DEVELOPMENT OF LEVEL SET APPROACH FOR GRADIENT AND NON-GRADIENT BASED STRUCTURAL TOPOLOGY OPTIMIZATION

• Main Author: Nathan
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/39125
• Institution: Institut Teknologi Bandung

Topology optimization plays an important role in structural design process where none of the guess can be predict in the early stage of design. There are two approach methods to solve topology optimization, namely gradient-based method and non-gradient based method. Gradient-based method have the advantage to solve a problem with thousands of number of design variables with only hundreds of finite element function evaluations. On the other hand, non-gradient based requires thousands of finite element function evaluations, but can solve highly nonlinear, multimodal, and noisy problems such as crashworthiness problem. In this paper, level-set function (LSF) is proposed for topological representation, then carrying out the classification based on the value of the function relative to a threshold. New approaches of radial basis function (RBF) and kriging interpolated level set (KILS) are used to form the LSF by interpolating at knot points to maintain a reasonable number of design variables that is independent from the mesh size. Hybrid genetic algorithm (GA), covariance matrix adaptation - evolution strategy (CMA-ES), and pattern search are then used to solve the optimization of non-gradient problems. Capabilities to approach gradient based result with non-gradient based method are further demonstrated on several test problems.