1. Introduction
Modern automobiles powered by conventional internal combustion engines (ICEs) are subjugated to two major challenges: firstly, to comply with the increasingly stringent emission legislation, and secondly, to have the lowest possible carbon footprint by operating in a manner of best fuel efficiency. These challenges are being addressed either by replacing the existing power-train in battery electric vehicles (BEVs) or by operating ICEVs with so-called carbon-neutral fuels such as: bio-fuels or low carbon fuels such as hydrogen, natural gas, liquid petroleum gas, methanol, and synthetic fuels such as E-fuels, etc. Moreover, the complete replacement of traditional power-trains (ICEVs) by battery-operated engines (BEVs) is not foreseeable in the near future, rather, multiple power-trains will coexist [
1], with more and more focus on the usage of carbon-neutral fuels on existing power-trains [
2,
3,
4,
5]. Thereby, it is still imperative to enhance incylinder technology, also for the new fuels, together with the deployment of more advanced exhaust gas after-treatment systems (EGAS). This way, a definite incentive is associated with the development of an advanced EGAS to meet the prescribed existing and future emission norms.
A crucial part of an EGAS is the catalytic converter, in which the harmful species in the exhaust gas are converted or reduced into less-toxic pollutants. In general, a modern catalytic converter consists of ceramic or metal monoliths, which has a honeycomb structure with channel densities of 64–124 cells/cm
, and the catalytic substance is distributed on the channel surface in porous form [
6]. Thus, a large interface between exhaust gas and the catalytic substance is ensured by such a structure and an enhanced chemical reaction can be expected for effective reduction of pollutants. Since the last few decades, the catalytic converter has been widely investigated at the atomic level to understand the influence of various catalytic substances (i.e., Pt, Pd, Rh, etc.) on the oxidation and/or reduction efficiency of EGAS [
7,
8], understanding the complex and relevant chemical reaction kinetics [
9,
10], analyzing the process parameters on EGAS performance [
11] or in designing the waste heat recovery system to utilize the thermal energy of the exhaust gas [
12].
With regard to the numerical modeling and design optimization of EGAS, computational fluid dynamics (CFD)-based analysis is becoming a popular tool, since it can provide a comprehensive and macroscopic understanding of the coupled thermal and fluid flow dynamics inside the monolith. Additionally, it has the potential to reduce the overall designing cost and product life cycle. However, development of CFD-based numerical models of such a complex system itself is a challenging task, which requires careful consideration of numerical models at the individual component level. Focusing on the monolith of the catalytic converter, many studies have been carried out using the volume-averaging Navier–Stokes (VANS) methodology in the context of the Reynolds-averaged Navier–Stokes (RANS) model, in which the monolith is treated as a continuous porous medium, while the flow is resolved in a time-averaging manner [
9,
13,
14]. The combination of VANS and RANS is able to provide a macroscopic view of the flow in the catalytic converter with low computational cost and simulation time. However, it does not allow a detailed understanding of the highly transient coupled heat and fluid flow phenomena taking place inside the monolith. In particular, the VANS/RANS approach experiences serious drawback while dealing with the transient evolution of velocity, pressure, and temperature fields with their respective fluctuations in the narrow monolith channels. Therefore, a better understanding of the coupled heat and fluid flow phenomena in the catalytic converter can be achieved only by resolving the heat and fluid flow structures also in the individual monolith channel, as pointed out recently in the work of Cornejo’s group [
15]. They carried out simulations with different channel shapes to study their influence on the pressure drop across the monolith. Additionally, they analyzed the influence of upstream turbulence on the pressure drop in the catalytic converter based on the large-eddy simulations (LES) of a single monolith channel [
16]. Thus, they demonstrated the importance of resolving the honeycomb structure of monolith and the applicability of the LES-based investigation of the complex EGAS.
