At IAM4RAIL’s final event in Madrid on 27 October, participants presented work connecting railway condition monitoring with maintenance decisions. The programme’s seven integrated demonstrators brought AI and digital twins into infrastructure research. DLR says it will pursue higher technological readiness with partners in the follow-on project.[1]
Separating wear from winter
An April study analysed 1.76 million current traces from 58 Dutch point machines—the mechanisms that move switch blades—over two years. An indicator built without fault labels adjusts for temperature while tracking extra mechanical resistance. Across the dataset, rank correlation with temperature moved from −0.357 for the original feature to −0.0083 for the indicator. That supports comparison across warmer and colder conditions.[2]
Twenty-eight-day averages provide a maintenance ranking. Changes following maintenance were manually marked for the study; operational use would need automatic detection or suitable repair timestamps.[2]
A repair changes the evidence
A July follow-up combined current traces, track geometry and maintenance records from 58 switches, covering January 2023 to mid-May 2026. Maintenance actions dominated changes in the measurements. Repairs can alter the track geometry and the conditions under which the point machine moves, so a lasting improvement must be assessed against both.[3]
The study exposed a sampling problem. Track-measurement trains visited a few times yearly; point-machine readings arrived with each movement. The authors call for daily track-condition information during ordinary operation to distinguish deterioration from changes in operating conditions.[3]
Sources & further reading
- IAM4RAIL final event DLR · 27 Oct 2026 · Accessed 29 Oct 2026
- Point-machine health indicator Intelligent Transportation Infrastructure · 24 Apr 2026 · Accessed 29 Oct 2026
- Multimodal railway switch analysis Applied Sciences · 29 Jul 2026 · Accessed 29 Oct 2026