Eventos Anais de eventos
COBEM 2021
26th International Congress of Mechanical Engineering
Harris hawks optimization approaches on the multivariable PID controller tuning
Submission Author:
Yan Lieven , AL , Brazil
Co-Authors:
Yan Lieven, Luiza Scapinello Aquino, José Henrique Kleinübing Larcher, Viviana Mariani, Leandro dos Santos Coelho
Presenter: José Henrique Kleinübing Larcher
doi://10.26678/ABCM.COBEM2021.COB2021-0055
Abstract
In this work, a performance comparison between the Harris Hawks Optimization (HHO) and four different created variants for the tuning of a decentralized proportional-integrative-derivative (PID) controller in a multiple-input multiple-output (MIMO) application is presented. The application has the objective to achieve optimal response on the control of a ball mill pulverizing system with a steady state decoupler, whose structure is composed of two inputs and two outputs. Metaheuristic approaches are suitable for this kind of problem due to its capacity to diversify the search space(exploration) and improve the quality of current solutions(exploitation). The HHO algorithm is a metaheuristic based on the cooperative hunting behavior of the Harris’s Hawks, that consists in a tactic called “surprise pounce”, in which each hawk attacks the prey from different positions and angles, cornering the target, exhausting it and finally diving to capture it when possible. Although the original HHO may present sufficient performance, it may be improved through addition of different techniques. To check for possible improvements, four HHO variants are implemented using different procedures, namely the use of cultural behavior based on normative and situational knowledges, oppositional-based learning on the second variant, covariance matrix learning on the third variant, and finally the application of quantum mechanics into the fourth one. The optimal parameter values for the PID controller are sought by minimizing the integral time squared error (ITSE) index of the response of the system. Simulations are performed using SIMULINK and MATLAB softwares. Statistical measures such as best, mean, median and standard deviation of the system response error for the tuned controllers are analyzed and compared over fifty runs. The obtained results have proven that the use of the previously mentioned proposals leads to improvements in the tuning efficiency of the HHO in this context, upgrading the performance of the PID controller for the control of the ball mill model.
Keywords
Proportional-Integrative-Derivative Control, Harris Hawks Optimization, Cultural Algorithm, Covariance Matrix Learning, Oppositional-Based Learning, Quantum Mechanics, Multiple-Input Multiple-Output Application
