Accepted Manuscript

An Innovative Plate Heat Exchanger of Enhanced Compactness

Francesco Vitillo, Lionel Cachon, Philippe Reulet, Emmanuel Laroche, Pierre Millan

PII: S1359-4311(15)00461-5
DOI: 10.1016/j.applthermaleng.2015.05.019
Reference: ATE 6618

To appear in: Applied Thermal Engineering

Received Date: 25 September 2014

Revised Date: 29 April 2015
Accepted Date: 2 May 2015

Please cite this article as: F. Vitillo, L. Cachon, P. Reulet, E. Laroche, P. Millan, An Innovative Plate
Heat Exchanger of Enhanced Compactness, Applied Thermal Engineering (2015), doi: 10.1016/
j.applthermaleng.2015.05.019.

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.
 ACCEPTED MANUSCRIPT
1 AN INNOVATIVE PLATE HEAT EXCHANGER OF ENHANCED

2 COMPACTNESS

3 Francesco Vitillo1, Lionel Cachon1,*, Philippe Reulet2, Emmanuel Laroche2, Pierre Millan2
1
4 DEN/CAD/DTN/STPA/LCIT, Bât.204 13108 Saint-Paul-lez-Durance, FRANCE
2

PT
5 ONERA – The French Aerospace Lab, F- 31055 Toulouse, FRANCE

6 Tel:* +33 (0) 4 42 25 7425 , Email:* lionel.cachon@cea.fr

RI
7

8 Abstract

SC
9 In the framework of CEA R&D program to develop the Advanced Sodium Technological Reactor for Industrial

10 Demonstration (ASTRID), the present work aims to demonstrate the industrial interest of an innovative compact

U
11 heat exchanger technology. In fact, one of the main innovations of the ASTRID reactor could be the use of a
AN
12 Brayton Gas-power conversion system, in order to avoid the energetic sodium-water interaction if a traditional

13 Rankine cycle was used. The present work aims to study the thermal-hydraulic performance of the innovative
M

14 compact heat exchanger concept. Hence, thanks to a trustful numerical model, friction factor and heat transfer

15 correlations are obtained. Then, a global compactness comparison strategy is proposed, taking into account
D

16 design constraints. Finally, it is demonstrated that the innovative heat exchanger concept is more compact then

17 other already industrial technologies of interest, showing that is can be considered to warrant serious
TE

18 consideration for future ASTRID design as well as for any industrial application that needs very compact heat

19 exchanger technologies.
EP

20 1. Innovative Concept Motivation and Description

21 In the framework of CEA R&D program to develop the Advanced Sodium Technological Reactor for Industrial
C

22 Demonstration (ASTRID) [1], an innovative compact heat exchanger technology has been proposed [2], [3] to
AC

23 provide solid technological basis for the use of a Brayton Gas-power conversion system.

24 In fact, compact heat exchangers deserve an increasing attention for 4th Generation nuclear reactors due to their

25 potential application in several component of a nuclear island. In this sense, the foreseen applications are for

26 GFRs using S-CO2 as thermal fluid for the power conversion cycle [4], [5], [6],[7], [8], for GFRs with traditional

27 gas Brayton power conversion cycle [9], [10] or in VHTRs [11], [12], [13], [14], [15], [16] and [17].

28 Applications to SFRs are more recent and still limited to S-CO2 Brayton cycle cases [18] or to the design of

29 auxiliary components such as the intermediate heat exchanger [19] or the DHRS heat exchanger [20]. While
 ACCEPTED MANUSCRIPT
30 studies are still ongoing on possible SFR power conversion systems ([21], [22] and [23]), work done at CEA

31 made it possible to design a first-of-a-kind sodium-gas heat exchanger based on PCHE technology [24]. The

32 ASTRID sodium-gas heat exchanger would couple the secondary sodium loop with the tertiary nitrogen loop

33 [25]. Tthis choice is of interest for SFRs since it allows avoiding the energetic sodium-water interaction if a

34 traditional Rankine cycle was used. However a high mechanical component is needed, since nitrogen is supposed

35 to have a 180 bar operating pressure, whereas the sodium side is typically at atmospheric pressure. The design

PT
36 of the gas-side (which determines the heat transfer resistance of the heat exchanger) of the Sodium-gas heat

37 exchanger motivated the present work. In fact, compact technologies are fundamental for the present application

RI
38 since the low heat transfer capacity of the gas foreseen (i.e. nitrogen) would result in a excessive heat transfer

SC
39 surface if conventional shell and tube heat exchangers were used. Hence it is worth studying innovative

40 geometries enhancing the heat transfer of the gas side and therefore the compactness of such a component.

41 However, the interest of the innovative geometry channel is not limited to the gas side of the ASTRID heat

42
U
exchanger: indeed it is of interest for all industrial applications needing compact heat exchanger technologies,
AN
43 (especially those working with fluids having poor thermal capacity) and high mechanical performance.

44 The innovative channel geometry is shown in Figure 1:

M
D
TE

45
EP

46 Figure 1- Superposed channels identification

47 The channel can be thought as the superposition of two single wavy channels in phase opposition (white and
C

48 yellow in Figure 1). The wavy channel are engraved on metal (i.e. any kind of steel) PCHE-type plates. Indeed,
AC

49 considering a plate consisting of several parallel wavy channels, the actual fluid stream is composed by several

50 wavy channels intersecting each other in several points, creating a fully 3D flow. An idea of the final

51 configuration is shown in Figure 2:

 ACCEPTED MANUSCRIPT

PT
52

53 Figure 2 - Example of the innovative fluid stream configuration (actual cross-section will be defined later)

RI
54 The innovative channel is composed by three elementary geometrical elements i.e. bends, straight channels and

55 mixing zones. If the bends are present in each of the two superposed wavy channels composing the innovative

SC
56 channels, the mixing zone can be thought as the region corresponding to the intersection of the two superposed

57 channels: they can communicate each other through an “open window” called mixing plane as Figure 3 explains:

U
AN
M
D
TE

58
C EP
AC

59

60 Figure 3- Mixing zone and mixing plane visualization

61 The straight duct links the bends with the mixing zones.
 ACCEPTED MANUSCRIPT
62 The innovative channel geometry can be identified by selecting four geometrical parameters: the half channel

63 cross-section geometry, the proper angle α with regard to the fluid principal direction, the straight duct length L

64 between two bends and the bend curvature radius R (Figure 4).

PT
RI
65

SC
66 Figure 4- Channel main geometrical parameters

67 Since the choice of these three geometrical parameters can be done arbitrarily, the reference half channel cross

68
U
section for the innovative channel in the present work is rectangular, with the shorter side D/2 equals to half the
AN
69 longer side D, i.e. to provide a global squared cross-section of height D. Other cross-section geometries will be

70 eventually studied and evaluated. The other geometrical parameters α, R and L will be specified for each
M

71 geometry of interest. The innovative geometry creates a strong fluid deformation: is has been shown [3] that the

72 flow is very three-dimensional with a continuous detachment and reattachment of the boundary layer. Even if
D

73 preliminary results suggest that the complexity of the flow is responsible for heat transfer enhancement [3], no
TE

74 performance evaluation has been done so far. Hence, the aim of the present work is to provide design basis and

75 motivation for the industrial interest of this innovative channel concept.

