Soot loading and regeneration of diesel particulate filters
Dr. Christof Hinterberger & Dr. Mark Olesen, ArvinMeritor Emissions Technologies GmbH

During the regeneration of Diesel Particulate Filters (DPF), high thermal loads can arise that are particularly detrimental to the lifespan of the filter ceramic. Extreme temperature peaks can lead to disintegration of the filter material. High temperature gradients cause thermal stresses that can result in micro-cracks in the brittle filter material that adversely affect the filter efficiency. To optimize the durability, function, cost and weight of DPF layouts, a novel simulation technique based on STAR-CD has been developed at ArvinMeritor Emissions Technologies that allows a highly realistic threedimensional simulation of the thermofluiddynamic behaviour of diesel particulate filters during transient soot loading and regeneration.

Modeling Background

The DPF filter has a structure of individual channels that are open on one end and plugged on the other. The exhaust gas enters a channel, passes through the porous sintered channel walls, and exits via adjacent channels. In the process, flow-borne soot particulates deposit in and on the walls (Fig. 1). The DPF thus exhibits three interconnected flow regions: the inlet channels, the wall flow and the outlet channels. Variations in the local wall filtration areas within the DPF segments, and the presence of the cement strips between DPF segments, influence the local flow distribution, and hence the soot deposition. Instead of four filtration walls, channels that border the segment corners or the cement strips have only two or three filtration walls, respectively (Fig. 2b).

Since a detailed resolution of the flow within the DPF is impracticable (for calculation speed and meshing reasons), and since the wall flow is not a continuum, a macroscopic modeling approach is used. In the technique developed at ArvinMeritor, the DPF filter is partitioned into distinct inlet and outlet channel flow domains and a solid domain (Fig. 3), all of which are interlinked by special source terms. The inlet and outlet channel flow fields are modeled using anisotropic porosities, with the contraction and expansion losses at the entrance and exit of the DPF being modeled via baffles. The wall filtration mass flux is reflected through mass sources and sinks within the porosities and is proportional to the local pressure differential across the wall. The local flow resistances are affected by the soot deposition. The DPF material (including cement strips) is modeled as a separate solid that includes heat exchange to the surroundings. This approach allows the capabilities of STAR-CD, augmented with custom user subroutines, to solve the governing equations directly, without resorting to co-simulation. The resultant coupling of the flow fields inside and outside of the filter ensures an efficient numerical procedure.

Meshing / Model Preparation

The meshing of the cement strips and the core, perimeter and corner regions of the DPF segments is specified via a parameter control file that includes mesh and geometry parameters. From an initial global mesh that spans the entire flow domain – without regard to the DPF internals – an immaculate surface geometry can be extracted for subsequent meshing of the DPF segmentation. Intersecting the DPF base segmentation mesh with the extracted surface yields the final DPF region (Fig. 2a). For this operation, the  capabilities provided by CD-adapco's automatic meshing module are leveraged. Additional scripting is used to re-embed the DPF subdomain in a suitably modified global mesh, to set the numerical parameters, and to prepare the final simulation model for job submission. In addition to the significant time-savings afforded by the automated scripting (typical times for meshing and embedding a new DPF layout are in the order of minutes), potential user errors are reduced and the reliability of the relatively complex simulation process is vastly improved.

Simulation Results

A detailed three-dimensional soot distribution is shown in Figure 4. The local soot deposition is coupled to the mass flux through the wall, and hence to the local pressure difference across the filter wall. The effect of the higher flow resistances in the channels adjacent to the cement strips is evident in the reduction of the soot loading in these channels. The partial flow deceleration and the flow impingement on the face of the cement strips diverts the flow locally and promotes soot deposition in the neighbouring cells that have four filtration walls.

An example of a triggered filter regeneration process in the presence of excess oxygen is shown in Figure 5. After the initial warm-up phase (after 90s) with moderate regeneration, a significant increase in the reaction rates initiates an uncontrolled, accelerated regeneration (after 120s). The exothermic heat that is transported downstream by the flow causes a distinctive, accelerated, localized regeneration in the rear third of the DPF. The extreme local heat release spreads radially and uncontrolled until the rear of the DPF is burned free of soot. The resulting thermal loads are especially critical for the filter ceramic durability. During the final phase (after 180s), the temperature distribution is relatively homogeneous within the entire DPF and the remaining soot in the front of the filter continues to burn down slowly.

Closure

Directly coupling the flow fields inside and outside the filter yields a very efficient simulation process. For example: a complete, transient soot loading calculation with 1/2 million cells requires approximately 10 CPU hours with a 3.4 GHz Xeon processor, or three hours with four processors. The finer temporal resolutions required for the regeneration physics (typically 1-2 sec time-steps) result in significantly more computationally intensive simulations – approximately 20 CPU hours per minute real time. However, with as few as eight CPUs, regeneration simulations can be conducted overnight.

The newly developed technique provides fundamental insights into the soot loading and regeneration behaviour of diesel particulate filters – insights that are the prerequisite for founded decisions about optimizing DPF layouts. The detailed conclusions that can be drawn about soot loading and thermal characteristics during the regeneration help assure DPF function is maintained and DPF failures are avoided over the operational lifetime. By harnessing the capabilities of STAR-CD and pro-STAR, rapid turnaround times are possible – enabling CFD considerations to direct the product design, even in the earliest concept and design stages.

Figures
01: Schematic of the flow through a DPF
02: Mesh structure of the DPF sub-domain
a) Numerical grid showing cell types in the DPF sub-domain
b) The underlying physical DPF segmentation
03: Layout of the computational model showing the partitioning of regions
04: Three-dimensional soot distribution
05: DPF Regeneration