基于KPCA-ICEEMDAN-IWT的特高拱坝变形监测数据预处理方法OA

To address the challenges of multicollinearity and complex noise interference in deformation monitoring data for ultra-high arch dams,this study proposes a joint preprocessing method integrating hybrid kernel principal component analy-sis(KPCA),improved adaptive noise-complete set empirical mode decomposition(ICEEMDAN),and improved wavelet thresholding(IWT).This method first constructs a comprehensive deformation index via Bayesian optimization of mixed-ker-nel KPCA to eliminate redundant information and collinearity effects.Subsequently,ICEEMDAN is employed for adaptive signal decomposition,enabling multi-criteria identification of high-frequency noise components.Finally,IWT filters out noise components and reconstructs the effective signal.Our case studies demonstrate that this combined model outperforms standa-lone ICEEMDAN,WT,and IWT methods in metrics such as root mean square error and correlation coefficient.It significant-ly enhances noise removal accuracy while effectively preserving useful high-frequency information,providing a more reliable data preprocessing technique for safety monitoring of ultra-high arch dams.