As already pointed out in previous paragraphs, the catalytic converter features a complex and coupled thermofluid flow phenomena, making the design and process optimization a challenging task. In this aspect, by following the second law of thermodynamics, the entropy-generation-based irreversibility analysis of such a system can be a useful tool for both detailed understanding of the underlying physical phenomena and in identifying the key parameters for efficient operation of the EGAS. Such analysis is essentially based on the fact that a reduction in the thermodynamic efficiency of a thermal device is essentially associated with increase in the entropy production of this system, also known as entropy generation [
17,
18]. Thereby, the analysis based on entropy generation minimization (EGM) can be readily adopted, first to characterize the real industrial systems such as cryogenics, heat transfer, storage, solar energy conversion, power plants, refrigeration plants, etc., and then to optimize them for the most energy-efficient operations [
19,
20]. As demonstrated also in our previous contributions [
21,
22], an analysis based on entropy generation is a suitable approach to provide deeper insight about the coupled thermofluid flow processes, identify specific flow features, and choose the optimal operating point with minimum energy lost. A detailed description of entropy generation analysis, in particular in the context of LES, can be found in [
21] and elsewhere. Several investigations based on entropy generation for different technical configurations and physical processes have been reported [
18,
22,
23,
24,
25,
26,
27,
28,
29,
30,
31,
32,
33]. However, a study based on the entropy generation analysis for the realistic automobile catalytic converter has not yet been reported.
The lack of numerical investigation on the heat and turbulent flow in the catalytic converter using LES, especially to resolve the heat and fluid flow in all channels in the monolith, together with the absence of a study based on the entropy generation for such configuration, provides motivation to the present study. Thereby, the objective of this work is to propose the methodology for the prediction of heat transfer and fluid flow effects on entropy generation in a realistic automobile monolithic catalytic converter using LES. Notice that chemical reaction is not considered in this work and left for the future research. In the present work, the thermofluid dynamics in a realistic three-way catalytic converter of a Lada Niva 21214 with/without (conjugate) heat transfer is studied using LES methodology. To validate the applied numerical method, experimental measurements were also carried out, in which the pressure drop between the inflow and outflow of the monolith is measured for different volumetric flow rates. The present work is organized as follows. The applied numerical methods is outlined in
Section 2. The geometry and operating conditions of the catalytic converter, as well as the numerical and experimental setup are described in
Section 3. In
Section 4, the results are presented and discussed in detail. Finally, the main findings of this study are highlighted in
Section 5.
2. Large-Eddy Simulation
In the present numerical study, the solution domain consists of two parts, namely a fluid and a solid domain. In accordance with the procedure described in our previous works [
33,
34], the fluid part is governed by the balance equations for incompressible Newtonian fluid flow with variable physical properties and Fourier heat transport, while in the solid part only the energy equation is solved due to nonexistence of any flow. Thereby, in the context of LES, the low-Mach number (
) formulation for the balance laws of mass, momentum, and energy are employed and given as (see [
35,
36]):
where
and
are the filtered and Favre-filtered quantities, respectively.
expresses the mass density,
the velocity,
h the sensible enthalpy,
P the mechanical pressure,
the molecular viscosity,
the thermal conductivity,
the isobaric heat capacity of the exhaust gas,
is the subgrid-scale viscosity, and
is the subgrid-scale Prandtl number. Regarding the modeling of subgrid-scale momentum transport, the Sigma model (
-model) proposed by [
37] is applied, while the linear thermal diffusivity model with constant subgrid-scale Prandtl number of
is employed to model the unresolved heat flux.
For the solid part, the classical heat equation is valid, and reads:
Here,
represents the temperature and
the thermal diffusivity of the solid domain. The solid and fluid regions are coupled via a thermal interface, where the temperature and the heat flux of both phases have to be equal, which leads to the following constraints at the fluid–solid interface:
where the terms with superscript
and
represent the quantities in the fluid and solid region, respectively. The thermal conductivity of the solid region is given by
, and
n is the direction normal to the solid surface.
