Eventos Anais de eventos
COBEM 2017
24th ABCM International Congress of Mechanical Engineering
USING MULTIOBJECTIVE OPTIMIZATION DESIGN FOR DRIVEN PENDULUM SYSTEM CONTROL
Submission Author:
Alisson Ravaglio Santos , PR , Brazil
Co-Authors:
Gabriel Herman Bernardim Andrade, Gilberto Reynoso Meza, Leandro dos Santos Coelho, Alisson Ravaglio Santos
Presenter: Alisson Ravaglio Santos
doi://10.26678/ABCM.COBEM2017.COB17-1531
Abstract
Almost all of the optimization problems found in science and engineering are represented by complex mathematical models with multiple parameters that require simultaneous optimization. This eventually leads to the issue of several constraints and conflicting objectives, requiring the use of advanced optimization techniques. Multiobjective optimization is the class of optimization problems which involves more than one objective function to be optimized simultaneously. It increases the problem complexity by requiring not only the computation of the optimal value for each objective but also analyze the tradeoff between each objective best value in order to compute the real global value for the required problem. The main contribution of this paper is to analyze the performance of Multiobjective Optimization Design techniques when applied to the optimization of the input values of a controller. We study the optimization of a multiple parameterized DMC (Dynamic Matrix Control) Controller, a model-based predictive control approach, applied to the control of a pendulum system. The controller parameters are optimized through the application of the sp-MODE (Spherical Pruning Multiobjective Differential Evolution), a Differential Evolution based algorithm which applies concepts of spherical pruning in order to improve diversity in the solution search space. Our approach validates the application of such optimization methods for control applications, since the optimized parameters of the controller worked neatly, making the pendulum system completely stable. The obtained resulting parameters were close to the truly ideal values.
Keywords
Multi-Objective Optimization, Dynamic Matrix Control, Pendulum System, M-dimensional visualization, Decision Making
