Sie sind auf Seite 1von 8

9078 Ind. Eng. Chem. Res.

2005, 44, 9078-9085

MATERIALS AND INTERFACES

Heat Transfer Characterization of Metallic Foams


Leonardo Giani, Gianpiero Groppi, and Enrico Tronconi*
NEMAS, Centro di Eccellenza per l’Ingegneria dei Materiali e delle Superfici Nanostrutturate,
Dipartimento di Chimica, Materiali ed Ingegneria Chimica “G. Natta”, Politecnico di Milano,
Piazza Leonardo da Vinci 32-20133 Milano, Italy

Gas-solid heat transfer coefficients were determined in open-celled metal foams as part of a
study aimed at evaluating the application of metal foams as catalyst supports in gas-solid
catalytic processes with short contact times and high reaction rates, typically controlled by
diffusional mass transport. Examples of such processes are found in the field of environmental
catalysis, including, for example, catalytic combustion, selective catalytic reduction of NOx by
NH3 (SCR-DeNOx), automotive exhaust gas after treatment, and also in the catalytic partial
oxidation of hydrocarbons for syngas or H2 generation processes. In this work, foam samples
made of FeCrAlloy and Cu with nominal porosities of 10 and 20 pores per inch (ppi) were
characterized by performing non-steady-state cooling measurements. Convective gas-solid heat
transfer coefficients were determined by applying a one-dimensional, heterogeneous model of
the cooling structure. The correlation Nu ) 1.2Re0.43Pr1/3 well described the dependence of the
dimensionless heat transfer coefficients on Re and Pr numbers for all the tests made in a range
of flow superficial velocities from 1.2 to 5.7 m/s, independently from the foam cell size (20 < Re
< 240). Such a correlation was derived assuming a prismatic idealization of the unit cell and
selecting the equivalent strut diameter as the characteristic size of the foams. This expression
satisfies the Colburn analogy with the correlation for mass transfer coefficients derived in a
previous investigation and resembles semitheoretical literature correlations for heat transfer
in flow across banks of tubes at low Reynolds numbers.

1. Introduction tortuous flow paths through the porous matrix that


promotes turbulence and enhances interphase heat
In a previous work1 we estimated gas-solid mass
transfer rates. In this way, heat can be removed from
transfer coefficients in open-celled metal foams2,3 by
or added to gases or liquids by letting them flow through
performing the catalytic oxidation of CO under diffu-
the foam and cooling or heating at the same time. An
sional control in a microreactor and by measuring the
apparent rate constant. Foams of different pore densi- example of such applications is the use of metal foams
ties were activated by deposition of a thin layer of as compact heat sinks for cooling of microelectronic
palladium on alumina.4 On the basis of a simple devices such as computer chips or power electronics.5
geometrical description of the metallic foam as a cubic Lu et al.6 developed a theoretical model to evaluate
framework of connected struts, all mass transfer data the overall heat transfer coefficient, considering both
were successfully correlated by the following expression: conduction through the solid and heat exchange from
the solid surface to the gas. Foams were modeled as
Sh ) 1.1Re0.43Sc1/3 (1) simple cubic unit cells consisting of heated slender
cylinders, making use of the analogy between flow
using the diameter of the struts as the characteristic through the foam and flow across a bank of cylinders.
length in dimensionless numbers. Heat transfer coefficients were determined experimen-
Equation 1 is useful in view of the application of metal tally by Younis and Viskanta7 for ceramic foams (alu-
foams as catalyst supports for fast, diffusion-controlled mina and cordierite) with pore diameters in the range
reactions. In many practical catalytic applications deal- of 0.29-1.52 mm. In their work, the authors estimated
ing, for example, with highly exothermic reactions, a volumetric heat transfer coefficient because the sur-
however, engineering calculations would call for the face area per unit volume of the foams was unknown.
evaluation of interphase heat transfer, too. Besides, A different dimensionless correlation with expression
open-celled metal foams can be used also as highly Nu ) CRem was obtained for each sample analyzed, with
efficient heat exchangers, taking advantage of the large the pore diameter chosen as the characteristic length
exchange surface area per unit volume and of the to calculate the Nusselt (Nu) and Reynolds (Re) num-
bers. The Reynolds exponent m was reported to vary in
* To whom correspondence should be addressed. Tel.: the range 0.42-0.96, while the Reynolds coefficient C
+39-02-2399 3264. Fax: +39-02-7063 8173. E-mail: took values in the range 0.139-0.638. The importance
enrico.tronconi@polimi.it. of an accurate determination of the geometrical param-
10.1021/ie050598p CCC: $30.25 © 2005 American Chemical Society
Published on Web 11/01/2005
Ind. Eng. Chem. Res., Vol. 44, No. 24, 2005 9079

