Title: Fault diagnostics on railway switches and crossings using a bat-optimised sequence-to-sequence neural model

Authors: Xiaoyun Deng

Addresses: School of Urban Rail Transportation, Liuzhou Railway Vocational Technical College, Liuzhou, 545616, China

Abstract: Railway switch and crossing requires large and complex structures, components, and technologies like electronic device control and switches to improve the railway network process. The large volume of components requires frequent monitoring and diagnosis systems to diagnose the faults. The fault diagnosis and identification system monitors the track circuits to ensure the railway network's safety and availability. The traditional fault diagnosis system requires fine-tuning methodologies to handle the complex switch-crossing information. Therefore, this paper uses the bat optimised sequence to sequence neural model (BOS2SNM) to handle the task according to the available measurement signals. The neural model identifies the faults from the temporal and spatial dependences, and the model learns from the data. During this process, the network function is optimised by applying the bat optimisation algorithm that minimises the deviation between the output values. The introduced system efficiency is analysed using the publically available dataset information and respective performance metrics. The experimental outcomes of the suggested BOS2SNM attain a high fault detection rate of 98.5%, fault diagnosis accuracy rate of 97.9%, prediction rate of 96.7%, precision rate of 95.6%, and F1-score rate of 94.5% compared to existing techniques.

Keywords: railway switch and crossing; fault diagnosis; bat algorithm; sequence to sequence model; neural networks.

DOI: 10.1504/IJSNET.2024.138516

International Journal of Sensor Networks, 2024 Vol.44 No.4, pp.226 - 236

Received: 20 Oct 2023
Accepted: 31 Oct 2023
Published online: 08 May 2024 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article