APPLICATION OF GENETIC ALGORITHM-SUPPORT VECTOR MACHINE (GA-SVM) FOR DAMAGE IDENTIFICATION OF BRIDGE

摘要

A support vector machine (SVM) optimized by genetic algorithm (GA)-based damagenidentification method is proposed in this paper. The best kernel parameters are obtainednby GA from selection, crossover and mutation, and utilized as the model parameters ofnSVM. The combined vector of mode shape ratio and frequency rate is used as theninput variable. A numerical example for a simply supported bridge with five girders isnprovided to verify the feasibility of the method. Numerical simulation shows that thenmaximal relative errors of GA-SVM for the damage identification of single, two andnthree suspicious damaged elements is 1.84%. Meanwhile, comparative analyzes betweennGA-SVM and radical basis function (RBF), back propagation networks optimized by GAn(GA-BP) were conducted, the maximal relative errors of RBF and GA-BP are 6.91%nand 5.52%, respectively. It indicates that GA-SVM can assess the damage conditionsnwith better accuracy.