<p>Water inrush is one of the most severe geological disasters during underground engineering construction. To investigate the variation of flow rate during tunnel water inrush, this paper establishes a mathematical model for groundwater flow, derives an analytical solution for the flow rate during water inrush event, and solves it using Physics-Informed Neural Networks (PINN). To evaluate the effectiveness of PINN for this particular problem, the numerical solution obtained under the same parameters is compared to the analytical solution. As a case study, this paper examines a flow rate process that occurred in the Chenaju Tunnel of the Dianzhong Water Diversion Project, the variation of flow rate is analyzed and an inversion of hydrogeological parameters for the rock mass within the tunnel sections is performed, and the mechanism of water inrush is revealed. The results show that PINN can accurately solve the mathematical model for underground water flow during water inrush and this numerical solution is almost in agree with the analytical solution. The inversion value of the hydraulic conductivity coefficient (<i>a</i>² = <i>K</i>/<i>μ</i>_<i>s</i>) obtained from the PINN model is 431.339 m²/d, with R² equal to 0.859 which means a very high permeability during the water inrush event. The water inrush in Chenaju Tunnel is mainly controlled by factors like faults, sandification, and erosion. The high groundwater level and high hydraulic conductivity coefficient around the excavation section have resulted in large-scale and high-risk water inrush accidents. This study can provide valuable insights for the prediction and prevention of sudden water inrush in underground tunnels.</p>