Mechanism-data fusion driven failure analysis and safety evaluation of high arch dam: A synergistic work of physical-numerical modeling and GMM-CHMM

Mechanism and data-driven methods are significant approaches for failure analysis and safety evaluation of high arch dams. The respective representative for them is physical-numerical modelling and the Gaussian Mixture Model-Continuous Hidden Markov Model (GMM-CHMM). A synergistic work of the two methods is prosperous in accurately classifying the structural states’ evolution and clearly interpreting the physical mechanism. However, the relevant research remains lacking. Therefore, this study carried out physical and numerical model overloading test based on a 180 m-high arch dam; The dam failure process was obtained and analyzed by the comprehensive evaluation indicators composed of integral structural deformation, energy, and damage information; A structure-improved GMM-CHMM considering index weight and interpretation enhancement is proposed to assess the whole-process dam states evolution. The overload-variant states of the researched arch dam were clearly classified into 7 levels ranking from safe to failure, with the physical meanings endowed by mechanism analysis. The graded safety factors in forms of overloading degree were determined, validated, and proved to be more precise than the mechanism-based method. The accurate multi-level classification and interpretation of arch dam states were achieved in this work, providing new insights into failure analysis and safety assessment of arch dam structures.