Research On Fault Diagnosis For Track Circuit Based On Intelligent Optimized Deep Network

In recent years,with the rapid development of Chinese railroads,the importance of railroads in China has increased.Ensuring the safe and efficient operation of railroad trains is becoming more and more critical.As a part of railroad train operation control,the track circuit plays an important role in ensuring safe and efficient train operation,and the normal operation of the track circuit will directly affect the efficiency and safety of train operation.How to use artificial intelligence technology to improve the intelligent level of track circuit fault diagnosis is of great significance.In this thesis,we combine various artificial intelligence techniques to study the method of track circuit fault diagnosis based on intelligent optimization depth network.The main contents of this thesis is as follows:(1)Taking ZPW-2000 rail circuit as the research object,the basic structure and principle of rail circuit are introduced,and 20 common failure modes of rail circuit are summarized.The theoretical analysis of the rail circuit is carried out by the four-terminal network theory,and the simulation model of the rail circuit is established by using Matlab/Simulink software,and the validity of the simulation model is verified by comparing with the measured rail circuit data,and the fault data is collected by setting different fault parameters for each module.(2)A rail circuit fault diagnosis method based on intelligent optimized integrated learning is proposed in the supervised case.Organic ensemble learning is combined with intelligent computing and reinforcement learning to fully exploit the rail circuit fault characteristics and improve the performance evaluation index.Firstly,convolutional neural networks,long and short-term memory networks and multilayer perceptron deep learning models,as well as support vector machines and random forests traditional machine learning models are used,which together constitute the ensemble learning base learners and solve the shortage of single learning models,and the use of different base learners ensures the diversity of ensemble learning.From the perspective of automated machine learning,the improved sparrow algorithm is used to optimize the structure and parameters of this ensemble learning model to overcome the problem that its structure and parameters are difficult to determine;On top of this,reinforcement learning Q-learning is introduced to optimize the combined weights of each base learner in the ensemble model,and the combined weights of each base learner of ensemble learning are determined intelligently;Finally,the error is obtained by comparing the prediction results of the ensemble learning model with the real results,and then the BP neural network is used to compensate and correct the prediction results,which further improves the fault diagnosis performance evaluation index of the rail circuit.The effectiveness and superiority of the proposed method are verified by comparison experiments.(3)In the semi-supervised case,an intelligent optimized semi-supervised generative adversarial network-based fault diagnosis method for rail circuits is proposed.The semi-supervised generative adversarial network simultaneously utilizes a small number of labeled samples in the training set,a large number of unlabeled samples and simulated samples generated by the generative network to participate in the training process of the discriminative network,and achieves a high accuracy rate of fault diagnosis.To address the problem of blindness in network design,the semi-supervised generative adversarial network structure and parameters are optimized using the improved gray wolf optimization algorithm to further improve the fault diagnosis performance of the model.The experimental results show that the proposed method in the semi-supervised case has high performance in all evaluation indexes.