Modal identification based on the time–frequency domain decomposition of unknown-input dynamic tests

This paper proposes the time–frequency domain decomposition method for the modal identification of ambient vibration testing. The continuous wavelet transform of the matrix of response correlation functions is first computed from raw data and then decomposed using singular value decomposition, leading to singular values and singular vectors. The singular values are used for the identification of frequencies and damping ratios, while the singular vectors are used for mode shapes. Modal parameters are estimated over a stability interval of time, taking into account the influence of the edge effects of the continuous wavelet transform, the weak variation in instantaneous damped frequencies and the high correlation of successive instantaneous mode shapes. A full procedure for the method is proposed and validated using both simulated responses and recorded data from a real structure. The results show a good identification of frequencies and mode shapes for linear systems having close modes or even repeated modes. Damping estimates present however some discrepancies for lower modes in comparison with other operational modal analysis techniques.