcold,in n (29)
5.1.1. Effect of stack height
The equivalent minimum heat carrying capacity ((ṁeqncp)min)n/2
Fig. 9(a–c) show the effect of number of rows on effectiveness and
among both streams can be obtained as:
channel pressure drop. The results are presented for three different [(( ) ) ]
Reynolds number ranging from 5 × 103 to 20 × 103 corresponding to
(( ) ) ) ) ((
three channel diameters varying between 1 and 2 mm with a channel ṁeqn cp = min ṁeqn cp , ṁeqn cp
min n/2 cold n/2 hot n/2
length fixed at 1500 mm. Since the channels are arranged alongside each
(30)
other and based on the assumption of no flow maldistribution across the
channels in PCHE stack, the channel pressure drop shown on secondary Based on equivalent minimum heat carrying capacity and equivalent
y-axis is representative of the overall stack pressure drop for the hot side. conductance corresponding to n rows, the NTUeqn can be obtained as
In a sCO2 recuperator, the pressure drop on the hot side is more sig­ follows:
nificant than the cold side pressure drop [61] based on the operating
range considered in the current study. Since the counterflow arrange­
ment in PCHE stack houses equal number of hot and cold channels with

10
V. Pandey et al. Renewable and Sustainable Energy Reviews 134 (2020) 110091

5.1.2. Effect of channel Reynolds number

The effect on channel Reynolds number on PCHE performance is
shown in Fig. 11. The results are presented for a vertical row comprising
of 40 plates/rows. Consistent with the results presented earlier, Δphot
shows an increasing trend with increase in Recold,in. For a given d, if
Recold,in is increased, turbulence effects increase significantly resulting in
higher pressure drop. At a particular d, the pressure drop with Recold,in
varies according to Δphot ~ (μ2/ρ) (Recold,in)7/4 which can be derived

phot
from Darcy’s pressure drop model using Blasius correlation [62] for
friction factor. The trend of pressure drop with Reynolds number is
similar to the findings provided by Zhao et al. [63] for supercritical
nitrogen in a straight channel. They show that pressure drop increases
monotonically with increasing mass flux. At a particular Recold,in,
decreasing d increases the pressure drop according to Δphot ~ (1/d3
(μ2/ρ)).
Recold,in However, effectiveness ε decreases with increasing Recold,in for a
given d as observed from Fig. 11. It is to be noted that increasing Recold,in
d = 1mm d = 1.5mm d = 2mm n is analogous to increasing channel mass flow rate as d is fixed. At higher
d = 1mm d = 1.5mm d = 2mm L
channel mass flow rates (higher Recold,in), the effectiveness is lower as
Fig. 11. Variation of effectiveness and pressure drop with cold inlet Rey­ the channels have less time for heat exchange [64]. In other words, as
nolds number. channel mass flow rate is increased, the equivalent minimum heat car­
rying capacity ((ṁeqncp)min)n/2 increases as compared to the equivalent
( ) conductance (UeqnAht,eqn)n for a column of 40 channels. Hence, it reduces
(( ) ) ṁeqn Δhcold the NTU and consequently the effectiveness. On the other hand, at a
where ṁeqn cp =
n/2
and (31) specified Recold,in, increasing d decreases the effectiveness. This is
(ΔTcold )n/2
cold n/2 attributable to the fact that increasing channel size increases the thermal
( ) resistance. This argument is also supported by the results of Kim et al.
(( ) ) ṁeqn Δhhot [65].
(32)
n/2
ṁeqn cp =
hot n/2 (ΔThot )n/2 5.1.3. Effect of channel length
( ) In the above two cases, the results were presented for a maximum
Ueqn Aht,eqn n channel length of 1500 mm based on current manufacturing limits. For
NTUeqn = (( ) ) (33)
the completeness of the analysis, it is worth investigating the effect of
ṁeqn cp
min n/2
channel length to ensure if 1500 mm is an adequate length for a recu­
perator. The results presented in Fig. 12(a) shows the effect of channel
The reduction in the pressure drop with increasing number of rows as length on PCHE performance for varying channel diameters with a fixed
shown in Fig. 9(a–c), is the result of fluid property variation across the Recold,in of 104. Observing the trends of pressure drop and effectiveness,
channels of hot stream. As n is increased, the temperature field across it is obvious that longer channel lengths are associated with higher
the channel changes, affecting the local variation of thermo-physical pressure drop, but it is interesting to note that the effectiveness also
properties. For a fixed set of geometrical conditions and Recold,in, the increases with channel length and tends to plateau after a channel length
pressure drop varies according to Δphot ~ (μ2/ρ) as the number of rows is of 2000 mm for all the three channel diameters considered. To under­
increased. At a given number of rows and d, the effectiveness is higher stand the variation of effectiveness with channel length, the concept of
for lower values of Recold,in which can be observed from Fig. 9(a–c). The volumetric heat flux can be used.
higher effectiveness corresponds to lower values of Recold,in is explained Increasing channel length decreases the heat transfer rate (q”) per
in the next section. unit volume as it can be observed from Fig. 12(b). The is because, as
phot
q"

L L
d = 1 mm d = 1.5 mm d = 2 mm n = 40 rows and n Recold,in d
d = 1 mm d = 1.5 mm d = 2 mm Recold,in

Fig. 12. (a) Variation of effectiveness and pressure drop with channel length, (b) variation of heat transfer rate per unit volume with channel length.

