Performance of a Genetic Algorithm applied to robust design in multiobjective systems under different levels of fractioning

This paper studies the performance of a Genetic Algorithm (GA) to find solutions to problems of robust design in multiobjective systems with many control and noise factors, representing the output vector in a single aggregation function. The results show that the GA is able to find solutions that achieve a good adjustment of the responses to their corresponding target values and with low variability, even with highly fractional experimental designs, which provide a limited number of data points to be fed to the GA. This conclusion is important for the practical application of the GA to robust design studies. Generally, such studies are carried out using scarce resources and dealing with other limitations, which force the engineer to use few experimental treatments and gather a limited amount of data. Thus, knowing that the GA performs well under such situation expands its applicability.