MULTI-SCALE DAM DEFORMATION PREDICTION BASED ON EMPIRICAL MODE DECOMPOSITION AND GENETIC ALGORITHM FOR SUPPORT VECTOR MACHINES(GA-SVM)

Empirical mode decomposition algorithm is used to decompose dam deformation data.By obtaining the multi-scale deformation components,the characteristics of deformation components and their related influence factors are analyzed.According to those characteristics,the independent GA-SVM deformation component prediction models are built,used to combine construct the empirical mode decomposition based GA-SVM multi-scale deformation prediction model.By analyzing the dam deformation data from a real case with the empirical mode decomposition,the following views is proved: the empirical mode decomposition can effectively perform the multi-scale decomposition on dam deformation data;and the deformation components decomposed by empirical mode decomposition algorithm have more clear characteristics which can be easily applied to the influence factors analysis for deformation components and surely achieve an higher prediction accuracy.Since the empirical mode decomposition based on GA-SVM multi-scale deformation prediction model is composed of different deformation components of prediction models,it is capable of showing the intrinsic principles of dam deformation,and able to perform dam deformation prediction in different feature scales simultaneously.Comparing accuracies via multi-scale dam deformation prediction model and multiple regression,time series analysis,GM(1,4),BP neural network and GA-SVM dam deformation prediction model,the empirical mode decomposition based GA-SVM multi-scale deformation prediction model is proved as a valuable new method with higher accuracy for dam deformation prediction.