Remaining useful lifetime estimation on multi-modal failure in railway switches

Remaining useful lifetime estimation on multi-modal failure in railway switches

Hutten, Florian (TU Delft Mechanical, Maritime and Materials Engineering)

Pang, Y. (mentor)
Negenborn, R.R. (graduation committee)

Delft University of Technology

Marine Technology | Transport Engineering and Logistics

2019-01-16

In this day and age, failure of railway switches is managed using preventive and corrective maintenance, not predictive maintenance. These maintenance methods may allow for failure, where such a feailure could have been predicted. To combat this issue, this paper proposes a real-time predictive maintenance framework, which can identify the failure mode, as well as estimate the reamining useful lifetime of the switch. The failure mode identification is facilitated by a principal component analysis (PCA), which analyzes the features extracted from the power measurements of the electromotor. The RUL estimation is facilitated by the use of a hidden semi-Markov model, which is trained and tested on the same sensor data as the PCA. The framework is then implemented in a real-world environment to identify the failure mode and to predict the moment of failure in the future. The framework evaluation reveals that the model prediction show better results then a naive prediction.

HSMM
RUL
Rail
Markov models
Principal Component Analysis
Failure Prediction
Predictive Maintenance

http://resolver.tudelft.nl/uuid:a87e6d3c-f2d6-4144-aa12-f1e8e8e32f3a

2024-01-01

Student theses

master thesis

© 2019 Florian Hutten