Data-driven Experimental Modal Analysis by Dynamic Mode Decomposition

TL;DR

This paper explores the use of Dynamic Mode Decomposition (DMD) for experimental modal analysis, demonstrating its effectiveness and limitations in extracting modal parameters from linear mechanical systems.

Contribution

It introduces DMD as a novel approach for EMA, compares it with existing methods, and analyzes its robustness against measurement errors.

Findings

DMD accurately captures modal parameters with small measurement errors.

Large measurement errors impair DMD's ability to extract modal parameters.

DMD's modal parameters are comparable to traditional EMA methods.

Abstract

This paper discusses the application of Dynamic Mode Decomposition (DMD) to the extraction of modal properties of linear mechanical systems, i.e., experimental modal analysis (EMA). First, theoretical background of the DMD is briefly reviewed and its relevance to the Ibrahim time-domain method is discussed. Second, DMD is applied to a single DOF system and multi-DOF discrete system to discuss the applicability and interpretation of the DMD as a method of EMA. Furthermore, the effects of measurement errors on the results of DMD are discussed. It is shown that with relatively small measurement errors, DMD can capture modal parameters accurately. However, with relatively large measurement errors, DMD fails to capture modal parameters. Finally, DMD is applied to experimentally-obtained displacement field of a cantilevered beam, and its modal parameters are extracted. It is shown that the modal parameters extracted by DMD are as accurate as the ones obtained by the existing modal parameter extraction method.