Large scale structural optimization using genetic and generative algorithms with sequential linear programming

Abstract

This thesis explores novel parameterization concepts for large scale topology optimization that enables the use of evolutionary algorithms in large-scale structural design. Specifically, two novel parameterization concepts based on generative algorithms and Boolean random networks are proposed that facilitate systematic exploration of the design space while limiting the number of design variables. The presented methodology is demonstrated on classical planar and space truss optimization problems. A nested optimization methodology using genetic algorithms and sequential linear programming is also proposed to solve truss optimization problems. Further, a number of heuristics are also presented to perform the parameterization efficiently. The results obtained on solving the standard truss optimization problems are very encouraging.