Search

In design optimization problems, the robustness of a solution is essential to be considered when it is going to be implemented or used in real world. Robustness can be defined in different ways, but in this work, it is considered as the property of a solution’s performance to be as insensitive as possible to perturbations in that solution. Optimizing both quality and robustness is generally addressed as a multiobjective problem. This paper presents a complete framework aided by decision maker, which can be used to solve robust optimization problems with one or more objectives. Decision maker actively takes part of the process, defining a priori the robustness metric, and specifying iteratively a robustness factor that joins quality and robustness metrics into a single fitness function. The framework is very flexible allowing decision maker not only to specify the right robustness level for him, but also to exchange metrics and algorithms if he wants. The framework was tested and validated using two single-objective robust benchmarking functions, and in one new multiobjective robust function, created to challenge the framework, which tries to seek the most robust Pareto optimal front according to decision maker preferences. [ABSTRACT FROM AUTHOR]