Catalysis Today 216 (2013) 121–134

Contents lists available at SciVerse ScienceDirect

Catalysis Today
journal homepage: www.elsevier.com/locate/cattod

Heat transfer properties of metal foam supports for structured

catalysts: Wall heat transfer coefficient
Enrico Bianchi a,b , Tobias Heidig a , Carlo Giorgio Visconti b , Gianpiero Groppi b ,
Hannsjörg Freund a , Enrico Tronconi b,∗
a
Lehrstuhl für Chemische Reaktionstechnik, Friedrich-Alexander-Universität Erlangen-Nürnberg, Egerlandstr. 3, Erlangen, D-91058, Germany
b
Laboratory of Catalysis and Catalytic Processes, Dipartimento di Energia, Politecnico di Milano, Piazza Leonardo da Vinci 32, Milano, 20133, Italy

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

Article history: In this work we investigate the heat transfer within single phase tubular reactors packed with open-
Received 16 March 2013 cell metal foams. Special focus is on the role of the coupling between the foam packing and the reactor
Received in revised form 12 June 2013 tube wall. Steady-state heat transfer experiments were carried out in a tube of 28 mm inner diameter
Accepted 19 June 2013
packed with different open-cell foams using N2 or He at a flow rate ranging from 15 to 35 Nl/min. The
Available online 27 July 2013
foam samples were made of either FeCrAlY or aluminum with porosities in the range of 89–95% and
pore densities ranging from 10 to 40 PPI (nominal pores per linear inch). For this purpose, axial and radial
Keywords:
temperature profiles along the tested foam samples and along the tube wall were collected in heating and
Metal foams
Sponges
cooling runs within a temperature range from 400 to 800 K. The effective radial and axial conductivity and
Heat transfer the wall heat transfer coefficient were then estimated for each run by nonlinear regression, using a 2D
Process intensification pseudo-homogeneous heat transfer model. The collected data allowed deriving a correlation for the wall
Optimized supports heat transfer coefficient as a function of the process conditions and of the foam structural properties.
Structured reactors Complementary to these experimental investigations, a numerical study in a 3D geometry of a foam
sample reconstructed from X-ray micro-computed tomography data was carried out in order to analyze
the heat transfer process close to the wall. For this, the conjugated heat transfer in both the solid and
the fluid phase was investigated by finite volume analysis for selected experimental conditions. Both
experiments and simulations confirm that the wall heat transfer resistance depends strongly on the gas
conductivity and on the foam geometry, but more weakly on the flow velocity.
© 2013 Elsevier B.V. All rights reserved.

1. Introduction realized in the field of catalysis, even though the peculiar prop-
erties of metallic foams, including their outstanding mechanical
Open-cell foams, sometimes also referred to as sponges, are resistance, make these cellular structures very promising also as
receiving considerable attention because of their very good catalyst supports for non-adiabatic chemical processes in tubular
gas/solid mass- and heat-transfer properties [1–4] and, in the case fixed bed reactors(e.g., steam reforming of hydrocarbons, CO
of highly conductive samples, of their promising heat transfer hydrogenation processes, and selective oxidation reactions). One
features [5–7]. As a matter of fact, open-cell metal foams are nowa- possible reason is the lack of engineering correlations for heat
days extensively employed as multifunctional heat exchangers and mass transport properties of open-cell foam structures. In
[8] and compact heat sinks for microelectronic devices [9,10], and this paper, we focus specifically on the evaluation of radial heat
as receiver dishes in solar reforming of methane [11]. They have transfer in tubular fixed bed reactors loaded with catalytic foams
also found application as static mixers for new milli-scale plug for strongly exo- or endothermic gas/solid processes.
flow reactors [12–14]. Besides these applications, open-cell foams, While a 2D two-parameter approach using both an effective
similar to other monolithic structures such as, e.g., honeycombs, radial conductivity and a wall heat transfer coefficient [19] is a com-
are considered as prospective catalyst supports for chemistry and mon approach for the description of catalytic packed bed tubular
energy-related processes, as they feature extremely high external reactors [18],in many of the experimental and theoretical studies
surface areas and low pressure drops [15–17]. Nevertheless, to of the heat transfer within open-cell foams saturated with flowing
our knowledge no industrial applications of foams are currently fluid available in the open literature the wall resistance is simply
neglected [10,20–22], despite the fact that it can actually represent
the controlling factor for radial heat transfer [5].
∗ Corresponding author. Considering the most recent commercial products based on
E-mail address: enrico.tronconi@polimi.it (E. Tronconi). open-cell foams, the importance of the resistance for the heat

0920-5861/$ – see front matter © 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.cattod.2013.06.019
122 E. Bianchi et al. / Catalysis Today 216 (2013) 121–134

Nomenclature
Greek symbols
List of symbols ε bed porosity; (–)
A constant to be determined (static Nusselt number); εT total porosity of foam sample; (–)
(–) εH hydraulic porosity of foam sample; (–)
B constant to be determined (convective Nusselt  dynamic viscosity of fluid; (kg s−1 m−1 )
number); (–)  mass density; (kg m−3 )
cp specific heat capacity; (kJ kg−1 K−1 )  viscous stress tensor; (N m−2 )
C exponential factor to be determined (convective
Nusselt number); (–) Subscripts
CHT cooling run at the highest temperature; 0: static properties;
CLT cooling run at the lowest temperature; a: axial;
D inner tube diameter; (m) b: bulk;
dp average pore inner diameter; (m) c: cell;
dc average cell inner diameter; (m) Conv.: convective contribution;
h specific enthalpy; (J kg−1 ) D: based on tube diameter;
hW wall heat transfer coefficient; (W m−2 K−1 ) dc : based on cell diameter;
hW,0 stagnant term of the wall heat transfer coefficient; e: effective;
(W m−2 K−1 ) f: fluid properties;
hW,Conv. convective term of the wall heat transfer coefficient; H: hydraulic;
(W m−2 K−1 ) in: inlet;
HHT heating run at the highest temperature; p: pore;
HLT heating run at the lowest temperature; r: radial;
k thermal conductivity; (W m−1 K−1 ) s: solid properties;
kf fluid thermal conductivity; (W m−1 K−1 ) S: volumetric surface;
ks solid thermal conductivity; (W m−1 K−1 ) T: total;
ke,r effective radial conductivity of the foam bed; w: wall;
(W m−1 K−1 ) z: axial coordinate;
ke,a effective axial conductivity of the foam bed; −∞: undisturbed condition;
(W m−1 K−1 )
ke,0 effective stagnant conductivity of saturated
medium; (W m−1 K−1 )
transfer at the wall seems to be a well-known issue also for some
L length of foam bed; (m)
commercial foam manufacturers such as, e.g., m-pore GmbH [14]
NuD overall Nusselt number, UDk
; (–)
f and Porvair Fuel Cell Technology [22,23]. These companies are in
hw d
NuW Nusselt number at the wall, kf
; (–) fact looking for technical solutions to design cellular structures fully
hw,0 d integrated into a tube, i.e., either without or with a reduced ther-
NuW,0 static Nusselt number at the wall, kf
; (–)
mal discontinuity between the foam and the tube. This solution has
hw,Conv. d
NuW,Conv. convective Nusselt number at the wall, ; (–) been found to be promising in relation to the adoption of foams as
kf
heat sink [22,23], but as to our knowledge, it has not been tested
P pressure; (Pa)
yet for catalytic reactors.
PD average pore density; (PPI)
cp In one of the few papers explicitly taking into account such a
Pr Prandtl number, k f ; (–)
f wall resistance, Peng et al. [24] conducted heat transfer experi-
r radial coordinate; (m) ments with air in the range of 150–550 ◦ C and 3–20 Sl/min on a
R radius of tube; (m) single 30-PPI ␣-Al2 O3 foam fitted into a tube of 50 mm inner diam-
R̂ specific ideal gas constant; (J kg−1 K−1 ) eter (I.D.). From the measured axial and radial temperature profiles
Win D
ReD Reynolds number based on tube diameter,  ; (–) they derived a correlation for the wall heat transfer coefficient as a
Win dc function of the Reynolds number based on the volumetric surface
RedC Reynolds number based on cell diameter,  ; (–)
W area. Since the experimental campaign was carried out using a sin-
ReS Reynolds number based on volumetric surface, S in ;
V gle foam sample, however, no evidence could be collected in this
(–) paper on the potential dependency of the wall heat transfer coef-
Sv geometric interfacial area density, or volumetric ficient on the foam geometry. Wall heat transfer coefficients were
surface, of foam; (m−1 ) also measured by Edouard et al. [25] for polyurethane and ␤-SiC
ts average middle strut section; (m) foams, running experiments with air at low temperatures (<100 ◦ C)
tn average thickness of the conjunction section of in a reactor of 76.5 mm I.D. Estimates of the wall heat transfer coef-
struts; (m) ficient were obtained at different flow velocities over the two foam
T temperature; (K) samples, but were then averaged by the authors to obtain a single
Tb bulk temperature of fluid; (K) mean value.
Tz,r temperature at axial coordinate z and radial coordi- Moreover, in a comparative study of the heat transfer proper-
nate r; (K) ties of structured packings in two-phase flow [26,27], also the wall
Tz,w tube temperature at axial coordinate z; (K) heat transfer coefficient of an aluminum foam was reported. The
u velocity;(m s−1 ) open-cell foam sample was tested at different flow conditions, and
U overall heat transfer coefficient; (W m−2 K−1 ) the results suggest a role of the convective contribution to heat
Win specific mass velocity; (kg s−1 m−2 ) transfer of minor importance with respect to the other types of
z axial coordinate; (m) investigated packings (open/closed cross flow structures, knitted
wire, glass beads).
 E. Bianchi et al. / Catalysis Today 216 (2013) 121–134 123

