Research On Fault Diagnosis Methods Of Railway Point Machine Based On Vibration Signal Analysis

With the rapid development of China’s railways,the demand for the safe and reliable operation of railway signaling systems is also increasing.As one of the three major outdoor signaling equipment,railway point machines play an extremely important role in the safe transportation of railways.Because the railway point machine works outdoors,the application scenarios are complex,the action frequency and failure rate are both very high,it is necessary to perform fault diagnosis and health monitoring on point machine.At present,the maintenance of railway point machines is still mainly manual,and there are problems such as low efficiency,missed diagnosis and misdiagnosis.However,the railway point machine is a mechanical and electrical equipment,the mechanical vibration generated during its action can reflect the fault characteristics in a new way,making up for the lack of electrical signals.Therefore,based on the vibration signal data of the railway point machine,the fault diagnosis research is carried out in this thesis,which provides a new idea for the fault diagnosis of the railway point machine.With the development of artificial intelligence technology,intelligent fault diagnosis methods based on machine learning begin to gradually replace the way of manual maintenance and identification of railway point machine’s faults.However,how to effectively and efficiently extract the vibration signal characteristics is the key to fault diagnosis of railway point machines.In recent years,the deep learning fault diagnosis methods driven by big data have become more and more widely used due to its adaptive feature extraction and powerful learning ability.However,the fault data actually collected on site has certain limitations,and there are problems such as insufficient number of samples and unbalanced distribution.Due to the limitation of data,the training of many diagnostic method models is constrained and the effect is not good.Aiming at the above problems,this thesis carries out multi-directional fault diagnosis methods research,based on signal processing,deep learning,multi-sensor fusion and data generation.The specific research contents are as follows:(1)Taking the vibration signal data of the railway point machine as the research object,a fault diagnosis method based on ensemble empirical mode decomposition,improved time-domain multi-scale dispersion entropy and particle swarm optimization support vector machine is proposed,and it is proved by experiments.The proposed method outperforms traditional signal processing methods.(2)Aiming at the shortcomings of signal processing methods requiring prior knowledge,a deep learning fault diagnosis method based on short-time Fourier transform and improved convolutional neural network is proposed,which can automatically extract key features and diagnose.Finally,the experimental analysis is carried out,and the diagnostic effect of the model in the noise environment is shown.(3)Aiming at the defects of limited information features of a single sensor and the inability of the model to obtain time delay features,a fault diagnosis method based on the MACN-GRU model is proposed.Based on the multi-channel mechanism of color image extraction by convolutional neural network,information fusion is carried out with channel attention;after the convolution module,a recurrent neural network module is added to extract the time delay feature information.Experiments show that the proposed method can further improve the fault diagnosis performance.(4)For the problems of insufficient sample size and unbalanced sample distribution,the network structure and objective function of the generative adversarial network are optimized,and a railway point machine fault data generation method based on the CACWGAN-GP model is proposed,which is used to generate high-quality samples to expand the training set.And the effect of improving the diagnosis accuracy after supplementing the fault data is verified by experiments,which can provide a certain guiding significance for the fault diagnosis of the railway point machine.