In this work, the working fluids, namely the air and the exhaust gas, are treated as homogeneous single component Navier–Fourier gases. Thereby, the thermodynamic and transport properties are calculated by means of seven-coefficient NASA polynomials [
38], and transport properties are approximated based on Sutherland’s formula [
39]. The material properties of the air and exhaust gas are evaluated using the material laws of the mixture. A detailed description about the adopted methodology can be found in [
36]. The species composition and the thermodynamic properties of the exhaust gas are summarized in
Appendix A.
4. Results and Discussion
In this section, the obtained numerical results for nonreacting turbulent heat and fluid flow characteristics of an exhaust-after-treatment system of a Lada Niva 21214 are discussed in detail. First, in the case of isothermal flow conditions (cases 1–9 of
Table 2), predictions of the pressure drop are compared with the experimental data. Then, heat and fluid flow features inside the monolith are analyzed under nonreacting nonisothermal conditions, and finally, entropy generation rates due to viscous dissipation and heat transport within the catalyst are characterized (cases 10–12 of
Table 2). Notice that the heat and fluid flow inside the catalyst and the entropy generation rates are only analyzed using numerical methods since such quantities are obviously very difficult to quantify experimentally inside a three-way catalytic converter.
4.1. Comparison with Experimental Data (Case 1–12)
In an automobile engine, the exhaust back pressure is a key design parameter and plays a determining role in efficient engine operation, while the affordable back pressure of an engine largely depends of the pressure drop across various components along the exhaust after-treatment circuits. In this regard,
Figure 4 shows the comparison of measured and simulated mean pressure drops as a function of the volumetric flow rates for a three-way catalytic converter employed in Lada Niva 21214. The experimental data is presented in a black circle and simulation data are shown as crosses, while the red solid line denotes a best fit of the measured pressure drop based on the Darcy–Forchheimer law [
41]. The dashed red line shows the corresponding Darcy–Forchheimer law for the hot exhaust gas (
K). Additionally, the pressure drop between the inflow and outflow side of the catalytic converter can be modeled very accurately by means of the Darcy–Forchheimer equation. For the present catalyst the Darcy–Forchheimer equation reads [
41]:
with the fitted coefficients
K = 1.7077 × 10
and
. The first term on the RHS describes the pressure drop due to friction, while the second term is related to turbulent dissipation. By means of this equation, the experimental data can be extrapolated to the case with higher volumetric flow rates and nonisothermal conditions. Notice that the quantities for hot exhaust gas are evaluated based on the averaged bulk temperature in the monolith obtained by the numerical simulation with fine grid resolution.
As it can be seen in
Figure 4, predicted mean pressure drops obtained from the simulation compared well with the experimental data and also with the derived Darcy–Forchheimer relation. This holds true for both cold air flow and hot exhaust gas flow. Therefore, good agreements are achieved for all flow rates under consideration. However, it can be observed that the predicted pressure drops from the simulation are slightly lower than the measured ones. This is mostly visible in the results from the cases with coarse grid and higher flow rates.
Besides the total pressure drop of the catalytic converter, it is also of interest to compare predicted skin friction at the duct walls in the monolith, which is directly connected to entropy production related to viscous dissipation. In the case of fully developed laminar duct flow, the friction coefficient can be calculated analytically as:
with a Poiseuille number of
for a duct with square cross-section shape (see [
42]). This leads to a classical value of
for case 9 with a Reynolds number of
of the duct flow at the center of the monolith. The friction coefficient predicted by the LES simulation is found to be
, which is slightly lower than the analytical value.
In summary, deviations of LES predictions with the finest grid resolution are relatively small (∼10%) considering such a complex test case of a realistic catalytic converter. This allows us to conclude that the present numerical approach is suitable to predict the key operation parameters of the catalytic converter. It can therefore be used for further analyses of the heat and fluid flow properties and entropy production inside the catalytic converter.
