Bayesian operational modal analysis with multiple setups and multiple (possibly close) modes

Abstract

Operational modal analysis (OMA) is increasingly applied to identify the modal properties of a constructed structure for its high economy in implementation. Though great achievement has been made in OMA, it is still challenging in the scenario of multiple setup data with close modes, due to the need to assemble the global mode shapes and the intervention of closemodes, especially when the data quality is low in some setups. A Bayesian approach is developed in this paper to compute the most probable value (MPV) of modal parameters incorporating data from multiple setups and multiple (possibly close) modes. It employs an expectation-maximisation algorithm which admits an analytical update of modal parameters except the frequencies and damping ratios, thus allowing an efficient computation of the MPV, usually in the order of tens of seconds for each frequency band even when there are a large number of degrees of freedom and long data. A comprehensive study based on synthetic and field test data is presented to illustrate the performance of the proposed algorithm. Comparing with three existing algorithms, it shows the quality of the identified global mode shape is good and insensitive to the method used when the data quality is consistently high in all setups; However, only the proposed Bayesian approach yields consistently reasonable results when the data quality is low in some setups.