The purpose of this work to address the influence of opera- conditions adopted in each test. For example, the acronym HLT is
tive conditions, thermo-physical and geometrical properties on the used to identify a heating test (H) with the thermostatic chamber
single phase wall heat transfer coefficient inside a tube filled with kept at low temperature (LT). Tests at different flow rates between
open-cell foams. In particular, the focus is on the wall coupling con- 20 and 30 Nl/min for He or between 15 and 35 Nl/min.for N2 [5]
ditions typical of current catalytic reactors, i.e., with the packing not were carried out in the two modes, resulting in a total of 111 runs.
directly brazed with the tube wall. To achieve this goal, both heat Radial temperature profiles along the bed length were measured
transfer experiments and 3D simulations were performed to eval- at the three different radial positions at steady-state conditions
uate the conjugated heat transfer in the solid and the fluid phase. by means of the three thermocouples previously mentioned. The
axial temperature profiles of the gas stream fed to the foams was
2. Materials and methods also measured in each test by moving these thermocouples 5 mm
in front of the beginning of the actual foam bed. The tube skin
2.1. Investigated foam samples temperature was measured by a fourth K-thermocouple sliding
in a thermowell tightly connected to the external tube wall. Each
Open-cell foams made of FeCrAlY (Sample A and B) and Al- temperature profile was taken at least twice during each run to
alloy (Al 6101 T6, samples C, D and E), representative of metallic check its reproducibility.
materials with a very different thermal conductivity (from about
16 W m−1 K−1 for FeCrAlY to 218 W m−1 K−1 for Al-alloy [28] at 2.3. Mathematical analysis of the experimental data
room temperature), were investigated in this work. In particular,
cylindrical samples with a diameter of 28 mm (+0.01, −0.00 mm), Experimental temperature profiles were used to estimate the
cut out of larger panels by electro-erosion, were used in the heat heat transfer coefficients of the foams. A steady-state non-reactive
transfer measurements. Due to the limited thickness of the avail- 2D pseudo-homogeneous heat transfer model of the test tube as
able panels, the depth of the cylinders was 25 mm for sample A, described in [5] was used for this purpose. We here recall the
100 mm for sample B and 50 mm for samples C, D and E. Three partial differential equation (PDE) for the temperature field as
axial through holes of 3.28 mm diameter were then drilled in each derived from the enthalpy balance in temperature form written in
sample at three different radial positions, namely at the center- cylindrical coordinates (Eq. (1)), and the corresponding boundary
line and 7 and 9 mm from the center. These holes allowed for the conditions along the radial (Eq. (2)) and axial (Eq. (3)) direction:
tight insertion of 3 stainless steel thermo-wells protecting sliding    
2 2
K-type thermocouples for the measurement of the internal axial ∂T 1 ∂T ∂ T ∂ T
temperature profiles in the heat transfer experiments. Win cp = ke,r + + ke,a (1)
∂z r ∂r ∂r 2 ∂z 2
The geometrical properties used in this work to characterize the
tested samples are summarized in Table 1. Details on the experi- ⎧
⎪ ∂T
mental techniques used to evaluate such properties can be found ⎨ = 0, at : r = 0;
elsewhere [5]. ∂r
(2 a–b)
⎪
⎩ ke,r ∂T = hw (Tz,w − Tz,r=R ), at : r = R;
2.2. Experimental set-up and procedures ∂r

Heat transfer measurements were performed in a lab-scale rig

⎧
⎪ ∂T
available in the laboratories of Politecnico di Milano. A sketch of ⎨ ke,a = Win cp (Tz=0,r − Tz=−∞,r ), at : z = 0;
∂z
the facility and more details on the set-up can be found in our pre- (3 a–b)
⎪
⎩ ∂T
vious paper [5]. The major part of the unit is a horizontal test tube = 0, at : z = L;
(28 mm nominal I.D.) where one or more foam samples, in order ∂z
to reach a total length of 100 mm, were loaded with interference The meaning of the symbols in Eqs. (1–3) and in the remainder
fit, to force contact between the samples and the wall. The tube of this paper is explained in the Nomenclature section.
was placed into a thermostatic oven with air recirculation (Maz- On the radial coordinate, we assume a symmetry boundary con-
zali Thermotest) where it can be fed with up to 30 Nl/min of He dition at the centerline (Eq. (2a)), and describe the heat transfer
(99.999%) or up to 35 Nl/min.of N2 (99.995%) at atmospheric pres- between the outer edge of the structured support and the tube
sure. An additional cylindrical FeCrAlY foam, used as static mixer, external skin by means of a wall coefficient hW (Eq. (2b)). Along
is placed in front of the test samples in order to minimize the entry the axial coordinate, the thermal flux at the entry is proportional to
length effects and to ensure a uniform gas flow and temperature the difference between the undisturbed temperature of the fluid
distribution over the tube cross section. The temperature of the gas (Tz=-∞,r ) and the temperature at the zero coordinate (Tz=0,r ) (Eq.
fed to the foams can be controlled by adjusting the temperature of (3a)), while an adiabatic condition is imposed at the exit of the
a second preheating oven, where a coiled section of the feed line is bed (Eq. (3b)). The inlet radial temperature profile was described
located. by a 3rd order fitting polynomial.
He (kf = 0.147 W m−1 K−1 NTP) and N2 (kf = 0.026 W m−1 K−1 According to the model herein presented, the governing param-
NTP) were selected as representative of a gas with high thermal eters are thus the effective radial conductivity (ke,r ) and the
conductivity (such as, e.g., synthesis gas) and a gas of low thermal effective axial conductivity (ke,a ) of the bed, and the wall heat trans-
conductivity (such as, e.g., air), respectively, in order to carry fer coefficient (hW ), similar to what usually assumed for packed
out measurements representative of the operation of the most bed reactors. The estimates of these three parameters have been
common strongly exothermic and endothermic processes. Two obtained in each experimental test by least squares nonlinear
operating modes were explored: (1) heating mode (H), i.e. “cold” regression on the measured axial temperature profiles. In order to
feed gas enters the thermostatic chamber, which is at 300 ◦ C (low reduce statistical correlation between parameter estimates, based
temperature, LT) or 500 ◦ C (high temperature, HT); and (2) cooling on the results of a previous study [5] wherein the effective radial
mode (C), i.e. the feed gas is preheated to a higher temperature than conductivity was experimentally and theoretically investigated, ke,r
the thermostatic chamber, which is at 100 ◦ C (low temperature, LT) was constrained to be greater than the contribution associated with
or 400 ◦ C (300 ◦ C for Al foams) (high temperature, HT). Acronyms static conduction through the solid connected matrix, calculated
composed by three letters will be used to identify the process according to the Lemlich correlation [29].
124 E. Bianchi et al. / Catalysis Today 216 (2013) 121–134

Table 1
Relevant geometric properties of the investigated foam samples.

Foam Material 1−εH 1−εT dc (mm) dp (mm) ts (mm) tn (mm)

A FeCrAlY 6.0% 5.1% 5.09 2.20 0.45 0.85

B FeCrAlY 6.9% 6.3% 1.55 0.86 0.14 0.28
C Al-6101 11.0% 11.0% 2.03 0.86 0.28 0.45
D Al-6101 10.3% 10.3% 3.59 1.83 0.44 0.72
E Al-6101 5.5% 5.5% 2.00 0.81 0.27 0.40

At each iteration of the regression routine, the model PDE was In cooling runs (Fig. 1c and d) the tube skin temperature is lower
solved numerically applying finite differences of second order in the than that of the flowing gas. Therefore, the temperature profile
axial direction (21 nodes) and the orthogonal collocation method increases from the wall to the axis of the bed, while it decreases
in the radial coordinate (7 nodes). from the inlet to the exit of test tube. In these runs the thermal
losses of the feed gas between the preheater and the test section
decrease with increasing the mass flow (because of the decreased
3. Results and discussion residence time in the tube segment located between the gas pre-
heater and the foam bed), so the inner temperature increases with
3.1. Measured temperature profiles increasing gas flow rate. Again, as noted in the case of the heating
experiments, the high thermal conductivity of He (Fig. 1d) makes
Examples of typical axial temperature profiles measured over these thermal losses more evident than in the case of N2 , so that
foam sample B in heating and cooling runs carried out at two dif- at the same process conditions, for He the inlet temperature in the
ferent flow rates with either He or N2 are shown in Fig. 1. foam bed is lower and the radial temperature profile in the foam
In heating runs (Fig. 1a and b) the temperature of the tube skin bed is almost flat at the end of the bed.
(wall profile) is higher than that of the flowing gas, so the tempera- The dependency of the axial and radial temperature profiles on
ture profile decreases from the wall to the center of the bed, while the type of gas (N2 vs. He) is presented in Figs. 2 and 3 for heat-
it increases axially toward the foam exit (axial coordinate = 0.1 m). ing runs at 25 Nl min−1 with a set temperature of the thermostatic
Increasing the flow rate has then the effect to decrease the tem- chamber of 500 ◦ C (high temperature conditions). These tempera-
perature of the gas entering the foam bed (axial coordinate = 0 m), ture profiles refer to both FeCrAlY samples (Figs. 2a–b and 3a–b) and
because of the reduced residence time in the preheating section of Al-based foams (Figs. 2c–f and 3c–f). In the latter case, the effect of
the tube. When He is fed to the system (Fig. 1b), the gas heating the foam microgeometry has also been investigated by comparing
both in the preheater and in the thermostatic chamber is signifi- the performances of samples C and D with different cell size (2 vs.
cantly faster than in case of N2 (Fig. 1a). As a result, at least for the 3.6 mm) but very similar solid fraction (between 10.3 and 11.0%).
case of the lowest investigated inlet flow rates, the temperature of In both Figs. 2 and 3 the axial temperature profiles (measured at
the foam approaches the skin temperature in the second half of the r/R = 0.00, 0.50 and 0.64 and at the reactor external skin) are plot-
reactor. ted on the left, while the radial temperature profiles (measured at