Table 1. Geometric Properties of the Metal Foam


Samples Investigated in This Work
cell
pore size  density Sv
samplea material a (m) (%) (ppi) ds (m) (m2/m3)
A Fe-Cr alloy 4.7 × 10-3 92.7 5.4 0.82 × 10-3 352
B Fe-Cr alloy 2.0 × 10-3 93.7 12.8 0.33 × 10-3 767
C copper 4.6 × 10-3 91.1 5.6 0.80 × 10-3 449
a Foams A, B, and C correspond respectively to foams B, D, and

F in ref 1.

Considering this model, the cell volume Vo can be


calculated as

Figure 1. Prismatic cell model for anisotropic foams.


Vo ) a2b (2)

The overall volume of the struts Vs can be expressed as


eters of the foam was emphasized by showing that the a function of the foam void fraction :
difference in the pore size declared by the manufacturer
and the measured value resulted in two different Vs ) (1 - )Vo ) (1 - )a2b (3)
correlations for the same foam.
Schlegel et al.8,9 have also determined heat transfer Alternatively, Vs can also be calculated as the overall
coefficients experimentally in ceramic foams with pore volume of the cylinders included in the unit cell,
diameters from 1.02 to 5.20 mm (nominal porosities of considering that each strut is shared among four cells:
10, 20, 30, and 50 ppi). The identified dimensionless
correlations, which took into account the pore diameter dS2
as the characteristic length, differed depending on the Vs ) π (2a + b) (4)
sample as in the work of Younis and Viskanta.7 Re- 4
ported Reynolds exponents were in the range of 0.42-
By combining eqs 3 and 4, one obtains an expression
0.47 for 10, 20, and 30 ppi samples and 0.27 for a 50
that relates the strut diameter ds to a and :
ppi sample, while C coefficients varied from 0.801 to 1.6.

[ ]
Geometrical data were provided by the producer or
b (1 - ) 1/2
measured by the authors. dS ) 2a (5)
In this work, heat transfer coefficients have been π (2a + b)
investigated by performing non-steady-state cooling
The specific area per unit cell volume is then computed
temperature measurements of hot structures in an air
according to eq 1.
stream. The experimental results have been analyzed
The specific area per unit cell volume Sv can be
by means of a transient mathematical model of the gas-
eventually expressed by consideration of eq 5 as a
solid heat transfer process in order to derive a general
function of the foam porosity  and of the pore diameter
correlation for gas-solid heat transfer coefficients.
a:
Finally, results of both the heat and the mass transfer

[ ]
investigations are herein discussed comparatively in 4(1 - ) 2 (2a + b)(1 - ) 1/2
order to check their consistency with the Colburn Sv ) ) π (6)
analogy in reticulated foam structures. dS a b