EP

76 2. Numerical model formulation

77 Numerical modeling has been identify as a primary mean to investigate the innovative channel flow, since a
C

78 wide experimental campaign covering all possible design options was considered as unpractical. Hence, the
AC

79 Anisotropic Shear Stress Transport (ASST) [2], [26] model has been identified to simulate the fluid flow. The

80 ASST model is supposed to improve SST model results especially for secondary motions [26], the SST model

81 having found to be already able to reproduce a PCHE as well as the innovative channel fluid flow fairly well

82 [27], [28]. The ASST model is coupled to a Simple Gradient Diffusion Hypothesis modeling approach for the

83 full thermal-hydraulic model. Its formulation consists of a nonlinear formulation of the Reynolds stress tensor as

84 follows:
 ACCEPTED MANUSCRIPT
2 1 1
ρ ρ Ω Ω Ω Ω
3 3 3

ρ Ω S Ω S ! #1$

%&' %&) %&' %&)

85 where , Ω and where τ is the turbulent time scale
%() %(' %() %('

86 , with β*=0.09.
*∗,

87 The closure coefficients C1, C2 and C3 are evaluated as follows:

PT
-.
#2$
# -./ -.0 ∙ $

RI
-.
#3$
# -.0 -.0 ∙ $

SC
-.
#4$
# -./ -.0 ∙ $

With S τ4 S56 S56 #5$

U
AN
88

89 The values of the five coefficients are:

M

90 CNL1=0.8, CNL2=11, CNL3=4.5, CNL4=1000, CNL5=1.

91 The eddy viscosity is evaluated as follows:

8
#6$
D

8=
9:; < ,@ ∗A
>
TE

92 Where:

1 /
∗
#7$
EP

8 : ∗ ∗ ∗
! #8$
C

7.407
> #9$
H
AC

J
93 The value of the A1 coefficient is H
K
.

94 The F2 function is the same as the SST model [29].

95 The model is closed through the k and ω transport equations:

L L L L
# $ ! N P
LM L; L; O L;

R
Q S ∗ = #10$
 ACCEPTED MANUSCRIPT
L L L L=
# =$ = ! N P
LM L; L; O, L;

Q, S= T, #11$

96 The turbulence kinetic energy production term can be expressed in the proper form:

LU
Q #12$
L;

PT
97 Finally the specific dissipation rate is expressed in the original form i.e.
V
Q, Q #13$

RI
98 The other parameters not mentioned so far maintains the same definition as the original SST model formulation

99 [29]. Closure constants are:

SC
O , 1.176, O , 1.0, O,, 2.0, O,, 1.168,

S∗ 0.09, : 0.31, S , 0.075, S , 0.0828

100
U
Regarding the thermal modeling approach, a Simple Gradient Diffusion Hypothesis (SGDH) model is used. It
AN
101 consists of modeling the turbulent heat flux ′ X as a function of the temperature gradient, i.e.:

Y LΘ
′ X′ #14$
O L;
M

102 with È being the average temperature and è’ being the fluctuating temperature. The σ\ is the turbulent Prandtl
D

103 number, which is assumed to be constant and equal to 0.85.

TE

104 3. Experimental database acquisition

105 To acquire a wide experimental database on the flow behavior, three test sections have been used. They are
EP

106 briefly discussed hereafter, referring to reference [3] for a thorough description and analyses.

107 3.1 – LDV measurement campaign

C

108 A 2-C LDV setup has been used to measure the principal and the radial mean and fluctuating velocities. The used
AC

109 technical means are a Spectra Physics Stabilité 2017 Laser, a TSI Colorburst Multicolor Beam Sepatator Model,

110 an ISEL displacing system, a TSI Colorlink Plus Multicolor Receiver Model 9230 and a TSI FSA4000 Multibit

111 Digital Processor. The visualization particles are Di-Ethyl-Hexyl-Sebacat (DEHS) droplets. The working fluid is

112 air at atmospheric pressure and temperature (i.e. p=1 bar and T = 20 °C). The inlet velocity is 13 m/s, to have a

113 correspondent inlet Reynolds number of 50 600. To evaluate the experimental uncertainty, three sources of

114 uncertainty have been generally identified: in particular, they are the uncertainty due to the data acquisition

115 chain, the uncertainty due to environmental conditions and the uncertainty due to the measurement volume
 ACCEPTED MANUSCRIPT
116 position. The final uncertainty is calculated adding the three obtained variances, calculated as the 3σ where σ is

117 the combined standard deviation determined by adding the variances of each source of uncertainty. The

118 evaluated uncertainties for the velocity fluctuations for the in-bend flow are given in Table 1 [3]:

Total 3σ
Uncertainty
Principal velocity U 0.31 m/s
Radial velocity W 0.15 m/s
Principal velocity

PT
0.07 m/s
fluctuation u’
Radial velocity fluctuation
0.09 m/s
w’
119 Table 1 - LDV fluctuating velocity uncertainty

RI
120 3.2 –PIV measurement campaign

SC
121 Differently from the LDV measurements, the inlet Reynolds number for PIV measurements was 12 000

122 (corresponding to an inlet velocity of 4.2 m/s), to acquire the database on a wider range of Reynolds numbers.

U
123 Measurements are done with air at atmospheric pressure and temperature (i.e. p=1 bar and T = 20 °C).
AN
124 The used laser is a double cavity 2x200 mJ pulsed QUANTEL EverGreen 200 YAG Laser, providing a 532 nm

125 (green) light sheet. Its pulse frequency is between 0 and 15 Hz. The camera was a PowerViewTM Plus 4MP
M

126 camera, with a frame-straddling time up to 200 ns.

127 The visualization particles are olive oil droplets. Droplets are created by a “TSI Oil Droplet Generator 9307”
D

128 atomizer, feed by pressurized air. The average olive oil droplet size map is typically 1 µm. Finally, data
TE

129 acquisition, pre and post-processing is done thank to the “TSI Insight 4G” software.

130 Again, three types of uncertainty are identified: the uncertainty due to the data acquisition, the uncertainty due
EP

131 to environmental conditions and the uncertainty due to the measurement plane position. The final uncertainty

132 (given in Table 2) is calculated as the 3σ total uncertainty, where σ is the uniform probability density function
C

133 standard deviation.

AC

Total 3σ Uncertainty

Velocity Magnitude 0.80 m/s

u’ Magnitude 0.38 m/s

v’ Magnitude Uncertainty 0.38 m/s

134 Table 2 - PIV fluctuating velocity uncertainty

 ACCEPTED MANUSCRIPT
135 To verify that the different Reynolds number between the LDV and the PIV measurements does not results in

136 major differences in term of flow characteristics, a comparison of the non-dimensional principal velocity in the

137 45° profile from the bend inlet is done, using the directly measured LDV data and PIV data extracted from the

138 measured bend flow velocity field. The comparison is illustrated in Figure 5.

PT
RI
SC
139

U
140 Figure 5 - LDV vs PIV bend 45° principal velocity profile comparison

141 See that the two profiles are practically the same, the highest difference detected (of around 0.05 m/s) being
AN
142 lower than the sum of the LDV and PIV uncertainties. This is of some importance to affirm that, once in a fully

143 turbulent flow regime, the occurring physical phenomena are practically the same no matter the specific
M

144 Reynolds number, at least in the a range not too far from the experimental one (i.e. 12 000< Re < 50 000).
D

145 3.3 – VHEGAS measurement campaign

TE

146 For thermal measurements, “1 mm - type K” thermo-couple are installed into the mockup (which is

147 geometrically different than the others to globally test multiple design geometries[3]) thanks to several holes
EP

148 machined on the mockup’s surface. Twenty-two TCs have been used, both to measure the inlet/outlet fluid

149 temperature and the wall temperature. The channel side and channel top/bottom TCs are located at 1 mm
C

150 distance from the channel wall. This has been done to approximate the actual local wall temperature by the

151 measured local temperature. The inlet/outlet TCs are located at channel cross-section center. The mockup is
AC

152 heated up by two heating plates, which are placed on the top and bottom surface of the mockup. The heating

153 plates provide a controlled thermal power of 75 W (this low value is chosen to neglect thermal compressibility

154 effects that are not taken into account by the numerical model). Injected thermal power is controlled by a

155 current-controller system. The mockup is placed inside a thermal insulation box. The flow inlet is given by 7-

156 bar pressurized air underexpanded to atmospheric temperature and pressure. The inlet Reynolds number was

157 12 000. A Serv Instrumentation Vortex flow-meter type 8800 is used to measure volumetric flow rate upstream
 ACCEPTED MANUSCRIPT
158 the mockup. The final goal of the VHEGAS test section is to measure the channel global heat transfer

159 coefficient, defined as follows:

^
] #15$
#_`a _bc $

160 Where ^ is the average heat flux on mockup wall (calculated by thermal power heating the fluid flow divided

161 by the known heat transfer surface), _`a is the average wall temperature (evaluated by the arithmetic average

PT
162 dbc
of nineteen TCs 1 mm far from the actual channel wall) and _ is the average fluid temperature. The latter is

dbc #;$ (x being channel curvilinear abscissa) obtained by the

calculated as the average value of a function _

RI
163

164 channel center temperature in three different points (i.e. channel inlet, outlet and between the two).

