Analysis of Seepage and Heat Coupling in Fine Soil: Experimental Investigation and PINN-Inverse Analysis

Seepage anomaly has been a persistent challenge in long-term earth–rockfill dam operations. Engineering practices have used temperature measurements to identify abnormal seepage patterns, yet the coupling of seepage and thermal conductivity within seepage pathways remains a subject of limited research. To address this, this study conducted an indoor model experiment to investigate the interplay between seepage and heat transfer within a fine sand seepage pathway embedded in compacted clayey soil. The findings indicate that temperature variations can effectively locate seepage paths, with the relationship between temperature and seepage influenced by temperature gradients and the consolidation process. To overcome the challenge of directly calculating thermal conductivity from model test data, a physics-informed neural network (PINN) is employed to solve the inverse problem related to the thermal conduction process, extracting the target physical parameters. By analyzing the regularity of pore pressure changes caused by temperature increase, a relationship between temperature change and pore pressure change is established. Based on the temperature change patterns under different seepage conditions, an equivalent thermal conductivity model under seepage influence is proposed, which exhibits a good fit with experimental data. Furthermore, considering the impact of saturation and porosity on thermal conductivity under different seepage conditions, a multifactor equivalent thermal conductivity model is established. These findings provide a more comprehensive understanding of seepage anomalies in earth–rockfill dams and offer valuable insights for their management and maintenance.