2. Experimental Section For isotropic foams, b equals a and the above equations
reduce to those presented in ref 1.
2.1. Metal Foams. Samples used in this research Estimates of the foam samples geometrical param-
comprised metal foams of two different nominal cell eters can be found in Table 1.
sizes (10 and 20 ppi) and were provided by Porvair as As reported, samples A and B were made of FeCr alloy
panels of 5% nominal relative density. The foams were while sample C was made of copper. The unit cell for
characterized by measurements of the pore size, by sample C is anisotropic: its pore size is a ) 4.6 × 10-3
image analysis, and by estimation of the open volume m in the transverse direction (defined with respect to
fraction , calculated from the apparent density of the flow) and b ) 3.3 × 10-3 m in the longitudinal direction.
foam (FFOAM) and the density of the foam struts (FS) as 2.2. Heat Transfer Runs. The present study was
described in more detail elsewhere.1 carried out in the same test rig used by Brautsch et al.10
Adopting the same approach developed in the previ- to determine heat transfer coefficients in metal honey-
ous mass transfer investigation,1 isotropic foams were comb-type structures at high flow rates. Cylinders of
described according to a cubic cell model of the struc- 75.0 mm diameter were cut from the metal foam panels
ture6 from which it is possible to estimate the average by electroerosion, producing a precise cut with no
struts diameter ds and the foam specific geometric area damage for the structure. The depth of the cylinders was
Sv as functions of the pore size a and of the porosity . 25.3 mm for the Fe-Cr alloy samples A and B and 50.0
However, also anisotropic foams have been herein mm for the Cu sample. Tests were made over each
investigated. In the case of anisotropic foams, pore sizes single structure and also placing more structures in
differ when either a transversal section or a longitudinal series inside the test rig, schematically depicted in
section is considered. Accordingly, in this work the foam Figure 2.
characteristic dimensions were derived by modifying the Air was fed by a blower at a constant flow rate in the
cubic model into a prismatic one as depicted in Figure range of 5.5-25.2 L/s (STP). Hydrogen was fed from an
1. independent line with two valves, one on/off suitable to
9080 Ind. Eng. Chem. Res., Vol. 44, No. 24, 2005

Table 2. Experimental Cases Studied


sample length (m) Re(ds)
A 2.53 ×10-2 48-239
A 2 × (2.53 × 10-2) 48-239
B 2.44 × 10-2 19-95
C 5.00 × 10-2 48-237

the combustion over the palladium catalyst and thus


letting cold air flow through the structure. The temper-
ature readings from the two bundles of thermocouples
were then recorded until the system reached laboratory
temperature.
Brautsch et al.10 reported that gas-solid heat transfer
coefficients in ceramic honeycomb monoliths measured
in the same apparatus during calibration runs deviated
less than 3% from well-established literature correla-
tions.
In this work heat transfer experiments were run on
each foam sample for eight different flow superficial
velocities: 1.2, 1.9, 2.5, 3.2, 3.8, 4.4, 5.0, and 5.7 m/s.
Figure 2. Schematic diagram of the test rig for heat transfer runs. The cases analyzed are reported in Table 2 where the
Reynolds number is referred to the struts diameter ds
turn off fast the hydrogen feed and the other one as defined in the Nomenclature section and on the
suitable to regulate the flow rate. A mass flow meter superficial empty tube velocity u.
was placed in the hydrogen line to control the flow and 2.3. Mathematical Model. Heat transfer between
therefore, the composition of the air-hydrogen mixture the foam structures and the fluid was described by a
