EARLY WARNING OF DAM SEEPAGE WITH COOPERATION BETWEEN PRINCIPAL COMPONENT ANALYSIS AND LEAST SQUARES WAVELET SUPPORT VECTOR MACHINE

摘要

Dam seepage often involves a large number of loads under complex nonlinear constraints. The motivation of this paper is to present an approach for predicting the dam seepage and establishing the early warning model. The factors affecting the dam seepage is studied comprehensively. Principal component analysis (PCA) is proposed to avoid the mutual influence of the affecting factors and reduce the computation cost. To describe the local and jumping signals in the dam seepage monitoring data precisely, least squares wavelet support vector machine (LSWSVM), which combines the advantages of wavelet analysis and least squares support vector machine, is adopted. The early warning model of the dam seepage is finally established based on PCA and LSWSVM. The applicability of this algorithm is illustrated using a typical dam in China. Compared with traditional methods, this algorithm has a lower computation time and superior prediction accuracy; thus, the proposed approach has great potential in practical engineering applications.