Design of Experiments for Simulation Modeling: Factor Screening, Prediction, and Optimization

Design of Simulation Experiments Overview

Simulation models often have many input factors and determining which ones have a significant impact on performance measures (responses) of interest can be a truly daunting task. The common approach of changing one factor at a time is very often incorrect, because for many models factors interact to impact on the responses. In this simulation course, we give a comprehensive presentation of design of experiments (DOE) specifically for simulation modeling, whose major goal is to determine which factors have the greatest effect on the responses, and to do so with the least amount of simulating. Other important uses of DOE are to develop a response surface (or metamodel) based on the important factors to predict the model response for factor combinations that were not actually simulated due to time constraints or to find the factor-level combination that optimizes the simulation response.

We discuss a simple and widely applicable approach to performing DOE in the context of simulation modeling, whereas commonly used methods based on classical statistics (i.e., ANOVA) make unrealistic assumptions such as constant variances and normally distributed residuals. Students will analyze experimental data in class using a leading DOE software package, and be prepared to apply their knowledge the following week at work.

Versions of this seminar have been presented to Lockheed Martin, NSA, and Sasol Technology (South Africa).