Abstract:As one of the three major outdoor components of the railroad signal system, the track circuit plays an important role in ensuring the safety and efficiency of train operation. Therefore, when a fault occurs, the cause of the fault needs to be found quickly and accurately and dealt with in a timely manner to avoid affecting the efficiency of train operation and the occurrence of safety accidents. This article proposes a fault diagnosis method based on multi-scale attention network, which uses Gramian Angular Field (GAF) to transform one-dimensional time series into two-dimensional images, making full use of the advantages of convolutional networks in processing image data. A new feature fusion training structure is designed to effectively train the model, fully extract features at different scales, and fusing spatial feature information through spatial attention mechanisms. Finally, experiments are conducted using real track circuit fault datasets, and the accuracy of fault diagnosis reaches 99.36%, and our model demonstrates better performance compared to classical and state-of-the-art models. And the ablation experiments verified that each module in the designed model plays a key role.

What problem does this paper attempt to address?

The paper proposes a fault diagnosis method for railway track circuits based on Multi-Scale Attention Network (MSAN). Railway track circuits are crucial in railway signal systems, and it is important to quickly and accurately locate and handle faults to avoid affecting train operation efficiency and safety. The paper utilizes Gramian Angular Field (GAF) to convert one-dimensional time series data into two-dimensional images, and takes advantage of convolutional neural networks to process image data. A new feature fusion training structure is designed to effectively extract feature information of different scales through spatial attention mechanisms. Experimental results show that the diagnostic accuracy of the model on a real railway track circuit fault dataset reaches 99.36%, outperforming traditional and state-of-the-art models. The key roles of each module in the model are verified through ablation experiments. The main contributions of the paper include: 1. Proposing a data preprocessing model based on GAF to convert one-dimensional time series data into image data, which preserves complete signal information and temporal dependencies and improves classification accuracy. 2. Designing a feature fusion method that extracts features of different scales through multi-scale convolutional networks, U-Net networks, and spatial attention networks, solving the time-consuming and inefficient problems of convolutional networks in fault diagnosis process. 3. Conducting fine experiments with real collected datasets to demonstrate the usability and accuracy of the method, and comparing it with other methods to show its superiority. The experiments section demonstrates multiple experiments using MSAN, with an average accuracy of 99.36%, and analyzes the reasons for misclassification through confusion matrices. These results validate the efficiency and accuracy of MSAN in railway track circuit fault detection.

Multi-scale attention network (MSAN) for track circuits fault diagnosis

An Efficient Approach for Guided Wave Structural Monitoring of Switch Rails Via Deep Convolutional Neural Network-Based Transfer Learning

Temporal-Spatial Attention Network: A Novel Axial Piston Pump Coupled Fault Diagnosis Method

An effective CNN-MHSA method for the fault diagnosis of ZPW-2000A track circuit

MSGAFN: Multi-scale graph attention fusion network for machine fault diagnosis

Attention-aware temporal-spatial graph neural network with multi-sensor information fusion for fault diagnosis

Fusion Innovation: Multi-Scale Dilated Collaborative Model of ConvNeXt and MSDA for Fault Diagnosis

Method for Fault Diagnosis of Track Circuits Based on a Time–Frequency Intelligent Network

Grounding Fault Diagnosis of Running Rails Based on a Multi-scale One-Dimensional Convolutional Neural Network in a DC Subway System

Multi-view graph convolutional networks based on multi-source information fusion for mechanical fault diagnosis

Multirelational Fusion Graph Convolution Network With Multiscale Residual Network for Fault Diagnosis of Complex Industrial Processes

A Two-Terminal Fault Location Fusion Model of Transmission Line Based on CNN-Multi-Head-LSTM with an Attention Module

Train bearing fault diagnosis based on multi-sensor data fusion and dual-scale residual network

Multi-scale convolutional network with channel attention mechanism for rolling bearing fault diagnosis

Fault Diagnosis Method and Application Based on Multi-scale Neural Network and Data Enhancement for Strong Noise

Intelligent Fault Diagnosis with Multi-scale Convolutional Dense Network

Bidirectional Shrinkage Gated Recurrent Unit Network with Multiscale Attention Mechanism for Multisensor Fault Diagnosis

Unsupervised rolling bearing fault diagnosis method across working conditions based on multiscale convolutional neural network

Bearing Fault Diagnosis Method Based on Attention Mechanism and Multilayer Fusion Network.

Multi-scale split dual calibration network with periodic information for interpretable fault diagnosis of rotating machinery

Global contextual feature aggregation networks with multiscale attention mechanism for mechanical fault diagnosis under non-stationary conditions