The structure safety of high arch dam is very important because of its characteristics such as large discharge, large discharge power, large discharge single width power and narrow valley under complex seismic acceleration. Therefore, taking the double-curved arch dam in Wudongde as the research object, the stress-strain characteristics of arch dam are studied. Firstly, the two methods of convection-solid coupling finite element method and PSO-BP neural network are reconstructed to form a research method. ANSYS Workbench is used to carry out fluid-solid coupling finite element simulation of the arch dam structure discharging water body, then the PSO-BP neural network analysis model was constructed. Through the comparative study of the prediction results of BP neural network and PSO-BP neural network, both prediction results meet the requirements, and under the same conditions, the prediction simulation convergence speed of PSO-BP neural network is faster, and the smaller the prediction error, the better the prediction ability. Finally, PSO-BP neural network analysis model is used to predict the stress and strain characteristics of the double-curvature arch dam in Wudongde, and the results are correct and reasonable. Based on the verification of the research process and results, the feasibility of the research method of reconstruction explored in this paper is verified.