Diseño e implementación de un controlador PID de ganancia programada usando técnicas de inteligencia artificial

Abstract

This paper analyzes three new control strategies that improve system response and its com- parison with classic PID control. The efficiency of the PID controller depends only on its gains Kp, Ki and Kd. Therefore, the strategies improve the control process by finding the optimal gains through the programed algorithm. The programmed gain control algorithms are based on artificial intelligence techniques such as fuzzy logic, the simulated annealing algorithm and the A* search method, resulting in the FGS-PID, SA-PID and A*-PID controllers. The control algorithms are tested in MATLAB and adapted on the STM32 core-144 embedded system, with the objective of controlling the water level of a multitank system. For this purpose, different performance tests are carried out in simulations and on the real system where the response of the plant to reference changes is evaluated. In addition, the settling time and percentage of overshoot are used as performance indexes. Also, the operation of the control system is tested by introducing possible failures in the plant, such as through leaks in the tanks caused by a hole in the bottom. Finally, the results show the effectiveness of the algorithms in reducing the settling time and overshoot in the response of the multitank system, furthermore, adding some tolerance to failures in the system. The aforementioned results allow the tracking of the reference in the presence of faults, which is not achieved with the classical PID control system.