11
V. Pandey et al. Renewable and Sustainable Energy Reviews 134 (2020) 110091

channel length is increased, the average heat transfer coefficient along manufacturing capabilities permit, higher channel length beyond 1500
the channel decreases. The heat transfer rate per unit volume can be mm can be preferred. Considering the additional length does not
expressed as follows: appreciably increase the pressure drop which is still below the 50 kPa
( ) pressure drop limit imposed during the design stage. It has been noted
Q̇ that smaller channel diameters and lower Recold,in tend to favour effec­
q˝ = total
(34) tiveness. However, small channel diameters and lower Recold,in results in
LWopt Hopt
higher pressure drop and significantly larger recuperator sizes,
where (Q̇)total is based on total mass flow rate (16.11 kg/s) considering respectively.
the heat transfer performance obtained from 40 rows. In the above
expression Hopt and Wopt are the optimum stack height and width, 5.1.4. Comparison with CFD simulation and correlation development
respectively corresponding to minimum rate of heat loss which will be A CFD study is undertaken to account for discrepancies in both
discussed in section 5.1.5. effectiveness and pressure drop obtained using standard correlations as
Increasing channel length can improve the effectiveness rapidly up to previously described in section 5.1.1. To correct the performance pa­
a certain length. Increasing L further does not increase the effectiveness rameters, CFD simulations are performed to derive correlations which
appreciably as volumetric heat flux approaches a nearly constant value are plugged back in the MATLAB code for the TRN stack. In this case a
as shown in Fig. 12(b). However, the effectiveness tends to fall rapidly unit cell approach is considered with prescribed channel cross section
for smaller channel lengths of less than 1500 mm. A channel length of having dimensions of 1.5 × 1.5 mm. The details of the boundary con­
1500 mm is selected for a 5.66 MW sCO2 recuperator considering the ditions for the unit cell are presented earlier in Table 6. Simulations
practical manufacturing difficulties. However, in future if using CFD solver ANSYS FLUENT® is performed for a 1000 mm channel

hot
T

CFD Result CFD Result

CFD Result CFD Result
TRN Result based on CFD correlation TRN Result based on CFD correlation
TRN Result based on CFD correlation TRN Result based on CFD correlation

Fig. 13. Variation of (a) temperature and (b) pressure drop along the channel length of hot and cold stream obtained from CFD simulation and TRN model using CFD
based correlation.

TRN Result based on CFD correlation

TRN Result based on CFD correlation
TRN Result based on Filonienko
TRN Result based on Filonienko
p

Fig. 14. Variation of (a) absolute difference between the local temperatures along the channel length obtained from TRN model using CFD correlation and the
Gnielinski correlation, (b) pressure drop along the channel length obtained from TRN model using friction factor derived from CFD simulation and the Filonienko
correlation.

12
V. Pandey et al. Renewable and Sustainable Energy Reviews 134 (2020) 110091

length. A comparison between the temperature profiles obtained from

CFD simulations and TRN model using correlation derived from the CFD
simulation data is presented in Fig. 13(a). Using the TRN framework 40
rows have been modelled. The temperature and pressure drop profiles
along the channel length of 21st hot channel and 22nd cold channel
(corresponding to the central location of the stack) have been used for
comparing the results with CFD simulations. The central channels are
selected because these channels mimic the periodic BCs at the top and
bottom surfaces of the unit cell model used for CFD simulations.
The temperature profile obtained from 1-D simulation using CFD

T
based correlation closely agree with the 3-D CFD simulations. The outlet
temperatures obtained from 3-D CFD simulations are 166.98 ◦ C and
310.05 ◦ C for hot and cold streams, in comparison to 168.38 ◦ C and
308 ◦ C obtained from 1-D simulations. The maximum relative error of
0.5% for the cold side and 0.8% for the hot side are observed between
the temperature values along the channel length. In Fig. 13(b), pressure
drop comparison between CFD result and TRN model using CFD based
friction factor is presented along the channel length. The pressure drop
along the channel length predicted by the TRN model agrees well with 3-
D CFD simulations except at the inlets of both hot and cold streams. CFD hot-1 hot-2 hot-3 hot-4 hot-5
results show a higher pressure drop at the inlet (within 0.02 m channel cold-1 cold-2 cold-3 cold-4 cold-5
length) of both streams on account of entrance effects. The hydrody­
Fig. 15. Temperature profiles for hot and cold streams along the channel
namic and thermal entrance length which are approximated as 0.015 m
length of ten straight channels.
[66] in the present case is smaller than the discretized length (0.02 m)
used in the simulation.
The Nusselt number for both hot and cold stream derived from 3-D The derived CFD correlations are used for correcting the stack length
CFD simulations data are expressed as follows: for a given set of performance parameters. For a channel length of 1500
for hot stream mm with a channel cross-section of 1.5 mm × 1.5 mm and channel mass
flow rate corresponding to Recold,in of 104, the corrected optimum
Nuh = 0.0252Re0.805 Pr0.723 (35) effectiveness and optimum pressure drop values across the stack are
85.2% and 12.1 kPa, respectively. On the other hand, for an effective­
for cold stream ness of 87.46% and pressure drop of 11 kPa which have been obtained
using conventional correlations, the corrected length using CFD based
Nuc = 0.020Re0.81 Pr0.338 (36)
correlations are found to be 1580 mm instead of 1500 mm. The corre­
The friction factor expression for both hot and cold stream derived lations deduced from CFD simulations have been used in the TRN model
from CFD simulation are expressed as follows: for the minimization of heat loss from the outer surfaces to attain op­
for hot stream timum stack dimensions.
0.46
fh = (37) 5.1.5. Selection of optimum stack dimensions based on minimum rate of
Re0.29
heat loss
for cold stream, The temperature profiles for 10 straight channels comprising of
alternately stacked 5 hot and 5 cold channels are shown in Fig. 15. The
0.41
fc = (38) average temperatures of the top, bottom and side surfaces using 2
Re0.27
channel and 10 channels are found to be different. Using two channels,
The above correlations are obtained for Reynolds number ranging the average temperature of top, bottom and side surfaces are calculated
from 5 × 103 to 20 × 103. After benchmarking the correlations derived to be 299.27 ◦ C, 184.78 ◦ C and 241.28 ◦ C, respectively. However, for 10
from CFD data, these have also been compared with existing conven­ channels the corresponding values are 331.43 ◦ C, 150.46 ◦ C and
tional correlations. 234.14 ◦ C, respectively. For attaining optimum stack dimensions
Fig. 14(a) shows the variation of absolute temperature difference resulting in minimum heat loss, the required average temperature of
between the local temperatures along the channel length obtained from outer surfaces should be selected based on a full stack PCHE model
TRN model (for 1000 mm channel length) using both CFD based cor­ instead of a two channel/two rows unit cell. Therefore, 40 rows based on
relations and conventional correlations. There are 40 rows considered in earlier observations described in section 5.1.1 are considered for esti­
TRN model, wherein, as described previously, 21st and 22nd rows are mating the average temperature of outer surfaces.
selected as hot and cold stream channels for comparison. It is evident For a temperature pinch of 10 ◦ C between hot outlet and cold inlet
from Fig. 14(a) that for the hot stream a maximum of 3.5 ◦ C and for cold (location where pinch occurs in the recuperator), a set of channel di­
stream a maximum of 3 ◦ C temperature difference is observed. Fig. 14(b) mensions (dh and L) is obtained for each Recold,in ranging from 5 × 103 to
shows the comparison of the frictional pressure drop along the channel 20 × 103. Contours of nopt based on minimum rate of heat loss discussed
length obtained from TRN model using CFD based and conventional in Section 4.2, is presented in Fig. 16(a). It is observed that as dh is
friction factor. The pressure drop along the channel length obtained increased for a fixed Recold,in, the nopt decreases on account of increase in
from CFD based friction factor and Filonienko [47] correlation are in channel mass flow rate resulting in a decrease in the total number of
close agreement. channels required to accommodate the entire stack mass flow rate. The

