AI-Assisted Wagon Inspection in Rail Freight

DB Cargo is testing a new AI-supported inspection system at the Munich North marshalling yard that combines underfloor cameras, microphones, and automated data analysis to improve technical wagon inspections in rail freight transport.

Inspection challenges in freight operations
Technical wagon inspection is a mandatory process in rail freight, performed after train composition and after loading or unloading. It ensures that wagons are safe to operate and compliant with technical standards. However, conventional inspections rely heavily on visual checks by personnel, which can be time-consuming and may miss early-stage defects, particularly in components that are difficult to access.

To address these limitations, DB Cargo is evaluating sensor-based technologies that provide a more comprehensive and data-driven view of wagon condition.

Full-view inspection architecture
The pilot system extends existing camera gate infrastructure, which already captures images of freight wagons from the side and above as they pass through inspection points. The new element is an underfloor camera installation that provides a detailed view from below the wagon for the first time.

The underfloor system consists of five camera modules positioned to image axles, brake rigging, wheelsets, and other undercarriage components. This configuration is intended to capture areas that are typically difficult to inspect visually during routine operations.

Acoustic sensing for wheel and running gear faults
In parallel with visual inspection, microphones installed in the inspection area record sound signatures as wagons pass through. These acoustic data are used to identify irregularities such as wheel flat spots or abnormal rolling noise, which can indicate developing defects before they become visible.

By combining visual and acoustic inputs, the system aims to detect both structural damage and functional anomalies affecting running gear and braking systems.

AI-based data evaluation
Images from the existing camera gates and the new underfloor cameras, together with audio recordings, are linked and analyzed using artificial intelligence models. The objective is to support inspectors with early, data-based assessments of wagon condition, enabling faster diagnosis and more targeted manual checks.

Rather than replacing human inspection, the AI is intended to act as an early warning and decision-support tool, highlighting potential damage before it leads to operational disruptions or safety incidents.

Collaborative research framework
The trial is part of the ASaG project (Automated Damage Detection on Freight Cars), a joint initiative involving DB Systel, DB InfraGO, CoDiVe, and the University of Wuppertal, alongside DB Cargo.

The project focuses on applying digital sensing and AI methods to freight rail operations, with the goal of improving safety, reducing inspection time, and minimizing unplanned downtime.

Implications for rail freight
If validated, the combination of underfloor imaging, acoustic monitoring, and AI analysis could significantly change how technical wagon inspections are performed. Earlier detection of defects may reduce damage escalation, improve fleet availability, and support more predictive maintenance strategies.

Within the broader context of rail digitalization, the ASaG project illustrates how sensor fusion and AI can augment traditional inspection processes, providing a more complete and consistent assessment of freight wagon condition in busy marshalling yards.