Detecting Wear in a Ball Screw Using a Data-Driven Approach

In this paper we present a novel method for detecting wear in a ball screw. A ball screw is a linear actuator that translates rotational motion to linear motion or vice versa. Monitoring ball screws in a timely manner is important to ensure availability of machines and therefore productivity. The method is developed in a data-driven approach by analyzing annotated measurement data of different ball screws from different manufacturers and identifying the significant differences between good and worn ball screws. This produces three meaningful indicators for wear, related to failure modes. One of the indicators is obtained via vibration analysis. Signals from an accelerometer are transformed to frequency space by fast Fourier transformation, and certain frequency bands are observed for abnormalities like a permanently increased spectral energy. Another indicator is obtained by observing the temperature of the spindle, which is affected by higher internal friction due to the damage. Increased friction also affects efficiency of a ball screw. Hence, the engine that drives a ball screw has to deliver more power to produce the same output power. Each of these three indicators delivers a characteristic value related to the fault state of a ball screw. Long term test measurements at different ball screws show the capability of the method.