Energy xxx (2015) 1e8

Contents lists available at ScienceDirect

Energy
journal homepage: www.elsevier.com/locate/energy

Heat transfer and pressure drop correlations for nned plate ceramic
heat exchangers
Helio Henrique Santomo Villanueva, Paulo Eduardo Batista de Mello*

rio da FEI, Department of Mechanical Engineering, Av. Humberto de Alencar Castelo Branco, 3972, Sa
Centro Universita ~o Bernardo do Campo, SP,
CEP 09850-901, Brazil

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

Article history: The use of ceramic materials is a good option to build heat exchangers for high temperature applications
Received 30 October 2014 when cost is a concern. This work presents heat transfer and pressure drop correlations for one nned
Received in revised form plate heat exchanger evaluated using CFD (Computational uid dynamics) simulations. One adequate
14 April 2015
turbulence model was used to include transitional Reynolds number range. The inuence of geometrical
Accepted 15 April 2015
Available online xxx
parameters is included into the correlations, following the same approach commonly used for offset strip
ns heat exchangers. For validation purposes, the CFD results are compared to experiments for one
particular geometrical conguration. The resultant correlations for the high temperature heat exchanger
Keywords:
HTHE
could be used for optimization purposes, considering possible applications. Analysis of simulation results
Ceramic heat exchanger revealed signicant heat transfer enhancement produced by a horseshoe vortex formed in the frontal
CFD part of the ns.
Biomass 2015 Elsevier Ltd. All rights reserved.

1. Introduction choice to substitute metals on HTHE due to its resistance to

HTHE(High temperature heat exchangers ) enable performance
enhancements in industrial processes and heat engines. Consid-
ering the case of heat engines, higher system efciency can be
obtained when the heat source presents higher temperature. The
EFGT(externally red gas turbine ) is one thermodynamic cycle that
could take advantage of the increased performance enabled by the
HTHE, with the extra benet of allowing the burn of biomass in gas
turbines [1].
When it comes to temperatures higher than 800
C, metallic
heat exchangers are not a cost-effective alternative. Experiments
with EFGT cycle [2,3], using metallic HTHE have demonstrated its
feasibility, but with a serious limitation on the work temperature
imposed to the heat exchanger. The TIT (turbine inlet temperature)
was maintained lower than 750
C in both cases. In a more recent
experimental study, TIT was increased to 850
C [4]. The inuence
of the TIT on the efciency of the EFGT cycle is presented in Ref. [5],
conrming that the use of one HTHE capable of supporting higher
temperatures would lead to efciency improvements. Information
in literature [6,7] states that ceramic materials are the technological
* Corresponding author.
E-mail addresses: heliohenriquevillanueva@gmail.com (H.H.S. Villanueva),
pmello@fei.edu.br (P.E.B. de Mello).
corrosion and temperature. Despite this, experimental data and
demonstration of feasibility studies involving ceramic heat ex-
changers are still scarse in the literature. Some recent de-
velopments in this area may be found in Refs. [8,9] and [10].
The use of nned plate heat exchanger conguration for the
construction of ceramic HTHE [1] provides the benet of high heat
transfer area per volume of the heat exchanger. The thermal per-
formance and pressure drop for the HTHE can be predicted using
the standard approach already used for compact heat exchanger
[11]. Experiments gathering data for the friction and Colburn fac-
tors have been conducted under laminar regime [10], for a ceramic
HTHE constructed using alumina (Al2O3).
CFD(Computational uid dynamics ) simulations can be used to
obtain the friction and Colburn correlations numerically, as a
function of Reynolds number, so that idealizations and geometry
changes can be easily studied in a short time. A CFD study that used
this approach, for one HTHE that could be constructed using nned
plates, is presented in Ref. [12].
This paper presents a numerical study of one ceramic nned
plate HTHE, as shown in Fig. 1. The objective is to gather data to
evaluate the inuence of geometrical parameters over the friction
and Colburn factors, considering laminar and transitional regimes.
The form of the correlations is similar to the one rst presented in
Ref. [13] for offset strip ns. The geometry of the ns is parame-
terized and a systematic investigation is conducted.

http://dx.doi.org/10.1016/j.energy.2015.04.017
0360-5442/ 2015 Elsevier Ltd. All rights reserved.