SC
165 The uncertainty evaluation has to be done for each measured temperature, and then propagated to derived

166 values, such as the absorbed thermal power and heat flux on channel walls.

U
167 Using a typical standard deviation σ of the used TCs of ±0.5 °C the total uncertainty is evaluated as 3 σ. The
AN
168 final 3σ uncertainty for inlet temperature, outlet temperature, and average wall temperature is shown in Table 3,

169 together with the final propagate uncertainties for the derived variables of interest.
M

Variable Mean Value ± Total 3σ Uncertainty

D

Inlet Temperature 33.3 ± 1.5 °C

TE

Outlet Temperature 67.2 ± 1.5 °C

Mean wall Temp. 64.4 ± 1.5 °C

EP

Volumetric Flow Rate 11.8 ± 0.2 m3/h

Absorbed thermal power 121.4 ± 6.2 W

C

Wall Heat Flux 3112 ± 159 W/m2

AC

170 Table 3 - Derived uncertainties of thermal variables

171 4. Model Validation

172 With the described database it is possible to validate the ASST model. To do it, computations on actual channel

173 used in the three experiments are done. A near-wall approach has been used to obtain proper Y+ values for the

174 first cell. A sensitivity study on mesh size is done in order to obtain mesh-independent results. Three
 ACCEPTED MANUSCRIPT
175 successively refined meshes (hereafter named as A, B and C according to the refinement level of the near-wall

176 first cell thickness) have been tested. This evaluation is shown in Table 4 for LDV channel case:

177

Configuration Y+ τw
A 2.5 1.001 Pa

PT
B 1.2 0.984 Pa
C 0.7 0.988 Pa
178 Table 4 - LDV channel meshing evaluation

RI
179 From values in Table 4, see that configuration C shows a converged wall shear stress solution. Based on these

180 trends, we retained configuration C as the reference meshing (difference with B configuration of 0.4%). Note

SC
181 that the same approach in terms of mesh evaluation is used for PIV and VHEGAS channel as well, assuring a

182 good mesh convergence for all the studied geometries.

183
U
In particular, see the ASST model results compared to PIV and LDV data in Figure 6 and Figure 7. As regard the
AN
184 thermal validation, only the heat transfer coefficient value measured at Re=12 000 is showed in Table 5.
M
D
TE
C EP

185
AC
 ACCEPTED MANUSCRIPT

PT
186

RI
187 Figure 6– Bend flow horizontal (up) and vertical (bottom) velocity fluctuation comparison between PIV

SC
188 experimental (left field) and ASST (right field) data [m/s]

U
189
AN
M
D
TE

190
EP

191 Figure 7 - Bend flow principal (left) and secondary (right) velocity fluctuation comparison between LDV

192 experimental (red points) and ASST (black line) results [m/s]
C

h2 efk h l
AC

g °j

VHEGAS exp. Value 218+45

ASST computed Value 187

193 Table 5 - VHEGAS and ASST heat transfer coefficient comparison

194 See a general good agreement between experimental data and ASST results with regard to the velocity

195 fluctuation fields, which are generally harder to get than a velocity field. The general trend is always well

196 described: in particular see that the ASST model can capture turbulence evolution, with the high turbulence
 ACCEPTED MANUSCRIPT
197 intensity at bend inlet due to the turbulence created in the previous mixing zone. Differences between

198 experimental and numerical results are generally spatially localized; in particular, the boundary layer velocity

199 fluctuations are sometimes under-estimated due to the lack of low-Re modifications in the ASST model, which

200 would provide a better description of the turbulent damping in the near-wall region. Other differences are present

201 in the fluid bulk especially for the radial velocity fluctuations. It is worth noting that the secondary motions are

202 responsible for flow swirling, which could be better described by a third-order formulation of the Reynolds stress

PT
203 tensor. This this-order formulation has not been implemented yet so far, hence it is not surprising to see local

204 under-estimation of the velocity fluctuations, especially in a complex flow like the present one. Moreover, the

RI
205 calculation of experimental and numerical heat transfer shows that the computed heat transfer coefficient lies

SC
206 within the experimental range, suggesting that the local differences between the experimental and numerical data

207 do not impact the model’s capability of correctly represent the macroscopic flow characteristics. As a last

208 comment, see that, even if only one heat transfer coefficient (corresponding to a Re = 12 000) has been

209
U
validated, results of Figure 5 let us state that validation is valid even for higher Reynolds numbers, at least not
AN
210 too far from the experimental Reynolds number range previously discussed.

211 5. Innovative channel performance Evaluation

M

212 Once validated the ASST-SGDH model, the actual performances have to be determined. In particular it is of

213 primary interest to obtain information about pressure drop and heat transfer capabilities of the proposed
D

214 geometry in order to demonstrate that the hypothesis made about heat transfer enhancement provided by the
TE

215 innovative channel are true and eventually compare it with other available technologies. Hence the aim of this

216 section is to analyze the innovative channel to calculate friction factor and Nusselt number correlations.
EP

217 Hence, firstly only a single channel like the one in Figure 1 is studied. Two geometries of interest have been

218 considered, whose geometrical characteristics are listed in Table 6:

C

Angle 45° 20°

AC

Straight distance 7 mm 14 mm
between bends
Bend radius of 2.5 mm 6 mm
curvature
Total channel height 2 mm 3 mm
219 Table 6– Reference innovative channel geometries for performance evaluation

220 As previously shown, meshing convergence has been evaluated by studying the wall shear stress values for

221 different y+ values. Table 7 provides details about the mesh convergence evaluation for the 45° channel case:
 ACCEPTED MANUSCRIPT
Configuration Y+ τw

A 1.1 28.1 Pa

B 0.7 27.7 Pa

C 0.5 27.5 Pa

222 Table 7 - Mesh convergence evaluation

223 From values in Table 7, see that configuration B shows a converged wall shear stress solution. Based on these

PT
224 trends, we retained configuration B as the reference meshing (difference with C configuration of 0.7%). See an

225 example of B configuration in Figure 8. Note that all the following computations are done after a meshing

RI
226 convergence evaluation like the one shown in Table 7, to ensure trustful results for any conclusions shown

hereafter. Moreover it is reassuring to find the same y+≈0.7 value both in Table 4 and in Table 7 to be used for

SC
227

228 different geometries, since it confirms the correctness of the meshing size evaluation for the fluid flow to be

229 studied.

U
AN
M
D
TE

230
231 Figure 8 - Innovative channel study reference mesh: bend (on top) and mixing zone (on bottom)
EP

232 A velocity inlet and a gauge pressure equal to 0 Pa pressure outlet (with regard to 180 bar operating pressure)

233 boundary conditions are used. A constant heat flux of 32.4 kW/m² is applied on channel walls. This value is a
C

234 best-estimate value for the ASTRID sodium-gas heat exchanger heat flux on the gas side. Flow inlet temperature
AC

235 is set at 350 °C. The working fluid is nitrogen. Nitrogen properties are taken from National Institute of Standards

236 and Technology (NIST) Thermodynamic and Transport Properties of Refrigerants and Refrigerants Mixture

237 database version 7.0 (REFPROP 7.0) available in the solver ANSYS FLUENT. Note that wall roughness in the

238 present analysis is neglected, even though it could play a role in the heat transfer (in this sense, its effects can

239 eventually be evaluated once shown the interest of the proposed geometry). The solver is Pressure-based one and

240 the Coupled pressure-velocity algorithm with pseudo-transient option is used. Gradients are evaluated through

241 the Green-Gauss node-centered method. Second Order Upwind Scheme is used for the spatial discretization of
 ACCEPTED MANUSCRIPT
242 momentum, turbulent kinetic energy, turbulent dissipation rate and energy transport equations. Friction factor is

243 calculated using the Darcy formula, i.e.

