A Hybrid Modeling and Data-Driven Approach for Vibration Analysis of the Ball Screw Feed Drive System

The ball screw drive system (BFDS) is known as a key component in modern CNC milling machines, which determines the quality of the finished product. Due to heavy and continuous operating conditions, the condition of the BFDS will deteriorate, with the leading causes including wear, fatigue, or cracking. In this study, based on the dynamic characteristics, the primary purpose is to propose a method combining model-based techniques and data-driven approaches to monitor the vibration modes of the BFDS. By using a suitable dynamic modeling method for the system in conjunction with the dynamic parameters published by the manufacturer, it is possible to determine the vibration frequencies corresponding to the ball screw vibration modes, including axial, torsional, and bending vibrations. In particular, the Lumped mass method is employed to represent the BFDS with multiple degrees of freedom, allowing for a more accurate investigation of axial and torsional vibrations compared to results calculated from empirical formulas and those obtained from experimental measurements. Additionally, the Euler-Bernoulli beam model is utilized to calculate the first-order bending vibration frequency of the ball screw. Furthermore, measurement experiments and the application of vibration signal analysis techniques, including Fast Fourier Transform (FFT) and Cumulative Calculation techniques, are conducted to determine the vibration frequencies of the BFDS. These results are then compared with the theoretical calculations to evaluate the accuracy of the proposed model. This analysis facilitates the determination and provision of detailed information about the actual dynamic behavior of the BFDS, laying the foundation for its development into a condition monitoring (CM) solution.