Prediction of Flows in an Automotive Catalytic

Converter
2023 Systems of Signals Generating and Processing in the Field of on Board Communications | 979-8-3503-9772-7/23/$31.00 ©2023 IEEE | DOI: 10.1109/IEEECONF56737.2023.10092061

N. G. Churbanova M. A. Trapeznikova
Moscow Automobile and Road Construction State Moscow Automobile and Road Construction State
Technical University (MADI); Technical University (MADI);
Keldysh Institute of Applied Mathematics (KIAM), RAS Keldysh Institute of Applied Mathematics (KIAM), RAS
Moscow, Russia Moscow, Russia
[email protected] [email protected]

A. G. Churbanov V. V. Emets
Keldysh Institute of Applied Mathematics (KIAM), RAS Moscow Automobile and Road Construction State
Moscow, Russia Technical University (MADI)
[email protected] Moscow, Russia
[email protected]

Abstract—In this paper, predictions of laminar isothermal pressure drop through the monolith, which allows to increase
exhaust gas flows in an automotive catalytic converter are essentially the converter efficiency.
studied. The coupled calculations of incompressible viscous flows
in the “porous medium−free flow” system were performed using In this paper, we performed calculations of the laminar
generalized Navier-Stokes equations. The calculation algorithm flow of an incompressible isothermal fluid in a configuration
for this model was implemented using the open software FEniCS. that is typical for automotive converters. For this purpose, we
A typical converter geometry was considered in the two- used the open-source code FEniCS, which allowed us to
dimensional planar formulation. Two cases with different implement, in our opinion, a more accurate mathematical
monolith properties were calculated and compared. In the first model and an efficient algorithm for its calculation.
case, the monolith was homogeneous and had the cell density of
400 CPSI (Cells per Square Inch) everywhere. In the second case, The geometry of the converter was considered in the 2D
the monolith had two spatial zones − central and peripheral − planar formulation. Two cases differing in the monolith
which had cell densities of 900 CPSI and 400 CPSI, respectively. structure were calculated and compared. In the first case, the
The calculations clearly demonstrated the possibility to control monolith was homogeneous with the fixed cell density. In the
the flow uniformity in the monolith by constructing it with zones second case, the monolith had two zones differing in the value
with different properties. of cell density in its cross-section. The central zone of the
monolith had a denser distribution of channels with a higher
Keywords—converter; porous medium; generalized Navier- hydraulic resistance, which led to evident changes of the
Stokes equations; FEniCS; flow prediction velocity field in the monolith. The calculations showed the
possibility to influence the flow uniformity in the converter in
I. INTRODUCTION order to increase its efficiency by using zones with different
properties in the monolith.
Catalytic converters are used to reduce the toxicity of
exhaust gases from car engines. Modern converters employ
monoliths made of different materials and having a II. MATHEMATICAL MODEL
honeycomb structure, i.e. consisting of a large number of The gas flow field in an automotive converter consists of
longitudinal channels parallel to each other. regions of a free flow and saturated porous medium
CFD (Computational Fluid Dynamics) approach [1−3], (monolith). To describe such a “porous medium−free flow”
based on the numerical simulation of thermal and system, the model [4, 5] was developed, which allows
hydrodynamic processes, is widely used for investigation and considering a single system of hydrodynamic equations for the
optimization of processes occurring in converters at different entire calculation domain. The equations of this model are
regimes of their operation. Calculations can be performed derived from the Navier−Stokes equations for an
using models that have different spatial scales. The incompressible fluid using the procedure of averaging over a
computational domain can include either one or several representative elementary volume of the porous medium. This
monolith channels, or the entire monolith using different model was designed to describe high-speed flows in media
models of the porous medium. The main goal of such with a high porosity and takes into account the spatial
calculations is to achieve flow uniformity and minimize heterogeneity of porosity and permeability. Moreover, it takes
into account the effect of solid walls contacting with the
saturated porous body.

