Sensitivity based solution down-selection for multi-objective optimization under parameter uncertainty

Abstract

This thesis proposes an approach to select attractive system designs out of a solution set for a multi-objective optimization problem concerning the robustness to parameter variations. Various uncertainties such as the system model error, the change in operational condition, and the increase/decrease in the cost elements are considered. Three different sensitivity-based down-selection methods – the sensitivity Pareto front, the sensitivity-threshold combination Pareto front, and the sensitivity-effect weighted sum – are introduced. The parameter variation effects on the Pareto optimal solutions are presented by comparing the objective function according to the uncertainty realization point about the deterministic optimization problem and meta-heuristic optimization. An engineering design case study demonstrates the validity and potential applicability of the proposed approach. The proposed methods are compared with the robust optimization approach in terms of performance and computational cost. A conceptual design framework for a missile system based on the proposed down-selection procedure is introduced. The down-selected solutions are compared with the solutions of robust optimization in terms of performance. The effectiveness of the proposed methods is demonstrated through statistical analysis for down-selected solutions and robust optimization solutions. As a result, this thesis concludes that the appropriate design solutions can be selected out of efficient solutions (Pareto front), considering the robustness to parameter uncertainties through the sensitivity-based down-selection methods.