Inverse analysis for estimating geotechnical parameters using physics-informed neural networks

Abstract

Physics-informed neural networks (PINNs) have been proposed for incorporating physical laws into deep learning. PINNs can output solutions that satisfy physical laws by introducing information, such as partial differential equations (PDEs), boundary conditions, and initial conditions, into the loss functions used during the construction of the neural network model. This study presents two cases in which geotechnical parameters were estimated through an inverse analysis of PINNs. PINNs were applied to simulate consolidation and unsaturated seepage processes. The inverse analysis of the PINNs helped estimate the coefficient of consolidation and the parameters related to the unsaturated soil hydraulic properties with sufficient accuracy. The inverse analysis of PINNs for geotechnical parameter estimation was found to be an effective approach that utilizes measurement data.