4.2. Predictions of Heat and Fluid Flow Features (Case 9 and 12)
Along with other factors, the efficiency of a catalytic converter is largely determined by how uniformly the flow and thermal fields are distributed upstream to the monolith. This allows the effective utilization of available surface area in the monolith for the intended purpose of particulate filtration, oxidation, reduction, etc. In this context,
Figure 5 shows a snapshot of the instantaneous magnitude velocity field of the isothermal case with
= 160.0 m
/h (case 9) together with the velocity distribution at several cross-sections (S1–S5) and the pressure distribution along the the monolith channel (see
Figure 5, top right).
A highly turbulent intake jet can be observed that impinges on the monolith, resulting in recirculation and separation flow in the front of the monolith. Therefore, the pressure around the stagnation point is high and decreases rapidly in the flow direction. Traveling further downstream, the flow gets laminarized inside the monolith and the pressure distribution becomes homogeneous in the radial direction. It is also interesting to observe that even the velocity is small at the stagnation region; the velocity inside the monolith is high around the center line and decreases in radial direction. This resultant velocity profile inside the monolith remains constant further downstream (see the velocity profile in sections S2–S4). This observation is analyzed quantitatively in
Figure 6 and
Figure 7 for the predicted mean and root mean square (rms) velocities in longitudinal direction, respectively, at the cross-sections S1–S5. Here,
D is the diameter of the inlet pipe and
r is the radial coordinate that points to the center (
is located at the duct or monolith center line). The quantities of the isothermal case (case 9) are plotted with black lines, while for the nonreacting nonisothermal case (case 12) they are plotted with red lines. Notice that dashed lines in the cross-sections S2–S4 represent the velocity profile inside the monolith region.
It can be seen in
Figure 6 that the mean velocity profiles out of the monolith (S1, S5) are not significantly influenced by the heat transfer. In contrast, mean velocity profiles for case 9 and case 12 differ significantly within the monolith. Therefore, the velocity profile in the nonreacting nonisothermal case is of uniform shape along the radial direction, while the velocity magnitude is maximal at center line (
) and decreases gradually in radial direction in case of isothermal conditions. These observations can be mainly attributed to the change in thermophysical properties, in particular the change in the viscosity of the exhaust gas due to temperature change. Moreover, these properties remained constant under isothermal conditions. In the case of velocity fluctuations, it is observed in
Figure 7 that the rms velocity is high at the impinging region (S1) decreases rapidly in the monolith and remains small in the entire monolith. This holds true for both isothermal and the nonisothermal conditions, which allows us to conclude that the laminarization of the flow in the monolith is driven largely by fluid flow effects rather than heat transport phenomena. It should also be noted here that the rms velocity obtained for isothermal conditions is higher than that of nonisothermal conditions, especially along the impingement section highlighting the impact of increased viscosity due to higher temperature in the nonisothermal case.
The temperature distribution inside the catalytic converter is shown in
Figure 8 (case 12) for the fluid regions, the solid parts, and for several cross-sections (S1–S5) (see
Figure 8, bottom). Similar to the velocity field, the hot turbulent intake jet impinges on the front side of the catalyst. Therefore, the evolution of the thermal field inside the monolith appears to be very similar to that of the velocity field (see
Figure 5). However, temperature gradients are significantly larger at the solid–fluid interface than at the fluid–ceramic interface. This holds true for all cross-sections in the monolith (see
Figure 8, bottom (S2–S4)). The corresponding mean and rms temperature profiles along the radial direction are plotted in
Figure 9 and
Figure 10 for the cross-sections (S1–S5), respectively.
The exhaust gas temperature is relatively low upstream the monolith due to intense heat transfer caused by recirculation (see also
Figure 9, S1). In contrast, inside the monolith, at the cross sections S2–S4, the exhaust gas is heated up due to heat transfer from the ceramic mass and the temperature profile is more homogeneous, especially around the core of the monolith. This tendency is maintained downstream to the monolith at section S5. Such temperature distribution is essentially responsible for the increased exhaust gas viscosity around the core region, while the viscosity is lower towards the solid–fluid interface region. This allows us to explain the homogeneous velocity profile for the nonreacting nonisothermal case, as observed in
Figure 6. The rms temperature profile is shown in
Figure 10. At S1, a significantly higher temperature fluctuations can be observed, while it is more or less negligible inside the monolith (S2–S4) and to certain extent also at S5. The higher temperature fluctuation is persistent along the solid–fluid interface at all cross-sectional locations (S1–S5) with negligible temperature variations in the solid region.