13
V. Pandey et al. Renewable and Sustainable Energy Reviews 134 (2020) 110091

Fig. 16. (a) Contours of nopt and (b) contours of channel length for various values of dh and cold inlet Reynolds number satisfying a temperature pinch of 10 ◦ C at the
cold inlet using straight channel-based PCHE.

T p n

Thot,in Phot,in Tcold,in pcold,in

cold hot L dh
cold hot L dh t
T and p field
µ cp k T p T Tpinch obtained Tpinch desired
p pobtained pdesired
HTC p
Rtc, Ra, Rc
T
Ax = b p

channel, L dh
Convergence check hcold overall
energy balance EF Qloss channel, L dh, t
hhot overall
nopt Hopt
EF

Fig. 17. The flow chart for optimal design methodology.

corresponding channel length variation with diameter dh is shown in stack size is presented in Fig. 17.
Fig. 16(b) for a temperature pinch of 10 ◦ C. Larger channel lengths are The procedure is illustrated for the sCO2 recupertor under consid­
associated with lower number of rows to limit the temperature pinch of eration having a channel hydraulic diameter of 1.5 mm and a length of
10 ◦ C. Lower values of nopt correspond to higher channel mass flow rates 1500 mm. The variation in effectiveness and pressure drop with cold
and hence, lower effectiveness. Therefore, Fig. 16 (a) and (b) can be used inlet Reynolds number corresponding to the optimum number of rows
to obtain the optimum stack dimensions corresponding to minimum obtained using the concept of minimum heat loss in presented in Fig. 18.
heat loss once the pinch temperature and channel mass flow rates are The Reynolds number on x-axis is uniquely associated with stack
fixed. The optimum stack dimensions can be subsequently used to obtain width and height. The variation of effectiveness and pressure drop are
the resulting “Overall Exposed Surface Area” (OESA) of the PCHE core consistent with earlier observations provided in section 5.1.2. For the
block. The OESA is the sum of top, bottom and two side surface areas. present case of a 5.66 MW recuperator, the optimum stack dimensions
The flow chart for a given temperature pinch and/or pressure drop, are found to be 1500 × 708 × 745 mm3 (Lopt × Wopt × Hopt) for a channel
channel dimensions and channel mass flow rate to obtain the optimal hydraulic diameter of 1.5 mm corresponding a Recold,in of 104. The

14
V. Pandey et al. Renewable and Sustainable Energy Reviews 134 (2020) 110091

and acceleration effect due to density changes. Pitch averaged method

discussed by Kim et al. [54] is used for development of correlations.
Fig. 19(a) compares the temperature profiles obtained from 3-D CFD
simulation and TRN model based on CFD correlations for a channel mass
flow rate corresponding to Recold,in of 5 × 103. The temperature profiles
for hot and cold streams obtained from both approaches are in good
agreement with deviations less than 1.5 ◦ C between the outlet temper­

phot,opt
atures of hot and cold streams. Fig. 19(b) shows the total pressure drop
opt

along the channel length for both hot and cold streams. The pressure
drops using CFD simulations are independently calculated for each pitch
variation of the zig-zag channel. The differences in pressure drops are
mainly observed near the exits. At the exit pitch of hot stream, the
pressure drop predicted by CFD simulation and TRN model are 0.087
kPa and 0.11 kPa, respectively. Similarly, at the exit pitch of cold
stream, the pressure drop predicted by CFD simulation and TRN model
are 0.016 kPa and 0.024 kPa, respectively. The difference between the
Recold,in pressure drops are due to relative error (<4% for hot stream and <3%
for cold stream) between exact CFD simulation data and correlation
L d data. These observations confirm that the proposed hybrid TRN-CFD
approach can be confidently used for other channel configurations as
Fig. 18. Variation of optimum effectiveness and optimum hot side pressure well to predict PCHE performance with reasonable accuracy.
drop with cold inlet Reynolds number. Using CFD based correlations, the variation of effectiveness and
pressure drop with number of rows for three values of Recold,in is
corresponding effectiveness and pressure drop are 85.2% and 12.1 kPa,
respectively as discussed in section 5.1.4. The dimensions of the cor­
rected PCHE stack obtained using the above procedure can be con­
Recold,in
structed to closely represent the actual built dimensions.
Recold,in

5.2. TRN model for non-straight channels Recold,in

Recold,in
5.2.1. Comparison with CFD simulation and correlation development
3-D CFD simulations are also performed for zig-zag channel
described in Section 3.2. The dimensions of the channel are listed in

phot
Table 5 in Section 3.2. Table 7 shows the Nusselt number and equivalent
Recold,in
Fanning factor correlations derived from 3-D CFD simulation data. The
equivalent Fanning factor includes the effect of friction loss, form loss
Recold,in

Table 7
Correlations derived from CFD for zigzag channel (5 × 103 < Re < 20 × 103).
Nusselt number Equivalent Fanning
factor n
Hot stream Nuh = 0.05Re 0.753
Pr 0.54
, 0.73 < Pr < 0.07 N = 40 L dh
fh = 0.00078 +
0.83 Re0.17
Cold Nuc = 0.047Re0.75 Pr0.39 , 0.8 < Pr <
fc =
0.061 Fig. 20. Variation of effectiveness and pressure drop with number of rows for
stream 1.47 Re0.15 zigzag channels for a cold inlet Reynolds number of 5 × 103.

CFD zigzag hot CFD zigzag hot

CFD zigzag cold CFD zigzag cold
TRN zigzag hot TRN zigzag hot
TRN zigzag cold
TRN zigzag cold
p
T

Fig. 19. Variation of (a) temperature and (b) pressure drop along the channel length of hot and cold stream obtained from CFD simulation and TRN model using CFD
based correlation.

