A Multi-Objective Post-Optimality Data Handling Approach to Robust Optimization
Kevin M. Ryan, Mark J. Lewis, Kenneth H. Yu · 2014
The use of a multi-objective based robust optimization method to solve single-objective optimization problems with environmental parameter uncertainty was investigated. Unlike commonly used robust optimization methods, the multi-objective method formulates an optimization problem such that post-optimality data handling techniques can identify multiple robust designs from a single solution set. This allows for comparisons to be made between different types of robust designs, thus providing more information about the design space. The method is presented in a multi-objective genetic algorithm context. An example problem with two design variables and one environmental parameter was used to illustrate the capability and nuances of the proposed method. In particular, a focus was put on analyzing the changes in the convergence and accuracy of the method with changes in the genetic algorithm initial conditions and data handling settings.