979-8-3503-9772-7/23/$31.00 ©2023 IEEE

Authorized licensed use limited to: Panimalar Engineering College - Chennai. Downloaded on August 06,2024 at 06:02:17 UTC from IEEE Xplore. Restrictions apply.
 Stationary laminar flows of an incompressible viscous III. SOLUTION PROCEDURE
fluid in the absence of thermal processes for the “porous For the numerical implementation of the above model, the
medium−free flow” system are governed by the following open source software FEniCS [6−8] was used. It was
generalized Navier−Stokes equations: developed for solving partial derivative equations of various
types by the finite element method.
ρ 1  1  2µ  The generalized Navier−Stokes equations (3), (4) were
∇ ⋅  u ⊗ u  = − ∇ ( φp ) + ∇ ⋅  ε(u)  − solved by the continuous Galerkin method with the
φ φ  φ  φ  (1) Taylor−Hood elements (P2−P1) [9]. To calculate the flow in
µ ρСF the converter, the following boundary conditions were used.
− u− | u | u, At the inlet of the inlet pipe, the parabolic velocity profile
K K
corresponding to the Poiseuille flow for a 2D planar channel
was specified. On all solid walls of the converter the no-slip
∇ ⋅ u = 0. (2) and no-permeability conditions were given. Modified “do-
nothing” conditions were imposed at the converter outlet [10].
To solve the system of discrete nonlinear equations, Newton's
Here ε(u) = 0.5(∇u + ∇uT ) is the strain rate tensor whereas method was employed, which showed a high calculation
ρ and µ stand for the density and dynamic viscosity of the speed.
fluid, respectively. The porous medium is described by the
parameters ϕ , K and СF , where ϕ is the porosity, K is the The computational meshes with triangular cells used in our
predictions were created with the open mesh generator Gmsh
permeability, and СF is the inertial resistance coefficient (the [11]. To visualize the obtained numerical results, the open
Forchheimer coefficient). It should be emphasized that in post-processing software ParaView [12] was utilized.
these equations u is the superficial velocity obtained by
averaging over the entire elementary volume containing both The numerical technique used in the present work was
the solid matrix and liquid, and p is the average pressure of the verified in our previous works [13, 14] on the well-known
liquid in the pore space [5]. The superficial velocity u is tests and showed a high efficiency and accuracy of the
related to the averaged velocity of the liquid in the pores u f calculations.
through the Dupuit−Forchheimer relation u = φ u f .
IV. NUMERICAL RESULTS
In the momentum conservation equation (1), the We consider a 2D planar catalytic converter consisting of
Darcy−Forchheimer drag term is nonzero only in the porous inlet and outlet tubes, between which there is a monolith and
medium. The variable porosity ϕ is given for the whole two diffusers adjacent to it on both sides. In the considered
“porous medium−free flow” system. It can have values in the configuration, both tubes have the same dimensions: width Dp
range 0 < ϕ < 1 in the porous medium and is equal to 1 in the and length Lp = Dp. The geometry of the converter has typical
free flow. Thus, the momentum equation in the free flow is values of the main parameters. Namely, both diffusers have
reduced to the Navier−Stokes equation, whereas in the porous the same angle of 40°, and the monolith has the width Dm =
medium we have the Brinkman−Forchheimer equation 2.5Dp and length Lm = 3Dp (i.e. Lc/Dp = 1.2). The geometry of
augmented with the convective terms. the converter is shown in Fig. 1 via the map of the variable
porosity φ where the monolith (black) has the typical value
In the case of forced convection, using reference scales of 0.78 whereas in the free stream (blue) we have ϕ =1.
the length L and velocity U0, equations (1), (2) can be written
for dimensionless quantities (denoted by the upper index "*")
as follows:

1 1  1  2 
φ φ  φ
( )
∇ ⋅  u* ⊗ u*  = − ∇ φp* + ∇ ⋅ 
 φRe
ε(u* )  −
 (3)
1 СF
− u* − | u* | u* ,
DaRe Da

∇ ⋅ u* = 0, (4)
Fig. 1. Geometry of the converter shown via the porosity map.
2
where Re = ρU 0 L / µ and Da = K / L are the Reynolds and The steady-state flow with Re = 1500, defined using the
Darcy numbers, respectively, and the value ρU 0 was used to2 width of the inlet pipe and the mean velocity at its inlet, was
calculated. This low enough Reynolds number allows us to do
normalize the pressure. calculations without turbulence models, which are still
unvalidated for flows in porous media [15].

