Stress Estimation of Concrete Dams in Service Based on Deformation Data Using SIE–APSO–CNN–LSTM

The stress behavior of key parts concrete dams is related to the safe operation dam. However, sensors in are susceptible aging and failure with increasing service life. Estimating structural under sensor or data loss scenarios for essential complex. This study presents a estimation method driven by observation data. Firstly, one-to-one correspondence exists between dam deformation reflecting load effect stress. simulating effects more convenient when it hard simulate complex directly. Therefore, based on observed before failure, spatial–temporal relationship structure multi-point deformations developed using convolutional neural networks (CNN) long short-term memory (LSTM). An improved particle swarm optimization algorithm combined information entropy (SIE–APSO) proposed simultaneously tuning network’s hyperparameter accelerating convergence. Finally, target part obtained. case shows that valid feasible. RMSE decreased approximately 21–58%, MAPE 19–58%, ARV 22–94% compared load-stress model.