Currently, data collection on the shop floor is based on individual resources such as machines, robots, and Autonomous Guided Vehicles (AGVs). There is a gap between this approach and manufacturing orchestration software that supervises the process of creating single products and controls the ressources' interactions. This creates the need to save resource-based data streams in databases, clean it, and then re-contextualize it, i.e., by connecting it to orders, batches, and single products. Looking at this data from a process-oriented analysis point of view enables new analysis prospects. This paper utilises these prospects in an experimental way by creating BPMN models for the manufacturing of two real-world products: (1) a low volume, high complexity lower-housing for a gas-turbine and (2) a high volume, low complexity, small tolerance valve lifter for a gas turbine. In contrast to the resource-based data collection, 30+ values are modeled into the BPMN models and enacted by a workflow engine, creating execution logs in the XES standard format. Conformance checks are carried out and interpreted for both scenarios and it is shown how existing classification and clustering techniques can be applied on the collected data in order to predict good and bad parts, ex-post and potentially at run-time.

Top