KNN-FSVM for Fault Detection in High-Speed Trains - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2018

KNN-FSVM for Fault Detection in High-Speed Trains

Résumé

Prognostics and health management can improve the reliability and safety of transportation systems. Data collected from diverse sources provide a chance and at the same time a challenge for data-driven PHM methods and models. The data often exhibit challenging characteristics like imbalanced data on normal and faulty conditions, noise and outliers, data points of different importance for the data-driven model, etc. In this paper, a k nearest neighbors-based fuzzy support vector machine is proposed for reducing the computational burden and tackling the issue of imbalance and outlier data, in fault detection. Fault detection is mathematically a classification problem. In this paper, the reverse nearest neighbors technique is adopted for detecting outliers and the k nearest neighbors technique is used to identify the borderline points for defining the classification hyperplane in support vector machines. Considering the position of each data point and the distribution of its nearest neighbors, a new method is proposed for calculating their estimation error costs. A real case study concerning fault detection in a braking system of a high-speed train is considered.
Fichier principal
Vignette du fichier
08448688.pdf (1.05 Mo) Télécharger le fichier
Origine : Fichiers éditeurs autorisés sur une archive ouverte
Loading...

Dates et versions

hal-01989032 , version 1 (08-02-2019)

Identifiants

  • HAL Id : hal-01989032 , version 1

Citer

Jie Liu, Enrico Zio. KNN-FSVM for Fault Detection in High-Speed Trains. 2018 IEEE International Conference on Prognostics and Health Management (ICPHM), 2018, Chongqing, China. ⟨hal-01989032⟩
41 Consultations
93 Téléchargements

Partager

Gmail Facebook X LinkedIn More