You are on page 1of 10

Chemical Engineering Science 267 (2023) 118309

Contents lists available at ScienceDirect

Chemical Engineering Science


journal homepage: www.elsevier.com/locate/ces

CFD analysis on the optimization of POCS performances under


randomical transformations: a bridge with open-cell foams
A. Vespertini ⇑, A. Della Torre, G. Montenegro, A. Onorati, I. Nova, E. Tronconi
Politecnico di Milano, Department of Energy, via Lambruschini 4, 20156 Milano, Italy

h i g h l i g h t s g r a p h i c a l a b s t r a c t

 We implemented an algorithm to
design differently randomized POCS.
 We investigated the effect of
randomization on the POCS mass
transfer properties.
 Identification of the randomization
process as bridge between ordered
POCS and foams.
 Proposal of a merely fluid-dynamic
approach to evaluate the geometrical
tortuosity.

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

Article history: Periodic Open Cell Structures (POCS) have been considered as possible catalytic substrates, thanks to
Received 25 June 2022 recent developments in the Additive Layer Manifacturing (ALM) technique. In this work, the impact of
Received in revised form 18 October 2022 the POCS cell shape on mass transfer has been investigated through CFD simulations. An algorithm able
Accepted 12 November 2022
to increase the wet surface of a simple cubic POCS has been developed, keeping constant the mean pore
Available online 21 November 2022
diameter and with negligible porosity variations. Catalytic performances of the substrates have been
evaluated, suggesting the possibility to enhance the mass transfer properties of the POCS up to that of
Keywords:
foams by continuous randomical transformation of its shape. A dedicated CFD multi-region solver has
POCS
Foams
been used for this purpose, accounting for heat/mass transfer and catalytic reactions. Moreover, a
Microstructure fluid-dynamic approach has been proposed in order to estimate the value of tortuosity, investigating
CFD its dependency on the POCS randomization level and its effect on the flow field.
Randomization Ó 2022 Elsevier Ltd. All rights reserved.

1. Introduction (Hutter et al., 2011), it was demonstrated that the addition of a


commercial metal foam in an empty tube can rise the heat transfer
Since the beginning of the 21st century, the range of possible performances up to ten times higher. Besides, introducing a
applications of open cell foam structures has widely expanded, pseudo-convective heat transfer coefficient, the impact of the pore
both in research and industrial applications (Banhart, 2001; diameter and the wall connection was investigated on the com-
Faure et al., 2011). In particular, mass and heat transfer properties mercial foam performances. Thanks to the recent advancements
of foams made them a very promising subject to investigate. In in the Additive Layer Manufacturing (ALM) technique, the possibil-
ity to control the most important geometrical parameters of the
structure increased the interest on periodic open cell structures
(POCS). In particular, creating customizable foams in order to
⇑ Corresponding author.
match specific constraints has made possible the characterization
E-mail addresses: andrea.vespertini@polimi.it (A. Vespertini), augusto.dellator-
re@polimi.it (A. Della Torre), gianluca.montenegro@polimi.it (G. Montenegro),
of pressure drop and heat transfer properties with respect to the
angelo.onorati@polimi.it (A. Onorati), isabella.nova@polimi.it (I. Nova), enrico. variation of cell and strut dimensions (Bastos et al., 2018): within
tronconi@polimi.it (E. Tronconi).

https://doi.org/10.1016/j.ces.2022.118309
0009-2509/Ó 2022 Elsevier Ltd. All rights reserved.
A. Vespertini, A. Della Torre, G. Montenegro et al. Chemical Engineering Science 267 (2023) 118309

Nomenclature

CFD Computational Fluid Dynamics UD Darcean velocity


uMAX Interstitial velocity k Permeability
km Mass transfer coefficient l Fluid viscosity
POCS Periodic Open Cell Structure C Form coefficient
KC Kelvin Cell L POCS length
IG Groppi performance index s Tortuosity
R% Randomization percentage parameter U Mass flow rate
Ri Reaction rate V_ F Volume flow
gi;j Diffusion rate Re Reynolds number
DP Pressure drop Sh Sherwood number
ALM Additive layer manifacturing Sc Schmidt number
Ji Mass transfer for the i-th specie dP Pore diameter
j Molar flow D Diffusivity
Ui Thermal diffusion Xi Mass fraction of the i-th specie
kR;i Kinetic parameter aV Specific surface area
qMAX Maximum displacement radii A Area-related porosity
Sw POCS surface area a Thermal conductivity
 Porosity
lS Strut length