Please cite this article in press as: Villanueva HHS, de Mello PEB, Heat transfer and pressure drop correlations for nned plate ceramic heat
exchangers, Energy (2015), http://dx.doi.org/10.1016/j.energy.2015.04.017
2 H.H.S. Villanueva, P.E.B. de Mello / Energy xxx (2015) 1e8

Fig. 2. : Geometry of the CFD domain with boundary conditions.

boundary conditions inuence the results or cause convergence

issues. Fig. 2 also shows the boundary conditions applied on the
CFD domain. The inlet boundary condition sets the temperature,
which the uid enters the domain and the mass ow based on the
desired Reynolds number. The outlet boundary imposes a zero
gradient normal to the face and a relative pressure. The wall
boundary is a no-slip wall with a xed temperature. The other
boundary faces of the computational domain are symmetry planes
(adiabatic and zero gradients). Only half of the n height is included
in calculation domain due to symmetry.

2.1. Geometrical parameters

The three geometrical parameters created are l, d and g. These

parameters are dened by 1e3, with the n dimensions presented
in Fig. 3. The n height is the variable H.
Fig. 1. High temperature heat exchanger constructed using nned ceramic plates.
D1
l (1)
H
In addition, the validation through comparison with experi-
ments with a ceramic HTHE prototype is presented. Considering D1
d (2)
the technological interest in HTHE and that the ceramic heat L
exchanger imposes geometry signicantly different from standard
heat exchangers, the level of agreement between experiment and
CFD predictions is considered the most relevant information pre-
sented in this paper.

2. Geometry

Correlations for the Colburn and friction factors are available in

Ref. [11] for many different geometrical congurations. These
geometrical congurations are not suited for ceramic heat ex-
changers because, ceramics require particular manufacturing pro-
cesses and geometrical design. In general, small thickness plates,
easily obtained with metals, would result in an excessively fragile
ceramic plate. Due to this, the correlations are not available for
geometries as the one shown in Fig. 1, which demonstrated its
feasibility using ceramics [10]. Following the approach used in
Ref. [13], geometrical parameters that dene the ns have been
created. As demonstrated by Ref. [12] a simulation including three
rows of ns is sufcient to obtain practically identical Colburn and
friction factor curves, if compared to the simulation conducted with
six rows. This simplication reduces the computational costs, as the
number of volumes in the mesh decreases drastically. As shown in
Fig. 2, the domain is extended on the inlet and outlet regions so the
ow can evolve properly and minimal effects of the idealization in Fig. 3. : Geometrical parameters that dene the n base.

Please cite this article in press as: Villanueva HHS, de Mello PEB, Heat transfer and pressure drop correlations for nned plate ceramic heat
exchangers, Energy (2015), http://dx.doi.org/10.1016/j.energy.2015.04.017
 H.H.S. Villanueva, P.E.B. de Mello / Energy xxx (2015) 1e8 3

Table 2
D Characteristics of each geometrical conguration simulated.
g 1 (3)
D2 Code Dh (mm) s a (m2/m3) Afr (mm2) Aff (mm2) A (mm2)

