2513100 – Seminar Data-driven Simulation for Industrial Systems (Master)

Data-driven simulation is about utilizing real-world data to support the extraction of simulation models in nearly real-time. Data-driven simulation is the enabler for Digital Twins. It involves data collection, preprocessing, model extraction, calibration, and validation. Benefits include improved accuracy, adaptability, reduced assumptions, and faster insights. One of the key methodologies that support data-driven simulation is process mining. This seminar aims to provide Master students with a particular understanding of data-driven simulation and its application in identifying bottlenecks, optimizing processes, and analyzing production Key Performance Indicators (KPIs). The seminar will guide participants through a step-by-step process, beginning with researching and defining data requirements based on the predefined scenarios and the provided AnyLogic model. Subsequently, participants will establish a data pipeline to collect the relevant data generated by the ground-truth model, resulting into a streamed event log. Once the data pipeline is operational, the third step involves utilizing process mining tools, such as pm4py Python library, to extract valuable information, including the process structure represented as a Petri Net. The Petri Net will be enhanced with pertinent details to facilitate system analysis aligned with the scenario objectives. For instance, distribution fitting can be incorporated to model stochastic activities. Finally, the mined simulation models in the form of stochastic Petri nets will be simulated using a Python library, for which the details will be provided in the seminar.