Research On Condition Rgcognition Of High Speed Train Based On Deep Learning And Classification Ensemble

China's high speed train develops rapidly,which has become the world's leading.However,with the high speed train running at a high speed,the performance of the bogie is down,which brings a huge hidden danger to the safe operation of the train.Bogie is an important part of high speed train,which plays an important role in the safety of the train and the comfort of passengers.One of the main techniques for monitoring the running status of the train is to collect and analyze the vibration signals by mounting the sensor on the train bogies.It has important research value to excavate the useful characteristic information from the monitoring data and realize the effective recognition of the typical condition.High speed train vibration signals are non-stationary and non-linear signals,which have the characteristics of complex and difficult identification.To solve the problem about the incompletion of feature extraction and inaccuracy of the identification performance on traditional methods,a multi-view classification ensemble method(MV-CE)for condition recognition of high speed train is proposed.The method combines with multi-view feature extraction and classification ensemble technology.For condition recognition,firstly,the wavelet energy,spectral coefficients and Ensemble Empirical Mode Decomposition fuzzy entropy are extracted.Further,we use the Fisher ratio to perform feature selection for the frequency domain features of the high speed train vibration signals,then the feature of the three views are constructed collectively.After that the Least Square Support Vector Machine(LSSVM)and the K-Nearest Neighbor(KNN)classifiers are used to recognize each view.Finally,the output results of multiple classifiers are integrated by using the Classification Entropy Voting Principle.The comparative results illustrate the completeness of the features extracted by the method and the validity of the ensemble model with diversity.Deep Belief Network(DBN)learns the features of the raw data automatically,and develops a new idea for the study of condition recognition of high speed train.Combining deep learning and classification ensemble technology,this thesis presents a novel DBN hierarchical ensemble model for high speed train condition recognition.Firstly,Fast Fourier Transform(FFT)coefficients of the vibration signals are extracted as the state of the visible layer of the model,and then DBN is used to learn the hierarchical features of the vibration signals automatically.The features of each layer learned by DBN are used to train Support Vector Machine(SVM),K-Nearest Neighbor(KNN),and Radial Basis Function(RBF)Neural Network respectively.Finally,the Majority Voting(MV),the Classification Entropy Voting Principle(CE),and the Winner Takes All(WTA)ensemble strategies are used for combination to get the final results.The experiments results show that the condition recognition rate of the proposed model is much higher than 10 contrast condition recognition methods.In addition,unlike the DBN model,the proposed model is affected slightly by the number of network layers and the size of hidden units.The vibration signals of different channels of the high speed train are both complementary and redundant.In order to make full use of the complementary information of multi-channel vibration signals,this thesis proposes a method of channel selection based on similarity ratio,and constructs a Multi-channel Deep Belief Network model(MDBN)to recognize conditions.Firstly,FFT coefficients of the HST vibration signals of all channels are extracted.Then the similarity ratio of FFT coefficients of each channel is calculated,and a number of channels with large similarity ratio are selected.Finally,MDBN model is constructed to learn the features of the selected multi-channel data and recognize conditions,and the feature fusion of multi-channel data is realized in the common layer of MDBN.The experimental results show that the feature extraction ability of MDBN is better than DBN model,and the condition recognition rate of MDBN is higher than DBN and DBN hierarchical ensemble model.