Busca do Conjunto Estabilizante para um Controlador PI Baseada em Deep Learning

Kurios Iuri P. de M. Queiroz; Samaherni M. Dias; Alex M. da Costa

doi:10.20906/SBAI-SBSE-2023/3893

Kurios Iuri P. de M. Queiroz Programa de Pós-Graduação em Engenharia Mecatrônica, Universidade Federal do Rio Grande do Norte, RN
Samaherni M. Dias Programa de Pós-Graduação em Engenharia Mecatrônica, Universidade Federal do Rio Grande do Norte, RN
Alex M. da Costa Programa de Pós-Graduação em Engenharia Mecatrônica, Universidade Federal do Rio Grande do Norte, RN

DOI: https://doi.org/10.20906/SBAI-SBSE-2023/3893

Keywords: PI Controllers, Stabilizing Set, Signature Method, Deep Learning, Artificial Inteligence, Data Generator

Abstract

This paper presents a Deep Learning approach for computing the stabilizing set of PI (Proportional Integral) gains. We considered a linear time-invariant and non-minimum phase system with two poles, one zero, stable, and with a single input and single output (SISO). The training process used data obtained from the Signature Method, a technique developed by researchers from Texas A&M University that computes the stabilizing set of PID (Proportional Integral Derivative) controller parameters. The learning curves are presented for two different neural networks, showing good results for both, where 0, 03 was the best value for the loss function (Binary Cross Entropy) and 0, 99 for the accuracy (Binary Accuracy).