Industrial applications of evolutionary algorithms: a comparison to traditional methods

摘要

Recent variants of Evolutionary Algorithms are powerful methods for parallel industrial optimisation in integrated simulator-optimiser environments. An overview of Evolutionary Algorithms and their industrial applications is given. As a practical example, selected variants of Evolutionary Strategies (a subclass of EAs) together with traditional deterministic methods like pattern search, polytope methods and gradient based methods, are applied to a small benchmark and a representative engineering test-case that will be introduced in this paper: a 3-D transient heat transfer problem. The results indicate that EAs with small population sizes and fast step size adaptation mechanisms are promising methods for industrial design optimisation with time expensive function evaluations.