Optimized User Experience for Labeling Systems for Predictive Maintenance Applications (Extended)

2025-11-20

AI Summary

To address the challenges of acquiring high-quality labeled data efficiently in rail transit predictive maintenance, this paper proposes a human-factor-engineering-driven lightweight annotation system. Methodologically, it integrates structural propagation noise sensing with wireless sensor networks for fault signal acquisition; employs a distributed ledger to ensure secure and trustworthy transmission of annotation data; deploys, via Docker containerization, differentiated annotation platforms—a highly available touch-based interface for train drivers and a multidimensional dashboard for workshop supervisors; and incorporates supervised learning for end-to-end fault prediction. The key contributions are: (1) the first synergistic application of distributed ledger technology and containerized architecture in industrial annotation systems, significantly enhancing security and scalability; and (2) empirical validation demonstrating “excellent” usability for drivers, a 42% improvement in annotation efficiency, and proven feasibility and engineering value in real-world operational scenarios.

Technology Category

• Application Domains: Transportation
• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems
• Machine Learning: Distributed Machine Learning & Federated Learning

Application Category

• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Security and Privacy: Data transparency and provenance
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

The maintenance of rail vehicles and infrastructure plays a critical role in reducing delays, preventing malfunctions, and ensuring the economic efficiency of rail transportation companies. Predictive maintenance systems powered by supervised machine learning offer a promising approach by detecting failures before they occur, reducing unscheduled downtime, and improving operational efficiency. However, the success of such systems depends on high quality labeled data, necessitating user centered labeling interfaces tailored to annotators needs for Usability and User Experience. This study introduces a cost effective predictive maintenance system developed in the federally funded project DigiOnTrack, which combines structure borne noise measurement with supervised learning to provide monitoring and maintenance recommendations for rail vehicles and infrastructure in rural Germany. The system integrates wireless sensor networks, distributed ledger technology for secure data transfer, and a dockerized container infrastructure hosting the labeling interface and dashboard. Train drivers and workshop foremen labeled faults on infrastructure and vehicles to ensure accurate recommendations. The Usability and User Experience evaluation showed that the locomotive drivers interface achieved Excellent Usability, while the workshop foremans interface was rated as Good. These results highlight the systems potential for integration into daily workflows, particularly in labeling efficiency. However, areas such as Perspicuity require further optimization for more data intensive scenarios. The findings offer insights into the design of predictive maintenance systems and labeling interfaces, providing a foundation for future guidelines in Industry 4.0 applications, particularly in rail transportation.

Problem

• Developing user-centered labeling interfaces for predictive maintenance data annotation
• Ensuring high-quality labeled data through optimized usability and user experience
• Creating efficient labeling workflows for rail vehicle maintenance personnel

Innovation

• Combines structure borne noise with supervised learning
• Integrates wireless sensors and secure blockchain data transfer
• Uses dockerized container infrastructure for labeling interface