Arrow DownArrow ForwardChevron DownDownload facebookGroup 2 Copy 4Created with Sketch. linkedinCombined-Shape mailGroup 4Created with Sketch. ShapeCreated with Sketch. twitteryoutube

Frontier decision-making is here

Find out more

Case study

Saving fleet maintenance costs for a leading train manufacturer

We pioneered forecasts of train breakdowns for a leading European train manufacturer. The forecasts allow for better scheduling of fleet maintenance, resulting in lower servicing costs, greater safety and increased reliability for train operators.


A leading European train manufacturer whose wider activities support more than 50,000 jobs in the UK economy.


In recent years train operating companies have moved away from ownership of the rolling stock towards a business in which trains are leased from manufacturers. As a result, maintenance of the fleet is largely carried out by manufacturers as part of the lease contract.

If manufacturers can use artificial intelligence to forecast breakdowns with greater accuracy, they can schedule maintenance more efficiently. This, in turn, would result in lower servicing costs and greater safety and reliability for train operators.

Considering the global rail transport industry as a whole, it is estimated that 1% of operational improvement would result in £20 billion of savings for companies in this sector.


The goal of the project was to build a machine learning (ML) system capable of identifying which train components are likely to break down soonest. We trawled through a data set consisting of diagnostic codes and failure notifications generated by 38 trains over three years. The interpretation of diagnostic codes was not specified, and no ‘dictionary’ was provided for translating numeric values into human-friendly status reports. Using phylogenetic analysis, we succeeded in classifying diagnostic codes into ‘families’ with specific meanings.

We then developed a neural network to predict failures of train components across the fleet.



We pioneered the use of deep learning for forecasting breakdowns in the rail industry. Our model enables the interpretation of 1.8 million diagnostic codes every year, whose meaning was previously not known or understood. For each train, this is equivalent to 5,000 additional data points that can be used to upgrade the safety and reliability of the fleet. For example, the number of doors that needed to be inspected to find a fault was reduced from 15 to three.

A conservative estimate of savings obtained by our neural network model is £1 million per year in passenger compensation alone. More efficient use of engineers’ time will add to this significantly.

Breakdown forecasting is set to radically change the way train manufacturers manage and price their servicing operations.



To find out more about what Faculty can do
for you and your organisation, get in touch.