Physics-informed neural network for modelling the thermochemical curing process of composite-tool systems during manufacture

We present a Physics-Informed Neural Network (PINN) to simulate the thermochemical evolution of composite material on tool undergoing cure in an autoclave. In particular, we solve governing coupled system differential equations -- including conductive heat transfer and resin kinetics by optimizing parameters deep neural network (DNN) using physics-based loss function. To account for vastly different behaviour thermal conduction cure, design PINN consisting two disconnected subnetworks, develop sequential training algorithm that mitigates instability traditional methods. Further, incorporate explicit discontinuities into DNN at composite-tool interface enforce known physical directly function improve solution near interface. train with technique automatically adapts weights terms corresponding PDE, boundary, interface, initial conditions. Finally, demonstrate one can include problem as input model resulting surrogate provides real-time simulation range settings use learning significantly reduce time similar trained model. The performance proposed is demonstrated multiple scenarios thicknesses boundary