This scientific paper examines how large language models (LLMs) can be used to verify cybersecurity compliance in the field of operational technology (OT) more efficiently, particularly in accordance with IEC 62443 and the rail-specific IEC 63452.

In addition to a basic architecture, the researchers are developing an extended Parallel Compliance Architecture (PCA) that incorporates additional regulatory contexts. This significantly improves the accuracy, quality of reasoning and reduction of hallucinations in the models.

OpenAI GPT-4o and Claude 3.5 Haiku, among others, were tested. The results show that retrieval-based LLM approaches can accelerate compliance assessments and improve quality – particularly relevant given the shortage of skilled workers in cybersecurity for critical infrastructures.

The QUDIS project is researching how Deutsche Bahn's digital control and safety technology can be protected against threats from future quantum computers. Classical cryptography will no longer be sufficient in the long term, as quantum computers could break many current methods.
QUDIS is therefore investigating the integration of post-quantum cryptography (PQC) into rail-specific security architectures. This includes the analysis and implementation of new cryptographic methods and associated protocols, as well as the development of crypto-agility to enable secure migration at a later date.

The focus is on real-world railway applications, including prototypes, demonstrators (e.g. OPC UA-enhanced, PQC-secured data transmission) and step-by-step evaluation on test tracks. Project partners include the University of Regensburg, Deutsche Bahn/DB Systel, RheinMain University of Applied Sciences, the University of Konstanz, and industry partners such as genua and INCYDE.

The project investigated how the currently standard chemical vegetation control on railway tracks, which often uses herbicides containing glyphosate, can be replaced by alternative, chemical-free methods. The focus was on plant protection and maintenance measures to ensure the functionality and safety of the track bed area without the use of glyphosate or other chemical agents.

Glyphosate-based herbicides are traditionally used to suppress vegetation growth in the ballast bed. However, operators such as Deutsche Bahn AG have already begun to voluntarily refrain from using glyphosate and are looking for alternative methods of vegetation management – against the backdrop of regulatory restrictions and growing environmental requirements.

Since autumn 2020, the research project has initially analysed scientific, technical and biological principles for vegetation control. From a total portfolio of 18 chemical-free individual methods, the three most promising approaches were identified, combined in a test vehicle and tested in practice:

• mechanical weeding,

• pressurised water application,

• electroweeding (using electrical energy to damage plants and roots).

These methods were evaluated both individually and in combination in terms of their efficiency and feasibility in field trials.

The combination of the three identified methods showed potential as a chemical-free vegetation management method in the track area. These alternatives are fundamentally feasible, but depend on specific operational requirements, local conditions and efficiency/cost aspects compared to conventional chemical methods.

As part of a technical feasibility study, the German Centre for Rail Transport Research (DZSF) investigated various non-destructive testing methods for railway prestressed concrete sleepers to enable early detection of hidden cracks. Such cracks can originate within the sleepers, significantly reduce structural strength long before becoming externally visible, and in extreme cases lead to sleeper failure and safety-critical incidents such as derailments.

The study examined the mechanisms of crack formation and ageing in prestressed concrete sleepers, with particular focus on previously underexplored damage mechanisms. In addition, a camera-based system was developed that enables significantly improved detection and resolution of surface defects compared to existing solutions. A complementary X-ray-based inspection method (X-ray backscatter tomography) was also analysed to identify changes in the material structure at the sleeper surface, which may indicate larger internal defects.

The results highlight the potential of advanced inspection technologies for early damage detection and improved maintenance strategies in rail infrastructure. A further part of the study addressing contactless acoustic methods for detecting hidden cracks is expected to be published soon.