Wrinkle-Free Membrane Design Based on Physics-Informed Neural Network

Due to extreme flexibilities, membranes are prone to wrinkling, which adversely impacts their optical and electrical properties, thereby restricting their applications in aerospace and flexible electronics. Achieving wrinkle-free membranes necessitates the accurate evaluation of the wrinkling capabilities of membranes and the implementation of suitable algorithms for optimal structural designs. To tackle the challenges in solving the complex partial differential governing equations, a physics-informed neural network (PINN) is employed to determine the stress distribution in non-uniform/irregular membranes under thermal and mechanical loads, and the related wrinkling capability is assessed based on the stress-based wrinkling criterion. To solve the complex multi-parameter optimization problem, the non-gradient particle swarm optimization (PSO) algorithm is adopted to obtain optimal parameters and then regulate the stress distribution to achieve wrinkle-free membranes. The accuracy of the PINN model and the wrinkle-free performance of the optimized membranes are validated through both finite element analysis (FEA) and experiments. This study demonstrates that the PINN model is capable of effectively characterizing the stress distribution in a stretched membrane under multi-physics loads, while the structural optimization yields optimal designs with wrinkle-free properties, providing useful guidance for the design of wrinkle-free membranes.