The evolution pattern of dam deformation reflects its structural response and operational state. Analyzing this pattern enables effective identification of the probability of deformation anomalies. Deviation reflects the extent to which dam deformation deviates from its expected evolution pattern and serves as an important basis for identifying deformation anomaly behavior. However, traditional deformation anomaly assessment methods overlook the distribution of extreme values within the deviations and the complex dependencies between measurement points, limiting the reliability of deformation anomaly assessment results. To address these limitations, this study proposes a regional deformation anomaly assessment method considering extreme‐value distribution of deviations. Initially, the improved temporal fusion transformer (ITFT) prediction model is employed to capture the temporal evolution pattern of dam deformation and compute the deformation deviations at measurement points. Subsequently, extreme‐value theory (EVT) is applied to establish a generalized extreme‐value distribution for the deviation extremes, and these distributions are used to correct the probability density function of deviations estimated by kernel density estimation (KDE), and this process determines the deformation anomaly rates for single measurement points. Finally, measurement points with similar deformation patterns are clustered using Ward’s hierarchical clustering algorithm, while the Frank copula model captures intraregion nonlinear dependencies for regional deformation anomaly assessments. The engineering application verifies that the proposed method accurately captures the extreme‐value distribution of deformation deviations and the complex dependencies between measurement points. This enhances the reliability and effectiveness of arch dam deformation anomaly assessment, providing a scientific basis for arch dam safety monitoring.
Chen et al. (Thu,) studied this question.