Managing Predictive Maintenance Workflows in the Era of Industry 4.0

Universität Ulm

Fakultätsöffentliche Vorstellung des Dissertationsvorhabens

Vortragender: Burkhard Hoppenstedt

Ort: Online-Videokonferenz

Zeit: 1. Juli 2020 um 10:15 (10 Uhr c.t.)

In the era of Industry 4.0 the amount of machine sensor data has been increasing by orders of magnitude. These big data open up new avenues for machine condition monitoring and predictive analytics. With predictive maintenance, companies want to monitor and predict the states of their machines with the goal to optimize maintenance cycles depending on contextual factors (e.g., machine utilization).  However, any predictive maintenance approach necessitates a sound workflow that covers phases like data collection, data preprocessing, and data modeling. Moreover, machine learning offers the potential to create insights into production processes by analyzing both historical and current machine data. Visualizing the collected data as well as the insights created directly at the production site, in turn, requires a correct mapping of information to machine components as well as suitable interaction patterns for machine operators during maintenance processes. This work aims at the operational support of sophisticated predictive maintenance workflows. For this purpose, the visual analytics process, which combines modeling techniques and visualizations, is adopted to create synergy effects by applying visual analytics and machine learning in combination with each other. In this context, the state-of-the-art on predictive maintenance is analyzed, sophisticated Industry 4.0 prototypes based on machine learning and visual analytics are developed, and the suitability of the interaction approaches involving advanced interactions devices (e.g., smart glasses) is evaluated. In this context, various use cases are addressed, such as outlier detection in hierarchical production systems, clustering of high-dimensional data, and object detection of produced parts. Moreover, we evaluate the usability of immersive analytics with smart glasses to enable the machine operator to apply sophisticated analytics at the production site.  Altogether, sophisticated support for predictive maintenance workflows is provided by enabling immersive analytics on complex data at the production site and by developing specific algorithms for machine data analytics.