The dimensions D1 and L2 are xed at 4 and 17 mm respectively, 1-2-2 4.1071 0.556 541.07 9.0 5.0 82.78
for all simulations. It is reasonable to consider that would be hard to 2-1-2 5.3647 0.556 414.23 13.5 7.5 95.07
2-2-1 4.6067 0.500 434.15 12.0 6.0 88.57
construct much ner ns with ceramics, so that D1 is xed at 4 mm.
2-2-2 5.2541 0.556 422.95 13.5 7.5 97.07
This value is feasible, as shown in Ref. [10]. The investigation of the 2-2-3 5.7973 0.600 413.99 15.0 9.0 105.57
geometry inuence was conducted for three different values of 2-3-2 5.1481 0.556 431.66 13.5 7.5 99.07
each parameter, maintaining each one bellow unity, as shown in 3-2-2 6.1069 0.556 363.89 18.0 10.0 111.35
Table 1. A code for each geometrical conguration was created to
name them.
It is important to note that the variations in geometry inuence " #
some calculation parameters, as the hydraulic diameter 4, total G2 ni  n 
A nm
2 o
frontal area Afr, minimum free-ow area Aff and heat transfer area A. DP 1s 1 f (9)
2 ni sAfr ni
All these parameters are dened as in Ref. [11] and are shown in
Table 2. Constant values for thermophysical properties are used in all
simulations. It is the correct way to obtain correlations for friction
4Aff and Colburn factors. Thermophysical properties must be included
Dh L (4)
A in calculation during design (or simulation) of a particular heat
For HTHE design purposes, the formulations for the Dh, a and s exchanger. Due to this, the friction factor simplies to 10.
are essential. These parameters, as a function of l, d and g are given
2DPrsAfr
by (5e7). f (10)
G2 A

   1
L2 1 p p 1 1 Temperature is imposed in the domain inlet and walls as
Dh 2L2 1  g lg  1 1
gD1 d 4 2 d l boundary conditions. The resultant temperature at the domain
(5) outlet permits to obtain the heat exchanger effectiveness 11, for the
particular case where the wall has constant temperature. The
    temperature values imposed as boundary conditions for Tw and Ti
L2 1 p p 1 1 1
a 2lg  1 1 L (6) do not inuence the heat exchanger effectiveness and Colburn
gD1 d 4 2 d l 2
factor for simulations with constant thermophysical properties.

Aff To  Ti
s 1g (7) (11)
Afr Tw  Ti
The number of transfer units NTU is evaluated from effective-
ness using (12).
3. Methodology
NTU ln1  (12)
The next two sections discuss how the friction and Colburn The overall heat transfer coefcient is determined from NTU
correlations are obtained from the CFD simulations and the equa- using (13).
tions relating the laminar, transition and turbulent ow regimes.
C$NTU
Uh (13)
A
3.1. Friction and colburn factors
Finally, the Colburn correlation is obtained with 14.
Pressure drop is evaluated from the CFD results with 8. The
pressure is xed at the outlet of the domain (boundary condition) Uh $s$Afr $Pr 2=3
j (14)
and the resultant average pressure at the inlet permits the pressure _ p
m$c
drop determination.
The Reynolds number, following methodology developed in
DP Pi  Po (8) Ref. [11], is dened by 15.

Friction factor for compact heat exchangers is dened in Ref. [11] G$Dh
using 9. Re (15)
m
The ow-stream mass velocity G depends on the mass ow rate
Table 1 and can be evaluated using 16.
Geometrical congurations simulated numerically.

Code H L D2 l d g m_ rU
G (16)
1-2-2 4 12 9 1.000 0.333 0.444
s$Afr s
2-1-2 6 10 9 0.667 0.400 0.444
2-2-1 6 12 8 0.667 0.333 0.500 For each simulation, one value of Reynolds to be investigated is
2-2-2 6 12 9 0.667 0.333 0.444 xed and mass ow rate is imposed in the inlet of the calculation
2-2-3 6 12 10 0.667 0.333 0.400 domain using 15 and 16. The ratio of free-ow area to frontal area s,
2-3-2 6 14 9 0.667 0.286 0.444 used in equations (10), (14) and (16), is dened by equation (7). This
3-2-2 8 12 9 0.500 0.333 0.444
parameter is used to obtain one estimate of velocity in cross-

