Exploring Pareto Frontiers in the Response Surface Methodology

Abstract

Multiple response optimization problems have many optimal solutions that impact differently on process or product. Some of these solutions lead to operation conditions more hazardous, more costly or more difficult to implement and control. Therefore, it is useful for the decision-maker to use methods capable of capturing solutions evenly distributed along the Pareto frontier. Three examples were used to evaluate the ability of three methods built on different approaches for depicting the Pareto frontier. Limitations of a desirability-based method are illustrated whereas the consistent performance of an easy-to-use global criterion gives confidence to use it in real-life problems developed under the Response Surface Methodology framework, as alternative to the sophisticated physical programming method.

Multi-objective data-driven optimization for improving deep brain stimulation in Parkinson’s disease

“…Mathematically [35], for a system f : R n → R m , where X is a set of feasible inputs on R n , and Y is the feasible set of objectives in R m , such that:…”

Unique Parameters Selection Strategy of Linear Canonical Wigner Distribution via Multiobjective Optimization Modeling

“…It then follows the deterministic numeric values of and . Due to the theory of uniform design [50], the selection of parameters can be narrowed down to a set of representative experimental points. Traversing these points and ranking the derived numeric values thus reach the maximum values of and .…”

Alveolar ridge preservation: Complications and cost‐effectiveness

“…Here, the most relevant treatments are those that are not dominated in terms of both efficacy (ridge width and height maintenance) and cost by another other treatment. These relevant treatments lie on the “Pareto frontier” (optimal path) 142,143 . In other words, a treatment with an equal or greater efficacy while simultaneously providing an equal or lesser cost dominates other treatments, whereas other therapies that would yield a lesser or equal performance while with equal or higher costs are considered inadmissible (Figure 8).…”

“…These relevant treatments lie on the "Pareto frontier" (optimal path). 142,143 In other words, a treatment with an equal or greater efficacy while simultaneously providing an equal or lesser cost dominates other treatments, whereas other therapies that would yield a lesser or equal performance while with equal or higher costs are considered inadmissible (Figure 8). Lastly, when the model was fit to log-scale data for capturing relative returns on membrane costs (using changes in log ridge width and height as the outcomes with log reference costs for the membrane), a diminishing return was observed for an additional dollar spent for a barrier membrane beyond approximately US$50 and gaining additional performance in ridge preservation (Figure 9).…”