Eine Plattform für die Wissenschaft: Bauingenieurwesen, Architektur und Urbanistik
Eine Plattform für die Wissenschaft: Bauingenieurwesen, Architektur und Urbanistik
A machine learning approach for maintenance prediction of railway assets
With the shift from manual to computerized solutions, many railway agencies are storing and managing the immense amount of data relating to assets’ properties and their operational performance. Yet, the maintenance of these assets is still driven by available budgets, planned schedules, experts’ intuition, and abrupt failures. Railway agencies are lacking automated solutions that could make use of the available data and could assist in decision-making processes related to maintenance planning. In this paper, we try to limit this gap by making use of machine learning approaches to harness the data. We trained two machine-learning classifiers on the dataset of points and crossings with the aim to predict their maintenance need and specific maintenance treatment. Based on the ensemble approach, random forest classifier obtained 87% accuracy in predicting maintenance need and 83% accuracy to predict maintenance treatment. The main contribution of this paper is, in using the data generated from the well-known SAP ERP system, to develop classifiers that are able to assist infrastructure managers in future maintenance decision-making.
A machine learning approach for maintenance prediction of railway assets
With the shift from manual to computerized solutions, many railway agencies are storing and managing the immense amount of data relating to assets’ properties and their operational performance. Yet, the maintenance of these assets is still driven by available budgets, planned schedules, experts’ intuition, and abrupt failures. Railway agencies are lacking automated solutions that could make use of the available data and could assist in decision-making processes related to maintenance planning. In this paper, we try to limit this gap by making use of machine learning approaches to harness the data. We trained two machine-learning classifiers on the dataset of points and crossings with the aim to predict their maintenance need and specific maintenance treatment. Based on the ensemble approach, random forest classifier obtained 87% accuracy in predicting maintenance need and 83% accuracy to predict maintenance treatment. The main contribution of this paper is, in using the data generated from the well-known SAP ERP system, to develop classifiers that are able to assist infrastructure managers in future maintenance decision-making.
A machine learning approach for maintenance prediction of railway assets
Zaharah Allah Bukhsh (Autor:in) / Aaqib Saaed (Autor:in) / Irina Stipanovic (Autor:in)
17.04.2018
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
DDC:
690
| British Library Conference Proceedings | 2002
Railway assets: effective outsourcing the key
British Library Online Contents | 2005
Deep Learning for Automatic Railway Maintenance
| Springer Verlag | 2021
Urban Highways Maintenance - Fixing city assets
Online Contents | 2009