Banca de QUALIFICAÇÃO: ALEX MUNIZ DA COSTA

Uma banca de QUALIFICAÇÃO de MESTRADO foi cadastrada pelo programa.
STUDENT : ALEX MUNIZ DA COSTA
DATE: 20/07/2023
TIME: 08:30
LOCAL: Remoto (google meet)
TITLE:

AI Applications in PID Controller Tuning

KEY WORDS:

PID controller, artificial neural network, and deep neural network.

PAGES: 38
BIG AREA: Engenharias
AREA: Engenharia Elétrica
SUBÁREA: Eletrônica Industrial, Sistemas e Controles Eletrônicos
SUMMARY:

The PID controllers are present in approximately 96% of industrial control loops, but it is estimated that a large part of these controllers are not tuned. Nowadays, it is still a difficult task to tune controllers using classical techniques, which has motivated the development of new tools, mainly using artificial intelligence to adjust the parameters of the PID controller. This work aims to develop a tool for tuning PID controllers, based on the stabilizing set theory developed by Texas A&M researcher Professor Bhattacharyya. A program was created capable of, given a second-order plant, finding the region of stability and, given values for Kp and Ki, verifying whether or not they are within the region of stability, that is, it can be verified which values can lead the system to instability. To implement the tool, a supervised deep neural network was used to classify the Kp and Ki values, having as input a database with more than 10,000 pairs of values for Kp and Ki. There were some settings for the structure of the neural network, such as the number of hidden officials, and athletes by championships, among others. In the tests carried out, the system presented an accuracy of more than 99.6%, proving to be efficient in classifying the points as belonging or not to the region of stability.

COMMITTEE MEMBERS:
Presidente - 1577068 - KURIOS IURI PINHEIRO DE MELO QUEIROZ
Interno - 2566657 - SAMAHERNI MORAIS DIAS
Externo ao Programa - 347565 - ALDAYR DANTAS DE ARAUJO - UFRN