2∆o Tr
m #15$
p² s

244 Where ∆o is the channel pressure drop, ρ is the fluid mean density, V is the flow average velocity, Dh is the

245 channel hydraulic diameter and L is the linear distance between inlet and outlet.

PT
246 Nusselt number is evaluated by:

]Tr
t #16$

RI
u
247 Where h is the heat transfer coefficient, Dh the hydraulic diameter and u the fluid thermal conductivity. The heat

SC
248 transfer coefficient is calculated as follows:

^
] #17$
#_`a _bc $

U
AN
249 ^being the imposed constant heat flux, _`a v`a _w and _bc vxy cz{ _wp , dS being the unit heat
250 transfer surface and dV the unit fluid volume.
M

251 What is still to be determined is the number of bends necessary to obtain a fully converged solutions in terms of

252 friction factor and heat transfer coefficient. In fact it has been demonstrated [3] that the innovative channel flow
D

253 is fully developed after four bends. Nonetheless this is not sufficient to avoid entry effects when dealing with
TE

254 friction factor and heat transfer coefficient evaluation. Therefore preliminary computations have been run to

255 identify this number of bends for the present study. Results are shown in Figure 9.
C EP
AC

256

257 Figure 9 - Friction factor (left) and heat transfer coefficient (right) as a function of the number of bends
258 for the innovative geometry

259 See that there are no major differences in terms of heat transfer parameters as a function of the number of bends,

260 whereas the friction factor seems to slightly increase (around 2%/bend). This effect is essentially due to single

261 bend contribution to the total pressure drop. In fact, if the fluid flow is fully developed after 4 bends [2], the
 ACCEPTED MANUSCRIPT
262 single bend pressure drop from the 4th bend on will be the same. However, when the number of bends presenting

263 a periodic flow is comparable to the number of bends necessary to flow development, entry effects provide a bias

264 in the final pressure drop value, which is no longer linearly proportional to the number of bends. Therefore it

265 would be necessary a number of bends much higher than 4 to be able to neglect these entry effects, i.e. to have a

266 negligible increase of the friction factor value with regard to the number of bends. By the way, this would result

267 in a too much big computational domain for the purpose of the present work. Therefore the final choice is to use

PT
268 eight bends for the analysis, trying to obtain a converged friction factor and heat transfer variables without

269 resulting in excessive time-consuming calculations.

RI
270 It is worth defining properly the hydraulic diameter, since it determines the non-dimensional numbers and the

SC
271 parameters of interest for the following analysis. In fact, if there is no issue in defining the hydraulic diameter for

272 undulated channels, in the present case there are two different regions where the hydraulic diameter could be

273 defined differently:

274 1.
U
The bend region, where the hydraulic diameter should be defined based on the rectangular cross-section;
AN
275 2. The mixing zones, where the hydraulic diameter might be defined based on the squared cross-section, even

276 if the definition of cross-section and wetted perimeter would not be clear due to occurring mixing.
M

277 The innovative channel hydraulic diameter definition should possibly include both the previous definitions.

278 Remember the classical hydraulic diameter definition for a channel flow:
D

4H
Tr #18$
o
TE

279 With A the channel cross section and p the wetted perimeter. This definition can be generalized as follows:

4H s| 4p
EP

Tr ∙ #19$
o s|
C

280 Where Lc is the actual channel length (i.e. the developed curved length for an undulated channel), V is the

281 channel volume and S is the channel wetted (i.e. heat transfer) surface. For the innovative channel flow, it is
AC

282 more useful to use this formulation rather than the classical one, since in this way the hydraulic diameter can be

283 considered as the volume average of all the possible hydraulic diameter values depending on the channel

284 position (i.e. bend or mixing zone). Note that for straight or wavy channels the Dh final value do not change with

285 the general formulation with regard to the traditional one.

286 By the way the channel volume and wetted surface are generally not available information. Therefore it is useful

287 to look for an approximation capable to provide a quantitative evaluation of the hydraulic diameter. Based on

288 different innovative channel geometries (identified by the specific inclination angle and bend longer side), the
 ACCEPTED MANUSCRIPT
289 hydraulic diameter has been calculated with the generalized formula thanks to Computer Aided Design (CAD)

290 tools. It results always [2]

Tr } T #20$.
/
291

292 Hence, since D is generally known due to design consideration and plate width needs, we will use the

D• } D as the reference value for the innovative channel hydraulic diameter. It has to be
/
293 approximation

PT
294 immediately noted that, for the same fluid properties, plate width and flow velocity, the innovative channel

295 Reynolds number differs from that of an undulated or straight channel. Anyway it is of primary importance to

RI
296 use this formulation because it takes into account the actual wetted (i.e. heat transfer) area enlargement factor

SC
297 that potentially results in a better compactness of the component. In fact, for the same total component volume

298 V, the innovative channel presents a higher heat transfer surface than other geometries, resulting in the lower

299 hydraulic diameter definition. Note also that the found expression for the innovative channel hydraulic diameter

300
U
evaluation is strictly valid for single channel as that of Figure 1. This is not the actual case for the complete
AN
301 geometry (see Figure 2), where several channels meet in different zones. However, it is still useful to

302 approximate the hydraulic diameter through the found expression to better represent the higher surface per unit
M

303 volume corresponding to the innovative geometry. See that generally the difference between the actual hydraulic

304 diameter inferred from CAD means and the approximation proposed is around 6-7%. It is expected by the author
D

305 that this difference can be as high as 10% when dealing with the real complete geometry having several parallel
TE

306 channels.

307 As already explained, correlations will be extracted from numerical computation results. All of the shown
EP

308 correlations are to be supposed to be valid for the shown Reynolds number range (typically 20 000 to 60 000)

309 and for Pr=0.7. See that the used Reynolds number range is practically the same as the one used for model
C

310 validation, except for the upper limit, which does not vary too much from the experimental highest Reynolds
AC

311 number (i.e. Re = 50 000 for LDV campaign), hence keeping the benefits of the validated model. We will look

312 for two correlations as follows:

m H€• ‚ƒ #21$

t
J./K †€•* #22$
_bc
„… J./
_`a
 ACCEPTED MANUSCRIPT
313 If there is nothing to be noted for the diabatic friction factor correlation, see that the formulation for the heat

314 transfer correlation allows to:

315 • Be independent of the fluid (through the ratio to the Prandtl number);

316 • Take into account significant variations of the fluid properties with the bulk to wall temperature ratio [30].

317 The influence of thermal gradients for current computations is found to be practically negligible, the correction

318 term for any shown Reynolds number being never below the value of 0.990 and 0.985 for the 45° and 20°

PT
319 computed flow geometries respectively. This is also very important with regard to thermal calculation

320 correctness, as previously shown. Final correlations for the two test-cases are shown in Figure 10 and Figure 11.

RI
321 Note that the standard errors are of 0.002 for friction and 2.5 for heat transfer correlations in the 45° geometry

SC
322 and of 0.001 for friction and 3.5 for heat transfer correlations in the 20° geometry.

