A Dam Deformation Residual Correction Method for High Arch Dams Using Phase Space Reconstruction and an Optimized Long Short-Term Memory Network

Dam safety is an important basic part of national water network security. Building a dam deformation prediction model based on monitoring data crucial to ensure safety. However, traditional statistical regression methods have shortcomings, such as weak nonlinear fitting ability when constructing and models. The residual the results usually contains parts that cannot be effectively explained by linear method, highly variable noisy. In this study, phase space reconstruction method used smooth term eliminate noise interference. On basis, improved long short-term memory (LSTM) neural learn nonlinearity contained in regression. Considering impact parameter selection performance, gray wolf optimization (GWO) algorithm determine optimal parameters for better performance. A high arch case with multiple measuring points research objects. experimental show can high-frequency components remove addition, GWO hyperparameters LSTM network, thereby accuracy. combination models deep learning improve performance while preserving interpretability transparency.