15
V. Pandey et al. Renewable and Sustainable Energy Reviews 134 (2020) 110091

presented in Fig. 20. A difference of 5% between the plateaued effec­ The contours of nopt for various values of dh and Recold,in corre­
tiveness values of straight and zig-zag channels is observed for all Recold, sponding to a 10 ◦ C temperature pinch in the recuperator is shown in
in. This suggests that significant increment in the effectiveness is Fig. 22(a). The corresponding contours of channel length is presented in
obtainable within 40 rows for zig-zag channels as well. Modelling higher Fig. 22(b). For a given Recold,in of 104 and dh of 1.5 mm, the nopt obtained
number of rows reduces the pressure drop by ~9% compared to two from Fig. 22(a) is found to be 256 whereas, nopt obtained for the case of
channel pressure drop. However, for effectively using the hybrid TRN straight channel-based PCHE is found to be 254. However, the channel
model, a minimum 20 rows is recommended to characterize the overall length for zigzag and straight channel-based PCHE are 2430 mm and
PCHE performance. 3100 mm, respectively, for Recold,in of 104 and dh of 1.5 mm. The cor­
responding stack width and stack height for zig-zag channel-based PCHE
5.2.2. Selection of optimum stack dimensions based on minimum rate of are 695 mm and 768 mm, respectively. Corresponding values for a
heat loss straight channel-based PCHE are found to be 690 mm and 764 mm. For
The temperature profiles for 10 zig-zag channels comprising of 5 hot minimum heat loss, the optimum stack dimensions (except channel
and 5 cold channels is shown in Fig. 21. Following the discussion in length) and nopt for both straight and zigzag channel-based PCHEs are
Section 5.1.5, 40 rows are also considered for estimating the average nearly identical at a given Recold,in and dh.
temperature of outer surfaces for the case of zig-zag channel-based For the selected channel length of 1500 mm and channel diameter of
PCHEs. The average temperatures of outer surfaces are used to estimate 1.5 mm for the recuperator, the variation in optimum effectiveness and
the rate of heat loss (Eq. (17)) in a similar way as discussed in the case of pressure drop with Recold,in based with nopt is shown in Fig. 23. The op­
straight channels. timum effectiveness and optimum pressure drop values across the stack
are estimated to be 92.34% and 25 kPa, respectively for a Recold,in of 104.

phot,opt
T

opt

Recold,in
hot-1 hot-2 hot-3 hot-4 hot-5 L d
cold-1 cold-2 cold-3 cold-4 cold-5
Fig. 23. Variation of optimum effectiveness and optimum hot side pressure
Fig. 21. Temperature profiles for hot and cold streams along the channel
drop with cold inlet Reynolds number for zigzag channel.
length of ten zigzag channels.

Fig. 22. (a) Contours of nopt and (b) contours of channel length for various values of dh and cold inlet Reynolds number satisfying a temperature pinch of 10 ◦ C at the
cold inlet using zigzag channel-based PCHE.

16
V. Pandey et al. Renewable and Sustainable Energy Reviews 134 (2020) 110091

The optimum stack dimensions are found to be 1500 × 710 × 751 mm3 pressure drop and exit temperatures are compared with experimental
(Lopt × Wopt × Hopt). Comparing the results presented in Fig. 23 for a zig- measurements reported in Kim et al. [54]. The boundary conditions and
zag channel, with the data for the straight channel PCHE stack shown in geometrical dimensions for zigzag semi-circular channel reported by
Fig. 18, the performance enhancement on account of channel configu­ Kim et al. [54] are listed in Table 8 and Table 9, respectively.
ration variation is evident in both the performance and PCHE stack size. The correlations used by Kim et al. [54] for laminar flow in a
semi-circular channel are given below.
The Nusselt number correlation based on local pitch averaged
5.3. Experimental validation method is given by:

The TRN model presented in the paper has been numerically vali­ Nu = 4.065 + 0.00305.Re, 350 < Re < 800, Pr = 0.66 (43)
dated for square/rectangular flow channels. However, for semi-circular The fanning factor correlation based on global pitch averaged
cross-sections necessary corrections in convective and conductive ther­ method is given by:
mal resistances term (refer Table 3 in section 3.1.1) in the TRN model
needs to be incorporated. The corrections are on account of variation in f .Re = 16.51 + 0.01627.Re, 350 < Re < 1200 (44)
heat transfer area and pressure drop based on perimeter and hydraulic The exit temperatures of hot and cold streams obtained from 1)
diameter of semi-circular cross section. The corrected convective resis­ experimental data reported in Kim et al. [54], 2) a system analysis code
tance and hydraulic diameter for a semi-circular cross-section channel GAMMA developed by KAIST [67], and 3) present hybrid TRN model are
are listed below: summarized in Table 10. In addition, the table also provides the hot side
The convective thermal resistance for semi-circular PCHE can be pressure drop obtained from experiments and correlation proposed by
obtained as follows: Kim et al. [54] and TRN model using CFD based correlations proposed
1 by Kim et al. [54]. As observed from the table, the TRN predictions very
Rc(i,j) = (39) closely follow the experimental data. This confirms the ability of the
h(i,j) (psemi Δx)
proposed hybrid TRN model to design and predict PCHE performance
where psemi is the perimeter of semi-circular cross-section which can be with reasonable accuracy. In addition to performance predictions, the
defined as: hybrid tool can be effectively used to generate a range of PCHE core
configurations based on specific user defined/design requirements
psemi = dsemi (π/2 + 1) (40)
before undertaking full-scale CFD simulations.
The equivalent thickness between hot and cold channel is estimated
as [54]: 6. Conclusions

teqn = dsemi (1 − π/8) (41)

