Identification of in-process machine tool dynamics using forced vibrations in milling process

Abstract

An accurate description of machine tool dynamics is essential for health monitoring, chatter prevention, and improvement of manufacturing accuracy. The standard identification approach of experimental modal analysis at the standstill of the machine does not realize the possible variations in the dynamics due to operational conditions. The in-process identification methods in the literature, on the other hand, are associated with implementation difficulties in industrial environments due to the required complex and specially-designed setup, limited excitation forms, or excessive measurement efforts. This paper proposes an industrial-friendly method to estimate the in-process structural dynamics of machine tools, considering practical demands and implementation limits. Knowing that the operational dependency of the dynamics is associated with the machine and spindle structure, the machine–spindle and holder–tool structures are considered two distinguished subsystems. The cross Frequency Response Function (FRF) of the coupled structure, between the tooltip and spindle flange, is determined by measuring forced vibrations during milling operations. The milling forces are considered as excitation input and the process is designed to be stable and chatter-free so that the simulated forces can be accurately used in the identification. Then, the receptance coupling theory is utilized to develop an optimization algorithm. Given the model of the coupled structure, the evolutionary optimization tunes the modal parameters of the machine–spindle dynamics and joint parameters until the predicted cross FRFs match the experimentally determined FRFs. The direct FRF at the tooltip is predicted using identified in-process machine–spindle dynamics through the receptance coupling method. The identified in-process dynamics are used to predict stability diagrams for different tooling systems and, moreover, to design an augmented Kalman filter to estimate cutting forces. Comparison of estimated SLDs and cutting forces with chatter test results and dynamometer measurements validate the outcomes of the proposed identification method. This paper demonstrates that the system dynamics under operational conditions can be successfully identified without requiring costly setup, specially-designed workpieces or operations, and destructive chatter tests.