Towards adaptive topology optimization

•A new approach to adaptive topology optimization.•A higher resolution in the description of the optimal shape.•The approach proposed requires revisiting density and sensitivity filtering.•Applied to the optimization of 3D parts and structures.

This paper presents a new fully-automated adaptation strategy for structural topology optimization (TO) methods. In this work, TO is based on the SIMP method on unstructured tetrahedral meshes. The SIMP density gradient is used to locate solid-void interface and h-adaptation is applied for a better definition of this interface and, at the same time, de-refinement is performed to coarsen the mesh in fully solid and void regions. Since the mesh is no longer uniform after such an adaptation, classical filtering techniques have to be revisited to ensure mesh-independency and checkerboard-free designs. Using this adaptive scheme improves the objective function minimization and leads to a higher resolution in the description of the optimal shape boundary (solid-void interface) at a lower computational cost. This paper combines a 3D implementation of the SIMP method for unstructured tetrahedral meshes with an original mesh adaptation strategy. The approach is validated on several examples to illustrate its effectiveness.