Data flow simulation models are often used for the modeling and analysis of complex dynamic systems. Although traditional analysis methods such as design of experiments or optimization methods can be applied directly to data flow simulation models, applying these techniques to complex systems with large numbers of controllable inputs and performance measures may not be able to be completed in an acceptable amount of time. This research focuses on the development of optimization and analysis methods applied to graph-based meta-models of data flow simulation models. The goal of this research is to create a method that can efficiently determine the values of controllable system input variables that will yield user-specified system output performance measure values. The methodology utilizes an existing graph-based meta-modeling technique that elicits the graph structure of the underlying data flow simulation model. To enable goal-oriented optimization on the elicited graph, edge weights are determined by performing experimental sampling and utilizing a regression model to each of the nodes in the elicited graph. In the case of nonlinear input-output relationships, a method which provided piecewise linear edge weights is used. Finally, mathematical programming formulations are developed to conduct the goal-oriented optimization. An experimental performance evaluation is conducted and illustrates the capability of the method to efficiently provide estimates of system inputs that will result in desired values of system performance measures.
Library of Congress Subject Headings
Computer simulation; Data flow computing; Piecewise linear topology; Graph theory; Information visualization
Department, Program, or Center
Industrial and Systems Engineering (KGCOE)
Goldsmith, Jeffrey Harrison, "Analysis and optimization methods of graph based meta-models for data flow simulation" (2010). Thesis. Rochester Institute of Technology. Accessed from
RIT – Main Campus