Automated and Accurate Geometry Extraction and Shape Optimization of 3D Topology Optimization Results

Abstract: Designs generated by density-based topology optimization (TO) exhibit jagged and/or smeared boundaries, which forms an obstacle to their integration with existing CAD tools. Addressing this problem by smoothing or manual design adjustments is time-consuming and affects the optimality of TO designs. This paper proposes a fully automated procedure to obtain unambiguous, accurate and optimized geometries from arbitrary 3D TO results. It consists of a geometry extraction stage using a level-set-based design description involving radial basis functions, followed by a shape optimization stage involving local analysis refinements near the structural boundary using the Finite Cell Method. Well-defined bounds on basis function weights ensure that sufficient sensitivity information is available throughout the shape optimization process. Our approach results in highly smooth and accurate optimized geometries, and its effectiveness is illustrated by 2D and 3D examples.