Dynamic Cluster Zoning of Arch Dam Deformation Considering Changing Working Conditions

Abstract

Arch dam deformation has regional characteristics, and clustering is a common method of regional classification for arch dams. Traditional methods ignore the impact of dynamic changes in temperature and water level. Besides, the noise of deformation data is detrimental to mining potential information. The objective is to devise a dynamic cluster zoning method for arch dams, which considers the changing working conditions under the coupling of water level and temperature in this study. First, the deformation periods are classified by K-means clustering, and the arch dam deformation series are denoised using a sparrow search algorithm-optimized variational mode decomposition combined with wavelet threshold (SSA–VMD–WT) denoising method. The arch dam measuring points for different periods are then clustered. The engineering case study demonstrates that the SSA–VMD–WT denoising method improves the reliability of deformation data. The dynamic cluster zoning method reasonably describes the deformation regularity of the arch dam under different working conditions.