Predicting rail washaways from rainfall data

The challenge

Derailments cost Arc Infrastructure about $2M a year, and they put Arc's employees, rail operators and the public at risk.

One preventable cause is a washaway: abnormal or intense rainfall displaces the rail formation under the track, and a train passing over the damaged section can derail.

Arc's innovation workshops suggested that washaways might be predicted by analysing past washaway events against rainfall patterns. Arc engaged Interfuze to test that idea with a proof of concept that could become a production-ready solution if it proved viable. That meant identifying the high-risk locations, finding and collecting the relevant data, and developing new analysis techniques to predict or detect washaways before or as they occur.

Our approach

We mapped the relationships between rainfall, geography and historical track washaway events through an extensive data analysis exercise.

We then correlated every weather station with flood events and with the track's kilometre marks. That gave a base set of rainfall thresholds for each track segment on the network, and an estimate of the impact rainfall at each location has on those track sections.

On that foundation we built a washaway prediction engine. It learns changing thresholds over time, monitors incoming rainfall, predicts which sections are at risk and raises alarms as an early warning.

For the proof of concept, alerts were shown in Power BI over an Azure SQL database and pushed by email to distribution lists of the relevant users. The engine can generate hourly alerts for each track segment and each weather station, and it can scale into GIS applications alongside field data from IoT devices.

The outcome

Analysis of rainfall patterns predicted about 93% of historical washaway events.

Interfuze delivered the proof of concept, and Arc began taking it into production for operational use. Arc can now monitor weather events proactively, to warn rail operators about sections of track that have experienced a washaway and reduce the risk of crossing them.

The solution is becoming a foundation that Arc will scale to monitoring and detection alerts in other areas.