Damage Detection And On-line Monitoring In Frame Structures Based On Wavelet Analysis

Damage detection and on-line monitoring are becoming more and more important in civil engineering. The invention of the dynamic detection methods, compared with the static methods, make damage detection more economical and more convenient, make on-line monitoring more practical and more feasible. At present, the methods in frequency domain are developed most integrally and applied most widely, and which in time domain are the second. But the local analysis in time and frequency simultaneously may not be satisfied by the techniques in frequency domain or in time domain, especially for the non-stationary and time-variant signals. In sensitivity analysis, the method of the sensitivity of the wavelet coefficient, which is developed recently, is more sensitive to damage than the ones ago in theory, for it recurs to the function of the wavelets in zooming in local character, but more experiment is needed to verify. Unavoidable noises mixed in signals in sampling are an obstacle of developing of the dynamic detection methods. The noise, whose origin is comprehensive and characteristic is complicated, often interfere with conner to obtain the real and valuable information of the structure from the sampled signals. Aiming at the considered problems above, the main content of the thesis, which researches by wavelet analysis, is listed as follows:1. The equation of the response sensitivity is a second order differential equation, which can be obtained from libration equation of multi-degree-of-freedom. It can be solved by Newmark method, which is a numerical iterative arithmetic. Then the sensitivity of the wavelet coefficient will be obtained by transforming the response sensitivity by wavelets. In the thesis, the sensitive extent of the sensitivity of the wavelet coefficient and the capability of identifying the changes of the physical parameter are verified by detecting and analyzing a four-storey frame in laboratory. Response sensitivity and the sensitivity of the wavelet coefficient and many different scales of it are compared to find out which parameter is more sensitive to damage. 2. One kind of signal analysis method for on-line monitoring is proposed, by the local analysis function of wavelets in time and frequency simultaneously and the perfect effect of Fourier in disposing stationary signal. When the happening of the damage can't be identified directly for the disturbance and noise, it can be identified after dealing with the signals by denoising, extending scale or observe the scales in file. The reliability and the feasibility are verified by detecting and analyzing a four-storey frame in laboratory. The time-variant moment is identified by wavelet, and the change of frequency is identified by Fourier.3. It can be denoised by the multi-scale analysis of the wavelets. The noise can be decomposed into three different parts, according to the origin and characteristic. One kind of denoising way in classifying and by steps,"denoising accidental coloured noise by contrast signalsâ†'denoising proper coloured noise by zero-replacement in part of signalsâ†'denoising white noise by compressing signals", is proposed in this thesis. After sampling and analyzing the signals of a four-storey frame in laboratory, the signals which are proved containing more noise are denoised by comparing with the signals which are proved containing less noise. The results are satisfying.