Intercriteria analysis of relations between model parameters estimates and aco performance

Abstract

In this paper we apply the approach InterCriteria Analysis (ICrA) to establish the existing relations and dependencies of defined parameters in non-linear model of an E. coli fed-batch fermentation process. Moreover, based on results of series of Ant Colony Optimization (ACO) identification procedures we observe the mutual relations between model parameters and ACO outcomes (execution time and objective function value). We perform a series of model identification procedures applying ACO. To estimate the model parameters we apply consistently 11 differently tuned ACO algorithms. We use various population sizes—from 5 to 100 ants in the population. In terms of ICrA we define five criteria, namely model parameters (maximum specific growth rate, μmax; saturation constant, kS and yield coefficient, YS/X) and ACO outcomes (execution time, T and objective function value, J). Based on ICrA we examine the obtained parameters estimates and discuss the conclusions about existing relations and dependencies between defined criteria. The obtained here results we compare with the ICrA results achieved using Genetic Algorithms (GA) as optimization techniques. Thus, based on the results of ACO and GA (the worst, best and average estimates) we define more precisely in which group (negative consonance, dissonance or positive consonance) fall the given ICrA criteria pairs.