U
AN
M

323
D

324 Figure 10- Friction factor correlations for innovative channel

TE
C EP
AC

325

326 Figure 11 – Nusselt number correlation for innovative channel

327 Table 8 shows the differences in terms of friction factor and heat transfer coefficient for the two studied

328 geometries:

329
 ACCEPTED MANUSCRIPT
Re f 45° [Pa] f 20° [Pa] (f20°-f45°)/f45° % h 45° [W/(m2K)] h 20° [W/(m2K)] (h20°-h45°)/h45° %

20000 0.171 0.051 -237.9 2375 1531 -35,6

30000 0.183 0.045 -304.3 3339 1971 -41,0

40000 0.166 0.043 -287.9 4071 2615 -35,8

50000 0.162 0.040 -302.8 4861 3076 -36,7

60000 0.153 0.038 -299.6 5637 3578 -36,5

PT
330 Table 8 – Friction factor and heat transfer coefficient analysis for innovative channel with α = 45° and α
331 =20°

RI
332

333

SC
334 See that there is a dramatic decrease of friction factor for the 20° geometry with regard to the 45° geometry:

335 there is a factor of about 5 between the two, whereas the decrease in the heat transfer coefficient for the 20°

336

U
geometry is much slighter than that of the 45° geometry. Moreover, comparing the obtained correlations with
AN
337 classical straight tube Blasius’(for friction) and Dittus-Boelter-McAdams’ (for Nusselt number) correlations, it is

338 worth noting that the difference between the straight tube and the 20° innovative channel is essentially in the
M

339 multiplying coefficient both for heat transfer and friction factor: this means that the 20° innovative channel has a

340 friction factor about two times higher than that of a straight channel, whereas the multiplying coefficient of the
D

341 heat transfer correlation is only 20% higher. Comparison is more complicated for 45° innovative geometry.
TE

342 Anyway results of Table 8 show that the 45° geometry presents very important pressure drop and a slightly

343 higher heat transfer coefficient. In this sense, it is of primary importance to determine which geometry provides
EP

344 the most compact design for given design constraints, or whether the high pressure drop given by the innovative

345 geometry is acceptable or not. This analysis will be done in the next section.
C

346 These correlations can be used by designers to design compact heat exchangers. Nevertheless, as already

347 explained, these correlations have been obtained for a single innovative channel, whereas there are many parallel
AC

348 channels in the real case, where additional mixing zone are created when two innovative channels meet (as

349 already shown in Figure 2). To better account for the real case, computations have been done for a three-channel

350 geometry. The single channel geometry is the same as the previously described 45° geometry in Table 6; the

351 three channels are separated by a 2 mm distance between the straight parts between two following bends of the

352 same channel.

353 Pressure drops have been evaluated in the center channel, trying to account for enhanced mixing and decreased

354 wetted surface of the real channel case. Same for the heat transfer evaluation, for which average bulk and wall
 ACCEPTED MANUSCRIPT
355 temperature have been calculated for the center channel. Note that the two side-channels do not represent the real

356 case, since additional mixing zones are created only when they meet the center channel. To properly account for

357 additional mixing zone a large number of parallel channels should be used. Here, to keep the strategies used so

358 far in the work as well as to limit the model size, three channels have been used to provide some information

359 about actual performance for the real case heat exchanger. Differences between pressure drop (the channel

360 geometry being the same of the reference 45° single channel) and heat transfer coefficient from single and triple

PT
361 channel geometry simulations are shown in Table 9. Design correlations are shown in Figure 12 and in Figure

362 13. Again, the standard errors for the triple 45° geometry correlations are 0.003 for the friction and 2.8 for the

RI
363 heat transfer.

SC
(htriple-
h single h triple
Re ∆P single [Pa] ∆P triple [Pa] (∆Ptriple-∆Psingle)/∆Psingle % hsingle)/hsingle
[W/(m2K)] [W/(m2K)]
%

20000 7118 9679

U26.5 2375 2525 6.3
AN
30000 16180 22500 28.1 3339 3553 6.4

40000 26030 36645 29.0 4071 4334 6.5

M

50000 39790 57194 30.4 4861 5199 6.9

60000 54160 81066 33.2 5637 6001 6.5

D

364 Table 9 - Pressure drop and heat transfer coefficient analysis single and triple innovative channel with α =
TE

365 45°
C EP
AC

366

367 Figure 12 - Friction factor correlation for triple innovative channel with α=45° and D=2 mm
 ACCEPTED MANUSCRIPT

PT
368

RI
369 Figure 13 – Nusselt number correlation for triple innovative channel with α=45° and D=2 mm

370 The triple channel pressure drop is sensibly higher than the single channel pressure drop (around 30%). This is

SC
371 essentially due to additional mixing zones, where fluid streams meet and provide additional hydraulic resistance.

372 For the same reason, heat transfer coefficient is higher, but less than the relative pressure drop increase (around

U
373 6 to 7%). Again the fundamental arising question is whether the innovative geometry provides superior
AN
374 compactness for given performance.

375 6. Compactness Comparison Strategy

M

376 The previous section has shown how it is of primary importance to determine which geometry is the most
D

377 compact one for given performance, in order to demonstrate its industrial interest. Remember that Compactness

378 can be defined as thermal power per unit volume. It has been argued that traditional compactness comparison
TE

379 criteria cannot be used to quantitatively compare different heat transfer geometries [2]. It can be preliminarily

380 observed that Hesselgreaves states that the general compactness factor is determined by the ratio of heat transfer
EP

381 area to the unit volume S/V [32]. In this sense, it is worth noting that the innovative geometry is potentially more

382 compact than other existing geometries (i.e. PCHE wavy channel) by regarding definition of generalized
C

383 hydraulic diameter (Eq.20). Referring to Figure 14 (in particular to the gas side), define D as the reference
AC

384 height of a PCHE channel of squared cross section. To clarify the hypothesis already made in the introduction of

385 total squared cross section even for the innovative channel, it is sufficient to state that a different cross-section

386 would be more different to compare to a PCHE and potentially less compact in terms of gas plate total thickness.

387 Therefore is it straightforward verifying that for the PCHE channel, either straight of wavy, it results that Dh=D

388 (the exercise is obvious for the straight channel case, whereas the wavy channel case reduces to a straight

389 channel through the developed length). This is not the case for the innovative channel. Hence from Eq.20 it

390 results that:

 ACCEPTED MANUSCRIPT
# /p$ ‡‡yxa x{ |ra‡‡{ Œ
T 4
# /p$ˆ‰Š‹ } • 1 #23$
Tr 3

391 Therefore the innovative geometry is always more compact than a PCHE geometry. However, it is convenient to

392 verify this result by fixing the desired performance (in terms of exchanger thermal power and pressure drop and

393 mass flow rate), to quantitatively verify which geometry has the lower volume. This approach is formally

394 correct, since it takes into account designer’s necessity to determine the size of a component for a given load.

PT
395 Determining the size of a plate heat exchanger is an iterative process: in fact, if the designer wants to use a more

396 resistive technology (with pumping power fixed to a given value), for a given total mass flow rate and total

RI
397 length, the channel flow rate (i.e. the channel Reynolds number) is decreased through increasing the number of

SC
398 plates and hence the number of channels. However this increases also the heat transfer surface (resulting in a

399 potentially higher exchanged thermal power). Hence the total channel length (giving the pressure drop, see

U
400 Darcy formula) has to decrease to meet the desired thermal load. The final design point has the right number of

401 plates and total length giving the desired performance.

AN
402 Finally, trying to represent the design process, we will impose the following hypothesis for all the geometries to

403 be compared:
M

404 1. Same thermal power Pth exchanged;

405 2. Same pressure drop ∆p;

D

406 3. Same total mass flow rate Γ;

TE

407 4. Same total component width (i.e. dimension “b” in Figure 14);

408 5. Fixed distance between two channels “d”;

EP

409 6. Fixed thickness of the sodium plate δsodium.

