Adaptive Neural Fuzzy Control For Mimo Nonlinear System

This thesis presented the research and application of artificial intelligence with hybrid system that is the combination of fuzzy system and neural network based on structure of MISO ANFIS network (MISO ANFIS:Multi-Input Single-Output Adaptive Neural Fuzzy Inference System) to design controller using ANFIS that has Multi-Input Multi-Output configuration (MIMO) to inherit the inherent advantages of individual controllers (fuzzy or neural controller) serving in process control for MIMO nonlinear plant. Specifically, it is the combination of two structures of the MISO ANFIS network to create two-inputs two-outputs ANFIS controller (MIMO) to control for two-inputs two-outputs nonlinear plant (MIMO); its main task is to minimize the error signal during the operation process of the system. The selected control plant to serve for studying ANFIS controller is the liquid couple tank system that has two-inputs two-outputs configuration (MIMO); it is nonlinear plant and has dynamic change in the operation process by the influence of work condition. From the construction of mathematical models of selected liquid couple tank system and the designed ANFIS controller, based on the Matlab-Simulink software build the simulation model of control system of couple tank system to verify the correctness of the control algorithm of the designed ANFIS controller. The simulation results showed that the designed ANFIS controller responded well for controlling the liquid level of the liquid couple tank system with high accuracy; it has good adaptability and stability with variation of different setpoints (desired signal), the output signal (the height of the liquid level) tracked the given trajectory of setpoint (even with the impact of disturbance) with overshooting, response time and steady error of the output signal are small. Additionally, this thesis also designs a controller by using the MFA neural network (MFA:Model-Free Adaptive), it is used to control the same plant with the purpose of getting results to compare with the control results of the ANFIS controller. Through the comparison of the simulation results of the ANFIS controller and the MFA neural network controller has been proved that the control results of the ANFIS controller have quality better than the MFA neural network controller. This also showed the feasibility and effectiveness of the ANFIS controller in process control. From that, the application of this ANFIS controller can contribute to improve the efficiency of the production process to respond the needs of industrial practice today.