A real-time optimization method for thermo-chemical coupled curing process of composites with LSTM network

In this present work, a novel real-time optimization method is introduced for autoclave curing of carbon fiber reinforced polymer (CFRP) composites, which employs LSTM network to actively control the defects, i.e. temperature overshoot and uneven cure induced by curing process. Firstly, a finite element (FE) based thermo-chemical coupled model is developed to evaluate the temperature and DoC evolutions, and experimentally validated by a large-size T-stiffened composite panel. Then, the information of curing profiles and the corresponding temperature and DoC differences extracted from FE simulations are used for Long Short-Term Memory (LSTM) network training. Finally, a real-time control framework is proposed by integrating the LSTM network with Q-learning algorithm to minimize the temperature and DoC differences during the curing process by adjusting the curing profile. The optimized curing profile shows a significant improvement compared to the original two dwell profile, with the temperature difference and DoC difference in the thickness and length directions both reduced. This design of curing profile can provide more insights into the composite intelligent manufacturing.