ICPM 2024 Keynote - Multi-paradigm, Online, Hierarchical Simulations to Support Process Mining
DOI: https://doi.org/10.11583/DTU.27276591
This is the recording of the academic keynote given at ICPM 2024 https://icpmconference.org/2024/keynotes/.
In this keynote talk, multi-paradigm, hierarchical, and online simulations will be introduced and discussed to support effective process mining. First, an overview of multi-paradigm simulations, such as discrete event simulation (DES), agent-based modeling (ABM), and system dynamics (SD), will be provided. While the goal of these modeling paradigms is the same (i.e. representing a real system validly and credibly), key characteristics and differences will be explained. In DES, objects/entities are relatively passive, and their dynamics are driven by a top-down comprehensive process model. On the other hand, agents in ABM are relatively active, and their dynamics are driven by their bottom-up behavior model considering their interactions with other agents and the environment. Unlike DES and ABM, which represent systems involving discrete entities/agents and events usually in a discrete-time, SD represents systems using difference or differential equations. In this talk, we will demonstrate ABM and SD result in the same system equation using a simple M/M/1 (i.e. one server, one type of customers, infinite queuing capacity) service system, where ABM system equation is derived involving an expected value and probability and SD system equation is derived via solving a differential equation.
Second, an online simulation-based planning and control (SPC) approach is introduced, where a fast-running DES simulation is used as a predictive tool to evaluate decision alternatives at the planning stage, and the same DES model (a twin-simulation running in real-time) is used as a task generator to drive a real system at the control stage. The simulation models used in this SPC approach are automatically generated from a resource model (e.g. information contained in an ERP system for an enterprise). In this talk, the requirements for automatic simulation model generation will be discussed, which will be useful for the process mining community to mine and develop a process and simulation model from event log data.
Third, an extension of SPC to a highly complex system is discussed, which involves a dynamic data-driven adaptive multi-scale simulation (DDDAMS) framework. A key module in this framework enhances the computational efficiency of the system-level simulation considering available data, computational resources, and model validity/credibility via dynamic switching of fidelity of component simulations and information gathering during the simulation execution over time. In this talk, a few case studies (i.e. simple M/M/1 service, smart manufacturing, unmanned aerial/ground vehicles) will be used to illustrate the above-mentioned concepts and facilitate discussions.
Biography
Young-Jun Son is the James J. Solberg Head and Ransburg Professor of School of Industrial Engineering at Purdue University. He is a Fellow of the Institute of Industrial and Systems Engineers (IISE), and has received the Society of Manufacturing Engineers (SME) 2004 Outstanding Young ME Award, the IISE 2005 Outstanding Young IE Award, the IISE Annual Meeting Best Paper Awards (2005, 2008, 2009, 2016, 2018, 2019), and the Best Paper of the Year Award (2007) in International Journal of Industrial Engineering. He has authored/co-authored over 220 research publications. He is a Department Editor for IISE Transactions and serves on the editorial board for seven additional journals. He was the vice chair for the SISO Core Manufacturing Simulation Data Standard Product Development Group. He has served as co-Program Chair for ISERC 2007, the General Chair for INFORMS Annual Meeting 2018, and the General Chair for Winter Simulation Conference 2019.
