Simulation response function identification using a frequency domain approach

Abstract

Traditionally, simulation response functions have been identified through multiple runs of the simulation based on an experimental design. Originating from frequency domain experiments, where the simulation parameters are oscillated sinusoidally during each run of the simulation, the new technique presented here employs classical regression techniques in order to identify both the form of an adequate model of the response function and the coefficients of that model, all from a single run of the simulation. The technique is used to identify the steady state response functions of simple queuing models, and a more complicated simulation for the security checking of passengers at an international airport.