The paper presents a robust design tool for estimating the stack size
To estimate the local heat transfer coefficient and pressure drop, the of a sCO2 recuperator. The tool uses a hybrid approach comprising of a
hydraulic diameter is calculated as: 1-D Thermal Resistance Network (TRN) model and CFD based correla­
tions to effectively predict the variation in temperature and pressure
πdsemi along the channel length. The strength of the design tool is demonstrated
dh,semi = (42)
π+2 for a PCHE core comprising of straight and zig-zag channels. Thermo-
For the purpose of experimental validation, simulations using hybrid hydraulic analysis considering the effect of geometrical parameters
model are performed with helium as the working fluid and the resulting and Reynolds number are performed for straight channel using con­
ventional correlations. Nusselt number and Fanning factor correlations
Table 8 are derived using 3-D CFD simulations performed using a unit cell
Boundary conditions for experimental comparison Kim et al. [54]. approach. Thereafter, conventional correlations for straight channels
T (K) p Mass flow Working
are replaced with CFD based correlations for stack optimization studies
(kPa) rate (kg/s) fluid based on minimum stack heat loss. Stack optimization studies for zig-zag
Low High
channels are also performed and the results are compared with the
temperature temperature
straight channels. The results obtained using hybrid model for zig-zag
Hot 576.45 804.95 1900 1.6827 × helium
channel show good agreement with the available experimental data.
in 10− 5
Cold 335.05 354.42 1900 1.6827 × helium
The following inferences are drawn based on the case study of a sCO2
in 10− 5 recuperator designed and analysed using the hybrid tool.

a) TRN model is a unique tool which can be applied for thermo-

Table 9 hydraulic analysis of full-scale PCHEs with straight/non-straight
Geometrical dimensions of zigzag channel Kim et al. [54]. channels.
Channel diameter Width of Plate Angle of Pitch of b) The effectiveness increases and pressure drop decreases with number
length unit cell thickness turn zigzag of rows (representing stack height). After a certain stack height
channel
which corresponds to 40 rows, both pressure drop and effectiveness
742 mm 1.51 mm 2.62 mm 1.46 mm 150◦ 24.6 mm

Table 10
Comparison between TRN model results and experimental data of Kim et al. [54] for exit temperatures.
Sl. No. Low temperature study High temperature study

Experiment GAMMA calculation using local Nu TRN model calculation Experiment GAMMA calculation TRN model calculation
using local Nu

Hot exit temperature [K] 345.58 347.84 346.28 375.34 376.76 374.1
Hot exit temperature [K] 563.47 563.65 565.21 784.04 782.04 785.26
Hot side pressure drops [kPa] 11.76 kPa – 12.7 15.32 kPa – 17.9

17
V. Pandey et al. Renewable and Sustainable Energy Reviews 134 (2020) 110091

do not change appreciably for a given channel diameter, channel Credit author statement
mass flow rate and channel length.
c) The concept of minimum heat loss from the outer surfaces of the Vivek Pandey: Modelling, Writing - original draft, Pramod Kumar:
PCHE is used to arrive at the optimum number of rows (stack height Writing- Reviewing and Editing, Conceptualization, Methodology, Pra­
(H)) and stack width (W) for a given set of channel hydraulic dip Dutta: Editing, reviewing, Conceptualization
diameter, length and Reynolds number (corresponding to channel
mass flow rate). Declaration of competing interest

The current TRN model is an important tool for manifold optimiza­ The authors declare that they have no known competing financial
tion studies in PCHEs considering flow maldistribution as it can predict interests or personal relationships that could have appeared to influence
the equivalent thermo-hydraulic performance by accommodating vari­ the work reported in this paper.
able channel mass flow rate and inlet pressure-temperature conditions
across each channel. It must be noted that the current TRN is a simplified Acknowledgement
model which does not account for heat transfer rate in the side width
(thickness of wall between two channels in the same plate) of the col­ The work reported in this paper is supported by the grant (Grant No.
umn of channels. However, an extended TRN model can further reduce TMD/CERI/Clean Coal/2017/034 (IISc) (G) dated 13-09-2018) from the
the stack dimensions for desired thermo-hydraulic performance. Department of Science and Technology, Government of India.

Appendix A. Simplified stress analysis for thickness and side width estimation

Simplified stress analysis has been performed to evaluate t and w. The plate thickness criteria can be obtained based on the method of Dostal [68].
The plate can be assumed to be a thick wall cylinder, with an inner radius of d/2 and an outer radius of tp. For thick wall cylinders, the tangential stress,
σ t, is calculated as [69]:
[ ] [ ]
Pi (ro /ri )2 + 1 − Po (ro /ri )2 + (ro /r)2
σt = 2
i
(45)
(ro /ri ) − 1

Fig. 24. Illustration of typical dimensions associated with PCHE cross section.

where r is the radius, P is the pressure, and the subscript i and o refer to the inner and outer surface, respectively. The maximum tangential stress is
always occurring at the inner surface. For maximum tangential stress to be less than allowable stress σ D, the required radius ratio can be obtained by
replacing r by ri in the above equation. For the cases where the internal pressure exceeds the external pressure, the limiting ratio is given by,
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
ro σ D + Pi
≥ (46)
ri σD + 2Po − Pi

where ri and ro denote the inner and outer radii of the pipe. For the present case, the external pressure on the piping is atmospheric, it is reasonable to
assume zero external gauge pressure, i.e., Po = 0. Pi is the internal pressure, which is 21.22 MPa, ro is tp, and ri is d as mentioned in Fig. 24.
Hesselgraves [6] recommends the following formula for the minimum wall thickness/side width w:
1
w=( ) (47)
σD
ΔP
+ 1 NF

where ΔP is the pressure differential between the hot and cold stream. Nf is the number of fins per meter and σD is the allowable stress. For PCHE, the
number of fins per meter can be taken as equal to the number of channel walls per meter, which is given as [68]:
1 1
NF = = (48)
pm,square d + w

18
V. Pandey et al. Renewable and Sustainable Energy Reviews 134 (2020) 110091

where pm,square is the pitch of the heat exchanger channel and d is the edge of the square cross section as shown in Fig. 24. In the present case, SS 316 L is
used which is commonly employed material for the diffusion bonded PCHEs, due to its durability in high temperature and high-pressure environment.
Allowable stress of 80 MPa corresponding to a temperature of 480 ᵒC is taken for minimum plate thickness and wall thickness calculations [70].
According to the above criteria, the ratio of side width to edge length (channel hydraulic diameter) and the ratio of plate thickness to edge length are
given as follows:
w t
≥ 0.17; ≥ 0.16; (49)
d d
Based on the lower limit of w and t, the selected values for side width and plate thickness are assumed to be 1.5 mm covering the wide range of edge
length (side of the square cross section) studied in this work.

