A Fault Diagnosis System for a Mechanical Reducer Gear-Set Using Wigner-Ville Distribution and an Artificial Neural Network

This paper describes a fault diagnosis system for mechanical reducer gear-sets using Wigner-Ville distribution and artificial neural network techniques. Reducer gear-sets are used in various traditional and modern industries. In the production of a reducer, the vibration and noise signals of the gear-set are usually used to determine the defective products or defective positions. Unfortunately, conventional fault diagnosis by humans is limited effectiveness and has no numerical standards. In the present study, the vibration signal of the gear-set is used to evaluate the proposed fault diagnosis technique. In the experimental work, feature extraction by Wigner-Ville distribution is proposed for analyzing fault signals in the reducer gear-set platform. Artificial neural network techniques using both a general regression neural network and conventional back-propagation network are compared in the system. The experimental results show the vibration can be used to monitor the condition of the gear-set platform and the general regression neural network (GRNN) has a better recognition rate and less recognition time than the back-propagation neural network (BPNN).