Blind separation of structural modes in largely underdetermined cases using a principle of near-invariance under filtration

In the latest years, blind source separation (BSS) in the context of operational modal analysis (OMA) has gained significant attention in the scientific community, yet the identification of underdetermined scenarios (i.e. with more active modes than sensed responses) still remains a challenge. Recently a new technique based on the evaluation of complexity pursuit (CP) has been studied; it exploits the principle of near invariance of modes under filtration for the purpose of separation by using filtered versions of the measurement signal. The aim of this work is to propose a solution to the largely underdetermined case by employing a generalized version of CP (GCP) implemented in the frequency domain. Measurement signals are divided in sub-bands in which GCP is applied separately; a strategy based on the aforementioned principle is also proposed to artificially augment the number of measurement. Applications, on simulated and real data, and comparison of results proved this new approach to be fast and efficient and, above all, easy to use.