No Time to Crash: Visualizing Interdependencies for Optimal Maintenance Scheduling

With the digital transformation in manufacturing, Predictive Maintenance (PdM) is increasingly proposed as an approach to increase the efficiency of manufacturing processes. However, system complexity increases due to mass customization, shorter product life cycles, and many component variants within a manufacturing system. So far, PdM mainly focuses on a single component or system-level and thus neglects the complexity by not considering the interdependencies between components. In a Multi-Component System (MCS) perspective, models covering interdependencies between components within a complex system are established and used for the prediction. Even if the predictive accuracy is superior, modeling interdependencies is a complex and laborious task that prevents the broad adoption of the MCS perspective. A potential way to tackle this challenge is using visualizations to discover and model the interdependencies. This paper evaluates different visualization approaches for PdM in the context of MCSs using a crowd-sourced study involving 530 participants. In our study, we ranked these approaches based on the participant's performance that aimed to identify the optimal timing for maintenance within an MCS. Our results suggest that visualization approaches are suitable to identify interdependencies and that the stacked-area approach is the most promising approach in this regard.