Design, Modelling and Simulation | Predictive maintenance models for Sydney Trains - RMCRC
post-template-default,single,single-post,postid-221,single-format-standard,bridge-core-1.0.5,ajax_fade,page_not_loaded,,qode-title-hidden,qode-child-theme-ver-1.0.0,qode-theme-ver-18.1,qode-theme-bridge,disabled_footer_top,disabled_footer_bottom,wpb-js-composer js-comp-ver-7.6,vc_responsive

Design, Modelling and Simulation | Predictive maintenance models for Sydney Trains

Design, Modelling and Simulation

Predictive maintenance models for Sydney Trains

Sydney Trains (Tony Radosevic) / University of Technology Sydney (Bogdan Gabrys)

Predictive maintenance models for Sydney Trains

Research summary

The advent of cutting-edge machine learning and analytics provides rail operators with an opportunity to reposition asset maintenance interventions. This project incorporated research and development work to first assess the available historical data quality and suitability as well as highlight the benefits and potential for leveraging of very large volumes of data collected by Sydney Trains for asset condition monitoring and maintenance optimisation.

Start/end date

12 November 2018 to 31 December 2019

Total contracted budget (including in-kind)


Key achievements

  • A comprehensive, critical state-of-the-art review on modern prognostics and diagnostics approaches in transportation systems with identified successful implementation of machine learning and advanced analytics in railway systems worldwide.
  • Extensive analysis of Sydney Trains’ operational and maintenance data for identified use cases, including a large number of point machines, has been conducted and its suitability for predictive maintenance was assessed.
  • Extensive benchmarking and predictive models development has been undertaken on the collected and linked ST’s operational and maintenance data, as well as identified open-source datasets from the railway industry.
  • Interactive data visualisation and interrogation software has been developed for point machines use case featuring various statistical analyses and summarisations, as well as integration of the developed predictive models for verification and illustrative purposes.


Gabrys, B. Automated composition, optimisation and adaptation of complex predictive systems. An invited keynote talk presented in November 2019 at the Workshop on Learning and Mining with Industrial Data, 2019 IEEE International Conference on Data Mining (ICDM), Beijing, China, 2019.

Kocbek, S. and Gabrys, B. Automated Machine Learning Techniques in Prognostics of Railway Track Defects. 2019 International Conference on Data Mining Workshops (ICDMW), Beijing, China, 2019, pp. 777-784, doi: 10.1109/ICDMW.2019.00115.