Technical study on rapid estimation method of peak acceleration of dams

Aiming at the background that earthquakes may damage dams and the occurrence of earthquakes needs to be predicted at the first time to cause impacts on dams, this paper investigates the rapid estimation method technology of peak acceleration of dams. Among them, the model based on the random forest algorithm takes the historical data of the seismic network as the original data, and utilizes the Bootstrap sampling method to carry out random sampling, and uses the out-of-bag (OOB) data evaluation score oob_score as the tuning parameter of the random forest regression model, and calculates the optimal parameters of the model. Finally, the effectiveness of the proposed method has been verified by comparative experiments on the test set, which can timely assess the impact of earthquakes on dams, while the overall performance of the random forest algorithm is better than that of the multivariate nonlinear regression algorithm, with an improvement of 0.75 in R2 and a reduction of 0.48 in MAE for the prediction of peak acceleration values, and an improvement of 0.06 in R2 and a reduction of 0.67 in MAE for the prediction of trigger time.