in order to work below the lean explosion limit. Both dynamic, heterogeneous two-phase one-dimensional
gases were mixed in a static mixer (model SMV provided (1D) model used by Schlegel et al.8,9 and Brautsch et
by Sulzer) before entry into the test module. In this al.10 On the basis of the experimental setup shown in
section the mixture was ignited by a palladium catalyst Figure 2, the model assumed plug flow conditions, with
supported on a cylindrical metal foam 5 mm in depth, the fluid temperature (Tf) and the solid temperature (Ts)
which acted as a preheater with a fast response to varying only in the flow direction (1D model). In view
changes in the hydrogen supply. of the good insulation of the experimental test rig, which
The heated air flow passed through a wire screen allowed us to closely approach adiabatic conditions, a
consisting of five layers of stainless steel (mesh 100), constant radial temperature distribution (Ts) in the
which ensured plug flow conditions and minimized metal foam was assumed. Adiabatic conditions were
radiation effects that could affect the temperature experimentally checked at steady state: indeed, tem-
reading of the thermocouples placed further down- peratures measured along the radial direction and
stream. between the average foam inlet and outlet temperature
The temperature of the air flow in the inlet section of the gas differed by less than 3 °C.
(Tin) was measured immediately prior to the foam by a The following energy balances for the fluid (eq 9) and
bundle of five Cr-Ni alloy type K thermocouples with for the solid (eq 10) described the heat exchange
0.5 mm sheath diameter and exposed tips (tip diameter between solid and fluid phases:
) 0.1 mm) to ensure quick response time. They were
located along one diameter of the tube section in a ∂Tf ∂Tf Sv
central region of 30 mm where the influence of the tube ) -v +h (T - Tf) (7)
∂t ∂x Ffcf s
wall was negligible, as shown schematically in Figure
2.
∂Ts Sv λ s f ∂2 Ts
The temperature of the gas flow at the outlet section ) -h (Ts - Tf) + (8)
(Tout) was measured by an identical bundle of five ∂t (1 - )Fscs Fscs ∂x2
thermocouples. Neither the inlet bundle of thermo-
couples nor the outlet one were in contact with the The balance equations account for (a) the thermal mass
structure and measured the temperature of the gas of both the gas and the solid phase; (b) convective
stream. Another wire screen was placed further down- transport in the gas phase; (c) gas-solid heat transfer,
stream. with all the resistance assumed to be confined in the
The foam structure was wrapped with a thin fiber mat fluid film near the struts surface; (d) axial heat conduc-
to prevent flow bypass. Such a fiber mat was assumed tion in the solid phase. This latter term includes a factor
to have negligible influence on the heat transfer mea- f ) 0.4 in order to take into account the tortuous path
surements in the structure due to its low thermal of the conductive heat flux through the reticulated
capacity. structure. The estimate of f has been computed accord-
For a given air flow, the hydrogen feed was increased ing to a correlation by Fourie and DuPlessis11 for the
until the inlet temperature Tin reached about 300 °C. effective thermal conductivity in isotropic metal foams,
This temperature ensured limited heat losses to the omitting the negligible contribution of the coupled fluid
surroundings through convection and radiation. After conductivity. Thermal diffusion in the axial direction
steady-state conditions were reached, corresponding to in the gas phase was neglected (Pe . 1) and radiation
equal inlet and outlet gas stream temperatures, the effects were not considered since the temperature was
hydrogen supply was suddenly turned off, extinguishing kept below 300 °C in all tests.
Ind. Eng. Chem. Res., Vol. 44, No. 24, 2005 9081

