Bayesian optimization for process planning selections in automated fiber placement

Since the arrival of Automated Fiber Placement (AFP) in the 1980s and 1990s, the manufacturing technique has been progressively improving in quality and reliability of the process and resulting structures. The state-of-the-art of AFP is allowing for the manufacture of structures with increasing geometrical complexity. To adequately manufacture such structures, a large effort in process planning is required to reduce defect occurrence to the best extent possible. Many optimization methodologies have been established for AFP in terms of the design and process parameters. However, limited work has been accomplished in terms of process planning optimization. The work presented in the latter will develop a Bayesian Optimization (BO) technique for the optimization of process planning with the use of the Computer Aided Process Planning (CAPP) software and its connection with Vericut Composite Programming (VCP). A case study is performed on a doubly curved tool surface and the optimal solution is compared with as-manufactured results for validation. A predictive tool is also developed to rank layup strategies based on defect importance. Results show a substantial improvement in defect occurrence and allow a process planner to immediately see the performance of various layup strategies and starting point combinations.