Research Projects

Over the past two decades, research has focused on developing modeling and controller design approaches for controlling liquid levels in multivariable liquid level systems. In industrial processes, multi-input multi-output MIMO systems are commonly used. Controlling these systems is challenging due to the complexities involved in nonlinear MIMO systems. The main control objective is to maintain desired liquid levels while reducing coupling effects. In today's complex process industries, controlling multiple variables such as temperature, pressure, and level in multiple interacting loops can lead to degradation of performance. External disturbances and parametric uncertainties further deteriorate the system's performance. Most previous works consider known parameters for control design, but this proposal adopts a nonlinear system identification technique to capture unknown parameters. Old multivariable control designs often use a decentralized PID controller and a decoupler, but these methods fail to provide suitable control performance due to parametric uncertainties and disturbances. This work proposes a machine learning-based robust decentralized with active disturbance rejection controller design, which can yield superior performance despite uncertainties and disturbances. The proposed work aims to close the gap by developing a nonlinear hybrid multivariable tank system experimental setup, modeling the system using a nonlinear system identification technique, designing a dynamic decoupler to resolve possible loop interactions, developing robust decentralized controllers considering actuator saturation, and verifying the performance of the proposed controllers through simulation and real-time experiments on a coupled tank liquid level setup.