Gearbox Fault Diagnosis Based On Filtering Techniques And Particle Swarm Optimization

As the gearbox body causes vibration and complex ambient noise and other factors influence during operation, the vibration signal acquired by the sensor contains not only useful fault information but also outside noise interference, which is not conducive to analysis and diagnosis of fault signals. From this perspective, this article firstly introduces the background, significance and current status of diagnostic methods of gearbox fault diagnosis,then propose adaptive filtering technique and and particle filtering technique in the pretreatment of gearbox vibration signals,and extract fault feature information, Finally, use the BP neural network optimized by particle swarm optimization algorithm to diagnose the fault of gearbox, and a complete set of gear box fault diagnosis system is built in the laboratory,which shows that this method can make accurate diagnosis of gearbox fault.Based on the study of several major adaptive filtering algorithms, a LMS algorithm (CTanh-LMS algorithm) based on correlated hyperbolic tangent function is proposed. This algorithm is applied to extract fault feature of gearbox based on the simulation, the results show that this method can obtain obvious fault characteristics, in addition, the algorithm is simple, converges fast, and fully meets the de-noising requirements of gearbox vibration signals. This article also according to the actual engineering system more for nonlinear non-gaussian this present situation, proposed the particle filter algorithm with to signal de-noising and the basic steps by simulation experiments demonstrate that the particle filter algorithm can greatly improve the SNR, signal with noise signal to play a good de-noising effects.Aimed at the disadvantage of basic particle swarm optimization (PSO)algorithm that is easy to fall into local extremum facing complex issues, combine this way algorithm with BP neural network,use PSO algorithm to adjust and optimize global parameters and use the BP neural network learning method to optimize local parameters. Combine the advantage of both methods, the accuracy of gearbox fault diagnosis is increased. The results of experiments showed that the proposed will adaptive filtering de-noising technology, particle filtering de-noising technology both used to signal pretreatment methods and particle swarm optimization respectively, combined with neural network technology was applied in gearbox fault diagnosis on, can get ideal result of the fault diagnosis.