A Robust Operational Modal Analysis Method and Its Application to a Concrete Arch-Gravity Dam

This paper deals with a recently developed operational modal analysis algorithm with emphasize on its theoretical basis and application. The algorithm establishes a connection between the second order blind identification with the covariance-driven stochastic subspace identification by introducing a new parametric state-space system model. The output vector of the state-space model contains the modal responses identified from the second order blind identification. Instead of repeatedly adopting single-degree-of-freedom fitting methods, all the identified modal responses are adopted to form the block Toeplitz matrix of the covariance-driven stochastic subspace identification method. Therefore, accurate modal parameters are expected to be obtained. The proposed method is applied to a concrete arch-gravity dam, 10 accelerometers in the dam body with 30 channels are used for modal identification and six modes are successfully identified, the auto correlation functions exhibit monotonic exponential decay, and the corresponding power spectral density functions are clean and concentrated, showing that the modal responses are well separated and the modal parameters are accurately identified.