800
Wall Profile Wall Profile

700
Temperature, K

20 Nl/min 20 Nl/min

600

a) 30 Nl/min b)
500
30 Nl/min

1000
30 Nl/min c) d)
20 Nl/min
900
Temperature, K

800 30 Nl/min
20 Nl/min

700
Wall Profile Wall Profile
0.000 0.025 0.050 0.075 0.100 0.000 0.025 0.050 0.075 0.100
Axial Coordinate, m Axial Coordinate, m

Fig. 1. Examples of different thermal profiles measured inside the foam for the sample B, in heating and cooling conditions and different flow rates. Symbols: , wall; ,
9 mm from center;䊉, 7 mm from center;,axis. Colors: black- wall profile; blue solid symbols- 20 Nl min−1 , red open symbols- 30 Nl min−1 . (a) Heating N2 feed with set point
chamber at 500 ◦ C (HHT). (b) Heating He feed with set point chamber at 500 ◦ C (HHT). (c) Cooling N2 feed with set point chamber at 400 ◦ C (CHT). (d) Cooling He feed with
set point chamber at 400 ◦ C (CHT). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
 E. Bianchi et al. / Catalysis Today 216 (2013) 121–134 125

800
Wall Wall

Temperature, K
z/L
700
1.00
r/R
600 0.75
0.64
0.50
0.50 0.25
500
0.00
a) 0.00 b)

800
z/L Wall
Wall
Temperature, K

1.00
700 0.75
0.50
r/R 0.25
600
0.64
0.50
0.00
500 0.00 c) d)
800
Wall z/L Wall
1.00
Temperature, K

700
0.75
r/R 0.50
600 0.64 0.25
0.50 0.00
500 0.00 e) f)
0.00 0.25 0.50 0.75 1.00 0.00 0.25 0.50 0.75 1.00
Dimensionless axial coordinate (z/ L) Dimensionless radial coordinate (r/R)

Fig. 2. Temperature profiles for Sample B (a-b), C (c-d) and D (e-f) in the same heating experimental condition (HHT, TSP = 500 ◦ C), feeding N2 at 25 Nl min−1 . a,c,e): measured
(symbol) and estimated (dotted line) axial temperature profile at different radial position (r/R = 0.00–0.50–0.64–wall). b,d,f): measured (symbol) and calculated (line) radial
temperature profiles at different axial position (z/L = 0.00–0.25–0.50–0.75–1.00).

z/L = 0.00, 0.25, 0.50, 0.75, 1.00) are shown on the right. The exter- conductivity for the He runs are 7.93, 7.42, and 4.09 W m−1 K−1
nal skin temperature is shown in Figs. 2b, d, f and 3b,d,f by a short respectively, virtually the same of those for the N2 runs (7.73, 7.22,
horizontal segment of the right y-axis (at r/R = 1.0). In line with the 3.86 W m−1 K−1 ). The wall heat transfer coefficient, in contrast,
axial temperature profiles, in the case of the heating experiments shows always a broader and positive increase moving from N2 to He
the external skin temperature is always higher than the tempera- runs. In fact, it varies from 95 to 500 W m−2 K−1 for sample A, from
ture measured within the foam bed at the three considered radial 213 to 1086 W m−2 K−1 for sample B, from 180 to 790 W m−2 K−1
coordinates. for sample C, from 132 to 565 W m−2 K−1 for sample D and from
149 to 901 W m−2 K−1 for sample E.
3.2. Estimated heat transfer parameters The heat transfer parameter estimates derived from the cool-
ing experiments are reported in Table 3 for the FeCrAlY samples.
The effective radial (ke,r ) and axial (ke,a ) thermal conductivities Both the effective radial conductivity and the wall heat transfer
as well as the wall heat transfer coefficient (hW ) were estimated coefficients slightly increase in cooling runs, performed howeverat
by regression of each individual run (Section 2.3). In Table 2 the a higher average fluid temperature than the heating experiments.
ranges of these estimates are reported for heating experiments In the case of aluminum open-cell foams, an additional constraint
in N2 and He. The variation within each range accounts for the was to not exceed the solid melting temperature of 600 ◦ C. The
dependency of the parameters on the run temperature and gas effect of the enhanced thermal losses between the preheater and
flow rate. The effective radial conductivity of sample A increases the test section, increased by the lower thermal capacity of He,
from a maximum of 0.5 W m−1 K−1 in heating runs with 35 Nl/min prevented us to record informative temperature profiles in cooling
of N2 to 0.87 W m−1 K−1 with 30 Nl/min of He, and from 0.7 to experiments over such samples.
0.98 W m−1 K−1 for sample B for the same operating conditions. Calculated axial and radial temperature profiles in heating con-
In the case of the aluminum samples (C, D, and E) the effective dition (TSP = 500 ◦ C) for samples B, C and D at 25 Nl min−1 are

Table 2
Ranges of effective conductivities and wall heat transfer coefficient estimated in heating runs, as a result of a range of operating conditions under N2 feed (a) and under He
feed (b) described in Section 2.2. a Data taken from ref. [5].

Sample N2 a He
−1 −1 −2 −1
ke,r (W m K ) hW (W m K ) ke,r (W m−1 K−1 ) hW (W m−2 K−1 )
A 0.38–0.50 89–95 0.76–0.87 468–500
B 0.63–0.70 176–213 0.95–0.98 1041–1086
C 7.73–7.74 150–180 7.92–7.93 752–790
D 7.22–7.22 120–132 7.41–7.42 548–565
E 3.87–3.88 138–149 4.08–4.09 847–901
126 E. Bianchi et al. / Catalysis Today 216 (2013) 121–134

800
z/L Wall
Wall 1.00

Temperature, K
700 0.75
r/R 0.50

600 0.64 0.25

0.50
0.00 a) 0.00 b)
500
800
Wall 1.00 Wall

r/R
0.75
Temperature, K

700 0.50
0.64 0.25
0.50 0.00
600 0.00 z/L

c) d)
500
800
Wall 1.00 Wall
0.75
r/R
Temperature, K

700 0.50
0.64 0.25
0.50 0.00
600 0.00 z/L

e) f)
500
0.00 0.25 0.50 0.75 1.00 0.00 0.25 0.50 0.75 1.00
Dimensionless axial coordinate (z/L) Dimensionless radial coordinate (r/R)

Fig. 3. Temperature profiles for Sample B (a-b), C (c-d) and D (e-f) in the same heating experimental condition (HHT, TSP = 500 ◦ C), feeding He at 25 Nl min−1 . a, c, e): measured
(symbol) and estimated (dotted line) axial temperature profile at different radial position (r/R = 0.00–0.50–0.64–wall). b, d, f): measured (symbol) and calculated (line) radial
temperature profile at different axial position (z/L = 0.00–0.25–0.50–0.75–1.00).

presented for N2 (Fig. 2) and He (Fig. 3) runs. The quality of the fluid properties on the effective radial conductivity. In Section 3.1
fit can be evaluated by noting the good superimposition of the we have already shown that the gas heating is significantly faster
experimental data points with the calculated profiles. when He is fed to the system instead of N2 . In the case of FeCrAlY
foams, the higher thermal conductivity of He with respect to N2
3.2.1. Effective radial thermal conductivity (about +500%) positively affects the effective radial conductivity of
As for classical packed beds of particles [30,31], the effective the foams, resulting in increments of about +40% (see ke,r for sam-
radial conductivity of open-cell foams is given by the sum of three ples A and B in Table 2). Such an effect is less evident however
different contributions: (i) the stagnant conductivity of the system, in the case of Al-based foams where the fluid contribution to the
which in turn depends primarily on the effective conductivity of the stagnant effective conductivity is almost negligible (+3–5%) in view
bulk solid material; (ii) the mechanical dispersion, that is the con- of the important conductive contribution of the solid matrix. The
vective contribution associated with the fluid flow; (iii) the thermal dominant role of the stagnant conductivity of the solid phase over
radiation, that becomes effective when the local temperatures are that of the gas phase is further confirmed observing that in the case
high enough. of poorly conductive metallic alloys, even when working with a
By analysis of the thermal behavior of foams made of different highly conductive gas such as He, it is not possible to obtain flat
materials and with different micro-geometries, the influence of the radial temperature profiles (Fig. 3b) comparable to those observed
thermal conductivity of the solid phase was investigated in detail for the case of highly conductive foams (Fig. 2.d and f), even when
in our previous work [5]. As specified above, it has been herein feeding a poorly conductive gas such as N2 .
constrained accordingly in the regression of data. The estimates of the radial effective conductivity presented in
Instead, the comparative analysis of the heat transfer experi- Table 2 exhibit variations not only depending on the intrinsic solid
ments carried out with N2 and He offers insight into the effect of the and gas properties, but also on the specific process conditions. This

Table 3
Ranges of effective conductivities and wall heat transfer coefficient estimated in heating runs, as a result of a range of operating conditions under N2 feed (a) and under He
feed (b) described in Section 2.2. a Data taken from ref. [5].

