Enhancing Four-Tank System Through Model Predictive Control-Based Adaptive Neuro-Fuzzy System

Abstract The four-tank system is a common industrial setup used in water treatment, chemical processing, and pharmaceutical applications. This work proposes model predictive control (MPC) for precise controlling of liquid levels of the four tanks considering the flow rate limits. The proposed approach integrates MPC with an ANFIS to improve real-time state estimation by adapting to nonlinear and time-varying dynamics and effectively handling external disturbances and measurement noise. Additionally, adaptive particle swarm optimization is employed to fine-tune the ANFIS network parameters, ensuring accurate and adaptive modeling to ensure accurate state estimation. The research presents a comparative synthesis between linear and nonlinear MPC controllers under different operating conditions, disturbances, and uncertainties. The MPC approaches are implemented using the $$\mu $$ <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"> <mml:mi>μ</mml:mi> </mml:math> aoMPC and ACADO toolkits to accomplish real-time requirements to solve the optimization problem. Furthermore, processor-in-the-loop experiments have been utilized to validate the numerical results, confirming the real-time applicability and computational efficiency. The results demonstrate that the NMPC controller, supplemented by ANFIS, achieves superior tracking accuracy, faster disturbance rejection, and reduced steady-state error.