Hardware/AIGermany2024
Switch-monitoring IoT and predictive analytics for rail
Trackside sensor kits and ML that predict switch degradation before a failure closes a line.
- Client
- Konux
- Region
- Germany
- Sector
- Railway IoT + predictive maintenance
- Engagement
- 2024 · 22 weeks
- Team
- 2 firmware · 1 hardware · 2 ML
- Status
- Deployed
The challenge
Railway point machines fail with little warning, and an unplanned failure blocks a line and cascades delays. Konux needed rugged trackside sensing plus models that turn switch-movement signatures into an actionable maintenance forecast.
What we built
The full pipeline, end to end — not a blurb.
- 01Hardware
A trackside sensor kit — acceleration, current, and temperature — in an industrial enclosure with cellular backhaul.
- 02Firmware
Event-triggered capture of switch throws with edge feature extraction and power management.
- 03ML
Signature-based degradation and anomaly models with remaining-useful-life estimates.
- 04Cloud
Ingest, per-asset health, and a maintenance-planning API.
- 05Deployment
A field install kit, OTA, and fleet monitoring.
Results
12 days
average lead time on switch degradation before failure
0.88
anomaly recall on historical incidents
31%
fewer switch-related unplanned delay minutes
1,900
point machines monitored in the deployment
Feeds the infrastructure maintenance schedule.
Have a constraint like Konux’s? Bring us yours.
Next engagement
Mindpeak
Computational pathology / vision