Implementation of sensitivity analysis in discrete event digital simulation

Abstract

Discrete event dynamic systems have been analyzed by use of either analytical or simulation models. Both approaches have certain drawbacks. In this thesis a new approach which aims to reduce the number of computer runs required for the sensitivity analysis in discrete event digital simulations is studied. The proposed approach estimates the sensitivity coefficients of the system output with respect to various parameters from the results of a SINGLE simulation run. It can derive answer to the question "How would the total simulation time change if we repeated the experiment under the same conditions, except for small perturbations in the sample durations of one of the activities?" for some or all of the activities in the system while observing the experiment. The efficiency of this approach lies in the fact that it has a computational advantage of M to 1, where M is the total number of activities for which the above question is asked.