Structural damage detection with canonical correlation analysis using limited sensors

This paper presents a novel structural damage detection method using limited sensors for beam bridge subjected to a moving mass. In this method, the measured acceleration responses are cut out by a moving window with a length determined by the sampling frequency and the fundamental frequency of the bridge. The windowed responses are processed by the subspace analysis to extract the principal components. The correlation coefficients between the principal components in the window are obtained using the canonical correlation analysis (CCA). Finally, a new damage factor in the window is defined based on the output of the CCA. Moving the window, a time series of the damage factor is obtained. When the moving mass passes through the damage location of the bridge, the curve of the damage factor has a peak that the damage can be identified. To demonstrate the proposed method, a beam bridge is numerically modelled and a simply supported beam bridge model is tested in the lab. Both the numerical and experimental results demonstrate that this method can successfully identify the single and multiple damages with a limited number of sensors.