Stochastic multi-objective optimisation of composites manufacturing process

Abstract

This paper addresses the development of a stochastic multi-objective optimisation methodology and its implementation in the manufacturing of thick composite parts. Boundary conditions variability was quantified conducting a series of experiments and stochastic objects have been developed representing these uncertainties. The stochastic optimisation scheme takes into account the uncertainty of process parameters and boundary conditions and identifies optimal solutions that minimise process outcomes such as process duration and extent of defect formation and their uncertainty. The Kriging method was implemented to construct a computationally efficient surrogate model of manufacturing based on sample points selected by the Latin Hypercube Sampling (LHS) method and generated by a Finite Element (FE) model of the process. Response surfaces were constructed to test the accuracy of the surrogate model against the FE solution. A Genetic Algorithm (GA) was utilised to solve the multi-objective optimisation problem. The surrogate model was coupled with Monte Carlo (MC) and integrated into the stochastic multi-objective optimisation framework. The results show that a significant reduction in process duration and process induced defects variability in comparison with conventional processing conditions of up to 80% and 40% respectively can be achieved by the optimisation.