The following initial and boundary conditions were


considered:
At t ) 0, solid and fluid were assumed to be at the
same constant temperature To corresponding to the inlet
and outlet temperature measured before switching off
the H2 feed valve.

Tf ) TS ) T0 [t ) 0, 0 < x e L] (9)

At the inlet cross section of the foam structure, the


fluid temperature corresponded to the experimental
inlet temperature profile f(t), described by cubic spline
interpolation:

Tf ) f(t) [x ) 0, t > 0] (10) Figure 3. Typical experimental temperature evolution of the gas
phase at the inlet and the outlet of the foam.
No flux boundary conditions for the solid were assumed:

∂TS
)0 [x ) 0] (11)
∂x
∂TS
)0 [x ) L] (12)
∂x
The heat transfer coefficient h in eqs 7 and 8 was
obtained from the Nusselt number defined in eq 13 with
the strut diameter as the characteristic length:

hds
Nu ) (13)
λf

Based on the analogy with heat transfer in bundles of Figure 4. Inlet and outlet temperature profiles obtained with one
(b) and two (9) segments of foam type A.
tubes, the following correlation was tried for Nusselt
number calculation:
steady-state conditions were reached. Once the hydro-
m 1/3 gen feed was shut down, cold air entered the structure,
Nu ) CRe Pr (14)
which resulted in a sudden drop of the air inlet tem-
where the coefficient C and the exponent m are adaptive perature, as shown in Figure 3. Such a steplike inlet
parameters depending on the flow type and on the temperature profile could be obtained because of the low
geometrical characteristics of the tested foams. The thermal mass of the catalytic foam and of the wire
estimates of these two parameters were obtained by screen placed upstream from the first bundle of ther-
global nonlinear regression on the experimental tem- mocouples: accordingly, the response time of the rig was
perature profiles of the 32 runs covering all the struc- negligible as compared to the dynamics of the heat
tures and flow rates studied. For this purpose we used transfer in foams. The structure exchanged heat with
a robust multimethod nonlinear regression routine the air stream and higher gas temperatures were
developed by Buzzi-Ferraris.12 At each iteration of the obtained at the foam outlet until gas-solid thermal
regression routine, the model PDES, eqs 7-14, were equilibrium conditions were achieved and the inlet and
solved numerically by the method of lines, involving outlet temperatures of the gas stream were the same
axial discretization by backward finite differences and within experimental error. The relatively high thermal
integration in time of the resulting ODE system by mass of the investigated foams made the cooling tran-
Gear’s method.13 sient long enough to obtain a well-resolved outlet
Since the temperature varied almost 300 °C during temperature profile, quite distinct from the inlet profile.
each experiment, the physical properties of the fluid Accordingly, gas-solid heat transfer coefficients, which
phase (air density, heat capacity, viscosity and thermal play a key role in the cooling transient, could be
conductivity) were regarded as temperature-dependent, accurately estimated from the experimental data.
whereas the properties of the solid phase (either Figure 4 compares the temperature profiles obtained
Fe-Cr alloy or copper) were averaged between room at a given flow rate testing one and two in series
temperature and 300 °C. segments of foam type A. While the evolution of the inlet
temperature does not change, as the heat capacity of
Results the upstream section of the rig is always the same, the
cooling rate markedly decreases upon increasing the
Foam samples A, B, and C were tested in the heat number of foams as an effect of the doubled thermal
transfer test rig at different flow rates. As an example, mass.
Figure 3 shows typical cooling curves obtained for type Figure 5 illustrates the effect of the air flow rate on
A foam with a flow superficial velocity of 4.4 m/s. the temperature evolution for the two different foam
At the beginning, the same inlet and outlet temper- samples A and B. At low flow rate (Figure 5a,d) the foam
ature of around 300 °C was measured by both sets of cooling transient is quite slow: as a result, the outlet
thermocouples, confirming that practically adiabatic temperature takes from 15 to 25 s to reach the final
9082 Ind. Eng. Chem. Res., Vol. 44, No. 24, 2005

Figure 5. Inlet and outlet temperature profiles obtained for type A foams (panels a-c) and type B foams (panels d-f) at different inlet
flow velocities and fitting curves.

steady-state conditions at room temperature depending this case, the increment of the heat transfer coefficient
on the foam geometry. Shorter cooling time intervals on decreasing the foam strut diameter further enhances
are required upon increasing the flow rate (Figure 5c,f). the cooling rate of foam B.
In particular, when the flow rate is increased from 1.3 Figures 4 and 5 also show fitting curves obtained by
to 5.0 m/s, the cooling time interval decreased by a factor running a global nonlinear regression on all the 32 runs
of about 3. Such an effect is consistent with the larger herein performed. The following correlation has been
heat capacity of the cooling air associated with the obtained for Nu calculation:
higher flow rate. Besides, a higher gas velocity results
in higher heat transfer coefficients, which further Nu ) 1.2Re0.43Pr1/3 [20 < Re < 240] (15)
increase the cooling rate.
Figure 5 also illustrates the effect of the foam The percent mean error of the regression was 6.19% and
geometry. Foams A and B have similar thermal mass, the correlation index was 0.9989. The experimental
as the heat capacity of foam A is just 15% higher than results were described accurately as apparent from the
that of foam B; nevertheless, when the cooling curves good fit obtained independently from foam type and air
recorded for both foams at the same flow rates are velocity. A slightly delayed cooling is predicted for the
compared, the cooling rate is clearly faster in the case two foams in series (Figure 4), possibly due to an
of foam B with higher cell density. This is mainly due enhanced role of heat dissipation caused by the longer
to the higher specific surface of foam B (Table 1), which time interval and by the higher external surface. Indeed,
enhances the gas-solid heat transfer efficiency. Also in even stronger deviations were noted in other experi-
Ind. Eng. Chem. Res., Vol. 44, No. 24, 2005 9083

comparing simulations generated either maximizing


(f ) 1) or minimizing (f ) 0) the second term on the
right-hand side of eq 8. Differences with the outlet
T-profiles calculated assuming f ) 0.4 were limited to
about 3 °C at the worst, and generally much less, even
for the run over the copper foam C at the lowest flow
rate, where the role of heat conduction would be
expected to be greatest. Accordingly, we conclude that
axial heat conduction played a negligible role in our
experiments, and the foam cooling transients were
governed essentially by interphase heat transfer.

