Compressed Dynamic Mode Decomposition for Full-Field Modal Identification from Video Measurements

Abstract

Traditional sensor-based modal identification methods can be intrusive, costly, and limited by sensor numbers and placement. This research presents an end-to-end vision- based automated full-field modal identification framework that overcomes these limitations by combining a Phase-Based Optical Flow (PBOF) method and a Hankel Compressed Dynamic Mode Decomposition (HCDMD) approach. The proposed HCDMD approach automates the full-field modal identification process with high efficiency by employing a compressed sensing strategy to accelerate computation and a Local Standard Deviation (LSD) index to separate structural and noise mode shapes. A laboratory experiment was utilized to validate the effectiveness and efficiency of the proposed framework in measuring full-field displacement and extracting modal parameters. The results demonstrate the potential of this fully automated end-to-end solution for full-field modal identification from video measurements.