Font Size: a A A

Research On Locomotive Bearing Fault Diagnosis Based On Wavelet Packet And Support Vector Machines

Posted on:2012-04-23Degree:MasterType:Thesis
Country:ChinaCandidate:B XiangFull Text:PDF
GTID:2218330368476172Subject:Control theory and control engineering
Abstract/Summary:PDF Full Text Request
Railway locomotive bearing is a key component of transportation equipment, As a long time, with the high-load operation, the locomotive bearing is prone to injury. Lots of Railway locomotive bearing faults belongs to the existence or occurrence of failures, thence, the quality of bearing running locomotive superior performance of the entire play. In view of the locomotive bearing the special status, therefore, carring out on the railway locomotive bearing fault diagnosis have good practical significance.The thesis was mainly based on wavelet packet anaiysis and signal pcocessing support vector machine based intelligent fault diagnosis research in two areas. The main contents are as follows:Discussion and analysis of the locomotive bearing the vibration characteristics of vibration signals and the mechanism of failure, simulated bearing fault signals of different components on the basis of mastering the fault signal characteristics, which can provide the method and basis vectors of locomotive bearing fault diagnosis.Traditional signal processing method is only suitable for stationary, non-time-varying signal processing, and does not have the capability of local signal analysis, using the advantages of wavelet packet analysis in mutation detection and signal denoising to decompose wavelet packet for fault signal, so that the characteristics of fault signals of different frequency information can be discovered. In order to analyze the amount of bearing on whether the point of view thave fault or not while taking the complexity of computation into account, selecting bands of fault characteristics obviously, which combined with the idea of frequency band energy analysis to achieve the feature vector extraction of simulated fault signal's specific band on bearing'outer ring, inner ring and the rolling.Using least squares support vector machines for fault diagnosis training samples do not need a lot of trouble, which also has a good classification ability in the case of high dimension, however, the support vector machine parameters (penalty factor and the radial basis function) model association have a greater impact on the diagnostic classification rate, with the deep study of particle swarm algorithm and a large number of simulation experiments, in order to improve the classification accuracy of the model, in this thesis, the particle swarm algorithm was used to optimize the combination of parameters, which has much influence in model classification ability. The simulation results shows that the classification capabilities of ARPSO particle swarm model for least squares support vector machine have been improved significantly.Using wavelet packet analysis of the fault signal feature vectors and the ARPSO particle swarm model for least squares support vector machine diagnostics together, it can achieve the intelligent fault diagnosis of locomotive bearing, and have achieved very good results.Theoretical and simulation results shows that the wavelet packet analysis support vector machine combined with the diagnostic method is effective on the locomotive bearing fault detection and diagnosis.
Keywords/Search Tags:Locomotive bearing, Fault diagnosis, Wavelet packet analysis, Support vector machine, Particle swarm optimization
PDF Full Text Request
Related items