Fault diagnosis of gear transmission system based on hybrid intelligent algorithms

The fault diagnosis of gear transmission system is a complex process because of it's various influencing factors and variable performance. Further more, the gear transmission system often runs under variable load dynamic transient conditions. In order to extract the fault features from the complex running state accurately and improve the reliability and effectiveness of fault diagnosis, a novel approach based on hybrid intelligent algorithms is proposed in this paper to diagnose faults in gear transmission system under load-varying dynamic transient conditions. We applies an improved particle swarm optimization (IPSO) algorithm to construct a sufficient small set of typical fault prototype which can represent efficiently the distributions of the fault classes. Meanwhile, an evolutionary Elman neural network is employed to predict the reference values of fault feature variables under dynamic operating conditions. Real-time fault symptoms are calculated with the predicted reference values and the real values of the feature variables. Then, fault diagnosis can be accomplished by calculating the similarity between the real-time fault symptoms and the constructed fault prototypes with the K-Nearest Neighbor classifier. Finally, the approach is applied to diagnose faults of a variable working conditions gearbox. The obtained results demonstrate the validity of the proposed approach.