this article, the performances of eight different cubic cell iso- as a characteristic parameter, pointing out its major role in the
reticular POCS were evaluated, highlighting their capability as cat- correlation.
alytic substrates due to the fast heat transfer and low value of pres- In (Groppi et al., 2007) the possibility of using ceramic foams
sure drop. Another upgrade granted by the progress in the ALM has been investigated, proposing a correlation for mass transfer
technique is the possibility to investigate the flow and thermal with the use of interstitial velocity uMAX , considering variation of
transport properties directly at the unit cell level, as it was done porosity, pore density and Reynolds number. Significant efforts
in (Kaur and Singh, 2020). In this study four different unit cells have been also made to reduce the computational burden and to
have been tested pointing out that, although it has the higher pres- generate virtual open cell structures with similar properties to
sure drop among them, the octet unit cell can grant an higher heat those of real open-cell foams. A possible solution is represented
transfer, concluding that it could be a promising geometry when by the Kelvin-Cell (KC) structure, which are more regular than
reticulated in three dimensions. Furthermore, focusing more in amorphous foams and this results in an easier meshing process.
particular on the after-treatment systems for internal combustion These structures have been tested as catalytic substrates (Lucci
engines, several comparisons have been proposed in literature et al., 2014), with a complete momentum and mass transfer anal-
between the performances of open cell foams and traditional hon- ysis, and used to approximate the foam (Lucci et al., 2017), demon-
eycomb structures (Bach and Dimopoulos Eggenschwiler, 2011; strating that high porosities and low pore diameter KC structures
Giani et al., 2005), revealing the potential advantage of these cat- can have similar mass-transfer behaviour, resulting in a limited
alytic substrates. Generally, foams benefit from radial transport difference ( 15%) in the performance index IG . Another interest-
with respect to traditional honeycombs, in which the fluid motion ing approach is the one proposed by (Bracconi et al., 2017), which
is compartmentalized in different channels. However, only high demonstrated the feasibility of foam geometry reconstruction
porosity open cell structures exhibit a comparably low pressure through the Voronoi tessellation. Centers of packed spheres have
drop: this results in an unvaried or worst trade-off between the been used as seeds for the tessellation, defining a clone-skeleton
gain in mass transfer and the increase of pressure drop. In (Lucci for the foam; the geometrical and flow-field properties of recon-
et al., 2015) through a computational analysis, the performances structed foams have been validated, showing a good agreement.
of a Kelvin-Cell (KC) structure has been compared also with honey- In this work, we focused on periodic open cell structures, inves-
comb, showing that similar conversion activity can be obtained tigating the mere effect of their shape on heat and mass transfer
with a lower surface area, granting a save of noble metals for the behavior, keeping constant the porosity and the mean pore diam-
catalytic process. Furthermore, the role of porosity has been clearly eter. This represents a further step with respect to previous analy-
pointed out in the estimation of a global performance index, which sis, where the effects of the strut shape (i.e. cylindrical, triangular)
is a dimensionless number representing the compromise between on heat transfer and fluid-flow was studied (Ambrosio et al., 2016),
gain in mass transfer and the consequent loss in pressure drop. but without changing the cell shape of the POCS. This is done
Ò
These promising results underline the importance of a fundamen- through a Python routine for FreeCAD , created by taking inspira-
tal analysis of the fluid-dynamics inside open cell structures, in tion from (Azman et al., 2015), which takes as input a cubic POCS
order to better understand the aspects which can affect positively and is able to enhance the surface area exposed to the flow without
their performances. In (Della Torre et al., 2014), the different pos- altering its main geometrical properties. In particular, this analysis
sible flow regimes inside microstructures were investigated aims at pointing out the different mass transfer behaviour caused
through CFD simulations, pointing out the transition from laminar by the random transformation, showing a continuous variation
to turbulent regime using dimensionless coefficients. Furthermore, from the cubic cell POCS up to foam performances, referring to
in (Klumpp et al., 2014), a correlation based on the Ergun equation generalized correlations proposed by (Reichelt and Jahn, 2017).
was proposed: the pressure drop has been expressed in function of To account for chemical reactions, we adopt a numerical frame-
the geometrical properties of the structure, viz. the strut length work already proposed in (Della Torre et al., 2016) which includes
and cell diameter. The effect of the cell orientation on the pressure models for diffusion and catalytic reactions, enabling the possibil-
drop was also investigated by introducing an area related porosity ity to predict the light-off curve for coated catalytic substrates.
2
A. Vespertini, A. Della Torre, G. Montenegro et al. Chemical Engineering Science 267 (2023) 118309

The paper is structured as follows: at first the methodology will 2.1. Fluid-washcoat mass and heat transfer model
be explained, focusing on the randomization process, introducing
the randomization percentage parameter and reporting the main The CFD model for fluid-flow is based on conservation laws for
features of the numerical model. Then the test case will be mass, momentum, energy and species conservation, which are
described, detailing the mesh generation process and the most rel- described in details in (Della Torre et al., 2016). Within this CFD
evant fluid-dynamic conditions imposed. Finally, the different framework, as shown in Fig. 1b washcoat is modeled by the first
results will be discussed, focusing on the effects of POCS shape layer of cells nearby the solid domain, having a global thickness
h i
variation on the geometrical tortuosity of the structure and on ranging between 5  105 ; 7  105 m. For the coupling, a simple
the resulting heat and mass transfer and conversion efficiency.
model has been applied to account for mass transfer: the diffusion
is solved in this model by taking into account the gradient between
the concentration at the cell center with respect to the nearby cell
2. Methods walls.
h i
Mass transfer for a single specie J i kgs is evaluated for each cell
The CFD model adopted in this work, already described in (Della
in the coupling region with this equation:
Torre et al., 2021), is based on the open-source CFD code Open-
FOAM. As it is outlined in Fig. 1a and 1b, it consists of a multi- Di 2
J i ¼ qi  km  A ¼ qi  A ð2Þ
region solver in which the fluid domain is coupled to the reaction L
region, called washcoat, and through a mixed coupled boundary
where Di is the diffusivity of the species within the cell, obtained by
condition the heat flux towards the solid domain is taken into
solving the Chapman–Enskog equation, L and A are the cell length
account. In particular, this boundary condition assigns a value for
and boundary face area, respectively, and km is the mass transfer
the temperature at the interface in a region accounting for the tem-
coefficient.
perature and conductivity of the neighbouring one through the fol-  
Thermal diffusion U i W is calculated in a similar way with the
lowing equations: K
following equation:
  ai
kj Li
T w ¼ T i  k L þk Ti  Tj ; Ui ¼ A ð3Þ
j Lj L
ð1Þ
i i
ki Lj  
T w ¼ T j  k L þk
i i L
j j
Ti  Tj ; where ai is fluid thermal conductivity.