Please cite this article in press as: Villanueva HHS, de Mello PEB, Heat transfer and pressure drop correlations for nned plate ceramic heat
exchangers, Energy (2015), http://dx.doi.org/10.1016/j.energy.2015.04.017
4 H.H.S. Villanueva, P.E.B. de Mello / Energy xxx (2015) 1e8
sectional area restrictions where ow accelerates. It depends on
geometrical conguration of the heat exchanger only.
3.2. Correlations
Friction and Colburn factors are dependent on Reynolds number
and geometrical parameters l, d and g. In order to state mathe-
matically this dependence, appropriate equations are used (17e18).
This form of the correlations was rst suggested in Ref. [13].
 0:1
f a1$Rea2 $ 1 c1$Rec2 $la3 $da4 $ga5 (17)
 0:1
j b1$Reb2 $ 1 d1$Red2 $lb3 $db4 $gb5 (18)
The use of these power law equations are justied because in
both deep laminar and fully turbulent ow regimes, the variations
of f and j follow a constant-slope function at the logelog graph [13].
4. Numerical model
All the CFD simulations were conducted using the Ansys CFX 14
nite volume code. The description of the code is beyond the scope
of this paper but can be easily found in its documentation or in the
literature. Only relevant information is given hereafter.
4.1. Governing equations
According to the CFX documentation, for steady state simula-
tions with heat transfer, the code resolves numerically the con-
servation equations given by 19e21.
V$rU 0 (19)
V$rU5U VP V$t SM (20)
 
k
V$rUT V$ VT ST (21)
cp
The partial differential equations (19) and (21) express the
conservation of mass, momentum and energy respectively. In order
to solve them numerically, we used the boundary conditions dis-
cussed in section 2 and appropriate numerical schemes.
4.2. Turbulence model
The results presented in this paper cover ows from laminar to
turbulent regimes, modelled with the transitional SST (Shear Stress
Transport) model [14]. Using four additional equations, this tur-
bulence model can resolve both regimes and the transition. To
evaluate and conrm the accuracy of this model, some simulations
have been made comparing its results with laminar model (no
turbulence model) and SST model (for high Reynolds number). The
transitional model reproduces these extreme conditions satisfac-
torily. One detailed description of the model with an evaluation of
its capabilities to predict heat transfer in pipes is presented in
Ref. [15].
4.3. Solution schemes
For discretization of the advection terms, the CFX high-
resolution scheme was chosen. This scheme uses it's own algo-
rithm that works tending as close as possible to the central differ-
encing scheme CDS without introducing unphysical oscilations. As
Please cite this article in press as: Villanueva HHS, de Mello PEB, Heat transfer and pressure drop correlations for nned plate ceramic heat
exchangers, Energy (2015), http://dx.doi.org/10.1016/j.energy.2015.04.017
a result, in regions of the domain that smooth gradients are present,
the CFX high-resolution scheme tends to CDS. At regions near to
discontinuities, the high-resolution scheme tends to the upwind
scheme (rst order) to obtain a bounded solution. According to
software documentation, in a simplied one dimensional case, the
algorithm resembles to Total Variation Diminishing TVD scheme.
The convergence was evaluated monitoring the root mean square
(RMS) of the normalized residuals for all equations. We took care to
check the convergence of the simulation using additional monitor
points. Pressure and temperature were monitored during the so-
lution to guarantee that they reached a stable level. The conver-
gence criteria for the RMS of the normalized residuals was set to
1.0  108.
4.4. Mesh
Computational costs and quality of simulations are directly
connected with the mesh used to discretize the domain. Hexahe-
drons describe the domain with a lesser number of elements. To
achieve a good solution near walls, a block strategy with attention at
the Y value, the dimensionless distance to the wall, and the
expansion ratio was taken. In the CFX documentation, for the
transitional turbulence model, the Y must be between 0.0001 and
8, where the desirable optimal value is one. The expansion ratio
must be between 1.05 and 1.4 where the desirable value is 1.1. To
cover all cases, we made three different meshes: one for the laminar
regimes and one for each Reynolds number on the turbulent regime.
4.5. Mesh dependency tests
To ensure precise results independent of the mesh density, a
dependency study of meshes that satised the CFX documentation
criteria was developed for Reynolds number of 10,000. The results
are presented on Table 3. These results show that once the criteria
from the CFX documentation are satised, the density of the mesh
does not inuence the results of f and j correlations witch are the
main results of interest.
Applying a block strategy with the expansion rate between 1.1
and 1.2 for all geometries, the maximum Y veried was in the
range of 0.76e4.45. Details about the meshes are shown in Table 4.
A typical mesh used in the simulations is shown in Fig. 4.
5. Experimental results
The experimental results used for validation were obtained with
a procedure detailed in Ref. [10]. The heat exchanger is constructed
with ceramic plates, as shown in Fig. 1. In summary, measured air
ow rates are supplied to the heat exchanger, one of them is heated
before entering the heat exchanger. The air ow rates are controlled
in order to obtain the same Reynolds number in hot and cold sides.
Inlet and outlet temperatures are measured using thermocouples.
The Colburn j factor is determined by these measurements, as
described in section 3.1, but with the number of transfer units NTU
Table 3
Mesh dependency results.
Volumes  103 Y j f
106 3.09 0.00407 0.01968
195 3.10 0.00403 0.01946
456 3.11 0.00417 0.01984
1669 3.13 0.00421 0.01979
2146 3.16 0.00417 0.01990
3039 3.22 0.00440 0.02028
4515 3.23 0.00441 0.02032
 H.H.S. Villanueva, P.E.B. de Mello / Energy xxx (2015) 1e8 5

