{"id":8409,"date":"2026-09-10T05:25:01","date_gmt":"2026-09-10T03:25:01","guid":{"rendered":"https:\/\/www.scs.ch\/uncategorized-en-us\/trustworthy-ai-for-the-measuring-wheelset\/"},"modified":"2026-09-20T07:38:41","modified_gmt":"2026-09-20T05:38:41","slug":"trustworthy-ai-for-the-measuring-wheelset","status":"publish","type":"post","link":"https:\/\/www.scs.ch\/en\/case-studies\/trustworthy-ai-for-the-measuring-wheelset\/","title":{"rendered":"Trustworthy AI for the Measuring Wheelset"},"content":{"rendered":"

Initial Situation<\/h2><\/div>

An ICN measuring train regularly measures the dynamic forces between wheel and rail on the high-speed curve sections of the SBB network, for example, along the southern foot of the Jura mountains. This is done using measuring wheelsets with sensors on the rotating wheels. These sensors are maintenance-intensive. It would be simpler to calculate the forces from acceleration sensors on the bogie. However, the physical relationship is complex and cannot be easily converted analytically.<\/p>\n<\/div>

\"Close-up<\/span><\/div><\/div><\/div>

Project Overview<\/h3><\/div>

Client<\/strong>
SBB<\/p>\n<\/div>

<\/div><\/div>

Sector<\/strong>
Data & High Performance<\/p>\n<\/div>

<\/div><\/div>

Solution<\/strong>
Explainable AI for Force Measurement on Wheel and Rail.<\/p>\n<\/div>

<\/div><\/div>

System Environment <\/strong>
\nICN measuring train, measuring wheelsets, acceleration sensors on bogie, RCM-PP data processing platform<\/p>\n<\/div>

<\/div><\/div>

Partner<\/strong>
\nBern University of Applied Sciences (BFH)<\/p>\n<\/div><\/div><\/div><\/div><\/div>

SCS’s Contribution<\/h2><\/div>

To achieve this, SCS, in collaboration with the Bern University of Applied Sciences, specifically trained a model for the conditions of the SBB network and designed it so that the calculated values can be reviewed and understood by SBB specialists at any time.<\/p>\n<\/div><\/div><\/div>

<\/i><\/div>

Explainable AI Instead of a Black Box<\/h4><\/div>
<\/div>
\n

The algorithm is designed so that the results are explainable and can be verified as plausible by SBB specialists.<\/p>\n<\/div><\/div><\/div>

<\/i><\/div>

Specifically Trained<\/h4><\/div>
<\/div>
\n

Designed for real-world conditions in the SBB network, curve radii from 250 m and lateral accelerations up to 1.8 m\/s\u00b2.<\/p>\n<\/div><\/div><\/div>

<\/i><\/div>

Risk Minimized<\/h4><\/div>
<\/div>
\n

Risk analysis and plausibility check of sensor data, including detection of sensor drift by comparison with previous runs.<\/p>\n<\/div><\/div><\/div>

<\/i><\/div>

Seamlessly Integrated<\/h4><\/div>
<\/div>
\n

SCS modularized, refined, and integrated the original algorithm from the Bern University of Applied Sciences into the automated RCM-PP data processing platform.<\/p>\n<\/div><\/div><\/div>

<\/div><\/div><\/div><\/div>
\"Drei<\/span><\/div><\/div><\/div><\/div><\/div>
<\/div>

Result<\/h2><\/div>

Wheel-rail forces can be calculated from simple acceleration sensors on the bogie. The maintenance-intensive sensors on the rotating wheels can largely be eliminated. The calculated values are traceable and reliable. The algorithm’s reliability is being tested, and measuring wheelsets are still occasionally used for validation.<\/p>\n<\/div><\/div><\/div>

from 250 m<\/h3><\/div>

covered curve radii in the SBB network.<\/p>\n<\/div><\/div><\/div>

up to 1.8 m\/s\u00b2<\/h3><\/div>

covered lateral acceleration.<\/p>\n<\/div><\/div><\/div>

Wheel sensors largely saved<\/h3><\/div>

only simple acceleration sensors on the bogie.<\/p>\n<\/div><\/div><\/div><\/div><\/div>

Services Provided<\/h2><\/div><\/div><\/div>

Algorithms, ML & AI<\/h3><\/div><\/div><\/div><\/div><\/div><\/div>

Data & High Performance<\/h3><\/div><\/div><\/div><\/div><\/div><\/div><\/div><\/div>
\"The<\/div><\/div>

Your Contact Person<\/h2><\/div>
Pascal Iselin<\/div><\/div>

Algorithms, Analytics & AI
\n+41 43 456 16 00<\/a><\/p>\n<\/div>

Speak with the responsible specialist<\/span>Speak with the responsible specialist<\/span><\/div><\/a><\/div>

For questions regarding explainable AI, machine learning, and the processing of safety-relevant measurement data.<\/p>\n<\/div><\/div><\/div><\/div><\/div><\/div><\/div><\/div><\/div>

More Case Studies<\/h2><\/div>