Discussion
Mathematical model analysis of transient cooling
experiments allowed us to obtain a correlation for the
calculation of heat transfer coefficients that covers a gas
velocity range from 1.2 to 5.7 m/s in metal foams with
measured pore sizes from 2.0 to 4.7 mm and void
fractions from 91.1 to 93.7, corresponding to a 20-240
range of Reynolds numbers defined with the diameter
of the struts ds as the characteristic length and the
superficial velocity at empty tube.
Notably the Nu correlation (eq 15) is quite similar to
the one found for mass transfer in metal foams in a
previous investigation based on CO combustion tests
over Pd/Al2O3 washcoated metallic foams1:

Sh ) 1.1Re0.43Sc1/3 [10 < Re < 100] (1)

The consistency of eqs 15 and 1 with the Colburn


analogy provides further confidence in the results.
Notice that the correlations were obtained indepen-
dently and with totally different methods: mass trans-
fer coefficients were determined in tests where the
catalytic oxidation of CO was performed in a diffusion-
controlled regime on coated foams, while heat transfer
coefficients were determined by performing non-steady-
state cooling measurements over uncoated samples.
The agreement between the results of heat and mass
transfer studies also confirms the good quality of the
washcoating technique developed to produce the cata-
lytically active foams4 used in CO combustion experi-
ments. Moreover, the higher flow rates achieved in the
heat transfer tests allow to extend the validity of the
mass transfer correlation (eq 1) to higher Reynolds
numbers.
On a more general basis, it is worth emphasizing that
eq 15 closely resembles semitheoretical literature cor-
Figure 6. Inlet and outlet temperature profiles obtained for type relations for heat transfer in flow across banks of tubes
C, anisotropic copper foam and fitting curve for different flow
velocities: (a) 2.5 m/s; (b) 4.4 m/s; (c) 5.7 m/s.
at low Reynolds numbers:14

Nu ) 0.9Re0.4Pr1/3 [10 < Re < 100] (16)


ments with three foams in series, which were therefore
not considered in the final analysis and are not reported Note that the literature correlation (eq 16) was derived
herein. The same explanation may apply also to the on the basis of the interstitial velocity at the restricted
overestimation of the outlet T-profile apparent in Figure section of the bundles of tubes for Re calculation, while
5 for case d only, which corresponds to the experiment in this paper, the superficial gas velocity at empty tube
at the lowest flow velocity over the foam sample B was used. Considering the geometry of the investigated
associated with the highest specific surface. As shown foams, the maximum interstitial gas velocity through
in Figure 6, where experimental and calculated curves the reticulated structure is 1.4 times higher than the
obtained for foam C are reported, the model herein empty tube velocity. Accordingly, for fair comparison
developed accurately describes also the behavior of the with the literature correlations, coefficients in eqs 15
anisotropic copper foam, proving that a simple modifica- and 1 should be corrected to 1.0 and 0.9, respectively,
tion of the cubic cell model is able to catch the geometric which closely resemble the coefficient in eq 16. However,
characteristics of such anisotropic foams. since the void fractions of all the foams samples
A simple sensitivity analysis concerning the influence investigated herein and in our previous mass transfer
of axial heat conduction in the foams was conducted by work1 are very close, our data do not allow us to identify
9084 Ind. Eng. Chem. Res., Vol. 44, No. 24, 2005

correlation to higher Reynolds numbers (10 < Re < 240).