Authorized licensed use limited to: Panimalar Engineering College - Chennai. Downloaded on August 06,2024 at 06:02:17 UTC from IEEE Xplore. Restrictions apply.
 Mesh independence studies have been done to obtain high-
fidelity numerical results in our predictions. Preliminary
calculations on a sequence of refining meshes have shown that
the numerical results obtained on the mesh of 437,386
triangular cells can be considered as mesh-independent.
In our model of the porous medium we need to describe a
honeycomb monolith formed by a system of longitudinal
channels. Thus, it is necessary to take into account the
essential dependence of its thermo-hydrodynamic properties
on the chosen directions. This means that it is necessary to use
different values of permeabilities Kx and Ky in order to protect
in calculations the flows of the liquid through solid walls of
separate channels. For this purpose, we used the transverse
Fig. 2. Flow trajectories colored using the velocity magnitude.
permeability Ky = 10-3K0, whose value was significantly less
than longitudinal Kx = K0 . Here K0 is the permeability value
defining the Darcy number in the governing equations. Such
an adaptation of the porous medium model to its channel
structure is in common use [16, 17].
The main problem in treating the monolith as a porous
medium is related to the definition of the parameters in this
model. The structure of the monolith, which in the cross
section is a grid of cells formed by longitudinal channels,
differs significantly from the standard porous medium
consisting of pellets of different shape. Moreover, here we are
dealing with high-speed flows, for which it is necessary to
determine numerically or experimentally the values of the
model parameters and take into account the nonlinearity in the
description of the hydraulic resistance. Fig. 3. Distribution of the pressure.
The most common in the automotive industry are
honeycomb monoliths formed by longitudinal channels with a
square cross-section. They have different cell densities, but the
most used is the monolith with the density of 400 CPSI (Cells
per Square Inch). Exactly that monolith was considered in the
first calculation. For the flow with Re = 1500, the
dimensionless parameters determining the hydraulic resistance
of the porous medium had values typical for this cell density.
Namely, the linear resistance term had the Darcy number
value Da = 10-4, and the coefficient for the quadratic resistance
term was equal to 3.7.
Fig. 2 shows the calculated trajectories of the flow
particles in the converter. In the front diffuser located after the
inlet tube, recirculation zones are observed near the side walls, Fig. 4. Velocity vectors colored using the ux value.
and the jet flowing on the monolith from the inlet tube spreads
through all of its channels. Fig. 5 demonstrates the profile of the velocity ux(y) at the
monolith transverse mid-section. The presence of the jet in the
The calculated pressure distribution is presented in Fig. 3. central region of the monolith leads to noticeable
It is easy to see that the maximum pressure gradients are inhomogeneity of the longitudinal velocity profile in it. It is
located in the central part of the diffuser in front of the well-known that increasing the flow rate at the converter inlet
monolith, providing redistribution of the flow entering into the (i.e., increasing the Reynolds number) only worsens the
monolith. situation and leads to a greater curvature of the velocity profile
Velocity vectors colored using the value of the in the monolith (see, e.g., numerical and experimental data
longitudinal velocity ux are depicted in Fig. 4. We observe given in [18]).
here that in the central region of the monolith velocity values
are higher than in the peripheral region adjacent to its walls.

Authorized licensed use limited to: Panimalar Engineering College - Chennai. Downloaded on August 06,2024 at 06:02:17 UTC from IEEE Xplore. Restrictions apply.
Fig. 5. Profile of ux(y) at the monolith transverse mid-section. Fig. 7. Velocity vectors colored using the ux value (dual cell density).