2.2. Washcoat: kinetics and chemistry


where T i;j are the cell-centre temperature in region i and j, Li;j are
the distances between cell center and face center at the boundary Fluid-flow, mass and heat transfer are different time-scale phe-
and ki;j are the thermal conductivities. nomena with respect to reactions, therefore to keep into account
the whole physics and preserve the reliability a Quasi-Steady-
State (QSS) approximation is applied, imposing the reaction rate
Ri;j equal to the diffusion rate gi;j :
   
gi;j Y g;i;j ; Y S;i;j ; T g;i ¼ Ri;j Y S;i ; T S;i ð4Þ

where the dependence with respect to reactants (i) and products (j),
species concentration (Y i;j ), substrate (T S;i ) and fluid temperature
(T g ) was taken into account. Qualitatively, Eq. 4 prevents the accu-
mulation of reactants and products in the washcoat region.
In this work, as in (Lucci et al., 2015) the CO oxidation has been
considered as a model reaction and the reaction rate was estimated
with an Hinshelwood type reaction model (that can be found in
(Della Torre et al., 2021 and Koltsakis et al., 1997)) in which, for
a generic reaction i, the reaction rate is calculated as follows
^A p
p ^B
Ri ¼ kR;i  ð5Þ
T s Gi
^A;B are the ratio between partial
where kR;i is the kinetic constant, p
pressures of each reactant and the reference pressure, T s is the sub-
strate temperature and Gi is the inhibition term, accounting for spe-
cies competition and physical phenomena of absorption or
desorption to or from the reaction site.
The kinetic constant is estimated as follows:
 
ER;i
kR;i ¼ AR;i exp  ð6Þ
RT s
where AR;i and ER;i are Arrhenius coefficients for reaction i. It is
worth noticing that within this work species competition is not pre-
sent, being CO oxidation the only reaction considered. Furthermore,
Fig. 1. 1a Schematics of the heat (Q) and mass (J) transfer between the coupled
the equilibrium term is considered equal to 1, being the reaction
regions. 1b Mesh detail for a multi-region case: fluid region (light-blue), washcoat irreversible. Arrhenius coefficients used for CO oxidation within this
(red) and solid (grey). analysis are taken from (Della Torre et al., 2021).
3
A. Vespertini, A. Della Torre, G. Montenegro et al. Chemical Engineering Science 267 (2023) 118309

2.3. POCS randomizer

The POCS randomizer utility is developed in Python exploiting


the open-source FreeCAD environment. It takes as input a simple
cubic–cell POCS, built up on periodically spaced spheres connected
through cylinders.
This utility displaces each corner sphere to a new random posi-
tion which is moved up to a maximum user-defined distance qMAX
(Fig. 2). In this way, the orientation and position of each cylindrical
strut is randomly prescribed. Being ðxi ; yi ; zi Þ the original position of
the i-corner sphere and S~i the position vector after displacement,
the generated structure corner points position is given by:
Fig. 3. Detail at the cell level of the randomization process: by increasing R% , the
S~i ¼ ðxi þ qi cos hi sin /i ; yi þ qi sin hi sin /i ; zi þ qi cos /i Þ ð7Þ disparity between the cells volume in each POCS is incremented.

where qi 2 ½0; qMAX ; hi 2 ½0; 360 and /i 2 ½0; 180 are randomly
picked within their range of values. The network of corner spheres
constituting the starting cubic cell POCS can be considered as a sim-
ple Bravais lattice, described by the equation:

X
3
~
R¼ nj aj k~j ; ð8Þ
j¼1

where k~j are the primitive cartesian vectors, aj is reticular path and
nj is the number of elements in the j-direction. Consequently,the
randomization process can be thought as a transformation ran-
domly varying the local reticular path aj ! a ~j , but preserving its
mean value. The effect of the randomization process at the cell level
is highlighted in Fig. 3, where the same cell is highlighted for differ-
ent R% POCS.
Finally, each obtained structure has been cut with a cube of Fig. 4. An example of fluid domain for the POCS: duct length is twice the POCS
4 length L, creating two buffer regions at the inlet (left) and outlet (right). The inside
edge 0:02 m in order to perfectly fit in a duct of area 4  10 m 2
structure is an example of randomized POCS.
(Fig. 4). The cut has been performed in the same position for all
the structures (at distance l2S before and after the first and last
sphere of the R% ¼ 0 network) to avoid the dependency on the cut- strut length is generally increased by the randomization process, as
ting procedure. In this way, for each POCS we introduce a random- shown in Table 1, while the porosity remains almost unvaried:
ization percentage parameter R% which depends only on the these parameters have been calculated directly from the POCS
maximum displacement radius qMAX defined as follows: structure using the volume average function object utility in Open-
FOAM, which reconstructs cell-by-cell the whole volume of the con-
1 qMAX sidered solid. In the third column, the mean strut length for POCS
R% ¼  100; ð9Þ
2 lS with 200 cells has been reported: this value has been obtained
thanks to a dedicated python routine in order to estimate the
where lS is the original strut length and the factor 12 accounts for the
asymptotic average strut length as the number of cells increases.
random picking of (qi ; hi ; /i ) which is defined with a normal Proba-
bility Density Function (PDF) with fixed extremal values, hence the
most probable is the middle value. One thing is worth noticing: 2.4. Mesh design
being X the set of all structures that can be created with this utility
and Y the set of all possible values of R% , the function F : X ! Y is a The computational grid was generated using a cartesian mesh
non-bijective surjective function, which means that each value of generator (snappyHexMesh, included in the open-source code
R% defines an ensemble of different POCS characterized by the same OpenFOAM-8.x). The resulting mesh is predominantly hexahedral
value of qMAX . The surface area increases with R% because the mean with a small percentage of polyhedral elements near the boundary
walls. The base cell size is 1 mm and is refined up to 0:25 mm at
the interface between the solid and fluid domain to correctly
resolve the complex shape of the POCS (Fig. 5) and the species dif-
fusion near the reaction zone, as it is shown in Fig. 1b. Finally,
starting from the interface, the latest fluid cells around the solid
region constituted the reaction zone and they have been shared
by both domains, giving rise to the inter-region coupling explained
in Section 2.1.
A symmetric boundary condition has been imposed at the wall
describing the external duct. Before the CFD investigation, a grid
sensitivity analysis has been carried out for the pressure drop as
a function of the cell number, pointing out the lowest number of
cells granting a good resolution for the pressure drop. This trade-
Fig. 2. POCS randomizer working principle: each corner sphere can be randomly
displaced up to a distance fixed by the user (qMAX ) and the result will be a POCS off point was set to around 7  105 cells. The resulting mesh is rep-
having a well defined R% . resented in Fig. 6, where the main boundary patches are high-
4
A. Vespertini, A. Della Torre, G. Montenegro et al. Chemical Engineering Science 267 (2023) 118309

