A novel self-sensing-based framework for misalignment detection of shafting

Abstract Misalignment is one of the frequently occurring faults of shafting. Traditional misalignment detection methods become inapplicable due to the cost and deployment constraints of external sensors. Recently, the control signals (CS) of the motor have been utilized to detect faults in the load machine, called self-sensing of the machine. Existing misalignment detection methods rely heavily on stable CS with a high sampling frequency. However, due to limitations on calculation and storage, the motor control unit cannot collect CS at a high frequency. Besides, the operation of the shafting also determines that it is difficult to select a stable CS. Given this, we first derive the influence of misalignment on CS with a low sampling frequency by constructing the dynamic model. Then, we presented a novel locally stable information granulation density based anomaly detection method to detect the integral fluctuations caused by misalignment of the shafting from CS with a low sampling frequency. Finally, the results of the simulation and experiment show that this method can detect the misalignment of shafting from CS under variable speed and load conditions, even at the low sampling frequency of 1 Hz. CS collected from the engineering also validates its effectiveness. This framework can provide a novel solution for the self-sensing of misalignment in a shafting.