Client Background
The Class 395 “Javelin” is the UK’s first domestic high-speed train, introduced in 2009 by Southeastern and built by Hitachi. Operating on High Speed 1 between London St Pancras, Ebbsfleet, and Ashford, it reaches speeds of 140 mph (225 km/h). Designed with dual-voltage capability, it can run on both high-speed and conventional Kent lines, improving connectivity across the region. Maintained at Hitachi’s Ashford depot, the Class 395 has become a symbol of reliability and efficiency, boosting regional growth and passenger satisfaction in Kent.
source: Rainham Rail Enthusiast
Our Approach
We employed a range of analytical and statistical techniques and determined:
- A large amount of regular checks (or exams) were ineffective, unnecessary and, in some cases, detrimental
- There were significant potential savings in moving away from regular maintenance to more sophisticated statistical approaches based on equipment history and condition
- Overall, there was a potential saving of some 20%
Client Challenges
Maintenance of high-speed rolling stock is costly due to the demanding requirements of operating safely and reliably at speeds up to 225 mph. High-speed operation causes greater wear on wheels, brakes, bogies, and pantographs, necessitating frequent replacement and inspection. Trains use advanced lightweight materials, propulsion systems, and digital technologies that require specialist tools and expertise. Strict safety regulations enforce proactive maintenance, while advanced diagnostics add further expense. Hitachi was seeking ways to contain or reduce maintenance costs.
Project Background
Hitachi engaged Opturion to look for ways to reduce maintenance costs. The focus was to look for opportunities to:
- Eliminate inspections (or exams)
- Reduce the frequency of exams.
- Predict failures based on the time between exams or maintenance
- Predict failures based on conditions
For checks, we went through a process to determine whether an exam was necessary at all or if the frequency was excessive. For example:
- Does the exam ever find anything? If so, how often?
- Does an exam look for something that the train manager would observe easily?
- Has an exam caused a problem previously?
For predictive maintenance, we used Weibull analysis to determine failure rates for all major systems.
For condition-based maintenance, we used a Bayesian approach to determine the prior and post distributions of failure.
The Results
There is potential to redirect a significant proportion of the resources devoted to scheduled exams towards more useful interventions. This presents the opportunity to save money, increase reliability, or both. For example, in 80% of exams, some tasks are either unnecessary, duplicated or could be done at other times.
Analysis and investigation have shown that regular statistical analysis (On Line Analytical Processing or OLAP) can identify early indications of unexpected failures.
There is an opportunity to move away from the current scheduled maintenance regime to a data-driven operation, where maintenance work is identified, prioritised, planned, and executed systematically.
In all, there is a potential cost saving of about 20% across the operation.
