Vibration-based health monitoring of ball screw in changing operational conditions

Structural health monitoring (SHM) of the ball screw of machine tools relies on the repeated observation of the damage-sensitive features. The natural frequency is a characteristic inherent to the structure that is more robust than other signal features as a SHM index. However, a major problem is that regular changes in the operating conditions, such as the feed speeds, the worktable positions and other uncertain complex boundary conditions, also affect the natural frequencies, and these factors may mask the subtle variations induced by structural damage. Therefore, it is necessary to adopt an effective method to eliminate those effects to avoid false alarms during structural damage monitoring. In this article, Bayesian ridge regression modelling methods are developed to eliminate the effects of the worktable different positions and feed speeds on the natural frequencies. The proposed method can model the natural frequencies that are simultaneously affected by two dimensional operational factors and provide a rigorous quantitative assessment of the uncertainties associated with the complex boundary conditions of the ball screw drive system and the inevitable estimation errors. First, the different worktable positions and feed speeds were tested on the experimental bench to verify the proposed method. Then, two novel criteria were used as damage warning signals, which can reduce false alarm. Finally, this method was applied to monitor the wear of a ball screw of a CNC machining centre in an automobile factory. The validity of the method was provend using actual monitoring results.