The idea to control the flow properties in the converter These changes in the flow structure can be observed more
using zones with different cell densities in the monolith is far clearly in Fig. 8. The calculated longitudinal velocity profile
from new (see, e.g., [19, 20]). It seems reasonable to use cells ux(y) in the mid-section of the monolith has no maximum in its
with a higher hydraulic resistance in the central region of the central region. Obviously, using more zones with different
monolith to eliminate here the undesirable jet. Moreover, it is properties, it is possible to obtain the desired degree of flow
possible to use several zones with different properties, in uniformity.
which the density of cells would vary along the radius of the
monolith and thus provide the desired uniformity of the flow
[19]. For such a numerical optimization, it is necessary to use
advanced models that adequately describe the complicated
processes under the consideration and to define accurately the
parameters of the model.
In this regard, we calculated the converter where the
monolith had two zones with different cell densities. In the
peripheral zone near the walls the cell density was still 400
CPSI, whereas in its central zone it was 900 CPSI. For the
zone with the cell density of 900 CPSI, we specified the
porosity of 0.73, Da = 9x10-5 for the linear resistance term and
5 for the coefficient at the quadratic resistance term. Fig. 6
shows the porosity distribution in such a heterogeneous Fig. 8. Profile of ux(y) at the monolith mid-section (dual cell density).
monolith with double cell density.
V. CONCLUSION
In this paper, we performed calculations of laminar
isothermal flows in an automotive catalytic converter. For this
purpose, the open numerical platform FEniCS was used. This
makes possible to implement, in our opinion, a more accurate
mathematical model compared to commercial CFD codes. A
typical converter geometry was considered in the two-
dimensional planar formulation. Two cases with different
monolith structures were calculated and compared. In the first
case, the monolith was homogeneous and had the fixed cell
density of 400 CPSI. In this case, the formation of the jet in
the central region of the monolith was observed. As the result,
Fig. 6. Monolith with the dual cell density. the evident inhomogeneity of the longitudinal velocity profile
in its cross section was observed. In the second case, the
As one would expect, this change in the monolith monolith had two zones with different cell densities of
parameters resulted in the disappearance of the jet in its 400 CPSI and 900 CPSI. The central zone of the monolith had
central region. As can be seen in Fig. 7, now the maximum the cell density of 900 CPSI and therefore had a higher
velocity values are observed in two parts of the zone with 400 hydraulic resistance, which resulted in changing the velocity
CPSI, adjacent to the walls of the monolith. field in the monolith. Thus, the calculations demonstrated that
it is possible to influence significantly the flow uniformity in
the converter by using zones with different properties in the
monolith.

