The ball screw is a core component for achieving precise linear motion, and its operational state directly affects the positioning accuracy and dynamic performance of the transmission system. Therefore, fault diagnosis of ball screws is crucial for ensuring stable system operation and reliable performance. Currently, vibration signal-based analysis methods are susceptible to installation constraints and high hardware costs, while deep learning approaches relying on motor control signals, despite their strong feature extraction capability, suffer from poor model interpretability and high demand for training data. To address these issues, this paper proposes a fault diagnosis method for ball screws utilizing motor output shaft position signals and controller current signals. First, the lead current signal is extracted by combining the output shaft position signal and its corresponding current signal. Then, the spectrum matrix profile of the lead current signal is computed. Subsequently, the Z-score is employed as an anomaly score for the lead current signal to identify and localize faults on the screw. Finally, experiments are conducted on ball screws under both normal and faulty conditions. The results demonstrate that the proposed method can effectively detect faults and implement fault position on the screw from the current signal.

ball screw,fault diagnosis,Current signal,spectrum matrix profile