Study On The Structural Damage Identification Based On The Response And PSO-SVM

The damage of major civil engineering structures will cause catastrophic consequences.Therefore,it is necessary for us to take effective measures and technical means to carry out health monitoring of the existing civil engineering structures.In recent years,vibration-based structural damage identification methods have been developed rapidly.However,due to the influences of measurement noise,incompleteness of modal information,difficulty in obtaining system input information,and effect of environmental factors,the accuracy and robustness of the results of structural damage identification need to be improved.The structural damage identification method based on response analysis needs only to consider the outputs of the system,while the input information of the system is not needed.Therefore,in this paper,the structural vibration responses are analyzed to obtain the structural damage feature vector,and the PSO-SVM is employed to achieve structural damage identification.The main contents are as follows:(1)The state of the art of several vibration-based structural damage identification methods at home and abroad is mainly introduced,and their advantages and disadvantages are reviewed.(2)The basic principles of the models and algorithms used in this paper are introduced,such as AR model,PSO-SVM,GRNN neural network,Hilbert-Huang transform,random forest algorithm.(3)The AR model is widely used in structural damage identification based on time domain data,because its calculation is simple,its modeling speed is fast,and the information related to structural vibration characteristics can be reflected by extracting its fewer parameters.A structural damage identification method based on AR model and PSO-SVM is proposed in.Firstly,the acceleration response signals of the measured points under undamaged and damaged conditions are extracted,the AR model is used to fit the measured acceleration responses,and AR model coefficients are extracted.Then,two damage indexes are constructed on the basis of AR model coefficients: the differences of AR model coefficient and the change rate of AR model coefficients pre and post damage.Finally,PSO-SVM is employed to identify the damaged location.The influence of noise is also considered.The numerical simulation and model test of a five-storey three-dimensional frame structure show that the method is effective and withthe ability of noise-resistant.(4)Aiming at the characteristics of nonlinear and non-stationary signals under ambient excitation,a structural damage identification method based on Hilbert-Huang transform and PSO-SVM is proposed.Firstly,the acceleration response signals of the measured points under undamaged and damaged conditions are extracted,and the original signal is decomposed into a series of IMF components by MEEMD algorithm.Then,the Hilbert spectrum is obtained by Hilbert transform of IMF components,and the damage characteristic parameters are extracted based on the IMF components and the instantaneous amplitude of Hilbert spectrum,such as,mean,standard deviation,root mean square,kurtosis,margin factor,skewness,correlation coefficient,energy entropy.Then,the random forest algorithm is used for feature selection,and the first five features are selected as the best feature subset for classification according to the importance of features.Finally,the optimal feature subsets are input into PSO-SVM as the sample sets to identify the damage loaction.Whether excited by white noise or by Qian'an seismic wave,the numerical simulation and model test of a five-storey three-dimensional frame structure have achieved relatively better structural damage recognition effects,and the results show that the method is effective and with the ability of noise-resistant.