Sample N2 a He
−1 −1 −2 −1
ke,r (W m K ) hW (W m K ) ke,r (W m−1 K−1 ) hW (W m−2 K−1 )
A 0.54–0.95 84–111 0.88–0.95 474–512
B 0.67–0.97 237–245 1.05–1.11 1071–1140
C 7.74–7.74 223–260 – –
D 7.22–7.22 151–158 – –
E 3.87–3.88 186–242 – –
 E. Bianchi et al. / Catalysis Today 216 (2013) 121–134 127

is the reason why we present data in the form of intervals instead 900
of exact values. We have already discussed in [5] that, on moving HHT
from runs at lower temperature to runs at higher temperature, the HLT
effective conductivity grows because both the solid and gas conduc- CHT Sample C: Al
CLT
tivities are directly proportional to temperature, and also radiation He
effects become more relevant. Also, for a specific heating or cooling 600

-1
hW , Wm K
condition a variation of the parameter estimates is observed upon Sample A: FeCrAlY

-2
changing the flow rate, denoting some convective contribution, but
this effect seems to be less important than the temperature effect.
Worth of notice is that the intervals in Table 2a and b are wider (on
a relative basis) for FeCrAlY foams where, because of a low solid Sample C: Al
300
conductivity, heat transfer contributions due to fluid convection Sample A: FeCrAlY
and radiation are more important. No significant differences can be
observed in the extent of the intervals estimated for the tests car- N2
ried out in the presence of He and those carried out with N2 . This
evidence is consistent with the nature of the heat transfer mecha- 0
nisms relevant in these intervals, namely convection and radiation, 15 20 25 30 35
which do not depend on the nature of the gas. Flow Rate, Nlmin -1

3.2.2. Wall heat transfer coefficient Fig. 4. Wall heat transfer coefficient in all the operative conditions with N2 feed in
In our previous study [5] we have shown that the temperature function of flow rate (Section 2.3.2). HHT: heating high temperature; HLT: heating
of the gas at the outlet never reaches the skin temperature of the low temperature; CHT: cooling high temperature. CLT: cooling low temperature.
Sample A (filled symbols): FeCrY, εT ∼95%, 10 nominal PPI; Sample C (hollow sym-
test tube. This is evident even in Figs. 2 and 3 herein reported: a bols): Al-alloy, εT ∼89%, 40 nominal PPI.
thermal discontinuity exists at the interface between the foams
and the tube wall. The wall heat transfer coefficient accounts for
the resistance to heat transfer in this area. It is worth noticing that, clearly inversely proportional to the gas thermal conductivity. For
especially for the samples with high effective radial conductivities example, the highest wall heat transfer coefficients are calculated
(samples C and D) fed with N2 (Fig. 2d and f), almost the whole radial for sample B in He runs, while five fold lower values of hw are
temperature gradient is concentrated at the wall. This means that estimated for the same foam sample in N2 runs.
the wall coefficient is the controlling heat transfer parameter for The ratio between the estimates of the wall heat transfer coef-
some of the explored conditions. A detailed analysis of the influence ficient measured over the same foams in experiments carried out
of the process conditions and of the properties of both the solid and with N2 first and then with He is almost constant, and close to about
the fluid phase is carried out in the following paragraphs. 0.2 regardless of the sample. Interestingly, this roughly corresponds
to the ratio of the thermal conductivities of the pure gases N2 and
3.2.2.1. Influence of the properties of the solid phase. Data in He. This confirms that, once the foam geometry is fixed (and in
Tables 2 and 3 show that, contrary to what happens for the effective particular its cell diameter) and the mechanical coupling between
conductivities of the foam bed, the wall heat transfer coefficient is the foam and the test tube is set, for given process conditions the
independent of the material of the investigated foam, and thus of wall heat transfer coefficient is directly proportional to the thermal
its intrinsic conductivity. In analogy to what has been found by conductivity of the process gas.
some of us in the case of metallic honeycomb monoliths [32], this
suggests that the additional resistance to the heat transfer is phys- 3.2.2.3. Influence of the process conditions. As already commented
ically located in an area where the solid foam struts are absent, i.e., for the case of the effective conductivities, the estimates of the
in the “gap” between the foam and the tube internal wall, filled by wall heat transfer coefficient presented in Tables 2 and 3 not only
the fluid. In this regard it is worth noticing that, despite the tested depend on the properties of the foam cell diameter and on the fluid
foams were cut by electro-erosion to have a sharp cut, the match thermal conductivity, but also exhibit slight variations upon chang-
with the tube cannot be uniform because of the intrinsic macro ing the process conditions. In particular, hw shows a dependence
porosity of the solid foam. Thus, the “gap” is likely not constant, on both the gas flow rate and the average run temperature.
but of irregular nature and varying thickness. Concerning the dependency on the gas flow rate, a weak, pos-
Also, in the investigated foam porosity range (89–95%), no evi- itive dependency is observed in Fig. 4 both in the case of N2 and
dence is found for a dependency of the wall heat transfer coefficient in the case of He. More evident is the dependency of the wall heat
on the foam void fraction (compare for example samples C and E in transfer coefficient on the average temperature of the run. Such an
Tables 2 and 3). On the contrary, the foam cell size (and pore size) effect can be investigated by comparing the estimates of the wall
reversely affects the value of the wall heat transfer coefficient. Both heat transfer coefficient found in heating and cooling experiments:
in the case of N2 and He runs, in fact, Tables 2 and 3 show that the due to the design of our runs and of our experimental set-up, in fact,
lowest estimates of the wall heat transfer coefficients are obtained the average temperature in the foam bed during cooling experi-
for samples A and D, i.e. those characterized by the biggest cells, ments is higher than during heating experiments. The data shown
while the highest value is found for foam sample B, i.e. the sample in Fig. 4 for foam sample C clearly indicate that the estimates of
characterized by the smallest cell size. the wall heat transfer coefficient in cooling runs are higher com-
pared to those estimated from heating experiments. This effect can
3.2.2.2. Influence of the properties of the fluid phase. The be explained considering the temperature dependency of the gas
comparison between the axial and radial temperature pro- thermal conductivity, which increases with higher temperatures.
files measured in the experiments with N2 (Fig. 2) and
with He (Fig. 3), and the comparison of data reported in 3.2.3. Reliability of the model parameters estimates
Tables 2 and 3 enable to assess the effect of the gas thermal The simplifying assumptions in the regression model affect the
conductivity on the wall heat transfer coefficient. The thermal estimated values of the thermal model parameters. Despite the dif-
discontinuity in the gap between the foam and the tube wall is ficulty to account for the impact of some assumptions, such as, e.g.
128 E. Bianchi et al. / Catalysis Today 216 (2013) 121–134

constant thermal properties or local thermal equilibrium between 100

the two phases, the reliability of the fitting parameters is mainly N2: runs at higher temperature
dependent on the errors inherent in the measured temperature N2: runs at lower temperature
profiles. For this reason replicated sets of data were collected in He: runs at higher temperature
a
the same operative condition for each sample, in order to deter- Correlation from this work
mine the uncertainty connected with the positioning of the probes,
+35%

Nu w,dc
which was found to be less than 2%. The error associated with the
thermocouple measurement accuracy is instead only 0.4% of the 10
measured temperature. However in the experimental set-up it was
not possible to measure directly the temperature of the inner side of
the tube wall, but only the external tube skin temperature instead. -35%
Nevertheless, to avoid manipulation of experimental data, this tem-
perature profile was used as such in the estimation of the model
parameters. The potential error associated with this approach is
discussed in the following. The bias resulting from neglecting the 1
1 10 100
resistance of the tight coupling between the thermocouple and the
Re dc
external tube wall is difficult to quantify, but one can estimate
the thermal resistance connected with the radial heat conduction
Fig. 5. For all the samples tested, both in N2 (blue symbols) and He (green symbols)
inside the stainless steel tube. Considering the tube wall thickness feed, the dimensionless wall coefficient is reported in function of Reynolds number.
of 2 mm and the conductivity of stainless steel of 15 W m−1 K−1 , the The Nusselt number is based on cell diameter and fluid conductivity. : Sample
conductive resistance in the tube is roughly 1.25 × 10−4 m2 K W−1 . A;: Sample B; :Sample C;夽: Sample D;䊉: Sample E; a Correlation for wall heat
This additional resistance results basically in an underestimation transfer coefficient: Eq. (7). (For interpretation of the references to colour in this
figure legend, the reader is referred to the web version of this article.)
of the wall heat transfer coefficient. The relative weight of this
underestimation is proportional to the wall heat transfer coeffi-
cient, which actually results in larger errors for the He runs than for 3.3. Proposed correlation for the wall heat transfer coefficient
the N2 runs. A minimum error of 1% is obtained for the wall coef-
ficient of 84 W m−2 K−1 (sample A, cooling with N2 ; Table 3) and a 3.3.1. Development of the correlation
maximum error of 14% for the wall coefficient of 1140 W m−2 K−1 According to the results presented above, a correlation for the
(sample B, cooling with He; Table 3). Thus, the minimum relative wall heat transfer coefficient containing both a conductive (hW,0 )
error on the real value of the estimated parameters is the positive and a convective (hW,Conv ) term is proposed as follows:
sum of the 95%-confidence interval regression and the additional hW = hW,0 + hW,Conv. ; (4)
systematic error resulting from the above-described way the tube 
 

temperature is measured. This results in an error range of 2–10% Conductive Contribution Convection

