Italy's government-backed train operator, Trenitalia, is fitting its fleet of trains with sensors to enable 'predictive maintenance' of components, potentially saving hundreds of millions of Euros.

Trenitalia is the country's primary operator, with 241 high-speed trains, 800 regional trains and 600 cargo trains. 

In 2014 it embarked on a five-year internet of things project to provide real-time insight into the condition of its rolling stock.   

The plans centre around the creation of its dynamic maintenance management system (DMMS), which combines machine learning algorithms with SAP's Hana in-memory computation platform. 

"It is aimed at completely transforming the way we do maintenance at Trenitalia," said CIO, Danilo Gismondi, speaking at an event in Pietrarsa, Italy, last week.

He added: "To be able to anticipate events and maintenance operations will enable our company to reduce errors and make processes more efficient."

Through the project, Trenitalia hopes to realise an 8-10 percent saving in total maintenance costs. To put that in context, the operator currently spends around €1.3 billion on maintenance each year. There are also likely to be savings around service contract penalties as downtime is reduced due to unplanned maintenance.

Typically, trains are taken in for fixed-schedule maintenance checks either when a fault occurs or when they have travelled a certain distance. With the use of sensor data this will change to a "conditional maintenance" approach, said Trenitalia's head of technology, Marco Caposcuitti, with sensors measuring parameters such as temperature and pressure for damage or stresses. This allows Trenitalia to carry out maintenance "before the failure occurs", he explained. 

Data analytics

CIO Gismondi said that the trains will create an "incredible amount of data", with hundreds of sensors embedded in each train generating approximately 5,000 signals per second.

This data is transmitted to the Trenitalia data centre, which will have one petabyte of storage supporting its DMMS platform upon completion. It will also have a total of 24 terabytes of 'online' in-memory data within the SAP Hana platform.

Gismondi said that the data generated will be used throughout the organisation, feeding into its rolling stock management ERP system and eventually providing train information to customers.

"It is a part of a big programme where we put the customer in the centre of our thoughts," he said.

"By correlating the data generated by the sensors and sharing it throughout the company - with the traffic control room operations, the customer engagement front office, the back office - this project will allow us to keep everybody aware about the train status, about the conditions of the different components we have on board, and also to come back to our customer to keep them always informed about the train status."