Elsevier

Chemical Engineering Science

Volume 187, 21 September 2018, Pages 107-116
Chemical Engineering Science

Turbulence generation after a monolith in automotive catalytic converters

https://doi.org/10.1016/j.ces.2018.04.041Get rights and content

Highlights

  • Flow after a monolith was studied by LES and a discrete channel model.

  • Above certain Re, flow leaving a monolith generates turbulence.

  • Instability is caused by flow passing around the corners of the channels.

  • Tested RANS models did not give the same results than from LES.

Abstract

This work reports theoretical studies of flow behaviour in a monolith outlet zone for different Reynolds numbers covering laminar and transitional/turbulent flow regimes. Monolith type substrate is the core part of the automotive catalytic converter. Due to computational limitations, most numerical models of the converter represent the monolith as a continuum, averaging the effect of the solid and the open space on the flow. This strategy is useful to study the macro-structure of the flow, however, it does not capture the exact behaviour of an actual honeycomb type structure, especially at its entrance and exit. In this work, which is a continuation of the publication by Cornejo et al. (2018), a series of 3D LES and RANS simulations are performed using different discrete channel geometry to study and quantify the velocity fluctuations of flow leaving a monolith. The results show that above a certain Reynolds number the instability of the flow after the monolith is significant, leading to turbulence generation. The velocity fluctuations are mainly explained by the flow past the outlet of the monolith, and their magnitude is related to the Reynolds number based on the thickness of the walls between channels. An expression for this critical Reynolds number has been designed and verified against numerical simulations. Parametric studies are carried out to illustrate the influence of the Reynolds number on the appearance of flow fluctuations at the outlet zone of the monolith.

Introduction

Monolith based catalytic converters were initially developed for the automotive industry to reduce the pollutants present in the exhaust gas, however, due to its many advantages over other reactor types, such as a relatively low pressure drop and high external area, they are now used in other industrial applications (Xu and Moulijn, 1998, Heck et al., 2001, Marín et al., 2005, Tischer and Deutschmann, 2005, Schutt and Abraham, 2004). In recent years, there has been much work on the modeling of monolith reactors, (Sadeghi et al., 2017, Gu and Balakotaiah, 2017, Devatine et al., 2017, Martínez et al., 2016), phenomena inside micro channels Iwaniszyn et al., 2017, Avramenko et al., 2015, Avramenko et al., 2017 and model-based optimization Karamitros and Koltsakis, 2017. Given current computational limitations, numerical models of the converter usually represent the monolith as a homogeneous anisotropic porous medium, through a volume average approach (Bella et al., 1991, Baxendale, 1993, Kumar and Mazumder, 2010, Bertrand et al., 2012, Hayes et al., 2012). This approach often leads to an acceptable agreement with experimental velocity profiles right after the monolith (Hayes et al., 2012, Clarkson et al., 1993, Jeong, 2014), but it fails to represent some phenomena existing in an actual honeycomb structure, especially at the beginning and the end of the monolith (Cornejo et al., 2017). In particular, models utilizing the homogeneous porous medium approach, are unable to predict turbulence-monolith-turbulence interaction adequately. The flow entering the converter has a Reynolds number of the order of 104, it decreases to a magnitude of 102 inside of the monolith channels, and then increases back to 104 after leaving the monolith. Those Reynolds numbers imply flow regime transitions from turbulent to laminar and laminar to turbulent along the converter in a driving cycle. To obtain accurate results in modeling experimental data, decoupled from the effect of the turbulence, researchers carefully control the operating conditions to keep the flow in laminar regime (Holmgren and Andersson, 1998, Shuai and Wang, 2004, Gu and Balakotaiah, 2016, Engelbrecht et al., 2017). If those conditions are not met, the turbulence inside the channels might have a significant impact on the results (Khodadadian et al., 2018). Strom et al. (2011) studied the flow transition entering a monolith and its effect on the deposition of solid particles in the monolith entrance. Ekström and Andersson (2002) reported experimental evidence of the effect of the turbulence at the beginning of the monolith on the pressure drop along the substrate. Cornejo et al. (2017) analyzed the decay of the turbulence once the flow enters into a single channel under typical operating conditions, then proposed a damping of the turbulence for a two-equations eddy viscosity models that includes this behaviour in the continuum approach. Despite many works aimed at understanding the flow behaviour in the first part of the monolith, the laminar to turbulent transition at its last part, where the flow passes from inside the substrate into an open space, has not received much attention in the literature. It should be emphasized that this phenomenon is not appropriately represented by a homogeneous porous medium, because the turbulence generation is a consequence of the presence of the solid substrate. The turbulence after the monolith requires attention because it affects the overall pressure drop of the exhaust gas after-treatment system significantly, which results important for the engine operating efficiency and fuel economy. In addition, in close coupled monoliths, the outflow of the first impacts the performance of the second by changing its inlet conditions.

