Metamodeling for Discrete-event Simulaitons in Manufacturing (2025-US-PO-2537)

Our team has been developing discrete-event simulation models of process and packaging systems for several decades. In 2024, we adopted JMP to streamline and strengthen how we interpret statistical results from these models.

Since simulation models can be time-consuming and expensive to run, we now use JMP to build surrogate metamodels, based on designed experiments that dramatically improve the efficiency of extracting insights from simulations. The metamodels yield deeper insight into system behavior than traditional outputs, such as 3D visuals or detailed run statistics provided during a single model run.

Our first metamodel revealed a clear solution to a long-running analytical challenge: With multiple buffers on a packaging line, how big should each be to optimize the line performance given limited capital resources? We present several other metamodels, including comparing several layout options for a line with a merge-divert mechanism and the nonlinear effect on performance based on where it was located.

Finally we present a case study of a forklift simulation used to support an investment decision for whether to construct a direct connection between two warehouses. Surrogate modeling has become a vital part of our simulation workflow, and JMP has played a key role in making this approach more accessible, efficient, and impactful for both for our team and for our stakeholders