Akcay Wins Best Paper Award at IISE Annual Conference
MIE Associate Professor Alp Akcay received the Best Paper Award from the Modeling & Simulation (MS) Division of the Institute of Industrial and Systems Engineers (IISE) for his paper on “Data-Driven Resource Aggregation for Scalable Digital Twin and Simulation Model Generation”. The award recognizes the best paper presented in the MS track at the IISE Annual Conference and Expo, one of the leading conferences in industrial and systems engineering.
Together with MIE PhD student Yucheng Wu, Akcay received the award for research that advances the development of digital twins for manufacturing systems. Digital twins are virtual representations of real-world production systems that help manufacturers evaluate performance, test operational changes, and support decision-making through simulation. As manufacturing facilities become larger and more complex, creating and maintaining detailed simulation models for every machine can require significant time and effort. The award-winning research addresses this challenge by automatically constructing simulation models from historical production data and determining the appropriate level of detail without extensive manual modeling.
The researchers developed a method that learns how machines behave from historical production records and identifies machines that perform similar functions within a production system. Using inverse reinforcement learning, the approach infers the underlying decision logic governing factory operations and groups similar machines into representative resource clusters. These clusters can then be used to build simplified simulation models that are substantially smaller while maintaining high accuracy.
The method was evaluated on a flexible flow shop production system, a manufacturing environment commonly found in semiconductor back-end assembly. In a case study involving three production stages and 19 machines, the approach reduced the number of modeled resources from 19 individual machines to 7 representative clusters. Despite this significant reduction in model complexity, the resulting simulation model maintained errors below 2% for both production throughput and average flow time when compared with a detailed machine-level simulation.
By automating the construction and simplification of simulation models, the research has the potential to make digital twins easier to develop, maintain, and scale. This capability is particularly important in rapidly changing manufacturing environments, where production systems are frequently modified and simulation models must be continuously updated to remain useful.
The recognition highlights the growing impact of MIE research in the areas of manufacturing systems, digital twins, and simulation-based decision support.