Fast Bayesian ambient modal identification with separated modes incorporating multiple setups

In full-scale ambient vibration tests, many situations exist where it is required to obtain a detailed mode shape of a structure with a limited number of sensors. A common feasible strategy is to perform multiple setups with each one covering different parts of the structure while sharing some reference degrees of freedom (dofs) in common. Conventional non-probabilistic methods exist that assemble the mode shapes identified in individual setups to form a global one covering all measured dofs. Challenging situations do exist where the assembled mode shape is questionable, possibly due to smearing effect of some setups which are of poor quality. This paper presents a Fast Bayesian method for modal identification that can incorporate the information in different setups consistent with probability logic. Using the proposed algorithm, Bayesian modal identification incorporating multiple setups can now be performed practically even on site. The method is illustrated with synthetic and field data.