Dataflow-based Visual Analysis for Fault Diagnosis and Predictive Maintenance in Manufacturing

Abstract

Predictive machine maintenance, which monitors the current condition of a machine, can be much more efficient than maintaining it on a strict schedule or only as a reaction to actual breakdowns. Although sophisticated theoretical models exist, these are not always employed in practice, presumably in part due to their abstract nature. Introducing interactive visualization into the analysis process may facilitate the adoption of predictive maintenance. We apply a dataflow-based visual analytics approach to the analysis of diagnostic machine data on a real-world dataset and collect feedback from domain experts.