Table 4 Uncertainty propagation analysis was performed with EES for

Mesh parameters. each test. We obtained typical uncertainty 1.5% for Reynolds
Reynolds Limits Volumes  103 Y number, 3.5% for Colburn factor and 3.7% for friction factor. The
200 Max 447 0.99
uncertainty for friction factor was reduced, if compared to results
Min 384 0.76 presented in Ref. [10] with the use of more appropriate pressure
10,000 Max 502 4.45 transducers.
Min 475 2.50 Experimental and numerical results are compared in next
30,000 Max 1240 3.84
section.
Min 1233 2.82

6. Numerical results

The numerical results are presented in three sections. Initially,

one correlation for a baseline geometrical conguration is evalu-
ated. This baseline conguration, with code 2-2-2 in Table 1, is
named central condition.

6.1. Correlation for the central condition

The central condition, with code 2-2-2 in Table 1, is dened by

l 0.667, d 0.333 and g 0.444. Other congurations present
these parameters with higher or lower values if compared to this
central condition. In Fig. 6, friction and Colburn factors, f and j
respectively, as a function of Reynolds number is shown for this
central condition. From these simulation results, dependency of f
and j on Reynolds number is obtained, and correlations (22e23) are
evaluated.
Fig. 4. Typical hexahedral mesh used to discretize the domain.
 0:1
fcentral 2:5922$Re0:602 $ 1 6:3$Re14 (22)
determined for counter-ow conguration. Pressure drop is
measured in the cold side only. Each test takes approximatelly  0:1
3e5 h to reach steady state (one point in j  Re diagram). All jcentral 1:0495$Re0:646 $ 1 7:4$Re12 (23)
measurements are monitored graphically during the experiment to
guarantee that steady state was reached. The schematic represen-
tation of experimental setup is shown in Fig. 5.
Air ow rate is measured using two turbine ow meters with 6.2. Inuence of geometrical parameters
range 3.4e34.0 m3/h and uncertainty 1.0%. The pressure trans-
ducer used in Ref. [10] was substituted by two new ones, with full In order to evaluate the inuence of geometrical parameters, the
scale of 2.5 and 12.5 mbar, both with 0.8% FS uncertainty. Each congurations presented in Table 1 are separated in categories,
pressure transducer is appropriate to measure pressure drop in a
specic ow rate range. We considered the uncertainty of K type
thermocouples equal to 0.75% or 2.2
C, whichever is greater,
following recommendations in literature. This approach introduces
high uncertainty for temperature measurements close to ambient
conditions (ow rates are measured in ambient conditions). In or-
der to reduce the uncertainty in this range, we conducted a cali-
bration and reduced thermocouples uncertainty to 0.35
C for
measurements bellow 50
C.
Air thermophysical properties were determined using EES (en-
gineering equation solver). Following the methodology presented
in Ref. [11], Reynolds number in hot and cold sides are calculated
using thermophysical properties evaluated at mean temperature at
that side, obtained from inlet and outlet temperature
measurements.

