How User-Friendly Interfaces Are Revolutionizing Railway Maintenance

In the high-stakes world of railway transportation, where delays and malfunctions can cascade into significant economic losses and safety risks, a quiet revolution is underway. The federally funded DigiOnTrack research project has developed a predictive maintenance system that combines structure-borne noise measurement with supervised machine learning to monitor rail vehicles and infrastructure in rural Germany. This innovative approach aims to detect potential failures before they occur, reducing unscheduled downtime and improving operational efficiency. However, the success of such systems depends critically on high-quality labeled data, necessitating user-centered labeling interfaces that domain experts—train drivers and workshop foremen—can seamlessly integrate into their daily workflows. The study reveals that when these interfaces are designed with meticulous attention to usability principles, they can achieve "Excellent Usability" ratings, transforming how maintenance data is collected and analyzed in safety-critical industries.

Ology behind this breakthrough involved a comprehensive system architecture that integrated wireless sensor networks, distributed ledger technology for secure data transfer, and a dockerized container infrastructure hosting both the labeling interface and alarming dashboard. Researchers equipped two train cars and one rail section with structure-borne noise measurement sensors to monitor rail conditions and vehicle states. To apply supervised machine learning s, they needed expert-labeled fault data, leading to the development of two distinct labeling interfaces: one for locomotive drivers using tablets to label rail infrastructure faults, and another for workshop foremen using computers to label train car faults and repairs. The interfaces were designed following Jakob Nielsen's usability heuristics, emphasizing minimalism, consistency, and error prevention, with iterative testing involving five design experts from Hamm-Lippstadt University of Applied Sciences who performed task-based evaluations using both tablet and computer setups.

From the System Usability Scale (SUS) and User Experience Questionnaire (UEQ) evaluations were striking and statistically significant. The locomotive drivers' interface achieved a mean SUS score of 85.00 (SD = 15.71), categorizing it as "Excellent Usability," while the workshop foreman's interface scored 72.50 (SD = 15.10), falling into the "Good Usability" range. In the UEQ analysis, the locomotive drivers' interface outperformed the workshop foreman's interface across key dimensions: Attractiveness (0.83 vs. 0.50), Perspicuity (2.00 vs. 1.05), Efficiency (2.25 vs. 1.65), and Dependability (1.55 vs. 1.25). These differences highlight how interface design tailored to specific user roles and contexts can dramatically impact usability metrics. The locomotive drivers' interface particularly excelled in Efficiency, receiving an "Excellent" classification, suggesting that its minimalist design optimized for tablet use during shift transitions effectively supported rapid, accurate labeling.

Of these extend far beyond the railway industry, offering actionable insights for designing labeling systems in broader Industry 4.0 applications. The study demonstrates that even in data-intensive, safety-critical environments, user-centered design can produce interfaces that domain experts find intuitive and efficient, potentially increasing data quality and adoption rates for predictive maintenance systems. However, the research also reveals limitations, particularly in Perspicuity for the workshop foreman's interface, which scored "Below Average" due to its more complex data entry requirements. This suggests that as labeling systems scale to handle more variables, designers must balance functionality with learnability. The qualitative feedback further emphasized s with data editing capabilities and potential confusion in time entry fields, pointing to areas for future optimization.

Despite these limitations, the study provides a foundation for developing guidelines for labeling system design in predictive maintenance contexts. The use of distributed ledger technology ensured data integrity and secure cross-company collaboration, while the modular, containerized architecture supported scalability. Future research should involve larger participant pools including non-design experts and actual annotator groups to validate these in real-world settings. As predictive maintenance becomes increasingly central to industrial operations, this work underscores that technological sophistication must be paired with human-centered design to achieve reliable, efficient systems. The DigiOnTrack project thus represents not just a technical achievement, but a paradigm shift in how we approach maintenance in critical infrastructure.

Source: Hallmann, M., Stern, M., Henning, J., Franke, U., Ostertag, T., da Costa, J.P.J., & Voigt-Antons, J.N. (2025). Optimized User Experience for Labeling Systems for Predictive Maintenance Applications.