for the estimated effective radial conductivity, while the range is The dimensionless form of the conductive contribution to the
5–29% for the wall heat transfer coefficient. wall heat transfer coefficient, NuW,0 , is assumed to be a constant
It is worth stressing that the narrow error range on ke,r is mainly whose value can be determined at given foam cell size and fluid
a consequence of having constrained the contribution associated thermal conductivity:
with the static conductivity through the solid connected matrix,
hW,0 dC
that, as discussed above, plays a dominant role in most of the ther- hW,0 : NuW,0 = = A; (5)
mal runs. Such a constraint also minimizes correlation between kf
parameters, thus reducing the error range of hw , similarly to what The dimensionless form of the convective contribution to the
reported by Westerterp and co-workers [33] when lumping ke,r and wall heat transfer coefficient, NuW,Conv , is instead a function of the
hw in a single overall coefficient Uov for the analysis of heat transfer Reynolds number according to:
properties of packed beds.
hW,Conv. dC
When modeling continuous, highly conductive supports such as hW,Conv. : NuW,Conv. = = BReCdc ; (6)
kf
the open-cell metal foams investigated in this work, the effective
axial coefficient cannot be “a priori” neglected in the energy bal- As shown in Eqs. (5) and (6) on the basis of the dependency of the
ance. Accordingly, our model takes into account the contribution wall heat transfer coefficient on the foam cell diameter, the latter
of the axial conductivity in the energy balance, as shown in Sec- has been chosen as the characteristic length for the evaluation of the
tion 2.3. The estimates of ke,a , which are close to (but lower than) conductive and convective Nusselt numbers and of the Reynolds
the corresponding radial conductivities, suffer however from the number, in analogy to the pellet diameter for packed beds.
inaccuracies associated with the simplified boundary condition at The constants A, B and C in Eqs. (5) and (6) represent the influ-
the tube inlet, which in the model solutions are largely absorbed ence of the different heat transfer mechanisms on the wall heat
in the contribution of the effective axial conductivity. On the other transfer coefficient, i.e., the conduction in the fluid phase and the
hand ke,a estimates are basically uncorrelated with both the radial forced convection due to mechanical dispersion of fluid flowing
effective conductivity and the wall heat transfer coefficient. For through the porous foam structure. Their estimates have been
example, representative correlation coefficients between the fit- obtained by least-square nonlinear global regression on the axial
ting parameters are 0.96 between ke,r and hw , 0.40 between ke,a temperature profiles collected in the 111 available experimental
and ke,r and 0.28 between ke,a and hw (sample B, cooling run at runs. The resulting correlation is given in the following:
30 Nl min−1 of N2 ). In fact the estimates of ke,r and hw are almost 
kf
unsensitive to ke,a : forcing the effective axial conductivity to be hW = 7.18 + 0.029Re0.8
d ; 4 < RedC < 255 (7)
equal either to the effective radial conductivity or to the gas con- dC C

ductivity, which differ by 1 to 2 orders of magnitude, affected the In Fig. 5 the experimental estimates of the Nusselt number
estimates of the other two fitting parameters by less than 10%. For (NuW = NuW,0 + NuW,Conv. ) and the corresponding model predictions
these reasons, no specific analysis of this parameter was performed are plotted versus the Reynolds number in a bilogarithmic plot. The
here. average relative error is 16%.
 E. Bianchi et al. / Catalysis Today 216 (2013) 121–134 129

150 400

300

-1
hW , Wm K
100

-2
-1
hW, Wm K
-2

200

50 a
PU foam, Edouard et al. α -Al2O3 foam, correlation of Peng et al.
a

a
PU foam, correlation from this work α -Al2O3 foam, correlation from this work
b
b
β -SiC foam, Edouard et al.
b
100
β -SiC foam, correlation from this work 100 200 300 400 500 600 700 800
0 Redc
0 10 20 30 40
Fig. 7. Wall heat transfer coefficients estimated with the correlation developed in
Redc
this work and literature correlation in function of Reynolds number on the cell diam-
eter. Fluid phase is air: gas properties are estimated at 300 ◦ C. a Correlations are
Fig. 6. Wall heat transfer coefficients estimated with the correlation developed in applied in the range of Reynolds number validity of Peng et al. [24] and with the
this work and literature data in function of Reynolds number on the cell diameter. same ␣-Al2 O3 foam used by authors: dp of 0.826 mm and porosity equal to 87.4%.
Fluid phase is air: gas properties are estimated at 100 ◦ C. a Wall heat transfer coeffi-
cient reported in Edouard et al. [25] for polyurethane (PU) foam: dC of 2.8 mm and
εH of 92%. b Wall heat transfer coefficient reported in Edouard et al. [25] for ␤-SiC
foam: dC of 2.3 mm and εH of 88%. 3.3.3. Comparison with other literature correlations
As mentioned in the Section 1, to date the only correlation avail-
able in the literature to describe the wall heat transfer coefficient in
tubular reactors loaded with open-cell foams was proposed by Peng
In the investigated range of process conditions (4 < RedC < and Richardson [24] (Eq. (8)) on the basis of the results collected
255) the main contribution to the wall heat transfer coefficient, with air for a single 30 PPI ␣-Al2 O3 foam (dp = 0.826 mm, 152.4 mm
accounting for over 75% of its value, is given by the conduction long) inserted into a tube of 50.8 mm I.D.
through the thin layer of gas located at the interface between the
hW
foam periphery and the inner tube wall. This result is fully con-  = 0.0692Re0.48
S ; 2.86 < ReS < 14.4 (8)
sistent with the experimental evidence discussed in Section 3.2.2 SV kf
showing a direct correspondence between the ratio of the wall heat
According to Eq. (8) the mechanism for wall heat transfer is
transfer coefficients measured in N2 and in He and the ratio of the
purely convective. In order to compare the predictions of Eqs. (7)
thermal conductivities of the two gases.
and (8) in their range of validity, it is necessary to express RedC as
a function of ReS . Under the assumption that dp ≈ 0.55dC , which
is the approximate correlation between the equatorial diameter
3.3.2. Comparison with other literature data (dC ) and the window diameter (dp ) of a simplified truncated octa-
As mentioned in the Section 1, wall heat transfer coefficients hedron bubble foam [34], the following correlation can easily be
for polyurethane (PU, dC = 2.8 mm, εH = 92%) and silicon carbide (␤- derived:
SiC, dC = 2.3 mm, εH = 88%) foams loaded in tubular reactors have
ReS 4εH
been recently estimated by Edouard et al. [25] from data collected RedC = (9)
0.55 (1 − εH )
in heating experiments similar to those carried out in this work.
In such tests, the wall temperature of the reactor of 76.5 mm I.D, Considering that the hydraulic porosity of the 30 PPI ␣-Al2 O3
loaded with foams of a length of 230 mm, was set to 100 ◦ C and air foam tested by Peng et al. is 87.4% [24], the range of validity of Eq.
was fed to the foam bed with velocities from 0.018 to 0.32 m s−1 , (8) can be rewritten as 144 < RedC < 726, an interval only partly
corresponding to RedC in the range of 1.7–38.5. superimposed (at the lower Reynolds number bound) with the one
The comparison between data reported by Edouard et al. [25] investigated in this paper.
and the predictions of equation (7) is shown in Fig. 6. Even though The comparison between the predictions of Eqs. (7) and
the foams tested by Edouard et al. [25] are very different from those (8) applied to the 30 PPI ␣-Al2 O3 foam tested by Peng et al.
tested in this paper in terms of material (polymer vs. ceramic vs. (SV = 3.24 × 104 m−1 [24]) under air feed is reported in Fig. 7. Both
metal), a good match exists between the literature data and the cor- correlations predict the growth of the wall heat transfer coefficient
relation proposed in this work, especially in the case of the stiff SiC with increasing Reynolds numbers, and the predictions of the two
foams. Remarkably, the data by Edouard et al. [25] were measured models converge at high Reynolds numbers. On the contrary, when
at low RedC , that is in a region where, according to the results shown the gas flow rate decreases, the purely convective correlation pro-
in this work, the wall heat transfer coefficient is almost entirely con- posed in [24] predicts hW significantly lower than the correlation
trolled by the conductive term. This is confirmed both by the trends herein derived. It is worth noticing, however, that these compar-
shown by the model predictions, which are almost insensitive to isons have been obtained extrapolating Eqs. (7) and (8) outside the
RedC , and also by the trend shown by the experimental data, that range of conditions in which they have been derived, i.e. for the case
are substantially flat. of Eq. (7) not above RedC = 255 and for Eq. (8) not below RedC = 144.
This result gives independent evidence that the newly derived Two conclusions can be drawn from this comparison. First, in the
correlation, Eq. (7), can correctly describe the wall heat transfer model reported in [24] the conductive contribution to the wall heat
coefficient, especially when the conductive mechanism controls transfer coefficient is likely somehow masked within the “convec-
the coefficient. tive” term, which in fact lumps both conductive and convective
130 E. Bianchi et al. / Catalysis Today 216 (2013) 121–134

contributions. The second consideration deals with the numerical

value of the conductive term hidden in the lumped term. Indeed,
this term seems to be lower than predicted by Eq. (7). This can be
explained considering that the coupling between the ceramic foam
tested by Peng and Richardson [24] and the test tube was likely
worse than in the case of the metallic foams used in this work. It is
expected, in fact, that, mainly due to the brittleness of the ceramic
foams and their lower thermal expansion coefficient with respect to
the metallic foams, the contact between ceramic foams and the test
tube cannot be as tight as in the case of the metal foams examined
in this work.