Table 1 Table 2
Main POCS parameters dependence on R% : mean strut length (considering 43 cells The concentration of each specie at the inlet is reported in terms of the mass fraction.
POCS and 2003 cells POCS), porosity and surface area.
specie mass fraction X i
R% < lS > ½m < lS > ½m  Sw ½m2  at the inlet
[43 cells] [2003 cells]
CO 0:1
0 5:00  103 5:00  103 89:7 2:63  103 O2 0.12
0 5.00  10-3 5.00  10-3 91.4 2.44  10-3 N2 0:78
12 5.09  10-3 5.07  10-3 91.2 2.49  10-3
25 5.32  10-3 5.28  10-3 90.9 2.61  10-3
35 5.68  10-3 5.54  10-3 90.5 2.72  10-3
45 5.96  10-3 5.89  10-3 90.0 2.85  10-3 ber, given by ReP ¼ qUd
l . As stated in (Della Torre et al., 2014), for
P

50 6.11  10-3 6.10  10-3 89.7 2.92  10-3


Re < 150, the Darcy-Forcheimer law for incompressible flow
describes the flow of a fluid through a porous medium with the fol-
lowing relation:
lL
DP ¼ U D þ qCLU 2D ; ð10Þ
k
where C is named Forcheimer or non-Darcian coefficient and, being
included in the second term, is known as the inertial coefficient, as
it is stated in (Lenci et al., 2022). In this work, the local Forchheimer
coefficient was also expressed as a function of the local permeabil-
ity value with an inverse power relationship. For ReP > 150 at first
laminar wake oscillations appear, then the flow becomes fully chao-
tic. The turbulence modelling has been handled by adopting the
low-Re k-OmegaSST RANS model (Menter and Esch, 2010). From
Eq. 10, through interpolation of pressure drop curves against veloc-
ity, the permeability for different values of R% can be extracted. As
Fig. 5. Mesh details for the fluid(blue)-solid(red) regions: an higher refinement already mentioned, firstly a light-off analysis with respect to fluid
level is applied to the inter-region boundary faces. and washcoat temperature was carried out, looking for the transi-
tion temperature between kinetically and diffusion controlled
regimes and to the differences due to randomization level.
Finally, the analysis of randomized POCS enables us the possi-
bility to compare different correlations for the tortuosity s. In the
literature (O’Connel et al., 2010), a very simple, merely geometri-
cal, definition of tortuosity can be found, which can apply in vari-
ous fields:
A
sGeom ¼ ð11Þ
c
representing the arc–chord ratio of a circular path. Furthermore, in
(Duda et al., 2011) an alternative definition of tortuosity can be
given, defining V as the set of fluid cells belonging to the POCS void
space:
X jU~i j j~
Lt j
sPhys ¼ ~
’ ð12Þ
cells2V jU axial;i j
L
Fig. 6. Detail of the main boundaries of the mesh: inlet (red), outlet (black), fluid–
solid boundary (light-blue) and surrounding walls (light-gray). in which the ratio depends on the orthogonal to motion velocity
components respect to the axial ones, while the second equivalence
is given by (Civan, 2007) and represents the ratio between preferen-
lighted in order to comprehend the case setup, explained in the fol- tial fluid pathways and the porous media length. To compare these
lowing section. ~ in Eq. 12 we
two definitions, in order to evaluate velocity field U
removed the inertial term r  ð/U Þ from momentum equation: in
2.5. Case setup and fluid flow modelling this way, a fictitious Stokes flow condition is imposed for all Re
numbers considered, meaning that the flow direction is mainly
At first, the mass flow rate at the inlet patch was fixed at dependent on the microstructure geometry.
U ¼ 0:002 kgs and POCS with different R% have been tested, being
the fluid and solid regions initialized at different temperature con- 3. Results and Discussion
ditions. The concentration of each specie at the inlet is given in
Table 2 and the whole system has been put at ambient pressure. As a first step of the analysis, the dependency of the mass trans-
The aim of this first part was to find the light-off temperature of fer on the randomization index R% was investigated. Results are
the reaction and also the identification of the pure mass transfer reported in Fig. 7 considering the following operating conditions:
regime. Secondly, the temperature of each component is raised mass flow rate / ¼ 0:002 kgs , temperature of inlet flow and substrate
up to 1000K, ensuring that the conversion of CO is limited only in the range T 2 ½400; 900K and randomization percentage
by the mass transfer properties of the structure. Within this second R% ¼ ½0; 12; 25; 35; 45; 50. CO conversion percentage is calculated
analysis, we adopted a pore-based definition of the Reynolds num- with the following equivalence:
5
A. Vespertini, A. Della Torre, G. Montenegro et al. Chemical Engineering Science 267 (2023) 118309

Fig. 7. Light-off curves for different randomization percentage R% at constant mass


