Exploring how optimal composite design is influenced by model fidelity and multiple objectives

This paper explores how optimal configuration of a composite panel is influenced by the choice of analysis model – analytic or computational – and the consideration of multiple objectives. While past research has explored aspects of this problem separately – composite ply orientation, multiple load scenarios, and multiple performance objectives – there has been limited work addressing the interactions between these factors. Three loading scenarios are considered in this work, and it is demonstrated that for certain scenarios an analytical model likely over-predicts composite performance. Further, for complex loading scenarios it is impossible to develop an analytical model. However, this work also demonstrates that the use of analytical models can be advantageous. Analytical models can provide similar estimates to computational models for some loading cases at significantly reduced computational expense. More importantly, it is also shown how solutions from the analytical model, which can be relatively cheap to find computationally, can be used to seed the initial designs of a Finite Element-based optimization. Run time reductions as large as 80% are demonstrated when these informed seeded designs are used, even when the designs were created for a different set of loading scenarios.