Real-time control for autoclave curing process of CFRP composites considering tool-part interaction

An integrated real-time control methodology is introduced to mitigate process-induced challenges, namely temperature overshoot, uneven cure, and interfacial shear stress, during the autoclave curing of Carbon Fiber-Reinforced Polymer (CFRP) composites. First, a high-fidelity Finite Element (FE) model incorporating tool-part interaction is developed to reveal the curing process of the composites, wherein the interaction is characterized by friction interface modeling with experimentally measured cure-dependent friction coefficients. The accuracy of FE model is confirmed through experimental tests on a doubly curved T-stiffened composite panel. This validated model then generates a dataset of curing temperature profile and associated defect information, which is used to train a customized Long Short-Term Memory (LSTM) neural network. We culminate in a real-time control framework that actively optimizes the curing process by integrating LSTM-based state prediction with Q-learning-driven decision logic. The optimized thermal profile demonstrates a clear performance enhancement over the traditional multi-dwell approach, achieving marked reductions in temperature difference, Degree of Cure (DoC) difference and tool-part interface shear stress, which provides more insights for intelligent composite manufacturing.