Optimized User Experience for Labeling Systems for Predictive Maintenance Applications

TL;DR

The paper addresses the challenge of enabling reliable predictive maintenance in rural rail networks by designing a cost-efficient, web-based labeling UI integrated with structure-borne sound sensors and ML workflows. It emphasizes usability and user experience, grounded in Nielsen heuristics, and outlines a codesigned, minimalist interface implemented with Vue.js/Vuetify for role-based labeling by train drivers and workshop foremen. A two-part labeling process feeds clean, high-quality labeled data into downstream analytics, with a planned usability and UX study to validate the interface and explore correlations with user traits. The work aims to reduce maintenance costs, improve service reliability, and provide scalable insights for Industry 4.0 applications in rail infrastructure beyond the initial setting.

Abstract

This paper presents the design and implementation of a graphical labeling user interface for a monitoring and predictive maintenance system for trains and rail infrastructure in a rural area of Germany. Aiming to enhance rail transportation's economic viability and operational efficiency, our project utilizes cost-effective wireless monitoring systems that combine affordable sensors and machine learning algorithms. Given that a successful labeling phase is indispensable for training a supervised machine learning system, we emphasize the importance of a user-friendly labeling user interface, which can be optimally integrated into the daily work routines of annotators. The labeling system has been designed based on best practices in usability heuristics and will be validated for usability and user experience through a study, the protocol for which is presented here. The value of this work lies in its potential to reduce maintenance costs and improve service reliability in rail transportation, contributing to the academic literature and offering practical insights for research on effective labeling user interfaces, as well as for the development of labeling systems in the industry. Upon completion of the study, we will share the results, refine the system as necessary, and explore its scalability in other areas of infrastructure maintenance.

Figures (4)

• Figure 1: Dashboard for the Labeling of Train Cars for the Workshop Foreman: List with Train Vehicles on the Left, Entry and Exit Time at the Top, Lists with Labels for Labeling of Faults and Repairs in the Center
• Figure 2: Overlay for Creating a new Label: User is asked for the new Label, List with already available Labels and Input Field for Entering new Label
• Figure 3: Data Verification before submitting Labels: User is asked if following Data should be submitted, Train Vehicle Data with selected Labels is listed
• Figure 4: User Interface for Labeling of the Rails for the Train Drivers: On the Left is the List of Events, at the Top Date, Time, Train Identification and Event Location are shown, Below the Map the List with eligible Labels is positioned