Clean air and water are increasingly in focus, while the demand for energy-efficient solutions to address these issues is growing. Filter development must rise to this challenge. Filters should be used in a resource-efficient way. This can be achieved by precisely matching filters to the application, both in terms of particle size distribution and the conditions of the fluid flow to be treated. Conventional filters often consist of one or more layers of filter media, which are either woven or composed of tangled fibers. The efficiency of these filters depends mainly on the tightness of the fabric.

The development of additive manufacturing has made great progresses in recent years. Using this technology, filter manufacturing is no longer restricted to conventional methods. In addition, computers have become more powerful and can carry out numerical flow simulations in a short time. The limits of additive manufacturing in terms of material and resolution have not been reached yet. A great potential for the developed method is expected here in the future.

The aim of this work is to present two methods to optimize filter structures with respect to the filtration efficiency and the pressure loss: first a simple geometry-based shape optimization of initial structures with the adjoint method and second an Eulerian Multiphase approach inspired by the adjoint method.

Both approaches are based on the same workflow, but are different with respect to the cost function used. The first steps are application-dependent: Creating an initial geometry in Computer Aided Design (CAD), meshing the initial geometry in Computational Fluid Dynamics (CFD) and determining pressure loss and separation efficiency for e.g. pollen separation in car ventilation systems. The cells at the fibers are then varied by mesh deformation until an optimum is found. While the first method requires cost functions that can be derived to run the adjoint solver, the second method can use arbitrary functions. These cost functions have to account for the different separation mechanisms and the pressure loss. The combination of the sensitivities was realized by an algorithm. Lagrange and DEM simulations were performed to determine the percentage of impaction and interception in the total separation efficiency. Physical prototypes of the often bionic-looking optimized structures can be generated using additive manufacturing.

The whole workflow was successfully tested for ...