Authorized licensed use limited to: Panimalar Engineering College - Chennai. Downloaded on August 06,2024 at 06:02:17 UTC from IEEE Xplore. Restrictions apply.
 REFERENCES [12] “ParaView - Open-source, multi-platform data analysis and
visualization application.” Available: https://www.paraview.org/
[1] S. Khalil, K. Pathan, Z. Ilahi, and S. Khan, “CFD Analysis of an (accessed Jan. 09, 2023).
Automobile Catalytic Converter to Obtain Flow Uniformity and to
Minimize Pressure Drop Across the Monolith,” CFD Letters, vol. 12, [13] A. Churbanov, N. Churbanova, S. Polyakov, and M. Trapeznikova,
pp. 116–128, Sep. 2020, doi: 10.37934/cfdl.12.9.116128. “Coupled Calculations of Flows in Domains Including a Porous Medium
and a Homogeneous Fluid,” Proc. 2021 Ivannikov Memorial Workshop
[2] O. Haralampous, and G. Koltsakis, “3D Simulation Models for After- (IVMEM 2021), Sep. 2021, pp. 17–22. doi:
Treatment Systems”, in 1D and Multi-D Modeling Techniques for IC 10.1109/IVMEM53963.2021.00009.
Engine Simulation, A. Onorati and M. Gianluca, Eds. Warrendale, PA:
SAE Int., 2020, pp. 417-456. [14] A. G. Churbanov, N. G. Churbanova, and M. A. Trapeznikova,
“Coupled prediction of flows in domains containing a porous medium
[3] R. E. Hayes, and I. Cornejo, “Multi-scale modelling of monolith and free stream”, Matematicheskoe modelirovanie, vol. 35, no. 1, pp. 34-
reactors: A 30-year perspective from 1990 to 2020,” The Canadian 50, Jan. 2023, doi: 10.20948/mm-2023-01-03. ISSN: 0234-0879.
Journal of Chemical Engineering, vol. 99, no. 12, pp. 2589-2606, 2021,
doi: 10.1002/cjce.24144. [15] M. J. S. de Lemos, Turbulence in Porous Media: Modeling and
Applications, 2nd ed. London, UK: Elsevier, 2012.
[4] P. Nithiarasu, R.W. Lewis, and K.N. Seetharamu, Fundamentals of the
[16] H. Ibrahim, W. Ahmed, S. Abdou, and V. Blagojevic, “Experimental
Finite Element Method for Heat and Mass Transfer, 2nd ed. Chichester,
UK: Wiley, 2016. and numerical investigations of flow through catalytic converters,”
International Journal of Heat and Mass Transfer, vol. 127, p. 546, Jul.
[5] D. A. Nield, and A. Bejan, Convection in Porous Media, 5th ed. Cham, 2018, doi: 10.1016/j.ijheatmasstransfer.2018.07.052.
Switzerland: Springer Int. Publishing AG, 2017.
[17] S. Porter, A. K. Mat Yamin, S. Aleksandrova, S. Benjamin, C. A.
[6] “FEniCSx,” FEniCS Project. Available: https://fenicsproject.org/ Roberts, and J. Saul, “An Assessment of CFD Applied to Steady Flow in
(accessed Jan. 09, 2023). a Planar Diffuser Upstream of an Automotive Catalyst Monolith,” SAE
[7] A. Logg, K.-A. Mardal, and G. Wells, Eds., Automated Solution of Int. J. Engines, vol. 7, no. 4, pp. 1697-1704, Oct. 2014, doi:
Differential Equations by the Finite Element Method: The FEniCS 10.4271/2014-01-2588.
Book, vol. 84. Berlin, Heidelberg: Springer, 2012. doi: 10.1007/978-3- [18] D. Tsinoglou, G. Koltsakis, D. Missirlis, and K. Yakinthos, “Transient
642-23099-8. Modeling of Flow Distribution in Automotive Catalytic Converters,”
[8] L. R. Scott, Introduction to Automated Modeling with FEniCS, Applied Mathematical Modelling - APPL MATH MODEL, vol. 28, pp.
Illustrated edition. Chicago, IL: Computational Modeling Initiative LLC, 775–794, Sep. 2004, doi: 10.1016/j.apm.2003.12.006.
2018. [19] J. Y. Kim, and S. Son, Improving Flow Efficiency of a Catalytic
[9] J. Donea, and A. Huerta, Finite Element Methods for Flow Problems. Converter Using the Concept of Radially Variable Cell Density – Part I.
Chichester, UK: John Wiley & Sons, 2003. SAE Paper 1999-01-0769. Warrendale, PA: SAE Int., 1999.
[10] V. John, Finite Element Methods for Incompressible Flow Problems, [20] I. Cornejo, G. Garreton, and R. Hayes, “On the Use of Dual Cell Density
vol. 51. Cham, Switzerland: Springer Int. Publishing AG, 2016. doi: Monoliths,” Catalysts, vol. 11, p. 1075, Sep. 2021, doi:
10.1007/978-3-319-45750-5. 10.3390/catal11091075.
[11] “Gmsh: a three-dimensional finite element mesh generator with built-in
pre- and post-processing facilities.” Available: https://gmsh.info/
(accessed Jan. 09, 2023).

Authorized licensed use limited to: Panimalar Engineering College - Chennai. Downloaded on August 06,2024 at 06:02:17 UTC from IEEE Xplore. Restrictions apply.