Multi-objective Parameter Tuning and Performance Comparison of PI and FOPI Controllers for a Children’s Anesthesia Administration System

Abstract

This work compares the performance of PI and FOPI controllers for a children’s anesthesia administration system applied to a FOPDT model based on data of 46 patients. Parameter tuning is performed through a multi-objective differential evolution optimization algorithm that includes a spherical pruning mechanism to obtain a good distribution along the Pareto front. The design objectives were the median and interquartile range of the integral of the absolute value of error (IAE). The Pareto front of the solutions of FOPI dominated the ones of the PI controller, although the PI Pareto front had a less disperse distribution. The best option of each Pareto set was determined by testing each option with the 46 patients and looking for the case that complies with the most or all problem constraints. The results of the tests showed that PI had a stable behavior with all patients, while FOPI showed an undesirable behavior with 5 out of 46 patients that did not comply with the constraints of the problem. Despite FOPI had a time response 37.33% faster, PI had better control characteristics and complied with all the design constraints in all patients, hence a control scheme such as FOPI is not justified.