Configuration of a genetic algorithm used to optimise fibre steering in composite laminates

In a previous study, a genetic algorithm (GA) was applied to define the optimum orientation of fibres in a unidirectional laminate in which the fibre orientations were allowed to vary continuously across the domain. The results were positive, but computation time was excessively long. The present study was undertaken to address this issue. The elements used by the GA (encoded representation, fitness criterion, operators affecting the population) were examined and optimised to reduce computation time. It was found that the use of a strain-based fitness criterion was better than use of a stress-based criterion, regardless of whether the load was above or below the elastic limit. A good balance of elitism and mixing was necessary in the GA parameters to obtain the fastest convergence. Automatic imaging of the GA output was achieved using the SAMCEF software package.