Reduced Bias User Preference Methods for Determining the Pareto-Optimal Solution Point

Engineering design is filled with tradeoffs between competing objectives such as performance, mass, cost, and schedule. A designer must navigate these complex multi-objective problems and deliver the right solution for their application. Multi-objective optimization techniques are powerful and widely used; however, a key drawback to these techniques is that they often output a set of equivalent solutions called the Pareto Front. The designer must perform an additional multi-objective down selection on the Pareto Front to determine a single Pareto-optimal solution point for their design. Existing Pareto Front processing techniques either use traditional infinitely adjustable weights, which can yield results that are highly sensitive to the selected weights, or are no-preference methods that don’t account for the designers’ inputs on the relative importance of the objectives. These existing Pareto Front down selection methods do not provide a sufficient way for the designer to input their insight on the relative importance of the objectives without the technique being overly sensitiveto this input. This thesis proposes two new methods to find a Pareto-optimal solution point. These methods are called the Prioritize and Funnel algorithm and the Rank and Influence Based Weighting method. These methods build off existing strategies but vary from previously proposed techniques by accounting for the designers’ inputs through a designer ranking of the objectives. When applied to a series of example problems, these two new methods were demonstrated to efficiently perform the Pareto Front down selection and respond accordingly to changes in the designer preferences