Automating heat speed restrictions from weather forecasts

The challenge

Arc Infrastructure manages and operates 5,500 kilometres of rail network across the southern half of Western Australia. As part of an ongoing program to reduce train derailments to zero, Arc wanted to understand how predicted and actual temperatures translate to heat speed restrictions: how severe a restriction should be, and where to apply it.

Field staff imposed heat speed restrictions by hand. Arc wanted to know whether automated information from the Bureau of Meteorology and DPIRD could make restrictions more accurate, particularly when lifting them, and make the process more rigorous and uniform. Interfuze was engaged to test that idea with a proof of concept that could become a production-ready solution if it proved viable.

The first task was to work out which weather stations cover which parts of the track, from their longitude and latitude, then to predict the restrictions to apply from the forecast temperature and alert field staff when a heat speed segment crossed its thresholds.

Our approach

We modelled temperatures and mapped the relationships between weather stations and track segments (heat speed segments).

We then correlated every weather station with the heat speed segments. That gave an algorithm to calculate the forecast temperature for each segment on the network from its distance to nearby weather stations, and an alerting mechanism based on how far the actual temperature on the day varies from a threshold.

On that foundation we built a heat speed restriction prediction engine on the Microsoft stack. It takes in publicly available weather forecasts and actual temperatures across the network, suggests the restrictions to apply, monitors actual temperatures, predicts at-risk segments and raises alarms when a segment is over- or under-protected. It integrates with Arc's existing Heat Speed Restriction application to create or lift restrictions.

The outcome

The analysis of forecast temperatures proved that heat speed restrictions can be predicted, and real-time processing of actual temperatures alerts field staff when the track is under- or over-protected.

For field staff, manually checking rail temperatures has been replaced by an automated solution. It monitors incoming forecasts and suggests heat speed restrictions, which staff accept and apply or cancel, and it monitors actual temperatures to alert on track sections at risk of derailment from overheating. The result is safer trains on the Arc network.