Abstract
Simulation models of different levels of fidelity are often available for evaluating alternative solutions of a complex system. High-fidelity simulations generate accurate predictions but can be very time-consuming to run. Therefore, they can only be applied to a small number of solutions. Low-fidelity simulations are much faster and can evaluate a large number of solutions. But simulation results may contain significant bias and variability. We propose a novel ordinal transformation framework to exploit the benefits of both high- and low-fidelity simulation models to efficiently identify a (near) optimal solution. A two-stage simulation optimization method under the ordinal transformation framework is described. Through preliminary theoretical analysis and numerical experiments, we demonstrate the promising performance of ordinal transformation, which opens up a new and potentially fruitful research avenue.