Physical informed neural network for thermo-hydral analysis of fire-loaded concrete

In the event of a fire within a tunnel, the rapid and substantial increase in temperature can prompt swift fractures within the concrete lining. This situation can severely compromise the structural load-bearing capacity and overall safety of the tunnel. The intricate interplay of factors within the tunnel, including temperature, humidity, and pore pressure, necessitates the adoption of a thermo-hydro coupled model for comprehensive analysis. However, these highly coupled models exhibit profoundly nonlinear characteristics that cannot be readily solved through analytical means. In this study, a Physics-informed neural network (PINN) is harnessed to address this intricate multi-field coupled issue. Neural networks possess a significant advantage in their capacity to autonomously differentiate and directly capture variations within the spatiotemporal domain. Through a comparison with experimental results, the proposed method's reliability is effectively demonstrated. The research findings hold the potential to significantly aid engineering professionals in swiftly conducting fire resistance assessments and risk evaluations for tunnels, whether they are under construction or already in operation.