Abstract
In real-world engineering design problems we have to search for solutions that simultaneously optimize a wide range of different criteria. Furthermore, the optimal solutions also have to be robust. Therefore, this paper presents a method where a multi-objective genetic algorithm is combined with response surface methods in order to assess the robustness of the identified optimal solutions. The design example is two different concepts of hydraulic actuation systems, which have been modelled in a simulation environment to which an optimization algorithm has been coupled. The outcome from the optimization is a set of Pareto optimal solutions that elucidate the trade-off between energy consumption and control error for each system. Based on these Pareto fronts, promising regions could be identified for each concept. In these regions, sensitivity analyses are performed and thus it can be determined how different design parameters affect the system at different optimal solutions.
Acknowledgements
The software for this work is based on the GAlib genetic algorithm package written by Matthew Wall at the Massachusetts Institute of Technology (see Wall n.d.).