Fig. 5. Schematic representation of experimental set-up. Fig. 6. Results obtained with the central condition (code 2-2-2 in Table 1).

Please cite this article in press as: Villanueva HHS, de Mello PEB, Heat transfer and pressure drop correlations for nned plate ceramic heat
exchangers, Energy (2015), http://dx.doi.org/10.1016/j.energy.2015.04.017
6 H.H.S. Villanueva, P.E.B. de Mello / Energy xxx (2015) 1e8

depending on each of the parameter is under consideration. Firstly,

the inuence of g is evaluated using congurations with codes 2-2-
1 and 2-2-3. The results obtained with these congurations are
compared to the central correlation, as shown in Fig. 7.
As shown in Fig. 7, the parameter g has minimal inuence over
the friction and Colburn factors. Due to this, this parameter was not
included in the nal correlation. Following the same approach,
Figs. 8 and 9 shows the inuence of l and d. Clearly, these param-
eters have signicant inuence over friction factor, but Colburn
factor is most inuenced by the parameter l.
The nal correlations that summarize the numerical results for
friction and Colburn factors were obtained using the least squares
method 24e25. We decided to not include the parameter g in the
correlations due to its very small inuence over the results, as
shown in Fig. 7. The most important geometrical parameter is l, and
this can be observed both in the exponents (0.40 for f and 0.22
for j) and graphically in Fig. 8. In the other hand, the inuence of
d over Colburn factor is almost inexistent.
 0:1  l
0:40
fcentral 2:5922$Re0:602 $ 1 6:3$Re14 $
0:667
 0:11
d
$
0:333
Fig. 8. Inuence of parameter l.
(24)

 0:1  l
0:22
unsatisfactory, due to the use of inappropriate pressure trans-
jcentral 1:0495$Re0:646 $ 1 7:4$Re12 $
0:667 ducers. Besides, the Reynolds range was very narrow: between 200
 0:08 and 500. After improvements in the experimental setup described
d
$ in Ref. [10] and the construction of a second ceramic heat exchanger
0:333 prototype, new experimental results were obtained. The compari-
(25) son between CFD simulations and these updated experimental data
is presented in Fig. 10. The CFD simulations reproduced the
6.3. Comparison with experiments experimental friction factor with good accuracy, over the entire
Reynolds range considered. Deviations around 20% were observed
In order to validate the CFD simulations we conducted a com- for Colburn factor at the upper Reynolds range.
parison between numerical results and experiments reported in The geometry used in the experiments of [10] lead to geomet-
Ref. [10]. These experimental data, obtained with a rst prototype, rical parameters dened as l, d and g in current paper equal to
revealed some limitations. The friction factor characterization was

Fig. 7. Inuence of parameter g. Fig. 9. Inuence of parameter d.