C
AC

410

411 Figure 14 - Compact heat exchanger major dimensions

 ACCEPTED MANUSCRIPT
412 Regarding the “a” dimensions, it can be demonstrated that: ∝ t , where N is the number of gas side plate and δ

413 the gas side plate total thickness equal to D+ε. In fact the sodium side is neglected from the comparison strategy

414 definition since it has a minor role on the global thermal resistance. Hence take into account only the volume of

415 the gas side of the heat exchanger. However this assumption does not change the validity of the adopted

416 approximation about the “a” dimension: since sodium and gas plates are superposed each other, it results that

417 N=Ngas= Nsodium. Hence it would be: t t•y• cz •y• cz t# •y• cz $ ∝ t . Finally, it results that,

PT
418 the volume V of the component, which is equal to abL, is only a function of the total length L (given by the

419 desired pressure drop value) and of the number of plates of the heat exchanger gas side N, since the gas side

RI
420 plate total thickness δ can be taken as constant to a first approximation. Together with these two parameters,

SC
421 the total channel height D (remember that D=Dh for a straight or wavy channel but D∝Dh for the innovative

422 channel) is the third parameter to take into account since it gives the total channel thickness (i.e. δ= D+ε), the

423 number of channel per plate (i.e. Nc=b/(D+d, see hypothesis 5), the channel cross section A0 (i.e. A0=D² for a

424
U
squared cross section) as well as the Reynolds number values to be considered into friction factor and heat
AN
425 transfer correlations. It will be demonstrated that, knowing N, L and D it is possible to correctly compare

426 different geometries.

M

427 In fact the iterative design process of a compact heat exchanger essentially consists of varying the total length

428 and number of plates of the component, for a fixed total channel height D, to meet the design goals in terms of
D

429 exchanged thermal power Pth and pressure drop ∆p.

TE

430 As it will be shown, writing the pressure drop and thermal power equations, we can express these two equations

431 in terms of three unknowns, specifically N, L and D. One unknown is of particular interest: in fact D can be
EP

432 fixed by the designer in order to have a certain plate thickness. Therefore we can fix a given D value and finally

433 have two equations (pressure drop and exchanged thermal power) and two unknown to be determined (N and L).
C

434 The problem has always a physical solution, since the exchanged thermal power and pressure drop always
AC

435 assume non-zero values. Therefore the comparison strategy will consist of writing L as a function of the pressure

436 drop and number of plates i.e. s s#∆o, t$ and as a function of the exchanged thermal power and number of

437 plates i.e. s s#„ r , t$. Since the two design constraints must be respected at the same time, there must be a

438 design point corresponding to the intersection point (L,N) such as s#∆o, t$ s#„ r , t$, Pth ,∆p and D being

439 fixed.

440 In particular, it is expected that s#∆o, t$ is a function increasing with N: this is due to the fact that, for a fixed

441 pressure drop, increasing the number of plates results in decreasing the mass flow rate per channel i.e. the
 ACCEPTED MANUSCRIPT
442 channel Reynolds Number and hence the channel pressure drop. Since we want the pressure drop to be constant,

443 the only way to recover the desired pressure drop value would be to increase L. On the other hand, it is expected

444 that s#„ r , t$ is a function decreasing with N increasing: in fact, for a fixed thermal power to be exchanger,

445 increasing the number of plates would results in increasing the heat transfer area: hence, even if the channel

446 Reynolds number decreases, the thermal power would likely exceed the fixed desired value. Hence the

447 component length should be shortened to meet the design goal in terms of thermal power. Therefore write:

PT
„r U ‘a• s’_T #24$

448 Where U is the total heat transfer coefficient, Sgas is the heat transfer surface of the gas side of the heat exchanger

RI
449 (which can be calculated based on the generalized definition of hydraulic diameter) and LMTDor Logarithmic

SC
450 Mean Temperature Difference is a widely used parameter when dealing with heat exchanger. For the a heat
∆–— ‚∆–˜
451 exchanger with two ends A and B it is defined as LMTD ∆›— . Now write all the terms as a function of the
™š
∆›˜

U
452 unknown, hence, neglecting the high temperature gradients correction factor:

p‘a•
AN
„r U 4 ∙ s’_T
Tr

4tt| HJ s|
U s’_T
M

Tr

s|
4tt| HJ s
U s s’_T →
D

Tr

s
TE

„ r Tr
s|
s #25$
4tt| HJ ∙ U ∙ s’_T

t ∙ t| ∙ HJ ∙ s| is the gas total volume, Nc is the number of channels per plate, A0 is the single
EP

453 Where Vgas

.•
454 channel cross section i.e. equal to D² for a squared cross section and
.
is the ratio between the actual channel
C

455 length and the straight distance between heat exchanger inlet and outlet. It takes into account the heat transfer
AC

456 surface gain provided by wavy channels1.

457 Now the total heat transfer coefficient is equal to:

1
U #26$
1 ž ‘a• 1 ‘a•
]‘a• u• {{ • {{ ]•y• cz •y• cz

1
In fact, the ratio is equal to one for a straight channel, whereas for undulated channel it is larger than one, meaning that in a fixed linear distance between inlet

s
and outlet, an undulated geometry provides a higher heat transfer surface. With the hypothesis of bend curvature radius far smaller than the straight distance
between two bends, it can be easily demonstrated that |k 1k
s Ÿ ¡V.
 ACCEPTED MANUSCRIPT
458 See that ž is the plate’s residual thickness, • {{ 2t¢s is the total steel heat transfer surface, ]•y• cz is the

459 sodium side heat transfer coefficient and •y• cz tt|,•y• cz o•y• cz s is the total sodium side heat transfer

460 area. Note that o•y• cz is the perimeter of the single straight channel of sodium, t|,•y• cz is the number of

461 channel per plate of sodium and the hypothesis of Ngas=Nsodium=N is still used. Hence write:

s
„ r Tr
s|
s
4tt| HJ ∙ U ∙ s’_T

PT
s
„ r Tr
s|
#26$
¥ ¨

RI
1
4tt| HJ ∙ ¤ § ∙ s’_T
¤ 1 ž ‘a• 1 ‘a• §
£]‘a• u• {{ • {{ ]•y• cz •y• cz ¦

SC
462 Note that s ∝ 1kT , therefore it results that s#„ r , t$ decreases when N increases, as already motivated

463 previously.

464 The pressure drops can be also expressed as:

U
AN
. . «{²>² . «{ -¯° >²
∆o m p m H =
¬²©ª- ©ª ®
465 ©ª ©ª ¬²

s 1 €• ‚ƒ
s 1 ΓTr ‚ƒ
H H
M

Tr 2 Tr 2 tt| HJ

1 ² Γ ‚ƒ
1
Hs Tr‚# ²ƒ$
#27$
D

2 t| HJ t ‚ƒ
TE

466 Again, see that the combination of all the different term is specific of the considered geometry, due to the

467 dependency on the superscript α and on the fixed D dimension. Hence, these quantities are of primary interest for
EP

468 a quantitative comparison of different geometries.

469 Finally, write the total component length as:

2 ∆o
s#∆o, t$ t ‚ƒ
Tr ²ƒ ‚ƒ #28$
C

Γ
H ²³ ´
t| HJ
AC

470 Since 0<α<2, see that s#∆o, t$increases if N increases, as previously explained.

471 These considerations demonstrate the validity of the present approach. As a final comment before showing the

472 results of the proposed approach see that this is based upon actual performance of the heat exchanger. As it has

473 been demonstrated,s#„ r , t$ and s#∆o, t$expressions depend on actual geometries and performance and hence

474 any comparison which is not based on the actual geometries and performance results in a loss of information that

475 likely provide incorrect results. By the way, using this approach allows for comparing different geometries with

476 different D dimensions since all quantities (i.e. Dh, Nc, A0) can be known once fixed a value for D). Therefore,
 ACCEPTED MANUSCRIPT
477 the generality of this approach is something important to be pointed out. The final design point, not having an

478 analytical solution, needs to be found either graphically or numerically via bisection techniques.

479 7. Different channel geometries comparison ASTRID Sodium/gas heat

480 exchanger

481 With the correlations obtained in the previous section it is possible to compare the compactness of different

PT
482 technologies, in particular:

483 1. 45° innovative channel shown in Table 6;

RI
484 2. 20° innovative channel shown in Table 6;

SC
485 3. Three parallel innovative channels as the one in Table 6;

486 4. Straight channel with squared cross-section with D=2.5 mm;

U
487 5. 45° wavy channel with squared cross-section with D= 2 mm.

