Research and software development of compression buckling for laser-welded stiffened plates based on physics-informed neural networks

Abstract

Stiffened plate structures, which enhance overall strength and stiffness, are critical in engineering applications, with stability under compressive loads being a key design consideration. While numerical approaches like finite element analysis are effective, their high computational cost drives interest in data-driven methods for failure load prediction. However, such methods often face challenges related to data quality, generalization and physical interpretability. This study proposed a Physics-Informed Neural Network (PINN) for predicting the compressive buckling failure of stiffened plates, and then developed the SiPFLP software based on the PINN. Firstly, compressive tests on laser-welded stiffened plates, combined with electrical and optical measurement, were conducted to elucidate failure mechanisms. A validated post-buckling analysis was then used to examine how geometric and material parameters influence buckling failure loads. These qualitative physical laws were employed to guide the training of the PINN for failure load prediction. The PINN was subsequently integrated with PyQt5 to develop the SiPFLP software. This research enhances the understanding of buckling failure in laser-welded stiffened plates and provides an innovative, physics-guided and computationally efficient tool for predicting failure loads.