Structural Modal Time Domain Identification Method Based on the Bayesian Uncertain Quantification.

Based on the Bayesian framework, a time domain method is proposed for the uncertain quantification of structural modal identification. First, a theoretical prediction model is constructed from the state space model in modal space and then transformed into physical space using the modal basis. Second, taking into account the uncertainty of the identification results caused by measurement noise and modeling errors, the negative log-likelihood function is constructed using time domain measurement data and a theoretical prediction model based on the Bayesian system identification framework. Finally, an unconstrained quadratic function for the identification parameters is derived through matrix vectorization, and, by mathematically transforming the optimization problem, only the dynamic spectral parameters (the natural frequencies and damping ratios) need to be identified, while the spatial parameters (the mode shapes and modal contribution factors) can be analytically calculated from the spectral parameters, which greatly reduces the dimensionality of the identification parameters. In numerical examples, the identification of the modal parameters for a spring–mass system and high-speed pantograph was studied, and the identified modal parameters based on the simulation response's data were in good agreement with the theoretical values. Moreover, the modal parameters of the actual structure of the pantograph were identified based on the experimental data, and the identifying uncertainties were quantified by the coefficient of variation. [ABSTRACT FROM AUTHOR]