Eemd Characteristics Analysis Based On Monitoring Data Of The High-Speed Train Running Gear

With the rapid development of the high-speed railway, the train speed is greatly improved, and the security issues are more serious. High-speed train running gear is an important part of the safety of the train, and it’s necessary to carry out research on its safety assessment analysis. The thesis based on monitoring data of high-speed train running gear, proposes an EEMD cross-fuzzy entropy extraction method, and do evaluation on several entropies of EEMD. The method provides the basis for troubleshooting and security state assessment of running train.The thesis firstly introduces three types of high-speed train running gear damper conditions and the source of experimental data, then do time-frequency domain analysis of the vibration characteristics of the various conditions in different directions at different speeds.Then, characteristics and decomposition of EEMD are studied deeply. The thesis takes EEMD mutual fuzzy entropy algorithm to do feature extraction of the simulated data from high-speed train running gear monitoring data.:(1) Make de-nosing of the vibration signal wavelet packet, and do EEMD decomposition, then a series of intrinsic mode function (IMF) vectors are got; (2) Extract mutual fuzzy entropy of each pair of IMF vectors to compose feature vectors; (3) Use support vector machine to do classification recognition of the feature vectors.Lastly, based on monitoring data of high-speed train running gear, the thesis designs evaluations and do integrated features assessments of the method proposed in the thesis and EEMD approximate entropy, EEMD sample entropy and EEMD fuzzy entropy from aspects of space distance, probability distance, recognition rate and confidence intervals to validate the algorithm.Based on monitoring data of high-speed train running gear, the thesis proposes a feature extraction algorithm based on EEMD mutual fuzzy entropy, and studies the features of high-speed train running gear in different operating conditions, and does classification and recognition of the features extracted to validate the algorithm. Lastly, several EEMD entropy feature extraction algorithms are evaluated based on monitoring data of high-speed train running gear.