The derived correlation also shows a strong similarity
with literature expressions for heat transfer in tube
bundles, which validates the concept of modeling the
foam as a array of cylinders.
The experiments performed in this work do not allow
us to discriminate between the use of either the super-
ficial velocity (u) or the maximum velocity (umax) to
correlate heat transfer data, as all the investigated foam
samples had approximately the same void fraction. To
clarify this point, experiments with foams with signifi-
cantly lower void fractions ( ) 0.80) will be performed
in future investigations.
Figure 7. View of the foam structure.
Acknowledgment
the best criteria for the correlation of the experimental
results with respect to flow velocity. The results herein We gratefully acknowledge the help provided by
reported also closely resemble those obtained by Timothy Griffin, Dieter Winkler, and Regina Granacher
Satterfield and Cortez15 for mass (and heat) transfer in from Alstom Power Technology Center, Baden-Dättwill,
woven-wire screen catalysts (gauzes). On the basis of Switzerland, where the heat transfer measurements
mass transfer experiments results and on critical re- herein reported were performed.
evaluation of previous literature data, these authors
clearly point out that for high-porosity screens, a model Nomenclature
of parallel wires normal to the direction of flow would a ) cubic cell dimension equal to the pore diameter (m)
seem more physically realistic than a porous solid or b ) cubic cell depth for anisotropic foams (m)
packed-bed model. Similarly, our results confirm that cf ) specific heat of gas phase (J‚kg-1‚K-1)
the description of foam structures consisting of a con- cs ) specific heat of solid phase (J‚kg-1‚K-1)
nected network of struts (Figure 7) based on the analogy C ) Reynolds coefficient in heat transfer correlation
with bundles of tubes is more physically sound than the D ) diffusivity (m2 s-1)
one based on the analogy with packed beds of particles dS ) equivalent diameter of foam struts in the cubic unit
previously proposed in the literature.16 cell (m)
More complex and, possibly, more accurate geometric f ) effective heat transfer conductivity factor in foams
models than the simple cubic cell can be adopted, h ) heat transfer coefficient (W‚m-2‚K-1)
namely, for example, the tetrakaidecaedron cell.17-20 kmt ) mass transfer coefficient (m‚s-1)
However, the increasing complexity of the model seems L ) depth of foam sample (m)
contradictory with the relatively poor accuracy in m ) Reynolds exponent in the heat transfer correlation
determining key geometric parameters such as the Nu ) Nusselt number based on struts equivalent diameter
diameter of the pores and of the struts in the foam [Nu ) hdS λ-1]
structure. Pe ) Peclet number [Pe ) RePr]
Besides, it has been proven in this paper that a simple ppi ) foam pore density (pores per inch)
geometric model allows us to handle the foam anisotropy Pr ) Prandtl number [Pr ) cf µ λ-1]
simply by switching from a cubic to a prismatic descrip- q ) heat flux (W‚m-2)
tion of the unit cell. Re ) Reynolds number based on strut equivalent diameter
[Re ) FdSu µ-1]
Conclusions STP ) standard temperature (273 K) and pressure (1atm)
Sc ) Schmidt number [Sc ) µ F-1 D-1]
Gas-solid heat transfer coefficients were measured Sh ) Sherwood number [Sh ) kmt ds D-1]
in FeCrAlloy and Cu foams of different pore sizes by Sv ) external surface area per unit volume of bed (m2
performing transient cooling experiments. ‚m(bed)-3)
The cubic cell model previously used to describe the t ) time (s)
foam network of connected struts and to estimate the Tin ) inlet temperature (°C, K)
strut equivalent diameter in isotropic samples was Tout ) outlet temperature (°C, K)
extended to characterize anisotropic samples. In this Ts ) solid temperature (°C, K)
case the cell is represented as a prism, and expressions Tf ) fluid temperature (°C, K)
to estimate the surface area and the cylinder diameter T0 ) initial foam temperature (°C, K)
were derived. u ) superficial flow velocity (m‚s-1)
By use of the equivalent strut diameter as the v ) interstitial flow velocity (m‚s-1)
characteristic length, a single correlation was obtained, Vo ) foam volume (m3)
unifying all the tests for isotropic and anisotropic foams Vs ) solid volume (m3)
and thus confirming the accuracy of the modified cell W ) foam mass (kg)
model in the estimation of the foam geometrical proper- w/w ) weight percentage
ties. x ) axial coordinate (m)
Furthermore, the correlation satisfies the Colburn
analogy with the one for interphase mass transfer Greek Symbols
derived in our previous work,1 confirming the good  ) foam porosity
quality of the coating used in the mass transfer tests to λ ) thermal conductivity (W‚m-1‚K-1)
activate the foam surface. Moreover, the correlation µ ) gas viscosity (kg‚m-1‚s-1)
herein derived extends the validity range of the previous Ff ) gas density (kg‚m-3)
Ind. Eng. Chem. Res., Vol. 44, No. 24, 2005 9085

