A novel approach to wiring network diagnosis utilizing time domain reflectometry and one-dimensional convolutional neural networks
Résumé
The timely and precise diagnosis of wiring network faults is crucial for the reliable and safe operation of industrial systems. This study introduces an innovative approach to wiring network diagnosis, which employs Time Domain Reflectometry (TDR) and a One-Dimensional Convolutional Neural Network (1D-CNN). The principal advantages of the TDR-1D-CNN-based approach include the following: (1) the compact architectural configuration of the 1D-CNN, which solely performs one-dimensional convolutions, rendering it suitable for real-time fault detection and monitoring; (2) the cost-effective and practical implementation in real-time hardware; (3) the capability to function without any predetermined transformation, hand-crafted feature extraction, or feature selection; and (4) the capability of detecting, locating, and characterizing faults in wiring networks, as well as estimating the number of faults within the wiring network. The effectiveness and feasibility of the TDR-and 1D-CNN-based fault diagnosis approach are demonstrated on representative wiring network structures, achieving at least 99% accuracy and macro-average sensitivity, and surpassing the performance of existing diagnostic methods.
| Origine | Fichiers produits par l'(les) auteur(s) |
|---|---|
| Licence |
