Including filter media heterogeneities in the simulation of filtration processes

The design of filter media is an important issue to meet the requirements of todays filtration processes. The media have to be optimized in terms of pressure loss, capturing efficiency, and dust holding capacity. One well-known technique to reach these goals is the design of layered (or graded) filtering media. In order to distribute the dust more evenly over the depth of the medium, such a design usually combines relatively open pre-filter material(s) with one or several fine filter layers downstream. However, such a design can lead to undesired effects, e.g. internal cake filtration. This occurs when particles penetrate the upstream layer but cannot enter (part of) the fine filter region downstream because the pores are too small. Thus, on the interface between the two layers, an internal cake is built up. Another widely used technique to increase the dust holding capacity and the filter lifetime is the pleating of filter media. However, the manufacturing process can lead to heterogeneities in the filter material distribution in the pleat. Embossing, for example, leads to locally compressed areas with a higher solid volume fraction. Obviously, this introduces variations of the local flow resistivity in the filter pleat, which are expected to influence both fluid flow and particle deposition. This study is devoted to the development of models taking into account such material heterogeneities in filtration simulation on the macroscopic scale. To this end, a multi-scale approach is proposed for both applications: Simulations on the microstructure of the filter material provide the corresponding permeability and efficiency properties. In the case of pleated filters, CT-images are used to create a simulation model for the distribution of the filter material, which is translated into a permeability distribution. The effective properties obtained from the microstructures are used in the macroscopic simulation to predict the filter lifetime. It is shown that by accounting for heterogeneities, corresponding simulations are able to produce results that are closer to experimental observation.