Study On The Structural Damage Identification Based On Wavelet Entropy Under Earthquake Excitation

Generally,structural damage identification methods based on vibration assumed that the structural responses were the stationary signal.In fact,the civil engineering structures during the servicing underwent the non-stationary excitations,such as the earthquake,wind,waves,which might affect the results of damage identification.Therefore,the study on the structural damage identification based on the wavelet entropy of vibration responses under the earthquake excitation was conducted in this paper.(1)An overview on the structural damage identification methods based on the vibration response was accomplished at home and abroad.(2)The Fourier transform,the wavelet transform,the information entropy,wavelet entropy,the BP neural network and the theory of single value statistical control chart were described briefly in this paper.(3)Study on the structural damage alarming based on the wavelet entropy of vibration response and single value statistical control chart.Firstly,the acceleration response signals on the different points of the structure excitated by earthquake were segmented,and the wavelet entropy of each segmentation of acceleration response signal was calculated respectively;then,the structural damage alarming index D was constructed with the calculated wavelet entropy;finally,the damage warning index D was used as a sample of single value statistical control chart,and which was used to alarm the structural damage.The results of the numerical simulation of 5 layer three-dimensional frame structure and shaking table model test showed that the proposed damage warning method had certain feasibility and the ability of noise resistance.(4)Study on the structural damage location identification based on the wavelet entropy of vibration response and the BP neural network.Firstly,the wavelet entropy of structural acceleration responses signals under earthquakeexcitation were calculated directly;then,two kinds of structure damage localization indicators: the wavelet entropy curvature difference(WECD)and wavelet entropy difference ratio(WEDR)were constructed with the wavelet entropy;finally,the wavelet entropy curvature difference(WECD)and wavelet entropy difference ratio(WEDR)were used as the input vector of BP neural network for structural damage location identification respectively.The results of the numerical simulation of 5 layer three-dimensional frame structure and shaking table model test showed that the two proposed damage location identification methods had certain feasibility,and the influence of noise on the two damage location identification methods was also studied.