Machine learning-based process simulation approach for real-time optimization and active control of composites autoclave processing

Abstract

For manufacturing of composites, several parts may be processed simultaneously in an autoclave or oven. Depending on the equipment design, tool/part geometries, and tool nesting, convective heat transfer Boundary Conditions (BCs) may vary around parts in each load. As a result, temperature histories in some of the parts may not conform to specifications such as limits on maximum part temperature, or part temperature rate. To mitigate risk, in addition to conducting finite element simulations prior to fabrication based on assumed BCs, leading and lagging thermocouples, embedded in parts or placed in proxy locations, are used to monitor temperature history during processing. In this study, a recently developed machine learning framework, CompML (Composites Machine Learning) is used for active control of the autoclave. CompML uses TC data at the start of the autoclave processing for real-time inverse modeling of the thermochemical problem, and to identify BCs for all parts in each load. The results are then used for realtime optimization of autoclave cure recipe to the shortest cycle that satisfies specifications in all parts. A successful virtual demonstration of the approach for HEXCEL AS4/8552 parts processed on Invar tools is discussed in the paper.