Condition monitoring and trend analysis of railway turnouts based on in-situ accelerometer measurements

Condition monitoring and trend analysis of railway turnouts based on in-situ accelerometer measurements

Paper in proceedings outside WoS and Scopus

High dynamic forces at railway switches and crossings (S&C) are the primary cause of frequent defect formation. Regular acquisition of onsite sensory data aids condition evaluation or maintenance planning, which subsequently mitigates problems of unexpected malfunction of S&C components. Accelerometer data collected by in-situ sensors in UK and Czech Republic were used in this research for defining important metrics and validating prediction models. A number of metrics can be calculated from collected signals to provide information about the condition of S&C and its components. Change of these parameters over time is revealed by trend analysis and may signalize increased material deterioration or formation of a defect. Trend analysis methods span from simple regression to more advanced machine learning models for time series prediction and are listed in this paper. Evaluation of proposed models is performed on collected data, and validation metrics are discussed. This paper provides a baseline for the development of a S&C condition monitoring system and overviews techniques for analysis of large amounts of data collected by automatic sensory systems.

High dynamic forces at railway switches and crossings (S&C) are the primary cause of frequent defect formation. Regular acquisition of onsite sensory data aids condition evaluation or maintenance planning, which subsequently mitigates problems of unexpected malfunction of S&C components. Accelerometer data collected by in-situ sensors in UK and Czech Republic were used in this research for defining important metrics and validating prediction models. A number of metrics can be calculated from collected signals to provide information about the condition of S&C and its components. Change of these parameters over time is revealed by trend analysis and may signalize increased material deterioration or formation of a defect. Trend analysis methods span from simple regression to more advanced machine learning models for time series prediction and are listed in this paper. Evaluation of proposed models is performed on collected data, and validation metrics are discussed. This paper provides a baseline for the development of a S&C condition monitoring system and overviews techniques for analysis of large amounts of data collected by automatic sensory systems.

Railway Switches and Crossings, Accelerometer Sensors, Trend Analysis, Predictive Maintenance

Railway Switches and Crossings, Accelerometer Sensors, Trend Analysis, Predictive Maintenance

KRČ, R.; AMBUR, R.; HADAŠ, Z.; OLABY, O.; VUKUŠIČ, I.; PLÁŠEK, O.; ENTEZAMI, M.; DIXON, R.

2024

22.08.2022

Civil-Comp Press

Edinburgh, United Kingdom

PROCEEDINGS OF THE FIFTH INTERNATIONAL CONFERENCE ON RAILWAY TECHNOLOGY: RESEARCH, DEVELOPMENT AND MAINTENANCE

10

https://www.ctresources.info/ccc/paper.html?id=9502