Using designer confidence and a dynamic Monte Carlo simulation tool to evaluate uncertainty in system models

Abstract

As the use of distributed engineering models becomes more prevalent, engineers need tools to evaluate the quality of these models and understand how subsystem uncertainty affects predictions of system behavior. This thesis develops a tool that enables designers and engineers to specify their perceptions of confidence. These data are then translated into appropriate probability distributions. Monte-Carlo-based methods are used to automatically provide correct propagation of these distributions within an integrated modeling environment. A case study using an assembly tolerance problem is shown and different confidence modeling methods are compared. The methods benchmarked are: worst case; statistical; conventional Monte Carlo simulation; and the dynamic Monte Carlo tool developed in this thesis. Finally the dynamic Monte Carlo tool is used together with surrogate modeling techniques. Comparisons based on implementation time, model execution time, and robustness are provided.