flow rate / ¼ 0:002 kgs and initial boundary condition T fluid ¼ T solid ¼ T washcoat ¼ Fig. 8. Pressure drop plots against darcean velocity for different R% . The pressure
T 2 ½400; 900K. drop behaviour of a cubic cell POCS with the same porosity of the R% ¼ 50 one is
reported. The correlation proposed by (Klumpp et al., 2014) for non tilted cubic cell
POCS is also displayed. All boundary and bulk temperatures are set to
! T fluid ¼ T solid ¼ T washcoat ¼ 1000K.
C CO;outlet  C CO;inlet
gCO ¼ ð13Þ
C CO;inlet
Table 3
where every concentration of CO is calculated by averaging the val- Permeability for each POCS randomization percentage R% .
ues of the inlet and outlet patch, respectively. The plot highlights
that the light-off temperature is not influenced by the randomiza- R% permeability k form coefficient C
[m2 ] 1
[m ]
tion level. On the other hand, the randomization index R% has a sig-
nificant effect on the conversion under diffusion control regime: 0 6:70  107 6:31  101
this effect is related primarily on the enhancement of the mass 12 6:54  107 7:04  101
25 7
transfer induced by the randomized structure, secondarily by the 6:29  10 7:53  101
increase of the surface area, as pointed out in Table 1. After the 35 5:90  107 7:94  101
7
light-off study, a condition of pure mass transfer regime was tested: 45 5:21  10 8:53  101
the inlet flow and washcoat temperatures were set at 1000 K, while 50 5:17  107 9:62  101
the Re number was varied with the aim of investigating the differ-
ent flow regimes. The pressure drop generated across the POCS is
plotted against Darcean velocity in Fig. 8: in this graph, the effect is the concentration on the i-th face of the wall. The carbon monox-
of different randomization levels R% on the pressure drop is high- ide concentration at the wall is negligible because of the assump-
lighted, showing that the form coefficient C is increased by the ran- tion of Eq. 4 that ensures the absence of reactants nearby the
domization process (Table 3). This is due to the increasing surface walls. At low Re number the diffusion in the stream-wise direction
area exposed to the flow at higher R% and, as it is pointed out in is predominant with respect to convection and this implies that the
(Lenci et al., 2022), permeability and form coefficient behave inver- CO concentration is expected to drop exponentially through the
sely proportional. POCS. This fact ensures that averaging the local mass transfer coef-
To point out the mere effect of the randomization procedure on ficients on the whole domain will lead to inconsistent results. For
the form coefficient, the same quantities have been plotted for a this reason, the global mass transfer coefficient km had to be calcu-
R% ¼ 0 POCS having the same porosity of the R% ¼ 50, showing that lated following these steps:
porosity and wet surface play both a crucial role on the pressure
drop. Finally, in (Klumpp et al., 2014) a correlation without empir- 1. the fluid domain was split into four slices according to Fig. 9;
ical factors was proposed for the pressure drop estimation of non- 2. by adopting Eq. 15, the local mass transfer coefficient for each
tilted cubic open cell structures having different porosities: cell i has been calculated, using as bulk concentration its mean
DP A 1   qf  aV 2 value within the belonging slice;
¼   U ; ð14Þ
L 1  A  3 inl 3. for each slice, all local mass transfer coefficients km;i have been
averaged, giving as a result the mass transfer coefficient of slice
where A is an area-related porosity, defined by the authors and aV i;
is the specific surface area of the POCS. The correlation contained in 4. finally, the global mass transfer coefficient has been calculated
Eq. 14 has been added to the chart, showing that our result for by averaging its local value in the two central slices.
R% ¼ 0 fits well with the aforementioned correlation. By fitting
these curves, from Eq. 10 the permeability and the form coefficient The values representing last domain has not been considered
can be extracted for each R% from the Darcy-term, demonstrating because, at low Re number, the latest domain is not reached by
that the randomization process slightly decreases the POCS perme- the reactant as it can be seen in Fig. 10: this figure also proves
ability (Table 3) proportionally to the increment of the wet surface the need for the application of this procedure to calculate the mass
of the microstructure (Table 1). To quantify the mass transfer transfer coefficient. Furthermore, its value on the first slice was
enhancement related to the randomization process, separating the discarded because of the fluid-dynamic effects at the POCS
contribution of the increased wet surface, the local mass transfer entrance that significantly vary the result.
coefficient for the i-th cell km;i has been evaluated as follows: The calculated global mass transfer coefficients are plotted
ji ½CO against darcean fluid velocity in Fig. 11, where the increasing mass
km;i ¼  ; ð15Þ transfer properties due to the POCS shape are visible: higher R%
Sw;i  C f ½CO  C w;i ½CO
POCS exhibit better mass transfer properties in the whole fluid
where ji is the molar flow calculated within the i-th cell of the react- velocity domain. Furthermore, by confronting these results with
ing region, C f is the bulk concentration in the fluid domain and C w;i the pressure drop curves, it’s visible that the loss in porosity at

6
A. Vespertini, A. Della Torre, G. Montenegro et al. Chemical Engineering Science 267 (2023) 118309

qU 2D
IG ¼  lnð1  gÞ : ð16Þ
DP
Eq. 16 represents the trade-off between mass transfer and pressure
drop generated by the POCS: it is worth noticing that, as can be seen
in Fig. 12a and 12b, lower performances are reached by the R% ¼ 50
POCS, meaning that in this case the increase of pressure drop, asso-
ciated to the inertial contribution, is higher than the related
enhancement of conversion capability. This behaviour can be
explained by the fact that R% ¼ 50 POCS represents an extreme con-
dition, in which the displacement radius (Fig. 2) of the corner
sphere is equal to the strut length. This condition implies that
around R% ¼ 50 the performance index saturation point is reached.
Furthermore, the cubic cell POCS having the same porosity of the
R% ¼ 50 one shows worse performances then the others: this high-
Fig. 9. Detail of the fluid domain splitting for the R% ¼ 0 POCS to calculate an lights the effect of porosity and shows that the randomized POCS
average mass transfer coefficient. behaves better than the cubic cell one having the same properties.
Finally, it can be noticed that the present results are consistent with
the one obtained by (Papetti et al., 2018) for rotated cubic POCS, in
terms of both trends and absolute values of the mass transfer.
In order to better asses the different flow conditions for each Re
number, an analysis with non-dimensional coefficients, already
introduced in (Della Torre et al., 2014), has been carried out. In par-
ticular, the pressure gradient across the POCS can be expressed as a
function of the variables that likely influence it:
DP
¼ f ðU D ; q; l; dc Þ: ð17Þ
L
Using the Buckingham P-theorem, two different non dimensional
quantities can be derived:
 
DPdc
2
qU D dc
P1 ¼ ¼ f 1 Re ¼ ; ð18Þ
LU D l l
Fig. 10. Mass fraction of CO inside the R% ¼ 0 POCS against the position on the z-
axis at Re ’ 0:3; the position is taken with respect to the beginning of the POCS.

