Assessment of algorithms for the probabilistic optimisation of composite panels

This report investigates the utility of leading optimisation algorithms for the probabilistic optimisation of composite stiffened panels. Suitable algorithms were selected and applied to a mono-stringer subject to uncertainties in deterministic, robust-design and hybrid robust-reliable design formulations. Gradient, direct, evolutionary, genetic algorithm and combined algorithms were then rigorously compared for both speed and convergence properties. For simpler problems such as deterministic design, direct methods were found to dominate in both speed and accuracy. However, as problem complexity increased, exploratory methods were increasingly found to be superior for locating global optima. The recommendation for future research is therefore, to utilise hybrid uncertainty-based optimisation with a multi-island genetic algorithm to efficiently find lighter, stronger and more consistently performing designs for aircraft structures.