A combined parametric shape optimization and ersatz material approach

This article describes a parametric shape optimization approach using vertical or horizontal structures with a fine parametrization of their center lines and profiles. In this context horizontal means a lateral connection from left to right and vertical means a bottom-up connection. These structures are projected to a pseudo density field associated with a fixed mesh using a differentiable mapping. This enables the use of existing topology optimization tools with respect to the solution of the state problem based on the pseudo density field. The approach belongs to a class of geometry projection onto a fictitious domain methods. It therefore shares the property that sensitivity analysis is reduced to extend the well known gradient calculation from topology optimization by chain using the sensitivity of the mapping from shape variables to pseudo density. The contribution lies in the combination with our specific shape parametrization and the associated regularization. Optimization problems can be formulated concurrently in terms of shape variables and pseudo density. We discuss regularization, periodicity constraints, symmetry formulations and overhang constraints in terms of shape variables. Volume and perimeter constraints are easily formulated in terms of the pseudo density. We see our approach as being particularly beneficial for certain problem classes where it may be difficult to restrict the design space, e.g. restricting isolated structures or holes or where a strict control of solid to void transition is necessary. Consequently, we show examples for phononic band gap maximization, boundary driven heat optimization and perimeter maximization for a flow problem. We also present a formulation of overhang constraints for additive manufacturing in terms of shape variables.