Research on Risk Evaluation Method for Levee Collapse Pits and Sinkholes Based on Microtremor Wavelet Energy Spectrum and Fuzzy Comprehensive Evaluation

Collapse pits and sinkholes are among the common hidden defects in levee engineering, characterized by high suddenness and difficulty in early identification during their formation process. Based on levee model tests, this paper adopts a method of embedding ice bodies combined with internal heating and melting to simulate the evolution process in which the internal support structure is gradually lost and a cavity is formed. Continuously acquired microtremor signals are collected by deployed high-sensitivity acceleration sensors, and time-frequency energy spectrum features are extracted using wavelet transform. The results show that after the cavity is formed, the energy concentration zone of the acceleration wavelet energy spectrum exhibits a significant low-frequency shift, with an energy frequency shift of 6 Hz (from 8 Hz to 2 Hz). At this point, the ice body has completely melted and the support loss rate reaches 100%, while the ground surface remains continuous without macroscopic collapse, indicating that this method can capture early warning signals before disaster occurrence. On this basis, a risk evaluation index system for collapse pits and sinkholes covering four indicators across three dimensions of "load conditions–dynamic fingerprints–deformation response" is constructed. A game-theoretic combination weighting method combining the analytic hierarchy process and the entropy weight method is adopted to determine indicator weights, and quantitative risk level evaluation is achieved based on fuzzy comprehensive evaluation. Two typical stages, namely before heating and after cavity formation (before surface collapse), are taken as evaluation samples. The evaluation results show that the comprehensive score before heating is 95.0 (Level I, safe), and decreases to 37.9 (Level V, high risk) after cavity formation. Sensitivity analysis shows that after removing the dynamic fingerprint indicators, the comprehensive score at the post-cavity-formation stage rises to 53.8, and the risk level decreases from "high risk" to "medium risk," with a relative increase of 41.9%, directly demonstrating that the dynamic fingerprints extracted based on wavelet energy spectrum are sensitive to the identification of internal cavities in levees, providing technical support for pre-disaster warning of collapse pits and sinkholes. This study provides an approach for risk analysis of levee collapse pits and sinkholes integrating measured indicators and fuzzy comprehensive evaluation.