Author

Sachin Sumant

Abstract

In industry, simulation is one of the most widely used probabilistic modeling tools for modeling highly complex systems. Major sources of complexity include the inputs that drive the logic of the model. Effective simulation input modeling requires the use of accurate and efficient input modeling procedures. This research focuses on nonstationary arrival processes. The fundamental stochastic model on which this study is conducted is the nonhomogeneous Poisson process (NHPP) which has successfully been used to characterize arrival processes where the arrival rate changes over time. Although a number of methods exist for modeling the rate and mean value functions that define the behavior of NHPPs, one of the most flexible is a multiresolution procedure that is used to model the mean value function for processes possessing long-term trends over time or asymmetric, multiple cyclic behavior. In this research, a statistical-estimation procedure for automating the multiresolution procedure is developed that involves the following steps at each resolution level corresponding to a basic cycle: (a) transforming the cumulative relative frequency of arrivals within the cycle to obtain a linear statistical model having normal residuals with homogeneous variance; (b) fitting specially formulated polynomials to the transformed arrival data; (c) performing a likelihood ratio test to determine the degree of the fitted polynomial; and (d) fitting a polynomial of the degree determined in (c) to the original (untransformed) arrival data. Next, an experimental performance evaluation is conducted to test the effectiveness of the estimation method. A web-based application for modeling NHPPs using the automated multiresolution procedure and generating realizations of the NHPP is developed. Finally, a web-based simulation infrastructure that integrates modeling, input analysis, verification, validation and output analysis is discussed.

Library of Congress Subject Headings

Computer simulation; Poisson processes; Simulation methods--Mathematical models; Client/server computing

Publication Date

11-1-2003

Document Type

Thesis

Department, Program, or Center

Industrial and Systems Engineering (KGCOE)

Advisor

Kuhl, Michael

Advisor/Committee Member

Sudit, Moises

Comments

Note: imported from RIT’s Digital Media Library running on DSpace to RIT Scholar Works. Physical copy available through RIT's The Wallace Library at: QA76.9.C65 S76 2003

Campus

RIT – Main Campus

Share

COinS