2015
EMSS 2015 Proceeding
AMEBA ? Structural evolutionary optimization: method and toolbox development
Authors: Marko Corn, Maja Atanasijevi?-Kunc
Abstract
Evolution algorithms are optimization methods that mimic a process of a natural evolution. Their stochastic properties result in a huge advantage over other optimization methods especially when solving complex optimization problems. In the paper Agent Modelled Evolutionary Based Algorithm (AMEBA) is first presented which was developed and implemented in MATLAB as a Toolbox. AMEBA algorithm has several advantages over other evolutionary algorithms and in this article the advantage of custom designed initial solution is presented. Custom designed initial solution is a solution of the problem that is built on the base of the knowledge of the system and represents a solution which the AMEBA algorithm will try to improve. This capability is presented with the example of an evolvement of multivariable controller for the pressure ? level system that represents non-linear, multivariable system which is very stiff, with the property of weak inherent coupling. The AMEBA algorithm significantly improved the initial controller solution which shows that classical controller structures can also be automatically altered to increase quality of the solution.
I3M Scientific Sponsors
I3M Industrial Sponsors
I3M Media Sponsors
