Diseño e implementación de un algoritmo de control adaptativo tolerante a fallos activo basado en redes neuronales y lógica difusa

Abstract

This work present the design methodology of two fault tolerant proportional-integral-derivative (PID) controllers. The first controller proposes a passive fault tolerant control strategy (PFTC) through online calculation of controller parameters using an ADALINE neural network whose learning rate is adjusted with a diffuse inference system. The second controller is a Fuzzy Gain Scheduling PID whose ranges of variation of parameters are adjust online with the information from a fault detection and diagnostic (FDD) module. The proposed scheme’s architecture forms an active fault tolerant controller (AFTC). The detection and diagnosis of faults is achieved by parity equations. The validation of the control algorithms is performed in a multiple-input multiple-output system (MIMO) of three tank, including the leakages and obstructions models. The simulations allow us to compare the performance of the proposed approaches with the classic PID control algorithm.