Multiple Decompositions for Modal Identification of Structures in the Time Domain

Blind source separation (BSS) has proven to be an effective approach for modal analysis of a structure using the output responses. The main challenge of this approach is to solve a smaller number of measured signals than operational modes (an underdetermined problem) and to estimate the exact number of operational modes automatically. Recently, tensor decomposition-based methods have attracted considerable attention for solving many complex situations in this field. This paper proposes a procedure based on multiple decompositions of covariance tensors constructed from vibration measurements for mode analysis in the time domain. Unlike a single decomposition, the proposed procedure uses multiple decompositions allowing the automatic selection of the exact number of these modes and the identification of modal parameters. The efficiency of the proposed procedure is illustrated by simulations and an experiment for different cases, even at different noise levels.