Servo Motor Electrical Fault Diagnosis of Misalignment Based on GRU Neural Network

Servo motors have been widely used in automated industrial production such as high-precision Computer Numerical Control. Ensuring long-term and safe operation of servo motors and transmission systems poses great challenges to fault diagnosis and early fault warning. Typical mechanical faults in the motor systems include various types such as installation misalignment, bearing damage and looseness. The occurrence probability of installation misalignment faults in varying degrees in industrial production is as high as 68%, which is the most common fault among all kinds. Due to the weak signal fluctuations, traditional time-frequency domain analysis methods may be difficult to achieve good results. This article proposes a combination of Permanent Magnet Synchronous Motor (PMSM) speed signal and Gated Recurrent Unit (GRU) model to diagnose and classify two types of misalignment faults. The original signal is sent to the GRU model for iterative training after extracting the time domain and frequency domain features. The effectiveness of the proposed strategy is verified by a series of experimental results.