Structural Design Optimisation Using Genetic Algorithms and Neural Networks

This paper relates to the optimisation of structural design using Genetic Algorithms (GAs) and presents an improved method for determining the fitness of genetic codes that represent possible design solutions. Two significant problems that often hinder design optimisation using genetic algorithms are expensive fitness evaluation and high epistasis. Expensive fitness evaluation results in slow evolution and occurs when it is computationally expensive to test the effectiveness of possible design solutions using an objective function. High epistasis occurs when certain genes lose their significance or value when other genes change. Consequently, when a fit genetic code has an important gene changed this can have a dramatic effect on the fitness of that genetic code. Often the reduction in fitness results in failure of the genetic code being selected for reproduction and inclusion in the next generation. This loss of evolved genetic information can result in the solution taking considerable time to discover.