There are three main phenomena that might generate instability and turbulence in the flow leaving the monolith. These are the turbulence from inside the channels, the turbulence caused by the flow leaving a channel acting as a jet, and the instability from the stream around the last part of the solid walls between channels. The channels of the monolith are often approximately square and the turbulence inside rectangular ducts has been studied by Tosun et al. (1988). They reported a critical Reynolds number close to 1600 and a strong effect of the corners in the generation of turbulence. Xu et al. (2015) observed turbulence in single square jets with Reynolds numbers above 30,000. The flow around the end of a wall between channels is comparable to the flow around rectangular objects, which depends on the width of the object, which in the case of a monolith, is the wall thickness. For flow around rectangular objects, Bruer et al. (2000) reported that for Reynolds numbers above 300, based on the object width, turbulence is generated after the object. Also, between 90 and 300, the flow is unsteady laminar, meanwhile, below Reynolds 90, it is steady. Those studies considered a single jet and the flow around a single rectangular object. A monolith has thousands of channels running in parallel at a short distance from each other, where the flow might interact, increasing the instability and producing turbulence at lower Reynolds numbers. It is also notable that obtaining detailed experimental data inside and close to the monolith channels is non-trivial. Hettel et al. (2013) reported that the inclusion of probes inside a monolith channel might introduce up to 50% of error in the measurements. Hence, numerical experiments are preferable to study this phenomenon at a channel scale.

The objective of this work is the study and description of the velocity fluctuations and flow regimes after the monolith of an automotive catalytic converter. Several geometries in 3D commonly found in the literature were analyzed using large eddy simulation (LES) and Reynolds Average Navier-Stokes (RANS) models in a discrete channel geometry. The results were compared in terms of the average and standard deviation of the velocity, flow regime and turbulence kinetic energy. The main novelty of this work is that we confirmed numerically that at some given values of the Reynolds number, defined using the monolith wall thickness, the flow becomes turbulent. To the best of our knowledge, this problem is not addressed in the literature, and it is needed to design new computational models, that take into account the generation of turbulence after a monolith.

Section snippets

Computational model

The central point of this work is the characterization of a flow leaving a monolith, therefore, we considered the last part of a series of discrete channels followed by an open space as a domain. Although phenomena such as the heat produced by the chemical reactions and the irregularity of the washcoat might also affect the flow regime and distribution in an operating catalytic converter, we considered only isothermal unwashcoated channels. Under these conditions, it is reasonable to assume

Discretization and grid independence

The complete domain was discretized into a fully orthogonal and uniform grid, composed of four million hexaedra with 50 μm as characteristic length. The time step was set to obtain a maximum Courant number (CFL) of around 0.5, which is equivalent to time steps of the order of 10−6 s. Both, the characteristic size of the control volumes and the CFL set on the simulations are the result of a systematic study described later in this paper. The discretization schemes and solver settings are

Results and discussion

This section discusses the basic flow features occurring at the monolith outlet and their dependence on the Reynolds number. This work uses two different Reynolds numbers to analyze the flow regime, turbulent or laminar, after the monolith. First, the channel Reynolds number, based in the channel size, and second, the wall Reynolds number, referred to the wall thickness of the substrate. Their mathematical expressions are described in Eqs. (11), (12) respectively.Rec=ρuLcμRew=ρu2Lwμ

For both

Conclusions

The flow after an unwashcoated monolith at different conditions was successfully studied by LES with a discrete channel geometry. The results were analyzed in terms of the flow regime and average and standard deviation of the velocity. Flow instability downstream the monolith was observed and it generates turbulence under certain circumstances. It was found that the main source of instability is the flow going around the last part of the substrate and it is focused close to the intersections

Acknowledgements

I. Cornejo acknowledges the receipt of a Becas-Chile (CONICYT) scholarship. Other funding was provided by the Natural Sciences and Engineering Research Council of Canada.

References (39)

Cited by (19)

  • Optimization of a monolith reactor 3-D developed laminar flow model

    2021, Chemical Engineering Research and Design
    Citation Excerpt :

    Single channel models are considered insufficient when complex phenomena occur, such as varying rates of catalyst deactivation, thermal instabilities and interactions within and among channels, or varying diffusion-reaction rates. Examples of such phenomena are discussed in Zygourakis (1989), and Cornejo et al. (2018). Under these circumstances, reactor level model development is pursued.

  • Influence of upstream turbulence on the pressure drop inside a monolith

    2020, Chemical Engineering and Processing - Process Intensification
    Citation Excerpt :

    There is evidence in the literature that adding the section before the substrate is important to have a realistic velocity profile at the inlet of the channels [13,17]. The section after the substrate is also important, since turbulence can be generated there [14] affecting the total pressure drop. To quantify the influence of the upstream turbulence on the total pressure drop, cases with and without turbulence were analysed and compared.

  • CFD-Modeling of fluid domains with embedded monoliths with emphasis on automotive converters

    2020, Chemical Engineering and Processing - Process Intensification
    Citation Excerpt :

    This leads to an underprediction of turbulence production. Moreover, [31] show that even when some channels are fully discretized, capturing the details of individual jets coming out of the channels, the values of turbulence are now too large for the standard k-ε model. However, the usage of the k-ω SST model shows a similar level of turbulent kinetic energy as a LES (Large Eddy Simulation) but with a different spatial distribution.

  • A new approach for the modeling of turbulent flows in automotive catalytic converters

    2018, Chemical Engineering Research and Design
    Citation Excerpt :

    This study compares the flow leaving a monolith using a continuum model against the predictions of LES in a discrete channel geometry. All LES data at channel scale were taken from a previous work, to which the reader is referred for details (Cornejo et al., 2018b). In that work, the triggering of turbulence was analyzed by carrying out a series of systematic numerical experiments, where the percentage of kinetic energy resolved is above the 95% in each of them, and the source of such turbulence was carefully discussed.

View all citing articles on Scopus
View full text