488 The geometries are chosen to show two different innovative channel geometries, the real-case three parallel
AN
489 innovative channel geometry, the simplest possible straight tube and a wavy channel, which is as a third

490 geometry of interest. Note that correlations for wavy channel are taken from reference [31] and are the following
M

491 ones:

m 3.117€• ‚J. K
#29$
D

t 0.0513€• J.K/K „… J./ #30$

TE

492 .Reference heat exchanger performances are taken from ASTRID sodium-gas heat exchanger illustrated in [2],

493 for which design constraints are listed in Table 10:

EP

Exchanged Thermal
24 MW
Power Pth
C

Gas-side pressure
1 bar
drop ∆p
AC

Sodium Inlet
530 °C
Temperature
Gas Inlet
310 °C
Temperature
Sodium Mass flow
99 kg/s
rate
Gas Mass flow rate 101 kg/s
494 Table 10 - ASTRID Design Constraints

495 Results are shown in Table 11.

 ACCEPTED MANUSCRIPT
Single Triple Single
Straight WAVY
innovative innovative Innovative
channel channel 45°
channel 45° channel 45° channel 20°

Compactness
20 28 27 25 23
[MW/m3]

% with regard to

PT
/ +40% +35% +25% +15%
Straight channel

RI
496 Table 11 - Compactness comparison of different geometries for ASTRID Sodium-Has Heat Exchanger
497 conditions

498 The innovative channel geometry presents always the higher compactness. The obtained gain is impressive, up to

SC
499 25% for the 20° case, and 40% for the 45° case. Indeed it is worth noting that these 25% and 40% gains are not

500 supposed to be the real ones, since the comparison between the single and the triple innovative channel with 45°

U
501 angle clearly shows that the real case being more resistive, some percent of compactness gain could be lost.
AN
502 Hence, even if a value of 35% for the 45° innovative channel geometry and 20% (author’s best estimate) for the

503 20° geometry are foreseen as the realer ones, anyway it represents an important result. Note that the drawback of
M

504 additional pressure drop created by the additional mixing zones for the real channel case (see Figure 2) could be

505 eliminated if the distance between innovative channels is large enough to allow for having completely separated
D

506 channel. However, this presumably high distance would deteriorate the global component compactness. In any
TE

507 way, the innovative channel compactness improvement is even more impressive when looking at the wavy,

508 which is an existing technology more and more used in the industry when dealing with heat exchangers. The 45°
EP

509 wavy channel seems to provide a smaller gain, compared to the innovative channel geometry. See that it could

510 be considered as illogical that the innovative geometry, which is the most resistive among the studied

511
C

geometries, is always the most compact one. Indeed this is not surprising at all: in fact, a more resistive geometry

512 leads the designer to reduce the heat exchanger length (as shown) to meet the pressure drop value goal. This is
AC

513 obviously favorable if the heat exchanger can deal, in that lower volume, the same thermal load. The major

514 parameter of interest for a compact heat exchanger is therefore the heat transfer surface to volume ratio, which is

515 higher for the innovative channel, as shown by the hydraulic diameter evaluation. Hence it is clear that the

516 interest given by the innovative geometry is due to the fact that the global length has to be reduced to meet a

517 pressure drop value, reducing the total volume of the component keeping the right heat transfer surface. Even if

518 additional plates are to be added, the final design point is always more compact than any other known industrial

519 technology. This is obviously truer if the number of plates to be added is not excessive: in fact in that case, too
 ACCEPTED MANUSCRIPT
520 short plates would be designed, which could be a potential drawback for other design constraints such as

521 thermal-mechanical resistance. As a final comment see that, based on purely thermal-hydraulic considerations,

522 the most compact geometry is that of 45° D=2 mm. However, even a fairly less compact geometry i.e. the 20°

523 innovative channel results to be more interesting than the other technologies previously compared.

524 8. Conclusions and Perspectives

PT
525 In order to increase the heat transfer coefficient (and the global compactness) of a compact heat exchnager the

526 basic idea of this work is to design a channel were the fluid flow is as much three-dimensional as possible.

RI
527 In particular the channel can be thought as the result of the superposition of two PCHE-type single wavy

SC
528 channels in phase opposition. The innovative channel geometry has to be studied numerically and experimentally

529 to demonstrate its industrial interest and the actual compactness gain. Thanks to a validated numerical model,

530 correlations for friction factor and Nusselt number for various geometries have been obtained. With these

531
U
correlations, a compactness comparison strategy has been developed: it takes into account design constraints, in
AN
532 particular the fact that different compared technologies have the same exchanger thermal power and pressure

533 drop. With this strategy, it has been shown that the innovative channel is the most compact one among the most
M

534 important existing industrial wavy heat exchanger technologies with other channel geometries.
D

535 It is worth showing that the proposed innovative heat exchanger is of immediate industrial fabrication, since all

536 needed machining and welding processes are already available. In particular, the more convenient way to
TE

537 produce plates is to mechanically or chemically machine the channels on a metallic plate. Welding can be done

538 using diffusion bonding processes, which provide a final one-piece plate stack. Hence, it is to be stated that the
EP

539 present innovative technology is already available for the state-of-the-art fabrication capabilities.
C

540 However, to perform a geometry optimization, an extensive parametric study on the innovative channel

541 geometrical parameter should be done. Moreover, the evolution of different thermal fluid performance would
AC

542 easily widen the potential applications of the innovative geometry in terms of design availability.

543 It will also be particularly interesting to compare different innovative channel geometries. In fact, the innovative

544 channel can be thought as the superposition of very different single channel cross sections, i.e. triangular, semi-

545 circular, trapezoidal, squared, etc. Thermal-hydraulic behavior of such innovative channels is worth being

546 studied to determine the best candidate. Even if some cross-sections might be considered as “exotic” and

547 industrially unfeasible, the hemi-circular cross-section appears to be very interesting due to its possible

548 fabrication via chemical machining.

 ACCEPTED MANUSCRIPT
549 Last but not least, questions may arise regarding safety and reliability aspects of ASTRID gas power conversion

550 cycle and especially regarding the sodium-gas heat exchanger. Indeed compact heat exchanger technologies

551 propose new and relevant challenges in the qualification domain. In particular, the diffusion bonding process

552 might affect the critical size of metallic grains in the plate structure. This could potentially decrease the local

553 material strength, increasing the likelihood of a failure. This is even more important when dealing with

554 component having several thousand channels. The size and number of channels make also particularly difficult

PT
555 to inspect the component and to detect eventual defects. Hence, metallurgy and inspection techniques are

556 expected to provide solid solutions to these open issues: once these solutions accepted in a proper regulatory

RI
557 framework, Brayton cycle SFRs will finally offer a clean and safe source of energy.

SC
558

559 9. Acknowledgements

U
560 This work has been done under the CEA CFR founding for the PhD of the principal author.
AN
561 10. List of Acronyms
M

562 CEA: French Atomic and Renewable Energy Commission

563 ASTRID: Advanced Sodium Technological Reactor for Industrial Demonstration

D
TE

564 GFR: Gas-cooled Fast Reactor

565 S-CO2: supercritical carbon dioxide

EP

566 VHTR: Very High Temperature Reactors

C

567 SFR: Sodium-cooled Fast Reactors

AC

568 PCHE: Printed Circuit Heat Exchanger

569 DHRS: Decay Heat Removal System

570 3D: 3 Dimensional

571 ASST: Anisotropic Shear Stress Transport model

572 ρ: Fluid Density

 ACCEPTED MANUSCRIPT
573 Ui: i-component of the Velocity field

574 u'5 : i-component velocity fluctuation

575 k: Turbulence Kinetic Energy

576 ω: Turbulence Specific Dissipation Rate

PT
577 µ t: Eddy Viscosity

RI
578 SGDH: Simple Gradient Diffusion Hypothesis

579 LDV: Laser Doppler Velocimetry

SC
580 2-C LDV: 2-component LDV

U
581 PIV: Particle Image Velocimetry
AN
582 VHEGAS: Validation of Heat Exchange in GAS facility
M

