Extended Evolutionary Algorithms for Multiobjective and Discrete Design Optimization of Structures
Harald F. Langer · 2005
An optimization algorithm for multi-objective problems with continuous and discrete design variables is developed and investigated. Primary development goals have been to enlarge applicability to a very broad spectrum of tasks in the field of structural design while simultaneously providing sufficient efficiency to be viable for complex and computationally costly problems. The algorithm is based on the concept of evolutionary algorithms and features special adaptations for the simultaneous treatment of continuous and discrete variables as well as for the handling of multiple objectives. Core elements for increasing the efficiency are a grain-parallelization on a computer cluster as well as the extension of the conventional evolutionary algorithm by response surface methods. The performance of the algorithm has been evaluated by numerical experiments with suitable mathematical test problems as well as representative structural design problems.