Performance Degradation Evaluation of Gearbox Bearings Based on Wavelet Packet Singular Spectrum Entropy and LVQ Neural Network Clustering
In order to study the performance degradation evaluation of gearbox bearing,,the simulation vibra-tion signal training sample of gearbox bearing is decomposed by wavelet packet according to the relevant data of gearbox bearing and gear of high-speed train.The eigenvector formed by wavelet packet singular spectrum en-tropy is calculated,which is input into the learning vector quantization(LVQ)neural network clustering model to establish the performance degradation evaluation model.Secondly,the feature vector of the test sample is ex-tracted in the same way and input into the established model to evaluate the bearing performance degradation state.Then,the bearing life fatigue test is selected for analysis,and the feature optimization and fuzzy C-means clustering algorithm are selected for comparison.Finally,according to LVQ neural network clustering algo-rithm,the clustering centers of normal state and failure state in the training samples are determined,and the per-formance degradation evaluation model is established.The results show that the algorithm combined with wave-let packet singular spectrum entropy and LVQ neural network clustering algorithm can distinguish the different degradation states of gearbox bearings well and accurately represent the bearing performance degradation curve.Membership degree is calculated by membership function as an evaluation index of performance degradation,which can quantitatively characterize the state of performance degradation.By comparing the characteristics of time-domain indicators and frequency-domain indicators,it is verified that this research method is more effective and sensitive to the early degradation,and thus can detect early degradation in time and accurately evaluate the degree of degradation.
NETL
NSTL
万方数据
