基于EEMD高频差分转换和GRU算法的大坝变形异常值判定OA

We propose an outlier detection method for dam deformation based on ensemble empirical mode decomposition(EEMD)high-frequency differential conversion and gated recurrent unit(GRU)analysis.First,the EEMD technology is employed to decompose the original deformation monitoring data,yielding deformation components at different frequencies.Then,high-frequency components containing abnormal deformation information can be extracted for differential sequence conversion,and identification criteria for abnormal single-jump points are proposed.Finally,a high-frequency component time series model is established by GRU neural networks,which is then combined with the abnormal single-jump point identification pattern to recognize outliers.The results show that EE-MD technology and high-frequency sequence differential conversion can effectively isolate deformation anomaly information.The proposed single-jump point detection model aligns with the characteristics of dam deformation anomalies.Combining GRU and the 3cr criterion for time series modeling analysis of differential sequences,all deformation anomalies can be accurately identified,and achieve higher ac-curacy than traditional methods.