Application Of Wavelet Analysis On Structural Damage Identification

Taking the ASCE health monitoring Benchmark model as the research object, several basic problems about the wavelet transform method used in real-time online damage identification for structures had been studied in this thesis. A tentative new approach is also developed for the application practice of the real-time online damage identification for structures. The main works are as follows:Firstly, the wavelet analysis theory is introduced briefly, including the development history of wavelet analysis, continuous wavelet transform, discrete wavelet transform and the fast"Mallat algorithm"for wavelet analysis.Secondly, for satisfying the research purpose of this thesis, a numerical simulation program has been developed to fix several critical problems in real-time online damage identification for structures. This program is modified and updated based on the simulation program developed by ASCE Task Group on Health Monitoring to simulate the dynamic response of Benchmark model. That is, the source program is modified and updated to be able to 1) calculate the nonlinear dynamic response of the Benchmark model with"sudden damage"during a certain loading excitation, in which numerical simulation results shows that the high order single-stepβ(Wang-β) method developed by professor Wang Huanding is obviously much accurate in calculation of the nonlinear response than most commonly used other two methods: Newmark-βmethod and Wilsion-θmethod; 2) calculate the dynamic response of the Benchmark model with"sudden-damage"characteristic under earthquake excitations to satisfy the purpose of examining the influence of different type of load excitation to the wavelet transform method used in structural damage identification; 3) rectify the error using of the noise adding function in the source program to provide accurate input noise level for examining the noise effects in wavelet analysis for damage identification.Mounts of numerical simulation test results indicates that, through carrying on wavelet transform to the time history of the response acceleration, the accurate damage instant can be read from the appearing of the spike in the time axis as well as the damage location can be predicted from the height of related spikes.Following conclusions are obtained through abovementioned studies: (1) In numerical simulations of the real-time online damage identification for structures, high accuracy method such as Wang-βmethod should be used in order to obtain high frequency information from structural responses.(2) The ReverseBior wavelet should be used as the mother wavelet since it's much better than others when discrete wavelet transform is used for damage identification.(3) Generally, the severer of the structural damage, the lowest identifiable sampling frequency will be relatively lower by using the wavelet transform. As well, the slighter of the structural damage, the lowest identifiable sampling frequency will be much higher. At the same damage level, the lowest identifiable sampling frequency under wind loading is relatively higher than that of under earthquake in damage identification simulation. Here, a reasonable sampling frequency range from 1000Hz to 2000Hz is suggested to be used in wavelet analysis.(4) The height of the identified signal spike is strongly related with the load intensity, the severity of damage as well as the distance from the measuring point (sensor position) to the damage location. With the increase of the excitation intensity, the height of the spike increases; With the decrease of the load intensity, the height of the spike decreases; With the increase of severity of the structural damage, the height of the spike increases, Contrariwise, the spike height decreases; With the increase of the distance from the measuring point to the damage location, the height of the spike decreases, Contrariwise, the spike height increases.(5) The noise effects of the measuring signal are: for identifying of the minus damage at a relatively lower excitation level, the sensors with very high precision are required. If the distance from the measuring point to the damage location is relatively large, the spike representing the minus damage might be submerged by the noise. If the damage is quite severe or the excitation level is quite large, the influence of the noise to the identified spike will be very slight.