Appendix B. Mesh/element sensitivity study

Grid convergence index (GCI) method developed by Roache [71] based on the principle of generalized Richardson extrapolation theory is used for
quantifying the discretization error bands resulting from mesh refinement. Table 11 presents the mesh resolutions used for the GCI analysis for a 1 m
channel length. The details of mesh and solid material used in the simulation are presented in Table 4. The number of cells selected is five. The cell size
h for 3-D calculation can be defined as,

[ ]1/3
1 ∑N
h= ΔVi (50)
N i=1

where ΔVi is the volume of the ith cell and N is the total number of cells used in the computations which should not be confused with the
number of channels (N) discussed subsequently.
Table 11
Mesh resolutions used for GCI analysis.

SL No. Case-1 Case-2 Case-3 Case-4 Case-5

3
Mesh volume [mm ] 18,000 18,000 18,000 18,000 18,000
Number of cells 4706889 2100571 866,871 366,873 146,265
Cell size [mm] 0.1563 0.2046 0.2748 0.3660 0.4974

Based on the cell size, the grid refinement factor is defined as,
hcoarse h2
r21 = = (51)
hfine h1

In GCI analysis, three sets of grid resolutions are required at a time to determine the key variables such as apparent order p.
1
p= |ln|(ε32 / ε21 )| + q(p)| (52)
ln(r21 )
( p )
r − s
q(p) = ln 21
p (53)
r32 − s

s = sign((ε32 / ε21 )) (54)

where ε32 = ϕ3 – ϕ2, ε21 = ϕ2 – ϕ1, ϕk denotes solution variable for k number of cells.
The percentage approximate relative error is obtained as,
⃒ ⃒
⃒φ1 − φ2 ⃒
e21
a = ⃒ ⃒
⃒ φ ⃒ × 100 (55)
1

The grid convergence index is computed as,

( 21 )
e
21
GCIfine = p a (56)
r21 − 1
Fig. 25 shows that the GCI reduced below 0.3% for all solution variables as representative grid size is decreased below 0.28 mm. The average heat
transfer coefficient along the channel is represented by α. Discretization error around 0.2% is assumed acceptable. The corresponding percentage
approximate and extrapolated relative error is <0.3% and <0.5%, respectively for all solution variables.

19
V. Pandey et al. Renewable and Sustainable Energy Reviews 134 (2020) 110091

Fig. 25. Effect of cell size on discretization error.

The GCI analysis for 1-D TRN model is also conducted for the various number of elements along the channel length presented in Table 12. The cell
size h for 1-D calculation can be defined as,
[ ]
1 ∑N
h= Δxi (57)
N i=1

Channel length of 3 m as an extreme case is selected to ensure the sufficiency of the number of cells based on percentage GCI. The outlet tem­
perature and pressure drop of 21st hot and 22nd cold stream among 40 channels modelled using TRN are selected as solution variables for GCI analysis.

Table 12
Percentage GCI for various solution variables obtained from 1-D TRN model.

Number of elements GCIfine (%) [φ = Tco ] GCIfine (%) [φ = Tho ] GCIfine (%) [φ = Δpcold ] GCIfine (%) [φ = Δphot ]

250
200 0.024 0.14 0.13 0.086
150 0.12 0.75 0.7 0.43
100 0.07 0.44 0.41 0.26
50 0.10 0.63 0.61 0.38

Based on GCI<0.45% for all solution variables, 100 elements have been selected in the present analysis. The corresponding percentage approx­
imate and extrapolated relative error is <0.27% and <0.46%, respectively.

References [10] AM A, Sharma A, Srivastava A, Chaudhury P. Effects of wavy channel

configurations on thermal-hydraulic characteristics of Printed Circuit Heat
Exchanger (PCHE). Int J Heat Mass Tran 2018;118:304–15.
[1] Luu MT, Milani D, McNaughton A, Abbas A. Analysis for flexible operation of
[11] Meshram A, Jaiswal AK, Khivsara SD, Ortega JD, Ho C, Bapat R, Dutta P. Modeling
supercritical CO2 Brayton cycle integrated with solar thermal systems. Energy
and analysis of a printed circuit heat exchanger for supercritical CO2 power cycle
2017;124:752–71.
applications. Appl Therm Eng 2016;109:861–70.
[2] –>Monje B, Sanchez D, Savill M, Pilidis P, Sanchez T. Asme. A design strategy for
[12] Yoon SJ, Brien JO, Chen M, Sabharwall P, Sun X. Development and validation of
supercritical CO2 compressors. Proceedings of the ASME turbo expo 2014: turbine
Nusselt number and friction factor correlations for laminar flow in semi-circular
technical conference and exposition. Volume 3B: oil and gas applications; organic
zigzag channel of printed circuit heat exchanger. Appl Therm Eng 2017;123:
rankine cycle power systems; supercritical CO2 power cycles; wind energy.
1327–44.
Düsseldorf, Germany. ASME; 2014. June 16–20V03BT36A003.
[13] Li1 Y, Xie Y, Zhang D. The thermal hydraulic performance of wavy PCHE with
[3] Lopez A, Monje B, Sanchez D, Chacartegul R, Sanchez T. ASME. Effect of
different materials and geometric parameters. MATEC Web of Conferences 2018;
turbulence and flow distortion on the performance of conical diffusers operating on
207:04008.
supercritical carbon dioxide. In: Proceedings of the ASME turbo expo 2013: turbine
[14] Li YZ, Xie YH, Zhang D. Optimization investigation on wavy channel printed circuit
technical conference and exposition. Volume 8: supercritical CO2 power cycles.
heat exchanger for solar energy powered brayton cycle. IOP Conf Ser Mater Sci Eng
San Antonio, Texas, USA: Wind Energy; Honors and Awards; 2013. June
2019;556:012038.
3–7V008T34A002. ASME.
[15] Tsuzuki N, Kato Y, Nikitin K, Ishizuka T. Advanced microchannel heat exchanger
[4] Ahn Y, Bae SJ, Kim M, Cho SK, Baik S, Lee JI, Cha JE. Review of Supercritical CO2
with S-shaped fins. J Nucl Sci Technol 2009;46:403–12.
power cycle Technology and current status of research and development. Nucl.
[16] Carlson M. Measurement and analysis of the thermal and hydraulic performance of
Eng. Technol 2015;47:647–61.
several printed circuit heat exchanger channel geometries. MS Thesis: University of
[5] Garg P, Kumar P, Srinivasan K. A trade-off between maxima in efficiency and
Wisconsin Madison; 2012.
specific work output of super-and trans critical CO2 Brayton cycles. J Supercrit
[17] Shi HY, Li MJ, Wang WQ, Qiu Y, Tao WQ. Heat transfer and friction of molten salt
Fluids 2015;98:119–26.
and supercritical CO2 flowing in an airfoil channel of a printed circuit heat
[6] Hesselgreaves JE. Compact heat exchangers, selection, design and operation. New
exchanger. Int J Heat Mass Tran 2020;150:119006.
York: Pergamon; 2001.
[18] Yoon SH, No HC, Kanga GB. Assessment of straight, zigzag, S-shape, and airfoil
[7] Mylavarapu S, Sun X, Figley J, Needler N, Christensen R. Investigation of high
PCHEs for intermediate heat exchangers of HTGRs and SFRs. Nucl Eng Des 2014;
temperature printed circuit heat exchangers for very high temperature reactors.
270:334–43.
J Eng Gas Turbines Power 2009;131(6):062905.
[19] Ngo TL, Kato Y, Nikitin K, Tsuzuki N. New printed circuit heat exchanger with S-
[8] Huang C, Cai W, Wang Y, Liu Y, Li Q, Li B. Review on the characteristics of flow
shaped fins for hot water supplier. Exp Therm Fluid Sci 2006;308:811–9.
and heat transfer in printed circuit heat exchangers. Appl Therm Eng 2019;153:
[20] Li Q, Flamant G, Yuan X, Neveu P, Luo L. Compact heat exchangers: a review and
190–205.
future applications for a new generation of high temperature solar receivers.
[9] Kim IH, No HC. Physical model development and optimal design of PCHE for
Renew Sustain Energy Rev 2011;15:4855–75.
intermediate heat exchangers in HTGRs. Nucl Eng Des 2012;243:243–50.