4.3. Predictions of Entropy Production Rates
Based on the observations in the previous section, it appears that heat and fluid flow phenomena inside the catalytic converter are extremely complex and are essentially a coupled heat and fluid flow process together with turbulence transition inside the monolith. To obtain further insight about such complex and interacting processes, entropy generation-based analysis is carried out next.
In turbulent fluid flow with heat transfer, the entropy production rates can be separated into two parts, namely entropy production rate by viscous dissipation
and, due to heat transfer,
. According to [
21,
36], in the context of LES, the temporal averaged filtered entropy production rates
, and
can be calculated as the sum of resolved and residual part as:
with
the Obukhov–Corrsin constant [
43],
the Smagorinsky coefficient [
44], and
the filtered width.
The calculated time-averaged entropy production rate related to viscous dissipation
within the catalytic converter is provided in
Figure 11 for the isothermal test case (case 9). Therefore, a logarithmic color map is employed to visualize the wide range of entropy production scales.
As it is visible in
Figure 11 (for case 9), for the section S1, the higher entropy production rates
are located predominantly close to the wall where a steep velocity gradient exists, and it decreases rapidly towards the core flow region. However, considerably higher entropy generation rate
is observed inside the monolith (S2–S4). This observation can be attributed to the strong turbulent dissipation due to wall-bounded flow along the narrow individual monolith channels, which ultimately results in flow laminarization. Moving further downstream to the monolith (at section S5), the
profile features a classical pipe flow characteristic. In order to analyze the impact of heat transfer,
Figure 12 presents a comparison of
for isothermal (case 9) and nonisothermal (case 12) conditions. Notice that the dashed lines of S2, S3, and S4 represent values of
close to the ceramic wall. Outside the monolith (S1, S5), the distributions of
are very similar for isothermal and nonisothermal conditions. In contrast, values of
differ significantly for isothermal and nonisothermal conditions within the monolith (S2–S4). Therefore, values of
are significantly higher under isothermal conditons, which suggests that the irreversible laminarzation process is more intense for such conditions. Notice that the calculated
is higher in the core region of monolith under isothermal conditions (case 9, black dash line), implying the combined influence of carrier gas properties and flow profile (see
Figure 6) on the entropy generation rate due to viscous dissipation.
Next,
Figure 13 shows the time-averaged entropy production rates related to heat transport
for nonreacting nonisothermal conditions (case 12). The corresponding profiles of
are depicted in
Figure 14. Due to the intense heat transfer at the impingement region (see
Figure 13, S1 and
Figure 14, S1), strong temperature gradients occurs in the vicinity of the solid walls, which results in high values of
. In contrast, the value of
is considerably small inside the monolith (S2-S4), especially at the core of the monolith and less evenly distributed as compared with
. Furthermore, at location S5 (see
Figure 13 and
Figure 14), the entropy generation rate is primarily due to strong temperature gradients.
To provide a more global perspective of entropy generation in the catalytic converter, the integral values of
and
for the monolith region and pipe regions are listed in
Table 3 for both isothermal and nonisothermal cases under nonreacting conditions.
From
Table 3, it can be concluded that entropy production by viscous dissipation (
) occurs predominantly in the monolith region due to high-velocity gradients at the walls of the narrow monolith ducts, where a large amount of kinetic energy is dissipated through friction. This suggests that the laminarization inside the monolith is purely a fluid flow process. In contrast, irreversible thermal processes are more intense in the pipe regions, as the total value of
is quite high there. Furthermore, entropy production related to heat transport is approximately ten times higher than entropy production due to viscous dissipation for this specific catalytic converter.