Network Rail IT bosses plan to test a network of sensors that will predict when rail equipment needs maintenance as part of a programme to cut the company's costs by a fifth by 2014.
The project, which will be rolled out across the UK if it is successful, will help the company replace defective parts before they fail.
The company has pressed ahead with the scheme following criticism in January by the Office of Rail Regulation (ORR) for engineering delays over Christmas. The ORR has told the company it must increase the percentage of trains that run on time from 90.6% to 92.6% by 2014.
Network Rail plans to pilot the project, dubbed intelligent infrastructure, on the Edinburgh to Glasgow line by April next year.
Network Rail CIO Catherine Doran said, "We want to develop a model that enables us to predict that a set of points, for example, will need replacing or fixing in the next month. When an asset fails, it causes disruption and late trains. We want to eliminate that failure.
The company plans to install 250 senors to collect data on the performance of equipment on the line. They will detect whether points are opening and closing correctly, and measure temperature of electronic equipment to detect overheating.
Network rail plans to collect the data by sending it to wireless nodes on the track, or by collecting it from a wireless enabled train.
It plans to analyse the data to identify patterns of behaviour in equipment that fails. It will use these models to predict when maintenance is needed and fix equipment before it breaks and causes delays.
"I would hope we have some decent data to work with in the next financial year, said Doran.
The company is selecting suppliers for measuring devices. It is likely to start the pilot project in quarter three or four of the current financial year.
How Network Rail plans to predict failure before it happens
* Thermometers detect overheating in equipment
* Movement sensors detect if points are working correctly
* Sensors send signals to wireless nodes on the track
* Data is fed from the nodes to a fixed telecoms network, and then into the data centre
* Software looks for patterns and builds models that can be used to predict future failure.