Aiming at the problems of noise interference in the dam deformation data,weak mining ability of the key information in conventional deep learning prediction models and difficulty in determining the optimal parameters,the mo-nitoring data were processed by using variational mode decomposition(VMD)combined with wavelet thresholding noise reduction algorithm.And then the noise reduced deformation data were predicted by using dual attention-based gate re-current unit(DA-GRU).The dung beetle optimizer(DBO)was introduced to optimize the model parameters.Thus,a deformation prediction model based on DBO-DA-GRU was established.Taking the measured deformation data of an arch dam as an example,it is verified that DBO-DA-GRU has higher prediction accuracy and better robustness than BP,GRU and DBO-GRU models,which can provide certain reference value for dam deformation safety monitoring.
维普
万方数据