• JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
 
  Bookmark and Share
 
 
Master's Dissertation
DOI
https://doi.org/10.11606/D.3.2020.tde-08072021-140955
Document
Author
Full name
Victor Daniel Reyes Dreke
E-mail
Institute/School/College
Knowledge Area
Date of Defense
Published
São Paulo, 2020
Supervisor
Committee
Garcia, Claudio (President)
Alvarado, Christiam Segundo Morales
Potts, Alain Segundo
Title in Portuguese
Desenvolvimento de uma ferramenta de validação de modelos para controladores preditivos baseados em modelos.
Keywords in Portuguese
Controle de processos
Controle preditivo
Inteligência artificial
Abstract in Portuguese
Quantificar a qualidade de um modelo é um problema existente na área de Identificação de Sistemas. Esta atividade, também conhecida como validação, é fundamental nas aplicações onde se utilizam Controladores Preditivos baseados em Modelos, porque estes precisam de um modelo adequado para seu bom funcionamento. Baseado nisto, nesta dissertação são implementados três algoritmos de Inteligência Artificial capazes de predizer, de forma autônoma, quão adequado pode ser um modelo para este tipo de aplicação. Os algoritmos são Árvores de Decisão, Máquina de Suporte de Vetores e Redes Neurais Artificiais. Eles predizem a qualidade do modelo a partir de resultados de outras métricas de validação não existentes. As plantas para a implementação destes algoritmos são: (i) Planta de Clarke (simulada) e (ii) Planta Piloto de Neutralização de pH (real) do Laboratório de Controle de Processos Industriais da Escola Politécnica da Universidade de São Paulo. Em ambos os casos se usa um algoritmo Dynamic Matrix Control - DMC ou sua variante Quadratic Dynamic Matrix Control - QDMC (em caso de se ter restrições) para executar o controle. Como resultado deste trabalho obtiveram-se algoritmos capazes de predizer a qualidade do modelo com uma acurácia de 84,1%, 91,5% e 91,0% para a malha de controle da Planta de Clarke, e de Nível e de pH para a Planta Piloto de Neutralização de pH, respectivamente.
Title in English
Model validation tool for model predictive control applications based on artificial intelligence.
Keywords in English
Artificial intelligence
Dynamic matrix control
Model based predictive controller
Quadratic dynamic matrix control
Abstract in English
Quantifying the quality of a model is an existing problem in the Systems Identification area. This task, also known as validation, is fundamental in applications where Model Predictive Control is used, because they need an adequate model for their proper operation. Based on this need, in this dissertation, the author implements three Artificial Intelligence algorithms that are capable of autonomously predicting how suitable a model can be for this type of application. The algorithms are Decision Trees, Support Vector Machine and Artificial Neural Networks. They predict the quality of the model from the results of other non-existent validation metrics. The plants for the implementation of these algorithms are the Clarke Plant (simulated) and the pH Neutralization Pilot Plant (real) of the Industrial Process Control Laboratory of the Polytechnic School of the University of São Paulo. In both cases, a Dynamic Matrix Control - DMC algorithm or its Quadratic Dynamic Matrix Control - QDMC variant (in case of constrained problems) is used to perform the control. This work results are algorithms capable of predicting the model quality with an accuracy of 84.1%, 91.5%, and 91.0% for the Clarke Plant, and the Level and pH control loops pH of the Neutralization Pilot Plant, respectively.
 
WARNING - Viewing this document is conditioned on your acceptance of the following terms of use:
This document is only for private use for research and teaching activities. Reproduction for commercial use is forbidden. This rights cover the whole data about this document as well as its contents. Any uses or copies of this document in whole or in part must include the author's name.
Publishing Date
2021-07-08
 
WARNING: Learn what derived works are clicking here.
All rights of the thesis/dissertation are from the authors
CeTI-SC/STI
Digital Library of Theses and Dissertations of USP. Copyright © 2001-2024. All rights reserved.