4. Numerical 3D simulation of the fluid flow and heat

transport

The wall heat transfer coefficients calculated on the basis of

our experimental results show only a weak dependency on the
Reynolds number. This is at first an unexpected result, as it is
in contrast to the stronger dependency found in conventional
packed beds [35]. In order to identify the physical reasons for such
a discrepancy, the mechanisms responsible for the heat transfer
between the tube wall and the solid foam must be investigated in
more detail. For this, a finite volume analysis (FVA) was performed
for the conditions of selected experimental runs. The purpose of
Fig. 8. (a) Metal foam segment (solid phase from reconstruction of ␮CT data) for
such a numerical study at this point is to investigate numerically simulation, (b) computational domain of the foam segment including fluid phase
the wall heat transfer under well-defined and verifiable condi- mesh.
tions in order to support and explain the experimental findings.
To study both the effect of different gas conductivities and the
Nusselt dependency on the Reynolds number, we chose to per- converged to a stable solution. The final mesh consists of about 5.6
form first simulations for the same volumetric inlet flows of N2 million elements, 90% of which are hexahedra.
and He for one experimental condition, i.e. heating of 30 Nl/min
feed gas to a nominal set point of 500 ◦ C. In fact, on changing the 4.2. Equations and simulated conditions
gas, both the gas conductivity and the Reynolds number (mainly
because of the different density) are modified. A systematic inves- The simulation model implemented in OpenFOAM combines
tigation of a range of different flow conditions for different gases both heat conduction in the solid region and thermal flow through
and temperatures in terms of 3D simulations is an extensive work the fluid region. A thermal energy equation is thus solved in the
on its own, which therefore will be carried out in a future dedicated solid region, while mass, momentum, and energy conservation
study. equations are solved in the fluid phase. As the Reynolds numbers
The mesh generator and the solver for the FVA were imple- (calculated based on the cell diameter) of all our experiments are in
mented in the open source environment OpenFOAM v2.1.1 [36], the range of the Forchheimer and post-Forchheimer laminar regime
a general library of C++ classes for numerical simulation mainly (RedC < 300 [13]), the set of steady-state laminar Navier–Stokes
used for computational fluid dynamics (CFD) simulations. equations are used to model the flow. Moreover, it is assumed that
the fluid can be described by the equation of state for an ideal gas
(P = R̂T ).
4.1. Geometry The mass conservation equation for the fluid is given by:

The 3D mesh for the simulations was generated from X-ray ∇ u = 0; (10)
micro-computed tomography images of a scanned foam sample
where  is the fluid density and u is the fluid velocity. The con-
(details in ref. [5]), that had been investigated in the experimen-
servation of momentum, neglecting buoyancy effects, is given by:
tal set-up. In our previous contribution [5] we have investigated
the heat conduction in different foam samples of varying pore and

strut size, porosity, and material. In the present contribution, in ∇ u u = −∇ P + ∇ ; (11)
contrast, the phenomena of interest are directly connected to the
role of the fluid phase (i.e., velocity and conductivity). Therefore, where P is the pressure,  is the dynamic viscosity of the fluid and
in the simulation studies herein discussed only one foam geometry  is the viscous stress tensor, defined for a Newtonian fluid as:
was considered. The chosen foam is sample D, as its relatively high   T  2  
solid fraction and cell diameter allow for a reasonable mesh gen- = ∇ u + ∇ u −  ∇u
 I ; (12)
3
eration. The subdomain consists of one fourth of the experimental
tube, containing a segment of the cylindrical foam sample D of 5 cm Neglecting gravitational potential energy, internal heat gener-
length. The resolution of the micro-computed tomography scan is ation and viscous forces due to the low viscosity of the simulated
close to 50 ␮m (Fig. 8a). The selected subdomain does not include fluid (gas), the steady-state equation of energy conservation for the
the through holes for the thermocouples except for (one fourth of) fluid phase can be written in enthalpy form as:
the center hole. Both the solid and fluid phase were meshed (Fig. 8b,    
1 kf
solid in red and fluid in blue). The simulation results were checked ∇ h + |u|2 u = ∇ ∇h ; (13)
2 cp
to be mesh independent by refining the mesh grid until the results
 E. Bianchi et al. / Catalysis Today 216 (2013) 121–134 131

where h is the enthalpy per unit of mass, kf is the thermal con-

ductivity of the fluid and cp the specific heat capacity at constant
pressure.
The energy equation for the solid, without heat generation, is:

∇ (ks ∇ T ) = 0; (14)

where ks is the thermal conductivity of the solid.

The purpose of the numerical analysis is a direct comparison
with experimental data. For this reason, the boundary conditions
are fixed in a manner suitable to reproduce the experimental ones.
The selected experiments correspond to 30 Nl/min of inlet flow
of N2 or He, heated in the tube at the nominal set temperature
of 500 ◦ C. The thermodynamic properties of the solid phase (alu-
minum) are assumed to be constant. Also, the thermodynamic and
transport properties of the fluid phase are assumed to be constant
and calculated based on the average temperature of the simulated
run for either N2 or He.
As only one fourth of the cylindrical tube was reconstructed
for simulation, symmetric boundary conditions are imposed for all
the variables in correspondence with the cut faces parallel to the
tube axis The reason why symmetric and not periodic boundary
Fig. 9. (a) Axial slice (normal to x-axis) of fluid and solid mesh (Fig. 8) show-
conditions are usually applied for meshes generated form X-ray ing velocity (arrows) and fluid temperature (color scale) distribution for N2 feed
micro-computed tomography [37] is that the opposite mesh faces of 30 Nl/min with adiabatic boundary condition for the solid at the wall. (b)
of the reconstructed sample are not identical. For the same simulation, normal velocity in a section transversal to the flow
A constant and uniform fluid velocity is set at the inlet, which direction (Z).

seems a reasonable assumption because of a flow distributor placed

in front of the real test sample. So, a typical combination of Dirichlet
and Neumann boundary conditions for velocity and pressure are simulations, namely an adiabatic boundary condition for the solid
fixed at the inlet and outlet (normal to the Z axis) of the volume. foam in contact with the tube wall. This approach enables a qual-
No-slip conditions are then imposed for the velocity at the interface itative study, using the same mesh as before but simply changing
of the tube, central hole and solid, respectively, and a zero gradient the boundary conditions. To simulate the discontinuity between
for the pressure field at the same location. the solid foam phase and the wall (i.e., the gap) and at the same
The thermal boundary condition between solid and fluid is a time forcing the incoming heat flux to pass through the fluid
conjugated heat transfer, i.e. continuity of the heat flux and equal phase, the second case considers applying the above described
temperature at the fluid–solid interface [38]. This means that the adiabatic boundary condition at the interface between solid and
energy equations for solid and fluid are coupled to allow heat tube wall.
transfer between the two phases. The thermal coupling is realized It is worth emphasizing that even though the adiabatic assump-
setting automatically a fixed temperature value or a fixed gradient, tion at the solid-wall interface could be regarded as a standard
depending on the direction of the heat flux. Despite the fact that approach in the simulation of packed beds, it is instead not obvi-
in catalytic applications an additional coating on the surface of the ous for the case of consolidated structures like open-cell foams.
open-cell foam will be present, in these pure thermal simulations In fact, most of the heat transfer works in this area assume
the oxidation of the Al alloy is neglected, because the passivation a physical continuity between the solid phase and the wall
layer on the exposed aluminum surface is extremely thin, usually [20,22].
only about 4 nm in thickness [39]. At the inlet, the fluid tempera-
ture is set to be the same as measured in the experimental runs, 4.3. Results of numerical analysis
while the outlet of the domain is set adiabatic, as is the interface
with the axial hole. The simulations were run on the “lima cluster” of the High
At the interface between the mesh and the tube wall, two dif- Performance Computing Center of the University of Erlangen
ferent conditions have been considered. In the first case, the same (Regionales Rechen Zentrum Erlangen, RRZE). The numerical sim-
wall temperature profile as collected in the experimental runs was ulations were stopped after approximately 104 iterations, when
applied to both the solid and the fluid phase. This case represents the residual errors of all the parameters were found to be below the
a perfect contact between the foam and the tube wall. However, threshold value of 10−9 . Examples for a qualitative analysis of the
in the experiments there is most likely a small gap between the simulation results are reported in Fig. 9. In Fig. 9a the temperature
foam sample and the tube at the wall. Therefore, the real experi- profile for N2 feed is shown as a colored contour plot while the
mental conditions should actually result in a solid fully embedded velocity is depicted as vectors (arrows) in a section parallel to
into the fluid. Generating a corresponding mesh which matches the the tube axis. In Fig. 9b the magnitude of the velocity component
experimental gap distance, however, would be a “trial and error” parallel to the tube axis (Z) is presented in a normal section of the
approach, as in fact the thickness of this gap is unknown. Also, the tube.
gap may vary in different experimental runs owing to the corre- Most interesting is to observe the average axial profiles of fluid
sponding thermal expansion of the sample and the tube. Moreover, and solid temperatures. As the mesh is only one fourth of the
due to the global dimension of the reconstructed domain, it would real sample, and the exact locations where the experimental tem-
be difficult to generate a grid with a sufficiently high resolution in perature profiles were collected (the through holes) are excluded
the gap region. from the reconstructed domain, the comparison of the average
For these reasons, in addition to the above-mentioned sce- temperature profile is more appropriate than focusing on local tem-
nario of a perfect contact between the solid foam and the tube, peratures. The average profiles of FVA are based on the section area
another, opposite limiting case has been investigated in the for the solid temperature and on the bulk temperature for the fluid.
132 E. Bianchi et al. / Catalysis Today 216 (2013) 121–134

the total heat flux at the wall, the fraction entering the system
through the solid phase was found to be around 94% for N2 and
92% for He. Interestingly, the fraction of solid in direct contact
with the wall is only around 11.7% (see Fig. 8b). This means that
the prevailing heat flux path is through the solid phase and the
fluid is mainly heated by the interphase heat transfer. This result
is in agreement with the conclusion of the theoretical study of
Vafai and Lee [40] about the heat transfer in porous media. The
authors found that in the case of highly conductive supports, high
interphase coefficients and perfect contact with the wall, the con-
trolling resistance for radial heat transfer is the effective solid
conductivity.
However, this case is not representative of our experimental
data. As the experimental profiles are below the ones predicted by
the simulations it is reasonable to assume the presence of a resis-
tance at the wall, which could be responsible for this difference.
As a limiting case, this resistance was introduced in our simula-
tions assuming an adiabatic condition for the solid at the wall. The
temperature profiles calculated for this new condition, while all
the other assumptions are the same as in the previous simula-
tions, are presented in Fig. 10. The new profiles now underestimate
the experimental data, however they are closer than in the pre-
vious case (better visible in Fig. 10a, N2 ). Looking in detail into
the solid and fluid temperature profiles one can observe that at
the entrance again the solid temperature is higher than that of the
fluid. This means that even when not physically connected with
the wall, the main path for heat transfer is through the solid phase.
In fact, close to the wall heat enters the system through the fluid,
which is at higher temperature than the solid itself. However, as
soon as it is in contact with the much more conductive solid, the
heat flux shifts mainly through the solid phase and again from
the solid to the fluid farther away from the wall. This proves our
hypothesis that the real experimental configuration is in between
the two limiting cases of perfect contact and of adiabatic bound-
ary conditions, respectively. Accordingly, the thickness of the gap
Fig. 10. Axial average temperature profiles for heating condition at 500 ◦ C external becomes a fitting parameter to match the experimental data. A
set point and 30 Nl/min: (a) N2 ; (b) He. Experimental wall temperature and average preliminary study to estimate the thickness of the real gap was per-
of collected profiles (symbols) are compared with the average solid (continuous line) formed with a new mesh featuring an external radius of 14.05 mm.
and fluid (dotted line) temperature for FV simulations with both perfect contact and
In this new configuration the solid is completely embedded in
adiabatic boundary condition for the solid at the wall.
the fluid mesh. The temperature profiles corresponding to sim-
ulations with this mesh are slightly higher than the scenario of
solid with adiabatic conditions at the wall (Fig. 10), but not yet
enough to provide a suitable fit with the experimental data, sug-
The bulk temperature (Tb ) is defined as the average temperature gesting the necessity of assuming even a smaller gap. This seems
weighted by the mass flow (un ) crossing a section (A) normal to realistic and consistent with the possible physical coupling (foam
the axial coordinate (Z): sample-tube, with some direct solid-to-solid contact points) of the
  experimental set-up. This order of magnitude estimation implies

