Many Objective Visual Analytics: Decision Aiding Tools for Conceptual Design

Open Access
Woodruff, Matthew John
Graduate Program:
Industrial Engineering
Doctor of Philosophy
Document Type:
Date of Defense:
December 03, 2015
Committee Members:
  • Timothy William Simpson, Dissertation Advisor
  • Timothy William Simpson, Committee Chair
  • Arunachalam Ravindran, Committee Member
  • Russell Richard Barton, Committee Member
  • Ling Rothrock, Committee Member
  • Decision Aiding
  • Many Objective
  • Conceptual Design
Many Objective Visual Analytics (MOVA) is a novel process wherin one poses a series of optimization problems and uses data visualization to examine the resulting solutions in order to understand a conceptual design problem. This dissertation introduces and defines MOVA, along with its historical context in the intersecting fields of optimization, conceptual modeling, data visualization, and decision aiding. Furthermore, it develops three decision-aiding roles in which MOVA can be used: (1) design selection, (2) problem discovery, and (3) problem exploration. For design selection, MOVA helps identify new alternatives and context. In problem discovery, MOVA focuses attention on the questions of exactly what is being optimized and why, providing new insights into the system being designed. When used as a process for problem exploration, MOVA exposes the structure of the design space. This dissertation also contributes a comparison between optimization tools, based on their suitability for use across multiple optimization problems based on the same design analyses. This comparison concludes that the Borg multi-objective evolutionary algorithm (MOEA) is the most effective tool presently available for optimizing many problem formulations as required by MOVA. Finally, using global sensitivity analysis tools, this dissertation further explores the possibility of using optimization tools for model diagnostics.