Please cite this article in press as: Villanueva HHS, de Mello PEB, Heat transfer and pressure drop correlations for nned plate ceramic heat
exchangers, Energy (2015), http://dx.doi.org/10.1016/j.energy.2015.04.017
 H.H.S. Villanueva, P.E.B. de Mello / Energy xxx (2015) 1e8
Fig. 10. Comparison of the correlation with experiments. The original experiments
presented in Ref. [10] were updated to increase Reynolds range and improve friction
factor characterization.
0.667, 0.286 and 0.5 respectively. Following Table 1, this geometry
would have the code 2-3-1.
6.4. Heat transfer behaviour
Local heat transfer coefcient, dened by equation (26).,was
obtained numerically and is shown in Fig. 11. As expected, the heat
transfer coefcient near the inlet boundary is higher and decreases
along the ow direction. As the ow develops, the difference be-
tween uid temperature and wall temperature decreases with a
consequent decrease in the local heat ux qw. As a result, the local
heat transfer coefcient decreases in the ow direction because the
denominator in equation (26) is constant.
After the rst row of ns on Fig. 11, local heat transfer
enhancement with horseshoe shape is observed in front of the ns,
Fig. 11. Comparison of local heat transfer coefcient (in W/m2 K) obtained with simulations including three and six rows of ns.
Please cite this article in press as: Villanueva HHS, de Mello PEB, Heat transfer and pressure drop correlations for nned plate ceramic heat
exchangers, Energy (2015), http://dx.doi.org/10.1016/j.energy.2015.04.017
7
Fig. 12. Wall and ns coloured by heat transfer coefcient (in W/m2 K) and streamlines
in grey.
revealing that heat transfer distribution in the rst row is different
from the remaining rows. However, when average heat transfer is
considered, in the form of Colburn factor, insignicant difference is
observed when the result obtained with three rows is compared to
the case with six rows of ns. This behaviour was previously
observed in Ref. [12].
Comparing the results obtained with different mass ow rates it
is possible to state that the local heat transfer enhancement is
inuenced by the Reynolds number.
qw
hlocal (26)
Tw  Ti
Fig. 12 presents the streamlines at the frontal part of the n,
where heat enhancement is veried. The vortex formed in this
region increase local heat transfer and explains the horseshoe
shapped contours observed in Fig. 11. This behaviour is well known
and docummented in literature. Similar behaviour is observed in
numerical investigations [16] and experiments [17].
Pressure eld is shown in Fig. 13 and reveals that the pressure
decreases at each new row of ns. One adverse pressure gradient
may be observed in the leading edge of the ns and is the cause of
the horseshoe vortex observed in Fig. 12.
8 H.H.S. Villanueva, P.E.B. de Mello / Energy xxx (2015) 1e8

Fig. 13. Pressure distribution (in Pa).
7. Conclusions
Experimental data available for the ceramic HTHE considered
herein is rare. For validation purposes, we used updated experi-
mental data rst presented in Ref. [10], with increased Reynolds
range. The validation conducted in this paper permits to afrm that Nomenclature
CFD simulations are capable of predicting pressure drop and heat
transfer with good accuracy.
The correlations for Colburn and friction factors developed from
CFD simulation include the inuence of Reynolds number and two C: heat capacity rate, W/K
geometrical parameters of the ns. The inuence of the third cp: specic heat, J/(kg K)
geometrical parameter investigated has no signicance. It is D1: n thickness, m
important to note that the geometrical congurations investigated
present different values of compactness, as shown in Table 2 by the
parameter a. For some applications, compactness of the HTHE may
have higher importance than performance. Another concern is
related to feasibility of ceramic HTHE. As shown in Ref. [10], the rst j: Colburn correlation, dimensionless
prototype constructed with alumina did not supported the thermo
structural load and presented cracks in the end of experiments,
conducted at 890
C.
Acknowledgements

