Electrical Diagnosis of Gear Wear Fault in RV Reducer Based on Neural Networks

The Rotary Vector (RV) reducer, as one of the core components of the industrial robot execution unit, is widely used in industrial systems. Early pre-fault diagnosis and state detection are crucial and necessary for predicting device remaining life. Traditional vibration signal fault diagnosis methods based on vibration sensors are easily affected by factors such as environmental noises. However, neural networks technology can utilize its powerful feature self-learning ability to capture and process weak signal changes. The fault diagnosis of system or equipment is essentially a process of pattern recognition. The time-frequency domain features of the original signal are extracted and input into the neural network model for iterative training. In recent years, the pattern recognition function of neural network has been widely used in the field of fault diagnosis. This paper proposes a method for RV reducer gear wear fault diagnosis based on multiple neural network models combined with motor electrical signals and compares the application effects of various network models. The effectiveness of the proposed strategy is verified by a series of experimental results.