Reverse engineering of production processes based on Markov chains

Understanding and providing knowledge of production processes is crucial for flexible production systems as many decisions are postponed to the operation time. Furthermore, dealing with process improvements requires to have a clear picture about the status of the currently employed process. This becomes even more challenging with the emergence of Cyber-Physical Production Systems (CPPS). However, CPPS also provide the opportunity to observe the running processes by using concepts from IoT to producing logs for reflecting the events happening in the system during its execution. Therefore, we propose in this paper a fully automated approach for representing operational logs as models which additionally allows analytical means. In particular, we provide a transformation chain which allows the reverse engineering of Markov chains from event logs. The reverse engineered Markov chains allow to abstract the complexity of run-time information as well as to enable what-if analysis whenever improvements are needed by employing current model-based as well as measurement-based technologies. We demonstrate the approach based on a lab-sized transportation line system.