VISION-BASED MODAL ANALYSIS OF MACHINE TOOL SYSTEMS: PROGRESS AND PROSPECTS

Abstract

Modal analysis of machine tools involves estimating natural frequencies, damping ratios, and mode shapes from the vibratory response of the machine tool. Usually, modal hammers and shakers are used to excite, and accelerometers or laser vibrometers are used to measure the response. Though these procedures have become routine, sometimes the use of accelerometers can result in mass-loading that distorts the response, and though laser vibrometers are non-contact, their use is precluded by their high costs. To counter these issues, vision-based modal analysis methods have emerged as a viable and promising alternative. The spatiotemporal response is estimated by treating every pixel in every frame in the video of the vibrating machine as a motion sensor. Image processing schemes leveraged from developments in allied fields are then used to register motion from video. The method is noncontact, full field, and only needs a camera and post-processing on a computer, and as such, it offers advantages over the traditional measurement methods. Since vision-based methods are potentially paradigm-shifting, this paper reviews the recent progress to contextualize the prospects of the method. The review includes discussions on selection considerations of cameras and acquisition parameters, on using markers and the machine's own features to register motion, on the efficacy of different motion registration schemes, and workarounds for when motion is spatiotemporally aliased. The paper concludes by discussing challenges and prospects related to motion synchronization, measuring speed and time-varying dynamics, and technological trends that may aid the adoption of the method.