20
V. Pandey et al. Renewable and Sustainable Energy Reviews 134 (2020) 110091

[21] Yang Y, Li H, Yao M, Zhang Y, Zhang C, Zhang L, Wu S. Optimizing the size of a [46] Baik S, Kim SG, Lee J, Lee JI. Study on CO2 – water printed circuit heat exchanger
printed circuit heat exchanger by multi-objective genetic algorithm. Appl Therm performance operating under various CO₂ phases for S-CO₂ power cycle
Eng 2020;167:114811. application. Appl Therm Eng 2017;113:1536–46.
[22] Baik YJ, Jeon S, Kim B, Jeon D, Byon C. Heat transfer performance of wavy [47] Lookingbill DN. Design and preliminary evaluation of a supercritical carbon
channelled PCHEs and the effects of waviness factors. Int J Heat Mass Tran 2017; dioxide brayton cycle for solar dish concentrator clean energy production. Doctor
14:809–15. of philosophy in engineering – mechanical engineering. The University of Nevada,
[23] Lee SM, Kim KY. Multi-objective optimization of arc-shaped ribs in the channels of Las Vegas; 2018. October 29.
a printed circuit heat exchanger. Int J Therm Sci 2015;94:1–8. [48] Dutta A, Gupta A, Sathish S, Bandooni A, Kumar P. Simple recuperated s-CO2 cycle
[24] Jing Q, Xie Y, Zhang D. Thermal hydraulic performance of printed circuit heat revisited: optimization of operating parameters for maximum cycle efficiency. In:
exchanger with various channel configurations and arc ribs for SCO2 Brayton cycle. Proceedings of the ASME turbo expo 2019: turbomachinery technical conference
Int J Heat Mass Tran 2020;150:119272. and exposition. Volume 9: oil and gas applications; supercritical CO2 power cycles.
[25] Kim DE, Kim MH, Cha JE, Kim SO. Numerical investigation on thermal–hydraulic Phoenix, Arizona, USA: Wind Energy; 2019. June 17–21V009T38A002. ASME.
performance of new printed circuit heat exchanger model. Nucl Eng Des 2008;238: [49] Noriega JSB. Design method for s-CO2 gas turbine power plants Integration of
3269–76. thermodynamic analysis and components design for advanced applications, Master
[26] Xu X, Ma T, Li L, Zeng M, Chen Y, Huang Y, Wang Q. Optimization of fin of Science Thesis. Faculty of Mechanical, Maritime and Materials Engineering
arrangement and channel configuration in an airfoil fin PCHE for supercritical CO2 (3mE), Delft University of Technology; 2012. October 21.
cycle. Appl Therm Eng 2014;70(1):867–87. [50] Sidebotham G. Heat transfer modelling-an inductive approach, vol. 405. Springer
[27] Li XL, Tang GH, Fan YH, Yang DL, Wang SQ. Numerical analysis of slotted airfoil Cham Heidelberg); 2015. p. 43. ISBN978-3-319-14514-3.
fins for printed circuit heat exchanger in S-CO2 brayton cycle. ASME J of Nuclear [51] National Institute of Standards and Technology, 2009; Gaithersburg, MD, NIST
Rad Sci 2019;5(4):041303. Standard Reference Database 23, NIST.
[28] Ngo TL, Kato Y, Nikitin K, Ishizuka T. Heat transfer and pressure drop correlations [52] Gnielinski V. New equations for heat and mass transfer in turbulent pipe and
of micro-channel heat exchangers with S-shaped and zigzag fins for carbon dioxide channel flow. Int Chem Eng 1976;16(2):359–68.
cycles. Exp Therm Fluid Sci 2007;32:560–70. [53] Filonienko GK. Friction factor for turbulent pipe flow. Teploenergetika 1954;4(1):
[29] Saeed M, Kim MH. Thermal-hydraulic analysis of sinusoidal fin-based printed 40–4.
circuit heat exchangers for supercritical CO2 Brayton cycle. Energy Convers Manag [54] Kim IH, No HC, Lee JI, Jeon BG. Thermal hydraulic performance analysis of the
2019;193:124–39. printed circuit heat exchanger using a helium test facility and CFD simulations.
[30] Ma T, Li L, Xu XY, Chen YT, Wang QW. Study on local thermal–hydraulic Nucl Eng Des 2009;239(11):2399–408.
performance and optimization of zigzag-type printed circuit heat exchanger at high [55] Kalitzin G, Medic G, Iaccarino G, Durbin P. Near-wall behaviour of RANS
temperature. Energy Convers Manag 2015;104:55–66. turbulence models and implications for wall functions. J Comput Phys 2005;204:
[31] Jeon S, Baik YJ, Byon C, Kim W. Thermal performance of heterogeneous PCHE for 265–91.
supercritical CO2 energy cycle. Int J Heat Mass Tran 2016;102:867–76. [56] Churchill SW, Chu HHS. Correlating equations for laminar and turbulent free
[32] Kim IH, Sun XX. CFD study and PCHE design for secondary heat exchangers with convection from a vertical plate. Int J Heat Mass Tran 1975;18:1323–9.
FliNaK-Helium for SmAHTR. Nucl Eng Des 2014;270:325–33. [57] Lloyd JR, Moran WR. Natural convection adjacent to horizontal surfaces of various
[33] Figley J. Numerical modeling and performance analysis of printed circuit heat planforms. New York: ASME; 1974. Paper 74-WA/HT-66.