583 TCs: Thermo-Couples

584 Re: Reynolds number

D

585 Pr: Prandtl number

TE

586 Nu: Nusselt number

EP

587 11. References

C

588 [1] E.ABONNEAU et al., “The ASTRID Project: Status and Prospects towards the Conceptual Design Phase”,

589 Proceedings of the International Conference on the Advances of nuclear Power Plants 2014 – ICAPP’14,
AC

590 Charlotte, USA, April 6-9, 2014, Paper 14116, pp. 447 - 455

591 [2]F.VITILLO, “Experimental and numerical contribution to heat transfer enhancement in compact plate heat

592 exchanger”, ED MEGeP, Toulouse, November 2014

593 [3] F.VITILLO, L.Cachon and P.Millan, “Thermal-hydraulic Analysis of an Innovative Compact Heat

594 Exchanger Channel Flow”, International Journal of Heat and Fluid Flow, Under review, sent on April 2015
 ACCEPTED MANUSCRIPT
595 [4] V.Dostal et al., “Supercritical CO2 cycle for Fast Gas-Cooled Reactors”, ASME Turbo Expo 2004: Power for

596 Land, Sea and Air. Vol. 7: Turbuo Expo 2004 Vienna, Austria, June 17-17, 2004, pp.683-692

597 [5] K.Nikitin, Y.Kato, L.Ngo, “Printed circuit heat exchanger thermal-hydraulic performance in supercritical

598 CO2 experimental loop”, International Journal of Refrigeration, Vol.29, N.5, pp.807-814, August 2006

599 [6] N.Tsuzuki, Y.Kato and T.Ishiduka, “High Performance printed circuit heat exchanger”, Applied Themal

PT
600 Energy, Vol.27, N.10, pp.1702-1707, July 2007

RI
601 [7] D.E. Kim et al., “Numerical Investigation on thermal-hydraulic performance of new printed circuit heat

602 exchanger model”, Nuclear Engineering and Design, Vol.238, N.12, pp.3269-3276, December 2008

SC
603 [8] M.A.Pope et al., «Thermal –hydraulic challenges of Gas Cooled Fast Reactors with passive safety features »,

604 Nuclear Engineering and Design, Vol.239, N.5, pp.840-854, May 2009

605
U
[9] K.Gezelius, Design of compact intermediate heat exchangers for gas cooled fast reactors”, MSc Thesis, Dept.
AN
606 of Nuclear Science and Engineering, Massachusetts Institute of Technology, 2004
M

607 [10] I.H.Kim et al., “Thermal hydraulic performance analysis of the printed circuit heat exchanger using a

608 helium test facility and CFD simulations”, Nuclear Engineering and Design, Vol.239, N.11, pp.2399-2408,
D

609 November 2009

TE

610 [11] D.McCormack, “The application of Printed circuit heat exchanger technology in the pebble bed modular

611 reactor demonstration plant”, ASME Turbo Expo 2011: Power for Land, Sea and Air, Vol. 3: Heat Transfer,
EP

612 Electric Power, Industrial and Cogeneration, paper No 2001-GT-0473. New Orleans, Louisiana, USA, June 4-7,

613 2001.
C

614 [12] D.Aquaro and M.Pieve, “High Temperature heat exchangers for power plants: performance of advanced
AC

615 metallic recuperators”, Applied Thermal Engineering, Vol. 27, N.2-3, pp.389-400, February 2007

616 [13] E.S.Kim, C.H.Oh and S.Sherman, “Simplified optimum sizing and cost analysis for compact heat exchanger

617 in VHTR”, Nuclear Engineering and Design, Vol.238, N.10, pp. 2635-2647, Octobre 2008

618 [14] F.Pra et al., “Promising designs of compact heat exchangers for modular HTRs using the Brayton cycle”,

619 Nuclear Engineering and Design, Vol.238, N.11, pp.3160-3173, November 2008
 ACCEPTED MANUSCRIPT
620 [15] S.Mylavarapu et al., “Investigation of High-Temperature printed circuit heat exchangers for Very-High

621 Temperature Reactors”, J.Eng.Gas Turbine Power, Vol.131, N.6, July 2009

622 [16] I.H.Kim, H.C.No, “Thermal hydraulic performance analysis of a printed circuit heat exchanger using a

623 helium-water test loop and numerical simulations”, Applied Thermal Engineering, Vol.31, N.17-18, pp.4064-

624 4073, December 2011

PT
625 [17] J.Figley et al., “Numerical study in thermal hydraulic performance of a printed circuit heat exchanger”,

626 Progress in Nuclear Energy, Vol.68, pp.89-96, September 2013

RI
627 [18] Y.Lee, J.I.Lee, “Structural assessment of intermediate printed circuit heat exchanger for sodium-cooled

SC
628 fast reactor with supercritical CO2 cycle”, Annals of Nuclear Energy, Vol.73, pp.84-95, November 2014

629 [19] H.Mochizuki and M.Takano, “Heat transfer in heat exchangers of sodium cooled fast reactor systems”,

U
630 Nuclear Engineering and Design, Vol.239, N.2, pp.295-307, February 2009
AN
631 [20] V.Vinod et al., “Experimental evaluation of sodium to air heat exchanger performance”, Annals of Nuclear

632 Energy, Vol. 58, pp.6-11, August 2013

M

633 [21] H.Zhao and P.F.Peterson, “Multiple reheat helium Brayton cycle for sodium cooled fast reactors”, Nuclear
D

634 Engineering and Design, Vol.238, pp.1535-1546, 2008

TE

635 [22] G.D.Perez-Pichel et al., «Potential application of Rankine and He-Brayton cycles to sodium fast reactors »,

636 Nuclear Engineering and Design, Vol.41, pp.2643-2652, 2011

EP

637 [23] Y.Ahn and J.I.Lee, “Study of various Brayton cycle design for small modular sodium-cooled fast reactors”,

638 Nuclear Engineering and Design, Vol.276, pp.128-141, September 2014

C

639 [24] L.CACHON et al., “Echangeur de chaleur à plaques à échange thermique et compacité améliorés”,
AC

640 PATENT n°HD14482, 2014

641 [25] L.CACHON et al., “Innovative Power Conversion System for the French SFR prototype, ASTRID”,

642 Proceedings of the International Conference on the Advances of nuclear Power Plants 2012 – ICAPP’12,

643 Chicago, USA, June 24-28, 2012, Paper 12300, pp. 538 – 552

644 [26] F.VITILLO et al.; “An Anisotropic Shear Stress Transport (ASST) model formulation”, Nuclear

645 Engineering and Design, Under review, sent in February 2015

 ACCEPTED MANUSCRIPT
646 [27] S.M.Lee and K.Y.Kim, “A Parametric Study of the Thermal-Hydraulic Performance of a Zigzag

647 Printed Circuit Heat Exchanger”, Heat Transfer Engineering, Vol 35, N.13, pp.1192-1200, 2014

648 [28] F.Vitillo et al., « Turbulence Model Comparison for Innovative Compact Plate Heat Exchanger Design

649 Application », Proceedings of the International Congress on the Advances in Nuclear Power Plants 2014 –

650 ICAPP’14, Charlotte (NC) USA, April6-9, 2014 Paper 14143, pp.1930-1939

PT
651 [29] F.R.MENTER, M.Kuntz, R.Langtry, “Ten years of industrial experience with the SST Turbulence model”,

652 Turbulence, Heat and Mass Transfer 4, Begell House, Inc., 2003

RI
653 [30] N.E.TODREAS and M.S.KAZIMI, “Nuclear Systems Vol.1 – Thermal-hydraulic fundamentals”, CRC

SC
654 Press, 2011

655 [31] C.GALATI, Numerical and Experimental Study for ASTRID Sodium-gas Heat Exchanger Design

656
U
Application, McS Thesis, University of Rome « La Sapienza », January 2015
AN
657 [32] J.E.Hesselgreaves, “Compact Heat Exchanger: Selection, Design and Operation”, Pergamon, 2011
M
D
TE
C EP
AC