Study comparing optimization methods for problems with multiple responses

Abstract

This work aims to evaluate the performance of two agglutinating functions, Desirability and, Modified Desirability, performed with two search methods, the Generalized Reduced Gradient and, the meta-Heuristic Simulated Annealing. The performance of methods was assessed by the Absolute Distance and the Mean of Deviation Percentage. For the optimization methods implementation were selected three multiple-response cases from the literature. The results of methods assessment show a better performance for the modified Desireability function, regardless of the search method for the multiple-response optimization, especially when these responses are modeled by equations within quadratic terms, regardless of: the number of terms; the types of responses; and, the number of variables. The methods and the results presented in this paper aim to collaborate with the research, development and assessment of advanced techniques for multi-response optimization in order to increase the performance of industrial processes.