exchanger for very high- temperature reactors. M.S. thesis. The Ohio State [58] Incropera FP, DeWitt DP. Fundamentals of heat and mass transfer. third ed. New
University; 2009. York: Wiley; 1990. p. 539–40.
[34] Figley J, Sun X, Mylavarapu SK, Hajek B. Numerical study on thermal hydraulic [59] Goldstein RJ, Sparrow EM, Jones DC. Natural convection mass transfer adjacent to
performance of a printed circuit heat exchanger. Prog Nucl Energy 2013;68:89–96. horizontal plates. Int J Heat Mass Tran 1973;16:1025–35.
[35] Daouk A. Performance analysis and modelling of a printed circuit heat exchanger [60] Carlson MD, Kruizenga AK, Schalansky C, Fleming DF. Sandia progress on
with air and carbon dioxide as working fluids. Master of applied science in advanced heat exchangers for sCO2 brayton cycles. In: The 4th international
mechanical engineering 2016; Department of Mechanical and Aerospace symposium- supercritical CO2 power cycles september 9-10; 2014. Pittsburgh
Engineering Carleton University Ottawa. Pennsylvania.
[36] Guo J, Huai X. Coordination analysis of cross flow heat exchanger under high [61] Pandey V, Kumar P. Thermal resistance network Model for a recuperator in s-CO2
variations in thermodynamic properties. Int J Heat Mass Tran 2017;113:935–42. power cycle. In: Proceedings of the 16th international heat transfer conference;
[37] Kwon JS, Bae SJ, Heo JY, Lee JI. Development of accelerated PCHE off-design 2018. p. 4919–30. IHTC-16 August 10-15: Beijing, China.
performance model for optimizing power system operation strategies in S-CO2 [62] Massey BS. Mechanics of fluids. eighth ed., 7.5. Taylor & Francis; 2006. p. 254.
Brayton cycle. Appl Therm Eng 2019;159:113845. chapter 7.
[38] Ke H, Xiaoa Q, Cao Y, Ma T, Lin Y, Zenb M, Wang Q. Simulation of the printed [63] Zhao Z, Zhang X, Zhao K, Jiang P, Chen Y. Numerical investigation on heat transfer
circuit heat exchanger for S-CO2 by segmented methods. Energy Procedia 2017; and flow characteristics of supercritical nitrogen in a straight channel of printed
142:4098–103. circuit heat exchanger. Appl Therm Eng 2017;126:717–72.
[39] Utamura M, Nikitin K, Kato Y. A generalised mean temperature difference method [64] Mroue H, Ramos JB, Wrobel LC, Jouhara H. Experimental and numerical
for thermal design of heat exchangers. Int J Nucl Energy Sci Technol 2008;4:11–31. investigation of an air- to water heat pipe-based heat exchanger. Appl Therm Eng
[40] Neises T, Turchi C. A comparison of supercritical carbon dioxide power cycle 2015;78:339–50.
configurations with an emphasis on CSP applications. Energy Procedia 2014;49: [65] Kim W, Baik YJ, Jeon S, Jeon D, Byon C. A mathematical correlation for predicting
1187–96. the thermal performance of cross, parallel, and counterflow PCHEs. Int J Heat Mass
[41] Turchi C, Ma Z, Neises T, Wagner M. Thermodynamic study of advanced Tran 2017;106:1294–302.
supercritical carbon dioxide power cycles for high performance concentrating solar [66] Çengel YA, Cimbala JM. Fluid mechanics: fundamentals and applications (1sted.).
power systems– preprint. In: Proceedings of ASME. 6th international conference on McGraw-Hill Higher Education; 2006. 0–07–247236–7.
energy sustainability & 9th fuel cell science. Engineering and Technology [67] Lim HS, No HC. GAMMA multidimensional multicomponent mixture analysis to
Conference; 2012. predict air ingress phenomena in an HTGR. Nucl Sci Eng 2006;152(1):87–97.
[42] Guo J. Design analysis of supercritical carbon dioxide recuperator. Appl Energy [68] Dostal V. A supercritical CO2 cycle for next generation nuclear reactors.
2016;164:21–7. Cambridge: ScD Thesis, Massachusetts Institute of Technology (MIT); 2004.
[43] Bennett K, Chen YT. Printed circuit heat exchanger performance analysis using http://web.mit.edu/22.33/www/dostal.pdf.
non-uniform segmental design method. Appl Therm Eng 2019;153:69–84. [69] Crandal SH, Dahl NC, Lardner TJ. An introduction to the mechanics of solids.
[44] Bone V, McNaughton R, Kearney M, Jahn I. Methodology to develop off-design second ed. New York: McGraw-Hill Book Company; 1972.
models of heat exchangers with non-ideal fluids. Appl Therm Eng 2018;145: [70] American Society of Mechanical Engineers. ASME boiler & pressure vessel code.
716–34. New York: An International Code; 2007.
[45] Marchionni M, Chai L, Bianchi G, Tassou SA. Numerical modelling and transient [71] Roache PJ. Quantification of uncertainty in computational fluid dynamics. Annu
analysis of a printed circuit heat exchanger used as recuperator for supercritical Rev Fluid Mech 1997;29:123–60.
CO2 heat to power conversion systems. Appl Therm Eng 2019;161:114190.

21