Análise comparativa de controladores PID, lógica fuzzy e redes neurais aplicados ao controle de velocidade de motor de corrente contínua sob variações de carga

Abstract

This work presents the development and comparison of conventional PID, Fuzzy-PID, and Neural-PID controllers applied to the speed control of a direct current motor subjected to reference changes and load variations. The study aimed to evaluate whether adjusting the PID gains through fuzzy logic and a neural network could improve system performance compared with the use of fixed gains. Initially, the main electrical and mechanical parameters of the experimental setup were identified and used to obtain the mathematical model of the motor and to perform simulations in MATLAB. The controllers were evaluated in eight simulated scenarios involving different load levels, successive reference and disturbance changes, and parametric variations of the plant. Subsequently, the three methods were implemented in LabVIEW and experimentally evaluated in six scenarios using a data acquisition device, an analog output conditioning circuit, and a virtual load inserted into the control signal path. The comparison was based on metrics such as rise time, settling time, overshoot, maximum speed drop, recovery time, integral error indices, and control effort. In the simulations, the Fuzzy-PID and especially the Neural-PID showed advantages in disturbance rejection, damping, and robustness against plant variations. In the experimental tests, the conventional PID produced faster responses in several events, whereas the intelligent controllers showed specific advantages through overshoot reduction, lower control signal variation, and more damped responses. It is concluded that none of the methods was superior under all evaluated conditions. The PID controller proved suitable for operating conditions close to its nominal tuning and for applications that prioritize speed and simplicity, while the Fuzzy-PID and Neural-PID presented advantageous characteristics when smoother responses, gain adaptation, and lower sensitivity to certain operating changes are required