Title: Data acquisition systems for alternating current switch machine prediction and health management

Authors: Xiongsheng Wu; Hanqing Tao

Addresses: School of Railway Signalling and Communication, Liuzhou Railway Vocational Technical College, Guangxi 545616, China ' School of Railway Signalling and Communication, Liuzhou Railway Vocational Technical College, Guangxi 545616, China

Abstract: Alternating current (AC) switch health predictive maintenance is crucial for reducing downtime and improving efficiency. The system analyses operational data like pressure, temperature, vibration, and voltage to predict potential failures using various learning techniques. However, it faces challenges such as slow convergence, suboptimal accuracy, and high computational costs. These issues are addressed by the optimised neural model (ONM), which employs a sequence-to-sequence neural model and grasshopper optimisation. Data is processed through windowing and lag feature procedures, followed by feature engineering to extract domain-specific statistics. The optimised algorithm fine-tunes parameters and captures temporal dependencies, achieving 98.56% accuracy and a loss function of 0.012. This enhances prediction robustness and reliability, ultimately optimising maintenance schedules and operational efficiency.

Keywords: alternating current; AC; switch machine; predictive maintenance; optimised neural model; ONM; windowing; lag features; exploration-exploitation; robustness; reliability.

DOI: 10.1504/IJSNET.2025.145625

International Journal of Sensor Networks, 2025 Vol.47 No.4, pp.255 - 269

Received: 19 Nov 2024
Accepted: 11 Dec 2024
Published online: 09 Apr 2025 *

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