Damage Identification Of Time-Varing Structures Based On Improved EWT Algorithm

In the field of structural health monitoring system,it is very common to run up against the time-varying system with structural parameters changing.According to the vibration response characteristics of the time-varying system,a reasonable signal analysis method is needed after obtaining the structural response signals,so that the structural damage information can be obtained more accurately and damage identification can be identified.Time-frequency analysis method can provide the simultaneous distribution information of time-frequency domain,which makes the time-frequency analysis method equip with its unique advantages when facing timevarying system and nonlinear system.Among them,empirical wavelet transform(EWT)plays an important role.Recently,EWT has came to foreground for its ability to construct selfadaption wavelets.However,traditional EWT method defines the filtering boundary according to the segmented Fourier spectrum of the processed signals when dealing with non-stationary signals,which may be ineffective and inaccurate for non-stationary signal decomposition.The traditional EWT algorithm is improved in order to make EWT method applicable to time-vating structure damage identification.Combining with the subsequent independent component analysis method and optimization algorithm,the method is suitable for damage identification of time-varying structure.The main contents of this paper are organized as follows:1.An improved EWT algorithm is proposed,which combines synchroextracting transform with EWT.The time-frequency spectrum of the signal with more concentrated energy is extracted by SET.According to the time-frequency spectrum,the number of signal components is obtained and the the filtering boundary,namely the filtering boundary of EWT,is calculated.Based on this,filter banks are constructed and EWT is used to extract Intrinsic mode functions(IMFs),and each modal function is an AM/FM signal.The improved EWT algorithm is used to decompose a non-stationary signal,which solves the problem of over-decomposing signals,and verifies that this method has good anti-noise performance.2.Extracting damage features.After applying the improved experiment wavelet transform,each group of original signals is decomposed into diverse frequency components,and the highfrequency components sensitive to damage in each group of signals are extracted to form a high-frequency signal matrix.As for high-frequency signal matrix,RobustICA is implemented for independent component analysis to find characteristic components.Through it,on the one hand,the position of damage can be located preliminarily;on the other hand,the occurrence time of damage can be confirmed.3.Obtaining physical parameters of the structure and quantifying damage.After determining the occurrence time of damage by RobustICA,the original time history signal is identified by several time interval.The genetic algorithm and particle swarm optimization algorithm are selected as parameter identification method,comparing and verifying by numerical simulation.Among the numerical simulation,single damage condition and multiple damage condition are set.Finally,the particle swarm optimization algorithm is established as the parameter identification method by considering the best fitness value and parameter optimization results,the final identification error is within 1%.4.Analyzing by experimental verification.A five-layer frame experimental model is applied,which intact condition as well as damage condition are set up and the acceleration response signal of each layer are picked up respectively.The above time-varying structural damage identification algorithm is used to analyze and process the signals,and the final absolute error and relative error are 1.999 % and 3.748% respectively,which verify the feasibility of the algorithm proposed in this paper.