FS ) apparent density of the hollow struts [FS ) W/Vs] (11) Fourie, J. G.; Du Plessis, J. P. Effective and Coupled
(g‚m-3) Thermal Conductivities of Isotropic Open-Cellular Foams. AIChE
J. 2004, 50, 547-556.
(12) Donati G.; Buzzi-Ferraris G. Powerful method for Hougen-
Literature Cited Watson Model Parameter Estimation with integral Conversion
(1) Giani, L.; Groppi, G.; Tronconi, E. Mass transfer charac- Data. Chem. Eng. Sci. 1974, 29, 1504-1509.
terization of metal foams. Ind. Eng. Chem. Res 2005, 44, 4993- (13) Hindmarsh, A. C. Odepack, a systematized collection of
5002. ODE solvers in scientific computing, Stepleman, R. S., et al., Eds.;
(2) Gibson, L. J.; Ashby, M. F. Cellular Solids; Pergamon North-Holland: Amsterdam, 1983; pp 55-64.
Press: Oxford, U.K., 1988. (14) Incropera, F. P.; DeWitt, D. P. Introduction to Heat
(3) Ullmann’s Encyclopedia of Industrial Chemistry, 6th ed.; Transfer; J. Wiley: New York, 1996.
Wiley: Weinheim, Germany, 2003; p 21. (15) Satterfield, C. N.; Cortez, D. H. Mass Transfer Charac-
(4) Giani, L.; Cristiani, C.; Groppi, G.; Tronconi, E. Washcoating terization of Woven-Wire Screen Catalysts. Ind. Eng. Chem.
method for Pd/γ alumina deposition on metallic foams. Appl. Catal. Fundam. 1970, 9, 613-620.
B: Environ., in press; published online August 26, 2005, 10.1016/ (16) Richardson, J. T.; Peng, Y.; Remue, D.; Properties of
j.apcatb.2005.07.003. ceramic foam catalyst supports: pressure drop. Appl. Catal. A:
(5) Banhart J. Manufacture, characterisation and application Gen. 2000, 204, 19-32.
of cellular metals and metal foams. Prog. Mater. Sci. 2001, 46, (17) Gibson, L. J.; Ashby, M. F. Cellular Solids; Pergamon
559-632. Press: Oxford, U.K., 1988.
(6) Lu, T. J.; Stone, H. A.; Ashby, M. F. Heat transfer in open- (18) Boomsma, K.; Poulikakos, D. On the effective thermal
cell metal foams. Acta Metall. 1998, 46, 3619-3635. conductivity of a three-dimensionally structured fluid-saturated
(7) Younis, L. B.; Viskanta, R. Experimental determination of metal foam. Int. J. Heat Mass Transfer 2001, 44, 827-836.
the volumetric heat transfer coefficient between stream of air and (19) Fourie, J. G.; DuPlessis, J. P. Pressure drop modelling in
ceramic foam. Int. J. Heat Mass Transfer 1993, 36, 1425-1434. cellular metallic foams. Chem. Eng. Sci. 2002, 57, 2781-2789.
(8) Schlegel, A.; Benz, P.; Buser, S. Wärmeübertragung und (20) Kwon, Y. W.; Cooke, R. E.; Park, C. Representative unit-
Druckabfall in keramischen Schaumstrukturen bei erzwungener cell models for open cell metal foams with or without elastic filler.
Strömung, Wärme- Stoffübertrag. 1993, 28, 259-266. Mater. Sci. Eng. A 2003, 343, 63-70.
(9) Schlegel, A.; Benz, P.; Buser, S.; Bestimmung der Wärmeü-
bergangskoeffizienten in keramischen Schaumstrukturen, Tech-
Received for review May 20, 2005
nische Mitteilung (TM-51-91-05), Paul Scherrer Institut, CH-5232
Villigen PSI, Switzerland, 1991. Revised manuscript received September 5, 2005
(10) Brautsch, A.; Griffin, T.; Schlegel, A. Heat Transfer Accepted September 28, 2005
characterization of support structures for catalytic combustion. Int.
J. Heat Mass Transfer 2002, 45, 3226-3231. IE050598P

Das könnte Ihnen auch gefallen