Physics-informed neural networks for data-free surrogate modelling and engineering optimization – An example from composite manufacturing

A review on machine learning implementation for predicting and optimizing the mechanical behaviour of laminated fiber-reinforced polymer composites

“…Tobias Würth et al [ 38 ] leverage physics-informed neural networks (PINNs) to optimize the thermochemical curing processes of composite materials. The dataset, crafted through sophisticated mathematical modeling of the curing process, allows PINNs to train effectively without the extensive data typically required.…”

The Application of Physics-Informed Machine Learning in Multiphysics Modeling in Chemical Engineering

“…Procedures of PINN-based surrogate model training and parameter optimization: (a) airfoil manufacture, (b) thermochemical composite curing process, and (c) heat exchanger . [Panel (b) was adapted with permission from ref .…”

“…The results obtained from two illustrative examples demonstrated that the velocity and pressure fields exhibited small pointwise errors and the approach converged to optimal parameters. Similarly, Wurth et al 171 proposed the use of a PINN to construct a parametric, data-free surrogate model for the thermochemical composite curing process. As shown in Figure 15b, the process parameter (including operating and geometric variables) and material property can be employed as inputs to the neural network to train the surrogate model, which is then integrated into an iterative optimizer to find the minimum manufacturing cost.…”