Tun dA  also that heat transfer through the fluid phase takes place across

  the stagnant sublayer close to the wall, and therefore is mainly

Tb =  
A  ; (15)
Z  of conductive nature. In fact, while a larger annulus around the
un dA  foam sample could be a preferential way for the flow through the
 tube, a micrometric distance would instead result in a wall contact
A Z
resistance directly proportional to the ratio between gap thickness
Because the velocity distribution inside the experimental tube and fluid conductivity, and only weakly dependent on the velocity
is not available, the area averaged temperature of the collected pro- regime.
files is used. In any case, according to Figs. 2–3, the radial profile for A further analysis was carried out to calculate the heat transfer
aluminum samples is flat. coefficient. While the main point of interest is the wall coefficient,
In Fig. 10 the experimental temperature profiles (symbols) are it is difficult to isolate this parameter in the post processing of
compared with the results of the numerical analysis for both the numerical analysis. The overall heat transfer coefficient for
N2 (Fig. 10a) and He (Fig. 10b). In both cases, the numerical each simulation is much more easily accessible. In any case, the
results for the assumption of perfect contact between the solid global coefficient is dependent on the effective radial conductiv-
phase and the tube wall widely overestimate the experimental ity and on the wall heat transfer coefficient [41]. As the former
data. The average solid temperature results (continuous line) are term is independent of the Reynolds number (see Tables 2 and 3
almost parallel to the wall profile, while the fluid profile (dashed for Sample D), the trend exhibited by the overall coefficient
line) sharply increases from the inlet (minimum of axial coor- should be proportional to the heat transfer coefficient close to the
dinate) value until it overlaps with the solid profile. Analyzing wall.
 E. Bianchi et al. / Catalysis Today 216 (2013) 121–134 133

Table 4 literature on porous media and randomly packed beds made of

Overall heat transfer coefficients, Nusselt and Reynolds numbers as calculated from
particles [18,31,35]. Despite the fact that the general physical
the numerical simulation with adiabatic boundary condition for the solid phase in
contact with the wall. The modeled cases correspond to experimental heating runs, processes in open-cell foam beds are similar, however, a main dif-
with nominal T set point of 500 ◦ C. The dimensionless numbers are based on the fluid ference exists in the relative importance of the individual transport
properties at the average temperature and on the tube diameter (D) as characteristic processes in consolidated or unconsolidated porous structures. At
length. first, the void fraction of a particle random packing is not constant in
Fluid Flow rate (Nl min−1 ) ReD (–) U (W m−2 K−1 ) NuD (–) the radial direction, but it strongly increases approaching the tube
wall, with a more or less pronounced radial distribution dependent
He 30 116 378 38
N2 30 892 91 55 on the ratio of the particle to tube diameter [42]. For this reason
N2 3 89.2 74 45 the region close to the wall experiences a different hydrodynamic
condition, with higher average velocity. In contrast to this the void
fraction of the open-cell foam support is a design parameter and is
The overall average heat transfer coefficient (U, W m−2 K−1 ) can a constant value over the radial coordinate. Moreover, while both
be calculated as: open-cell foams and classical packings of spheres exhibit point con-

tacts with the wall, the conductive path in the radial direction of
qdS ⁄(T − Tb )
W the particle packed bed reactor is extremely weak, because it relies
only on point contacts between particles. In continuous supports
U=
S
 ; (16)
such as open-cell foams, on the other hand, heat conduction inside
dS the continuous solid matrix is extremely efficient and the solid-wall
S coupling exerts a substantial impact which is also relevant farther
from the wall. These features are responsible for the different role
where q is the heat flux (W/m2 ), dS the elemental face area on the of the convective contribution to the wall heat transfer coefficient:
wall and TW the specific tube wall temperature (varying along the it provides the main contribution to the wall heat transfer coeffi-
axial coordinate, as in the experiments). In the end, the average cient in the case of particle packings [31], whereas open-cell foams
overall Nusselt number can be easily calculated from Eq. (16), using exhibit a less pronounced dependency on the flow rate. This con-
the tube diameter as the characteristic length. sideration, shown in this work for single-phase flow, applies also
UD in case of two-phase flow [26,27].
NuD = ; (17)
kf

In Table 4 the overall heat transfer coefficients and Nusselt num- 6. Conclusions
bers are presented as a function of the Reynolds number (based on
the tube diameter) for the He and the N2 simulation at 30 Nl/min, Extending our previous work [5], metallic open-cell foams made
for the case of adiabatic boundary conditions of the solid at the wall. of FeCrAlY and Aluminum, from about 89% to 95% of porosity and
When changing the density, also the Reynolds number changes from 10 to 40 nominal PPI, were tested in a tubular reactor in order
from about 115 for He to about 892 for N2 . The average heat transfer to characterize the single-phase wall heat transfer. Axial and radial
coefficient is about 4 times higher with He, but the Nusselt num- temperature profiles within the foam bed were measured during
ber is higher than in the case of N2 . However, the increment of heating and cooling heat transfer runs under flow of both nitro-
the Nusselt numbers is limited in comparison to the correspond- gen and helium. By fitting a pseudo-homogeneous 2D heat transfer
ing increments of the Reynolds number (an increase of 40% with model to such data it was possible to estimate the effective radial
respect to an almost eightfold increment of the Reynolds number), and axial conductivities of the foam bed as well as the wall heat
confirming that the gas conductivity is the controlling factor. This transfer coefficients.
result respects the trend found experimentally for both the wall For the same foam samples the wall heat transfer coefficients
and the overall heat transfer coefficients at these conditions. In fact were found to be in the range of 89–260 W m−2 K−1 when using
for N2 and He runs (HHT, 30 Nl min−1 ) the overall heat transfer nitrogen, and in the range of 468–1140 W m−2 K−1 when using
coefficients [41], evaluated with the thermal parameters of Table 2, helium. Our results point out that the wall heat transfer coefficient
are 113 and 432 W m−2 K−1 respectively: such values are reason- is essentially proportional to the gas conductivity and to the foam
ably close to those obtained from the numerical study (Table 4). pore density, thus suggesting a dominant conductive mechanism
To further prove the weak dependency of the overall heat trans- at the investigated conditions. A positive effect associated with a
fer coefficient on the Reynolds number, the effect of changing only convective contribution to the wall heat transfer coefficient was
the flow velocity while keeping the gas properties constant was also also identified, but this was found to be of minor importance in the
investigated in a different simulation. In order to enable a compari- investigated range of Reynolds number (4 < Redc < 255). The pre-
son, the new case was based on the same boundary condition as for vailing role of heat conduction in the fluid phase was then proved
the simulation with 30 Nl/min of N2 (i.e. no solid contact with the simulating the conjugated heat transfer under the same condi-
wall) using the same thermodynamic and transport properties. But tions of selected experiments. Both the limiting cases of direct
this time the inlet flow rate was assumed to be 3 Nl/min. The aver- contact between solid phase and tube wall and of adiabatic bound-
age temperature profile was higher compared to the case at higher ary conditions for the solid phase at the wall were investigated
velocity, but the overall heat transfer coefficient decreased by only in the simulation studies. The dependency on the Reynolds num-
18%, and the Nusselt number varied by only 22% for a tenfold reduc- ber was numerically investigated by changing both the flow rate
tion of the Reynolds number. So, again, the role of heat conduction (30–3 Nl/min N2 ) and the fluid properties (N2 vs. He). In both cases
in the fluid phase in the case of no direct contact between the solid the overall Nusselt number, which is proportional to the Nusselt
foam and the tube wall is proved. number at the wall, was weakly affected, increasing only by 40% and
by 22% with respect to an almost eightfold and tenfold increment
5. Comparison with conventional packed beds of the Reynolds number, respectively.
The present experimental and simulation results confirm that
The effect of the gap and of the solid-wall contact on the open-cell foams are promising candidates as supports for catalytic
radial heat transfer has been extensively analyzed in the classical applications, especially when highly conductive metals (such as
134 E. Bianchi et al. / Catalysis Today 216 (2013) 121–134

