Structure Optimization with a Bio-inspired Method

Abstract

We explore a structure optimization strategy that is analogous to how bones are formed in embryos, where shape and strength are mostly defined. The strategy starts with a rectangular grid of elements of uniform thickness with boundary displacements and force conditions. The thickness of each element can grow or shrink depending on the internal strain, this process is done iteratively. The internal strain is found using the finite element method solving a solid mechanics problem. The final shape depends only on five parameters (von Mises threshold, thickness grow and shrink factors, maximum and minimum thickness). An evolutionary algorithm is used to search an optimal combination of these five parameters that gives a shape that uses the minimal amount of material but also keeps the strain under a maximum threshold. This algorithm requires to test thousands of shapes, thus super-computing is needed. Evaluation of shapes are done in a computer cluster. We will describe algorithms, software implementation and some results.