A wavelet-based system identification technique for structural damage assessment

In this study, a wavelet domain covariance-driven system identification technique is proposed to address structural damage assessment problem. Under the assumption that the ambient excitation can be modeled as a white-noise process, the output covariance is computed firstly to separate the effect of random excitation from the response measurement. The wavelet transform is then employed to convert the covariance vector in the time domain to the power scalogram in the time-scale plane. The wavelet coefficients along the energy concentrated curve are extracted and structural modal parameters and physical properties can then be estimated based on the amplitude and the phase angle of the extracted wavelet coefficients. As the wavelet transform has a capacity to capture both stationary and transient information from the original measurement, the proposed technique provides a promising approach for identifying structural modal parameters and physical properties under random excitations. After identifying structural stiffness matrix, structural damage position and severity can then be assessed by directly checking the change of the measured stiffness matrix. A numerical study on a 3-storey frame is performed to demonstrate the proposed technique and verify its accuracy. The results prove its capacity in both identifying structural physical properties and assessing structural damage conditions