Asymmetric-dot-pattern fusion fault identification of motor-driven belt transmission system in industrial robots

The motor-driven belt transmission component is one of the most important parts in industrial robots on smart manufacturing production line. The complex motion and progressive wear may reduce their reliability and potentially lead to significant operational losses. Meanwhile, it is difficult to extract effective features from vibration signal for fault diagnosis due to the inherent buffering characteristics of the belts. However, the current signals can compensate the loss of some fault information with their sensitivity to the abnormal change of transmission torque. Therefore, an asymmetric-dot-pattern (aSDP) vibration and current fusion diagnosis strategy is proposed to accurately identify various fault types of motor-driven belt transmission. The current and vibration signals are fused into a single aSDP image based on empirical mode components in the same frequency band. In order to characterize the features from different aSDP images, the similarity between the aSDP images of unknown and template faults is calculated by the fusion of perceptual and difference hash. Furthermore, a weighted similarity mechanism is proposed to address the inconsistent classification of the similarity feature in different bands. Motor-driven belt transmission experiments are conducted on an industrial robot to validate the proposed current and vibration fusion methods under different conditions. Experiment results show the average fault identification accuracy of the proposed method is 98.70%. It demonstrates that the proposed method is capable of fusing current and vibration signals effectively for diagnosing faults of motor-driven transmission with flexible components and is preferable of the superior performance when compared to existing methods.