rio da
The authors would like to acknowledge Centro Universita
FEI and Fundaa~o de Amparo a  Pesquisa do Estado de Sa
~o Paulo
(FAPESP) for the research support.
References
[1] Schulte-Fischedick J, Dreiigacker V, Tamme R. An innovative ceramic high
temperature plate-n heat exchanger for EFCC processes. Appl Therm Eng
2007;27:1285e94.
[2] Al-attab K, Zainal Z. Performance of high-temperature heat exchangers in
biomass fuel powered externally red gas turbine systems. Renew Energy
2010;35:913e20.
[3] Traverso A, Massardo AF, Scarpellini R. Externally red micro-gas turbine:
modelling and experimental performance. Appl Therm Eng 2006;26:
1935e41.
[4] Cordiner S, Mulone V. Experimental-numerical analysis of a biomass fueled
microgeneration power-plant based on microturbine. Appl Therm Eng
2014;71(2):905e12.
[5] de Mello PEB, Scuotto S, Ortega F, Donato G. Inuence of turbine inlet tem- d: geometrical parameter, dimensionless
perature on the efciency of externally red gas turbines. In: 11th Interna- l: geometrical parameter, dimensionless
tional conference on sustaintable energy technologies; 2012. p. 996e1007.
[6] McDonald CF. Recuperator considerations for future higher efciency micro- s: ratio of free-ow area to frontal area, dimensionless
turbines. Appl Therm Eng 2003;23:1463e87.
[7] Shah R, Thonon B, Benforado D. Opportunities for heat exchanger applications
in environmental systems. Appl Therm Eng 2000;20:631e50.
[8] Kee RJ, Almand BB, Blasi JM, Rosen BL, Hartmann M, Sullivan NP, et al. The
design, fabrication, and evaluation of a ceramic counter-ow microchannel
heat exchanger. Appl Therm Eng 2011;31:2004e12.
[9] Fend T, Vlker W, Miebach R, Smirnova O, Gonsior D, Schllgen D, et al.
Experimental investigation of compact silicon carbide heat exchangers for
high temperatures. Int J Heat Mass Transf 2011;54:4175e81.
[10] de Mello PEB, Scuotto S, Ortega F, Donato G. Heat transfer and pressure drop
in a plate and n ceramic heat exchanger. In: 8th world conference on
experimental heat transfer, uid mechanics and thermodynamics; 2013.
p. 16e20.
[11] Kays WM, London AL. Compact heat exchangers. New York, NY: McGraw-Hill;
1984.
[12] Monteiro DB, de Mello PEB. Thermal performance and pressure drop in a
ceramic heat exchanger evaluated using CFD simulations. Energy 2012;45:
489e96.
[13] Manglik RM, Bergles AE. Heat transfer and pressure drop correlations for the
rectangular offset strip n compact heat exchanger. Exp Therm Fluid Sci
1995;10:171e80 [Aerospace Heat Exchanger Technology].
[14] Menter F, Langtry R, Vlker S, Huang P. Transition modelling for general pur-
pose CFD codes. In: Mulas WR, editor. Engineering turbulence modelling and
experiments 6. Amsterdam: Elsevier Science B.V.; 2005. p. 31e48.
[15] Abraham J, Sparrow E, Tong J. Heat transfer in all pipe ow regimes: laminar,
transitional/intermittent, and turbulent. Int J Heat Mass Transf 2009;52:
557e63.
[16] Hemida H, Spehr F, Krajnovi S. Local heat transfer enhancement around a
matrix of wall-mounted cubes using passive ow control: large-eddy simu-
lations. Int J Heat Fluid Flow 2008;29:1258e67.
[17] Vintrou S, Bougeard D, Russeil S, Nacereddine R, Harion J-L. Quantitative
infrared investigation of local heat transfer in a circular nned tube heat
exchanger assembly. Int J Heat Fluid Flow 2013;44:197e207.
A: exchanger total heat transfer area, m2
Aff: exchanger minimum free-ow area, m2
Afr: exchanger total frontal area, m2
D2: lateral n offset, m
Dh: hydraulic diameter, m
f: friction correlation, dimensionless
G: exchanger ow-stream mass velocity, kg/(s m2)
hlocal: wall heat transfer coefcient, W/(m2 K)
H: high of the ns, m
k: thermal conductivity, W/(mK)
L: length of the n, m
L2: length of n cell, m
m_: mass ow, kg/s
NTU: number of transfer units, dimensionless
P: pressure, Pa
Pr: Prandtl number, dimensionless
qw: wall heat ux, W/m2
Re: Reynolds number, dimensionless
SM: source term in momentum equation
ST: source term in energy equation
St: Stanton number, dimensionless
T: temperature,
C
Ti: mean temperature at the exchanger inlet,
C
To: mean temperature at the exchanger outlet,
C
Tw: temperature of the exchanger wall,
C
U: velocity, m/s
Uh: overall heat transfer coefcient, W/(mK)
ni: specic volume at the exchanger inlet, m3/kg
nm: average specic volume, m3/kg
no: specic volume at the exchanger outlet, m3/kg
Y: dimensionless distance to the wall, dimensionless
Greek
a: ratio of total heat transfer area to total volume of the exchanger, m2/m3
g: geometrical parameter, dimensionless
: effectiveness, dimensionless
m: dinamic viscosity, kg/m2 s
r: density, kg/m3

Please cite this article in press as: Villanueva HHS, de Mello PEB, Heat transfer and pressure drop correlations for nned plate ceramic heat
exchangers, Energy (2015), http://dx.doi.org/10.1016/j.energy.2015.04.017