Calculation of Pore Pressure Dissipated from a Clay Layer Due to Groundwater Extraction Using Physics-Informed Neural Network (PINN) with Reference to Land Subsidence Analysis

PINN, standing for physics-informed neural network, is a relatively new method that is increasingly used in solving Partial Differential Equations (PDEs) to simulate various physical processes, including that of consolidation. In this study, Physics-informed Neural Network (PINN) was applied to solve the 1D consolidation equation to calculate pore pressure dissipated by groundwater extraction from an underlying aquifer, which will further used in calculation of land subsidence. This study also aimed to investigate effect of transfer function in PINN analysis as well as the interaction between the physics-informed constraints with the deep learning (DL) network. As a case study a location along the planned URMT line no. 2 in Hanoi was selected for subsidence analysis. A total of 12 PINN analyses with different hidden layers (5 and 10) and different neuron number per one hidden layer (5, 10 and 20) were conducted. The results of pore pressure calculated by PINN were compared with those obtained by a classic finite difference solution, showing a good match, validating the PINN approach used, and thus opening up a new data-driven and physics-informed avenue of land subsidence simulation and prediction in near future.