CFD to Optimise the Performance of Novel Diesel Particulate Filters

Filtration of a high flow rate, high temperature gas stream containing very fine particulate presents a major challenge. Most (carbon) particles in Diesel exhausts are ~10-100 nm in size, constituting a serious health hazard, since such particles can easily reach the lungs. The obvious way to remove them from a gas stream is to pass it through a structure with a fine, interconnected pore network, preferably having high tortuosity. Depending on momentum-related factors (think drag forces and Stokes numbers) particles may strike the solid surfaces and adhere. Since the particulate are very fine, then very fine filters are needed for their efficient capture. This requirement is at odds with the need to maintain high gas flow rates though the filter. This is necessary to prevent excessive pressure drops across the filter, which can generate large upstream back pressures that impair the performance of the engine. The objective of this project is to develop a new generation of composite structure for diesel exhaust filter applications. As well as being fine-scale and permeable, the structures must also be stable when subjected to high temperatures, high thermal gradients and thermal shock. This is particularly important during the regeneration process, when burning of the carbon inside the filter (which doesn’t occur uniformly) can raise the local temperatures to over 1000°C. At these temperatures, even if the phases concerned remain thermodynamically stable, the associated thermal stresses may be high (generating cracks).

3D architectures were obtained using computed X-ray microtomography. These geometries were meshed using the native meshinng capabilities of Simpleware's ScanIP +FE module. The meshes were imported into COMSOL multiphysics where full CFD simulations were conducted. Particle tracing capabilities, also available through COMSOL Multiphysics, were included in order to monitor the passage (and capture) of particulate.

Predicted permeabilities were in excellent agreement with measured values, conferring confidence in our modelling approach. The models were used subsequently to help optimise novel filter designs.