Research On Fault Diagnosis Of S700K Switch Machine Based On The Combination Of 1DCNN And BiLSTM

As the basic equipment of railway signaling,the S700K switch machine has a large number,high frequency of use,and complex working environment.When faults occur,whether it can be repaired in a timely manner has a great impact on the efficiency and safety of train operation.At present,the staff mainly use the action curve of the switch machine sampled by the signal centralized monitoring system as the basis for judgment,and use experience to identify and process faults of the switch machine.Problems such as low efficiency in fault identification,inaccurate positioning,poor real-time performance,and labor-intensive material resources cannot be avoided.With the rapid development of modern information technology,it has become a trend for switch machines to face automated and intelligent fault diagnosis.Based on the working principle and fault analysis of S700K switch machine,this thesis proposes a fault diagnosis method of switch machine combining One-Dimensional Convolutional Neural Network（1DCNN）and Bidirectional Long Short Term Memory（Bi LSTM）in view of the limitations of the current fault diagnosis methods of S700K switch machine.The main research work is as follows:Firstly,a detailed analysis is conducted on the mechanical structure,switch control circuit,and working principle of the S700K switch machine,studying the relationship between the output tension,output power,and working state of the switch machine.Based on the principle of collecting power curves,the normal operation process of the S700K switch machine is divided into 5 different stages,in order to conduct in-depth analysis and induction of its faults,and obtain 8 common types of switch machine faults on site and their corresponding fault power curves and causes.Secondly,in response to the difficulty in extracting effective fault features for S700K switch machine fault diagnosis,this thesis uses 1DCNN to directly apply it to the original signal to complete feature extraction.By detailing the network structure and important characteristics of 1DCNN,and explaining the Activation function and Dropout methods commonly used in deep learning,during simulation analysis,the visual analysis method based on t-Distributed Stochastic Neighbor Embedding（t-SNE）shows the characteristics of convolution layer extraction,verifies the effectiveness of automatic feature extraction ability of 1DCNN,and concludes that two-layer convolution has the best effect.Then,in response to the problem of difficult joint optimization between classification algorithms and signal processing in the S700K switch machine fault diagnosis method,this thesis combines 1DCNN and Bi LSTM.The original power curve is processed by Discretization and normalization,the processed data is input into 1DCNN layer to automatically extract important features,and then the extracted features are input into Bi LSTM layer through Flatten layer to further mine the correlation of features,and finally the Softmax function is used in the full connection layer to achieve intelligent fault diagnosis.Using the measured data provided by a certain telecommunications depot as the data validation for simulation,the model only needs to be optimized overall during training to determine the optimal hyperparameters.The simulation results show that the classification accuracy of the model reaches 97.88%,which is at least 1.08%higher than other models;The ablation experiment showed that 1DCNN and Bi LSTM were effectively combined,with accuracy,recall,and comprehensive evaluation index F₁values not less than 93.75%,93.33%,and 96.55%,respectively,proving that it can be used as an effective model for fault diagnosis of S700K switch machines.Finally,the S700K switch machine fault diagnosis system was designed and implemented.Based on model establishment and simulation analysis,a fault diagnosis system was constructed using Py Qt5 to further verify the feasibility of the proposed method and meet the real-time monitoring needs of switch machines on railway sites.