Damage detection of framed structures subjected to earthquake excitation using discrete wavelet analysis

This paper describes an application of discrete wavelet analysis for damage detection of a framed structure subjected to strong earthquake excitation. The response simulation data on each floor were obtained by non-linear dynamic analysis. Damage to the frame was introduced due to the non-linear behaviour of the columns and beams. In order for the structural members to reach the yield point or go slightly beyond yielding, the earthquake excitation was scaled up with the appropriate factor. Since the dynamic behaviour of an inelastic structure during an earthquake is a non-stationary process, discrete wavelet analysis was performed in order to analyze the simulation response data for each floor. It was shown that structural damage on a floor, and the time when this occurred can be clearly detected by spikes in the wavelet details of the response, acquired from the corresponding floor. Damage can be detected by observing the spikes directly from the wavelet details or following a statistical procedure. An automatic numerical procedure that can clearly distinguish between the spikes associated with damage and the spikes due to non-stationary excitation is proposed. The effects of noise were taken into account by adding a white Gaussian noise to the simulation response data. Damage to the element can also be detected again from the noised signal, if the level of details and the order of wavelets in the wavelet analysis of the response signal are increased. Numerical results show the effectiveness of the discrete wavelet approach to damage detection of framed structures.