Supporting decisions in production line processes by combining process mining and system dynamics

Abstract

Conventional production technology is static by nature, developments in the area of autonomous driving and communication technology enable a novel type of production line, i.e., dynamic production lines. Carriers of products are able to navigate autonomously through a production facility, allowing for several different “production routes”. Given such dynamic behavior, it is interesting for a production line manager to study in what type(s) of station(s)/resource(s) he/she needs to invest in. We can do so by analyzing the behavior of the autonomous production line, to calculate what change is most likely boosting performance. In this paper, we use historical event data, which are the actual execution of the process, to support the design of system dynamic models, i.e., a high-level predictive mathematical model. The purpose of our framework is to provide the possibility for production line managers to oversee the effects of the changes at the aggregated level in the production line, regarding different performance metrics. At the same time, we provide the freedom in choosing the level of detail in designing the model. The generated model is at a customized aggregated level. We evaluated our approach based on synthetic event logs in which we emulate the effect of policy changes, which we predict accordingly.