Correlated Decomposition for Analyzing Dynamic Stochastic Systems

点击翻译按钮获取中文摘要

9300058 Taaffe This research is concurrent with creating methodology and software for efficiently analyzing models of manufacturing systems, including complexities such as assembly, blocking, and time dependence. We obtain efficient analysis by combining the speed and efficiency of queueing network approximations with the details and generality of computer simulation experimentation. Our combined approach, which we refer to as Correlated-Decomposition Approximation and Simulation (CDAS), allows the strengths and weaknesses of these two methodologies to complement each other. The approximations and simulation can work together by using the approximations as control varieties or the our newly developed more general biased-control varieties, for the simulation. The goal is to improve analysis efficiency by two or three orders of magnitude compared to ordinary simulation experiments. The combined CDAS methodology should analyze the behavior of large complex manufacturing systems in a matter of seconds or minutes rather than the hours or days that simulation alone can require and to do so with accuracy beyond that possible in present queueing network approximation software.