Fig. 11. Mass transfer coefficient km for CO is plotted against darcean fluid velocity
for different R% . The mass transfer coefficient behaviour of a cubic cell POCS with
the same porosity of the R% ¼ 50 one is reported. All boundary and bulk
temperatures are set to T fluid ¼ T solid ¼ T washcoat ¼ 1000K in order to perform these
tests in pure mass transfer regime.

higher R% , inherent to the randomization procedure, has an higher


impact on the pressure drop than on the mass transfer coefficient,
meaning that the surface area increment plays a crucial role on the
value of mass transfer coefficient.
The purpose of introducing a randomization procedure is to Fig. 12. 12a Performance index represented with respect to fluid darcean velocity:
increase the mass transfer, in such a way as to achieve a positive R% ¼ 50 represents the worst trade-off point between mass transfer and pressure
drop for every Re value. 12b The gain in performance index with respect to R% ¼ 0 is
balance between mass transfer and pressure drop. To better point plotted against randomization percentage R% at different fluid velocity. All
out this balancing, a performance index has been established boundary and bulk temperatures are set to T fluid ¼ T solid ¼ T washcoat ¼ 1000K in
according to the definition reported in (Giani et al., 2005): order to perform these tests in pure mass transfer regime.

7
A. Vespertini, A. Della Torre, G. Montenegro et al. Chemical Engineering Science 267 (2023) 118309

 
DPdc 1 l
P2 ¼ ¼ f2 ¼ ; ð19Þ
LqU 2D Re qU D dc

and then plotted with respect to Re number. From P1 it is reason-


able to expect that all the POCSs with different R% collapse to a sin-
gle solution at low Re number, which represents the pure viscous
(Darcy) regime. Furthermore, two different linear trends should
be visible, representing the Darcy Forcheimer regime (for
10KReK150) and post-Forcheimer regime (for ReJ150). This is
what is shown in Fig. 13awhere it is visible that for ReJ10 the lin-
ear trend depends on R% , showing a steeper slope at higher random-
ization percentage. Furthermore, it is visible in Fig. 13b that for low
flow velocities (corresponding to Re  0) the different plots con-
Fig. 14. P2 is represented with respect to Re number in double logarithmic scale.
verge to a single solution, which represents the purely viscous
motion across the POCS: this solution coincides, for each value of
R% , because the increment in wet surface is negligible for the flow and foams:
condition. Then, for higher velocity, the Forcheimer inertial term
!0:43
gains more importance, leading to a marked distinction related to Re 1
Sh ¼ 0:91 2 Sc3 : ð21Þ
the different randomization percentage. In particular, a change of pffiffiffiffi ð1  Þ
3p
slope can be recognised around Re  200, marking the transition
from a laminar flow condition to a fully turbulent one. The same In those equations, the Sh number dependency on Re can be gener-
concepts are highlighted by Fig. 14: P2 at high Re number should
ally considered to be as Sh / Rea , where awire ¼ 0:283 and
asymptotically tend to the constant value, representing the fully
afoam ¼ 0:43. Sh number has been evaluated, for all the differently
turbulent regime, while at low Re number it behaves as lkL Re
1
. These randomized structures with the following equation:
two trend are shown in the plot and they are consistent with
Fig. 13a and Fig. 13b. km dP
Sh ¼ ð22Þ
An interesting feature of the randomization process is the pos- Di
sibility of changing, with a continuous transformation, the mass
transfer performances of cubic cell POCS up to those of foams. This where dP is the pore diameter and Di is the diffusivity of the i-th
can be demonstrated by adopting the general correlations pro- specie. Fig. 15 shows the dependency of the Sh on the Re number
posed by Reichelt and Jahn (2017) for wire-mesh structures for different randomization levels. At low Re number the differently
(whose cross sectional area is similar to the cubic cell): randomized POCS collapse to a single solution, according to what
 0:283 has been shown with the P numbers. This happens because the
Re motion inside the structure is governed by viscosity and the incre-
Sh ¼ 0:94 ð20Þ
 ment of the surface area does not have effects on the mass transfer:
these behaviours are in agreement with the findings reported in
(Meinicke et al., 2020 and Pallares and Grau, 2010). For ReJ10,
the behaviour becomes linear and the strut alignment becomes
negligible. The surface area exposed to the flux is incremented, dif-
ferentiating the curves: when the inertial effects become more
important (ReJ150), the POCS shape enhances the convection pro-
cess R% . Finally, it can be noticed that at high Re number the curves
approach the correlation proposed by (Ferroni et al., 2021) as the
randomization percentage increases.
To confront these values with the proposed correlations, only
data with ReJ20 have been used, or when the behaviour of the
curves become linear. By performing a linear interpolation of the
double logarithmic scale, for every R% values, we obtain the power
coefficients reported in Table 4, from which is visible the continu-

Fig. 13. 13a P1 is plotted against Re, while 13b is the same plot in double Fig. 15. The Sherwood number (Sh) is plotted against the Reynolds number (Re).
logarithmic scale, from which is visible the convergence at low value of Re. All Eq. 21 has been plotted to compare the different behaviours. All boundary and bulk
boundary and bulk temperatures are set to T fluid ¼ T solid ¼ T washcoat ¼ 1000K in order temperatures are set to T fluid ¼ T solid ¼ T washcoat ¼ 1000K in order to perform these
to perform these tests in pure mass transfer regime. tests in pure mass transfer regime.

