Research On Feature Extraction And Fault Diagnosis Method Of Vibration Signal Of Rolling Bearing In Train

Railway transportation is an important part of the integrated transport system, but also the booster of the "the Belt and Road Initiative". Safety is the theme of the railway transportation, especially in the time when the speed of the train is increasing and the high-speed railway is booming, the security problem becomes the focus of the whole world. Trains are the most important tools to complete the task of railway transportation, it is directly related to whether the railway transportation is safe and efficient. And whether the trains are able to run normally depend on the operating statuses of their key parts. However, the on-line identification, diagnosis, prediction and monitoring technology of the existing trains and their key parts in China are far from meeting the needs of the sustainable development of the integrated transport system.For this important problem, it is urgent to study and propose an effective method for identifying and monitoring the states of the key parts of the trains. For this reason, this paper carried out the research works on the rolling bearing of the trains as follows:(1) In order to deal with the demodulation problem of rolling bearing defect vibration signal in heavy noise, a new method based on time-delayed correlation algorithm and ensemble empirical mode decomposition (EEMD) was presented. After the discretization and unbiased estimation of the original signal’s autocorrelation function, denoising pretreatment was implemented by appending a rectangle window. Then an envelope signal can be obtained. After the EEMD decomposition, some interested intrinsic mode functions (IMFs) can be collected. By making the Hilbert transform of the IMFs, we can get the local Hilbert marginal spectrum from which the defects in a rolling bearing can be identified. Through simulation analysis and experimental data validation of motor bearing, the results show that the proposed method is more effective than direct modulation or only time-delayed correlation demodulation or combine time-delayed correlation with EMD demodulation in denoising and diagnosing the rolling bearing’s defect information.(2) When the rolling bearing fails, it is usually difficult to determine the damage degree. Aiming at this problem, a new fault diagnosis method was presented to achieve feature extraction and intelligent classification of different fault positions and damage degree of rolling bearing signal. To start with, MFDFA was used to compute the multifractal spectrum of the vibration signal of each status. Next, two kinds of multifractal spectrum parameters which are the most sensitive and stable were found and employed as fault feature values. Then feature values were regarded as the input of LSSVM based on PSO for judging rolling bearing fault position and its damage degree. Finally, the effectiveness of the method was verified by analyzing the data of motor bearing and the axle box bearing, and comparison was made with other related methods. The results show that, the presented method can accurately achieved the intelligent diagnosis of rolling bearing fault position and damage degree, has better generalization than LSSVM or support vectors machine (SVM) methods which not optimized by PSO, and is hopeful for solving practical engineering problems.(3) The relationship between ASD parameters and the degree of fault of rolling bearing is studied. A new intelligent diagnosis model based on ASD parameter estimation and PSO-LSSVM is proposed. To start with, estimate the four parameters of ASD of the vibration signals of each status. Next, two kinds of parameters which are the most sensitive and stable were found and employed as fault feature values. Then take the feature values into the PSO-LSSVM for judging rolling bearing fault position and its damage degree. Finally, the effectiveness of the method was verified by analyzing the data of motor bearing and the axle box bearing. The results show that, the presented method can accurately achieved the intelligent diagnosis of rolling bearing fault position and damage degree.(4) The relationship between MFDFA parameters and ASD parameters is studied, and a kind of intelligent diagnosis model based on feature fusion of MFDFA and ASD is proposed. Firstly, MFDFA was used to compute five MFDFA features while five ASD features were obtained by fitting the distribution to the vibration signals of each status and calculating the Probability Density Function (PDF). Secondly, Kernel Principle Component Analysis (KPCA) was used to achieve dimensionality reduction fusion of the combination of original features to gain the Kernel Principle Component Fusion Features (KPCFF). Thirdly, the KPCFFs served as the input of PSO-LSSVM to assess rolling bearings’ fault position and damage severity. Finally, the effectiveness of the method was validated by the data of motor bearing. The results demonstrated that the developed method can achieve intelligent diagnosis of rolling bearings’ fault position and damage degree and can yield better diagnosis accuracy than single feature method or corresponding single feature fusion method.