Engineering design problems often occur as multi-objective optimization problems related to the Pareto frontier determination problem. One of the objectives of this article is to clearly demonstrate the determination, meaning and characteristics of the Pareto frontier by means of several insightful examples of a two dimensional problem, both with and without constraints. In addition to its educational value, the main contributions of the article are summarized as follows. Firstly, we demonstrate the importance of presenting results both in the design variable and the objective space in order to become acquainted with all the data required for correct decision making and interpretation of optimization results. Secondly, we present an approach to explicit (analytical) determination of the Pareto frontier equation in the design space. In this case, the objective functions have to be differentiable. Thirdly, we propose a suitable modification of the Gsoal attainment method as an efficient way to determine the Pareto frontier for some classes of multi-optimization problems.
[1]
Eckart Zitzler,et al.
Evolutionary algorithms for multiobjective optimization: methods and applications
,
1999
.
[2]
R. Storn,et al.
Differential Evolution: A Practical Approach to Global Optimization (Natural Computing Series)
,
2005
.
[3]
R. K. Ursem.
Multi-objective Optimization using Evolutionary Algorithms
,
2009
.
[4]
Gabriele Eichfelder,et al.
Adaptive Scalarization Methods in Multiobjective Optimization
,
2008,
Vector Optimization.
[5]
椹木 義一,et al.
Theory of multiobjective optimization
,
1985
.
[6]
Kalyanmoy Deb,et al.
A fast and elitist multiobjective genetic algorithm: NSGA-II
,
2002,
IEEE Trans. Evol. Comput..
[7]
E. Polak,et al.
On Multicriteria Optimization
,
1976
.