copper or aluminum) are used to manufacture foams with a high [14] C. Hutter, A. Zenklusen, R. Lang, P.R. von Rohr, Axial dispersion in metal
relative density and high pore count (PPI). We have further shown foams and streamwise-periodic porous media, Chemical Engineering Science
66 (2011) 1132–1141.
that, even for the case of a small gap between the foam and the [15] A. Inayat, H. Freund, A. Schwab, T. Zeiser, W. Schwieger, Predicting the specific
tube wall, as encountered in tubular reactors packed with foams, surface area and pressure drop of reticulated ceramic foams used as catalyst
the wall heat transfer coefficient shows only a weak dependency on support, Advanced Engineering Materials 13 (2011) 990–995.
[16] A. Inayat, H. Freund, T. Zeiser, W. Schwieger, Determining the specific sur-
the convective contributions. In such a configuration the heat flux face area of ceramic foams: the tetrakaidecahedra model revisited, Chemical
is forced to pass through the fluid phase, which becomes the con- Engineering Science 66 (2011) 1179–1188.
trolling resistance at the wall. In fact, according to the simulations [17] A. Inayat, J. Schwerdtfeger, H. Freund, C. Körner, R.F. Singer, W. Schwieger,
Periodic open-cell foams: pressure drop measurements and modeling of an
the resistance between the tube and the solid foam is found to be
ideal tetrakaidecahedra packing, Chemical Engineering Science 66 (2011)
inversely proportional to the intrinsic conductivity of the saturating 2758–2763.
fluid, in line with the experiments. [18] G.F. Froment, Design of fixed bed catalytic reactors based on effective transport
models, Chemical Engineering Science 17 (1962) 849–859.
These identified characteristics and dependencies allow for
[19] C.A. Coberly, W.R.J. Marshall, Temperature gradients in gas streams flowing
the design anddvantageous use of open-cell foams as catalyst through fixed granular beds, Chemical Engineering Progress 47 (1951).
supports in compact tubular reactors for strongly exothermic [20] V.V. Calmidi, R.L. Mahajan, Forced convection in high porosity metal foams,
and endothermic gas/solid processes, where the existing, well- Journal of Heat Transfer 122 (2000) 557–565.
[21] J.P. Du Plessis, J.H. Masliyah, Mathematical modelling of flow through consoli-
established packed bed reactor technology is inherently limited by dated isotropic porous media, Transport in Porous Media 3 (1988) 145–161.
low radial heat transfer rates. [22] W. Lu, C.Y. Zhao, S.A. Tassou, Thermal analysis on metal-foam filled heat
exchangers. Part I: metal-foam filled pipes, International Journal of Heat and
Mass Transfer 49 (2006) 2751–2761.
Acknowledgements [23] C.Y. Zhao, W. Lu, S.A. Tassou, Thermal analysis on metal-foam filled heat
exchangers. Part II: tube heat exchangers, International Journal of Heat and
The authors from Politecnico di Milano acknowledge fund- Mass Transfer 49 (2006) 2762–2770.
[24] Y. Peng, J.T. Richardson, Properties of ceramic foam catalyst supports: one-
ing by the Italian Ministry of Education, University and Research, dimensional and two-dimensional heat transfer correlations, Applied Catalysis
Rome (MIUR, Progetti di Ricerca Scientifica di Rilevante Interesse A: General 266 (2004) 235–244.
Nazionale, prot. 2010XFT2BB) within the project IFOAMS (“Inten- [25] D. Edouard, T. Truong Huu, C. Pham Huu, F. Luck, D. Schweich, The effective ther-
mal properties of solid foam beds: experimental and estimated temperature
sification of Catalytic Processes for Clean Energy, Low-Emission profiles, International Journal of Heat and Mass Transfer 53 (2010) 3807–3816.
Transport and Sustainable Chemistry using Open-Cell Foams as [26] K. Pangarkar, T.J. Schildhauer, J. Ruud van Ommen, J. Nijenhuis, J.A. Moulijn, F.
Novel Advanced Structured Materials”). Kapteijn, Heat transport in structured packings with co-current downflow of
gas and liquid, Chemical Engineering Science 65 (2010) 420–426.
The authors from Erlangen University gratefully acknowledge
[27] D. Vervloet, M. Kamali, J. Gillissen, J. Nijenhuis, H. Van den Akker, F. Kapteijn,
the support of the Cluster of Excellence ‘Engineering of Advanced J. van Ommen, Intensification of co-current gas–liquid reactors using struc-
Materials’ at the University of Erlangen-Nuremberg, which is tured catalytic packings: a multiscale approach, Catalysis Today 147 (2009)
S138–S143.
funded by the German Research Foundation (DFG) within the
[28] ERG, Duocel® Foam Properties, in: http://www.ergaerospace.com/
framework of its ‘Excellence Initiative’. foamproperties/introduction.htm, retrieved 13 Feb. 2012.
[29] R. Lemlich, A theory for the limiting conductivity of polyhedral foam at low
density, Journal of Colloid and Interface Science 64 (1978) 107–110.
References
[30] R. Bauer, E. Schlünder, Effective radial thermal conductivity of packings in gas
flow. Part II. Thermal conductivity of the packing fraction without gas flow,
[1] I. Ghosh, Heat transfer correlation for high-porosity open-cell foam, Interna- International Chemical Engineering 18 (1978) 189–204.
tional Journal of Heat and Mass Transfer 52 (2009) 1488–1494. [31] V. Specchia, G. Baldi, S. Sicardi, Heat transfer in packed bed reactors with one
[2] L. Giani, G. Groppi, E. Tronconi, Heat transfer characterization of metallic foams, phase flow, Chemical Engineering Communications 4 (1980) 361–380.
Industrial and Engineering Chemistry Research 44 (2005) 9078–9085. [32] G. Groppi, E. Tronconi, Honeycomb supports with high thermal conductivity
[3] L. Giani, G. Groppi, E. Tronconi, Mass-transfer characterization of metallic foams for gas/solid chemical processes, Catalysis Today 105 (2005) 297–304.
as supports for structured catalysts, Industrial and Engineering Chemistry [33] J.G.H. Borkink, P.C. Borman, K.R. Westerterp, Modelling of radial heat trans-
Research 44 (2005) 4993–5002. port in wall-cooled packed beds–confidence intervals of estimated parameters
[4] G. Groppi, L. Giani, E. Tronconi, Generalized correlation for gas/solid mass- and choice of boundary conditions, Chemical Engineering Communications 121
transfer coefficients in metallic and ceramic foams, Industrial and Engineering (1993) 135–155.
Chemistry Research 46 (2007) 3955–3958. [34] R. Williams, The Geometrical Foundation of Natural Structure: A Source Book
[5] E. Bianchi, T. Heidig, C.G. Visconti, G. Groppi, H. Freund, E. Tronconi, An appraisal of Design, Dover, New York, 1979.
of the heat transfer properties of open-cell foams for strongly exo-/endo- [35] A.G. Dixon, D.L. Cresswell, Theoretical prediction of effective heat transfer
thermic catalytic processes in tubular reactors, Chemical Engineering Journal parameters in packed beds, AIChE Journal 25 (1979) 663–676.
198–199 (2012) 512–528. [36] OpenFOAM® , in, http://www.openfoam.com, retrived March 2013.
[6] A. Bhattacharya, V.V. Calmidi, R.L. Mahajan, Thermophysical properties of high [37] K.K. Bodla, J.Y. Murthy, S.V. Garimella, XMT-based direct simulation of flow
porosity metal foams, International Journal of Heat and Mass Transfer 45 (2002) and heat transfer through open-cell aluminum foams, in: Thermal and Ther-
1017–1031. momechanical Phenomena in Electronic Systems (ITherm), 2010, 12th IEEE
[7] R. Coquard, M. Loretz, D. Baillis, Conductive heat transfer in metallic/ceramic Intersociety Conference on, IEEE, 2010, pp. 1–9.
open-cell foams, Advanced Engineering Materials 10 (2008) 323–337. [38] A. Kopanidis, A. Theodorakakos, E. Gavaises, D. Bouris, 3D numerical simulation
[8] D.P. Haack, K.R. Butcher, T. Kim, T.J. Lu, Novel Lightweight Metal Foam Heat of flow and conjugate heat transfer through a pore scale model of high porosity
Exchangers, American Society of Mechanical Engineers, New York, NY, United open cell metal foam, International Journal of Heat and Mass Transfer 53 (2010)
states, 2001, pp. 141–147. 2539–2550.
[9] J. Banhart, Manufacture, characterisation and application of cellular metals and [39] T. Campbell, R.K. Kalia, A. Nakano, P. Vashishta, S. Ogata, S. Rodgers, Dynamics
metal foams, Progress in Materials Science 46 (2001) 559–632. of oxidation of aluminum nanoclusters using variable charge molecular-
[10] T.J. Lu, H.A. Stone, M.F. Ashby, Heat transfer in open-cell metal foams, Acta dynamics simulations on parallel computers, Physical Review Letters 82 (1999)
Materialia 46 (1998) 3619–3635. 4866–4869.
[11] T. Kodama, A. Kiyama, K.I. Shimizu, Catalytically activated metal foam absorber [40] D.-Y. Lee, K. Vafai, Analytical characterization and conceptual assessment of
for light-to-chemical energy conversion via solar reforming of methane, Energy solid and fluid temperature differentials in porous media, International Journal
and Fuels 17 (2003) 13–17. of Heat and Mass Transfer 42 (1999) 423–435.
[12] C. Hutter, C. Allemann, S. Kuhn, P. Rudolf von Rohr, Scalar transport in a milli- [41] J. Beek, Design of packed catalytic reactors, Advances in Chemical Engineering
scale metal foam reactor, Chemical Engineering Science 65 (2010) 3169–3178. 3 (1962) 203–271.
[13] C. Hutter, A. Zenklusen, S. Kuhn, P.R. von Rohr, Large eddy simulations of flow [42] A.G. Dixon, Correlations for wall and particle shape effects on fixed bed bulk
through a streamwise-periodic structure, Chemical Engineering Science 66 voidage, Canadian Journal of Chemical Engineering 66 (1988) 705–708.
(2011) 519–529.