Derivative-Free Trust-Region-Guided Explicit Level Set Topology Optimisation

This work investigates the use of explicit level set parameterisation for topology optimisation using a metamodel-based trust region strategy optimiser.The explicit level set parameterisation consists of building a uniform Design of Experiments using a Permutation Genetic Algorithm, followed by building the Level Set Function using Kriging.Through decoupling the parameterisation from the simulation physics, the use of sensitivity data becomes optional thus enabling computationally complex disciplines (where sensitivity data is not available, e.g.crashworthiness, electromagnetics) to be included.This is achieved through the use of a sequence of approximations to the functions of the original optimisation problems based on a trust region strategy.The method is demonstrated on a benchmark 2D topology optimisation problem to examine the effectiveness of the technique.