8
A. Vespertini, A. Della Torre, G. Montenegro et al. Chemical Engineering Science 267 (2023) 118309

ous increasing trend. It is worth noticing that the bottom limit, rep-
resented by the wire mesh (Eq. 20), has exactly the same power
coefficient of the R% ¼ 0 POCS. Furthermore, the value obtained
at higher R% does not approach the power coefficient found in
Eq. 21: this is probably caused by the fact that they linearly inter-
polated all the values between 1KReK1000, while the linear beha-
viour is evident only at higher Re number. Finally in this figure it is
pointed out that, as the randomization percentage increases, the
behaviours of the POCS approach the correlation for foams pro-
posed by (Ferroni et al., 2021 and Reichelt and Jahn, 2017).
Finally, the fluid tortuosity s was evaluated at different flow
conditions for each R% POCS. Results are shown in Fig. 16: the grey
curve is the one obtained with the procedure mentioned in Sec- Fig. 17. Details of fluid-flow through R% ¼ 0 POCS on a slice perpendicular to the x-
tion 2.5 and it was found to be independent of the Reynolds num- direction for different Re. The presence of eddies is highlighted and the deviation on
jU j
the y-direction, calculated as Devy ¼ jU y j, is represented.
ber as expected. In the flow regimes considered, the calculated inl

values of tortuosity depend on three aspects: the presence of iner-


tial effects, the formation of eddies and the randomization percent-
age R% . At first, from Re ¼ 1:5 to Re ¼ 55 the inertial effects become
important, because the flow is entering the Darcy-Forcheimer
regime: this results in lower values of tortuosity due to the lower
adherence of fluid flow at the surface. By further increasing the
Re number, the flow becomes turbulent, causing the presence of
eddies (Fig. 17) behind struts orthogonal to the fluid velocity. It
can be noticed that this effect is magnified for the simple cubic cell
POCS, while its impact progressively decreases as R% increases.
This behaviour can be understood by focusing on the geometrical
point of view because, as the randomization percentage R%
increases, the velocity component of the fluid flow orthogonal to
each single strut generally lowers.
It is remarkable that, in the imposed Stokes flow condition, the Fig. 18. Details of fluid-flow at Re ¼ 1:5 around different R% POCS on a plane
tortuosity of each POCS is fixed and independent of the Re number, perpendicular to the x-direction.
making this value related to the only geometrical properties of the
POCS. Besided, to prove the goodness of this value, its value for
erties of the randomized POCS, i.e. its intrinsic tortuosity s. To
R% ¼ 0 can be confronted directly with the one calculated with
focus more on this details, in Fig. 18 a comparison between the
the relation proposed by (Klumpp et al., 2014), which is
fluid flow inside the microstructure for R% ¼ 0 and R% ¼ 45 POCS
s ¼ A ’ 1:02, which is consistent with the intercept of the grey
is reported: it is evident the effect of the POCS shape on the fluid
curve. This means that the average path of the fluid flow through streamlines, heavily deviating the flow direction in the case of high
the microstructure can be associated only to the geometrical prop- R% values.

4. Conclusions
Table 4
Power coefficients for Sh-Re interpolation.
In this work a numerical study has been performed, aiming at
R% a investigating the effects of POCS surface area on its permeability
0 0:285 and mass transfer performances. To this purpose, a randomization
12 0:300 percentage parameter has been introduced, which represents a
25 0:325 quantitative way to figure out the degree of struts misalignment,
35 0:333
50 0:352
starting from a simple cubic cell POCS. This parameter is intro-
duced in order to depend only on the strut length, making this
analysis easily replicable on other cubic cell structures. Since a
value of R% defines an ensemble of POCS, a further important inves-
tigation has to regard the main performance parameters variability
with respect to a single R% value. The main achievements of this
paper are:

 for the same condition of mass flow and reacting surface tem-
perature, the conversion capability is found to be strictly depen-
dent on the randomization level of the POCS structure;
 it was numerically shown that a simple cubic cell POCS can be
transformed into a randomized one through a continuous pro-
cess, behaving more and more similarly to a foam in terms of
mass transfer properties. This could represent a possible theo-
retical bridge between periodic cubic cell structures and foams.
Nevertheless, to strengthen the transition between cubic cell
Fig. 16. Plot of tortuosity s vs randomization percentage at different Re number.
and foam, a further analysis on the other parameters (i.e. pres-
The dashed grey line represents the simulations carried at Stokes flow condition for
every Re number value. sure drop, tortuosity etc.) has to be carried. Furthermore, it has

