Predictive maintenance (PM) is the future of asset management in the rail sector, and a good data analytics project is the way to get there.

A successful data analytics project can analyse seemingly unrelated train data and predict future equipment and system failures with confidence levels as high as 80 per cent or more. However, orchestrating this kind of a project is no easy feat, explains Louisa de Vries of Downer’s Rollingstock Services.

“You need a detailed road map for investing in technology and data sets; and once you have validated your findings, you need to secure buy-in from stakeholders,” she said ahead of the AusRail PLUS 2019 Conference.

Ms. de Vries is managing one of Downer’s major PM software development projects – TrainDNA – which began in early 2018. The project is exploring vast volumes of historic train data – collected since 2012 – looking for correlations between factors like valve pressure and subsequent HVAC system failure.

“In analysing and making sense of this data we’ve had to dedicate several years, work with a strong team of experts, and manage the (oftentimes awkward) task of balancing budgets and stakeholder expectations – in a project which is, by nature, very unpredictable,” said Ms. de Vries.

“Unlike other projects – which tend to have a fixed budget and a discernible outcome – data analytics projects require flexibility and ongoing reinvention, depending on what sort of findings are revealed throughout the project.

“In some ways, it’s more like a science experiment. You have many terabytes of data and you’re not really sure what you’ll find. You may or may not find any correlation at all – so it’s difficult to map out an end point and make promises to stakeholders about what the outcome will be,” she said.

In overcoming these challenges, Downer has found an agile delivery model to be the most effective project management strategy.

Agile methodology places emphasis on team collaboration and regular stakeholder engagement. In doing so, it permits the scope of the project to adapt and change throughout its lifecycle.

“We have found that the best way to achieve stakeholder buy-in is through this incremental delivery model,” said Ms. De Vries. “Asking our stakeholders to wait a few years until the project’s completion would be too much of a trusting exercise. Everyone is much happier when we can articulate the findings at regular intervals and they can see the benefits along the way.”

“We have been running the project in four-week sprints. At the end of each sprint we release added functionality, which means stakeholders can start to take on board the findings as we go. It’s been an incremental process of research, report, release, repeat.

“Incremental change management has been another key benefit of the agile methodology,” she continued. “Through some of the findings we’ve articulated, we’re already starting to see benefits in our business.”

Downer’s first release was its pilot, at the end of year one. It then achieved a working version of the TrainDNA and, subsequently, began mapping out the next few four-week sprints.

“For our next release we are striving to have a more stable platform and get more capability through the analytics,” said Ms. de Vries.

“We have built up a number of reports already – looking at batteries, airbag pressure, HVAC systems – but the next milestone is to achieve greater stability in the platform.”

Hear more from Louisa de Vries and co-presenter Bertha Wai at this year’s AusRail PLUS 2019 Conference, due to take place 3-5 December 2019 in Sydney.

Learn more and register here.