9
A. Vespertini, A. Della Torre, G. Montenegro et al. Chemical Engineering Science 267 (2023) 118309

been proved that for the maximum randomization level (i.e. Della Torre, A., Lucci, F., Montenegro, G., Onorati, A., Dimopoulos Eggenschwiler, P.,
Tronconi, E., Groppi, G., 2016. Cfd modeling of catalytic reactions in open-cell
50%) the performance index decreases, as a consequence of
foam substrates. Comput. Chem. Eng. 92, 55–63.
the strong increase of pressure drop, which is not balanced by Duda, Artur, Koza, Zbigniew, Matyka, Maciej, 2011. Hydraulic tortuosity in arbitrary
the gain in mass transfer; porous media flow. Phys. Rev. E. 84.
 a method for the evaluation of the tortuosity has been proposed Faure, Raphaël, Rossignol, Fabrice, Chartier, Thierry, Bonhommea, Claire, Maıtre,
Alexandre, Etchegoyen, Grégory, Del Gallo, Pascal, Gary, Daniel, 2011. Alumina
resorting to a fluid dynamic approach, namely the evaluation of foam catalyst supports for industrial steam reforming processes. J. Eur. Ceram.
the Stokes flow condition, where inertial effects are neglected. Soc. 31, 303–312.
This enables the possibility of evaluating the mean path length Ferroni, C., Bracconi, M., Ambrosetti, M., Maestri, M., Groppi, G., Tronconi, E., 2021. A
fundamental investigation of gas/solid heat and mass transfer in structured
associated to the complex microstructure, which can be related catalysts based on periodic open cellular structures (pocs). Industr. Eng. Chem.
directly to the definition of tortuosity, adopting a simple fluid Res. 126, 1035–1047.
dynamic simulation instead of complex geometrical methods. Giani, Leonardo, Groppi, Gianpiero, Tronconi, Enrico, 2005. Mass-transfer
characterization of metallic foams as supports for structured catalysts. Ind.
Further studies could investigate the relation between the Eng. Chem. Res. 44, 4993–5002.
slopes of Sherwood curves and the tortuosity of the fluid flow. Groppi, Gianpiero, Giani, Leonardo, Tronconi, Enrico, 2007. Generalized correlation
for gas/solid mass-transfer coefficients in metallic and ceramic foams. Ind. Eng.
Chem. Res. 46, 3955–3958.
Hutter, C., Büchi, D., Zuber, V., Rudolf von Rohr, Ph., 2011. Heat transfer in metal
Declaration of Competing Interest foams and designed porous media. Chem. Eng. Sci. 66, 3806–3814.
Kaur, Inderjot, Singh, Prashant, 2020. Flow and thermal transport through unit cell
topologies of cubic and octahedron families. Int. J. Heat Mass Transf. 158.
The authors declare that they have no known competing finan- Klumpp, M., Inayat, A., Schwerdtfeger, J., Körner, C., Singer, R.F., Freund, H.,
cial interests or personal relationships that could have appeared Schwieger, W., 2014. Periodic open cellular structures with ideal cubic cell
geometry: Effect of porosity and cell orientation on pressure drop behavior.
to influence the work reported in this paper. Chem. Eng. J. 242, 364–378.
Koltsakis, G., Konstantinidis, P., Stamatelos, A., 1997. Development and application
range of mathematical models for 3-way catalytic converters. Appl. Catal. B:
References Environ. 12 (2–3), 161–191.
Lenci, A., Zeighami, F., Di Federico, V., 2022. Effective forcheimer coefficient for
Ambrosio, Giuseppe, Bianco, Nicola, Chiu, Wilson K.S., Iasiello, Marcello, Naso, layered porous media. Transp. Porous Media 144, 459–480.
Vincenzo, Oliviero, Maria, 2016. The effect of open-cell metal foams strut shape Lucci, Francesco, Della Torre, Augusto, Rickenbach, Janvon, Montenegro, Gianluca,
on convection heat transfer and pressure drop. Appl. Therm. Eng., Volume 103, Poulikakos, Dimos, Eggenschwiler, Panayotis Dimopoulos, 2014. Performance of
pp. 333–343. randomized kelvin cell structures as catalytic substrates: Mass-transfer based
Azman, Abdul Hadi, Vignat, Frédéric, Villeneuve, François, 2015. Design analysis. Chemical Engineering Science, 112 143–151, 2014.
configurations and creation of lattice structures for metallic additive Lucci, Francesco, Della Torre, Augusto, Montenegro, Gianluca, Eggenschwiler,
manufacturing. 14e Colloque National AIP-Priméca, La Plagne (73). Panayotis Dimopoulos, 2015. On the catalytic performance of open cell
Bach, Christian, Dimopoulos Eggenschwiler, Panayotis, 2011. Ceramic foam catalyst structures versus honeycombs. Chem. Eng. J., 264, 514–521.
substrates for diesel oxidation catalysts: Pollutant conversion and operational Lucci, Francesco, Torre, Augusto Della, Montenegro, Gianluca, Kaufmann, Rolf,
issues. SAE Int. https://doi.org/10.4271/2011-24-0179. Eggenschwiler, Panayotis Dimopoulos, 2017. Comparison of geometrical,
Banhart, John, 2001. Manufacture, characterisation and application of cellular momentum and mass transfer characteristics of real foams to kelvin cell
metals and metal foams. Prog. Mater Sci. 46 (6), 559–632. lattices for catalyst applications. Int. J. Heat Mass Transf. 108, 341–350.
Bastos, Núria F., Rebelo, Kari Anne, Andreassen, Luis I., Ríos, Suarez, Piquero, Juan C., Sebastian Meinicke, Konrad Dubil, Thomas Wetzel, and Benjamin Dietrich.
Camblor, Hans-Jörg Zander, Grande, Carlos A., 2018. Pressure drop and heat Characterization of heat transfer in consolidated, highly porous media using a
transfer properties of cubic iso-reticular foams. Chem. Eng. Process.: Process hybrid-scale cfd approach. Internal Journal of Heat and Mass Transfer, 149,
Intensif. 127, 36–42. 2020.
Bracconi, Mauro, Ambrosetti, Matteo, Maestri, Matteo, Groppi, Gianpiero, Tronconi, Menter, F., Esch, T., 2010. Elements of industrial heat transfer prediction. 16th
Enrico, 2017. A systematic procedure for the virtual reconstruction of open-cell Brazilian Congress of Mechanical Engineering.
foams. Chem. Eng. J. 315, 608–620. O’Connel, Barry M, McGloughlin, Tim M, Walsh, Michael T, 2010. Factors that affect
Civan, Faruk, 2007. Reservoir formation damage (second edition): chapter 3- mass transport from drug eluting stents into the artery wall. Biomed. Eng.
petrographical characteristics of petroleum-bearing formations. Fundamentals, Pallares, J., Grau, F.X., 2010. A modification of a nusselt number correlation for
Modeling, Assessment, and Mitigation. forced convection in porous media. Internal Commun. Heat Mass Transf., 37.
Della, Torre A., Barillari, L., et al., 2021. Numerical assessment of an after-treatment Papetti, V., Dimopoulos Eggenschwiler, P., della Torre, A., Lucci, F., Ortona, A.,
system equipped with a burner to speed-up the light-off during engine cold Montenegro, G., 2018. Additive manufactured open cell polyhedral structures as
start. SAE Technical Paper. https://doi.org/10.4271/2021-24-0089. substrates for automotive catalysts. Int. J. Heat Mass Transf. 126, 1035–1047.
Della Torre, A., Montenegro, G., Tabor, G.R., Wears, M.L., 2014. Cfd characterization Reichelt, Erik, Jahn, Matthias, 2017. Generalized correlations for mass transfer and
of flow regimes inside open cell foam substrates. Int. J. Heat Fluid Flow 50, 72– pressure drop in fiber-based catalyst supports. Chem. Eng. J. 325.
82.

10

You might also like