Robust Decentralized Controller Design for Coupled Tank Systems
Loading...
Files
Date
2022
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Department of Electronics and Instrumentation Engineering, NITS
Abstract
The majority of controllers in process industries is having either good set-point tracking or a good load disturbance rejection. While disturbance rejection is prioritized in pro cess industries, the controller’s dynamic performance is also important. The coupled tank system is one of the most common processes which is encountered in pharmaceutical, chemical, wastewater management, and petrochemical industries. The coupled tank system is a benchmark problem whose controller design is difficult since it is a non-linear, non-minimum phase TITO system. Moreover, the coupling effect of CTS will always offer several challenges for controller design. So the main goal of this work is to create a reliable decentralised controller for controlling the level of the liquid in coupled two-tank MIMO system that integrates both set-point tracking and the load disturbance rejection characteristics as well. For a simplified design and a simple analytical solution, the reduced-order model is obtained first by implementing model order reduction on the decoupled transfer function matrix. The decoupled matrix is of higher order and has a huge time delay. Hence, a FOPDT model of both decoupled tanks is obtained using the frequency curve fitting model order reduction technique. The decentralized PI controllers are designed using a variety of control strategies adopted from the literature. The Direct synthesis method in which the controller gain is analytically calculated by comparing the closed-loop transfer function of the plant and controller with the desired closed-loop transfer function. Next tuning of a decentralised PI controller gains based on desired Gain and Phase margin requirements is implemented for tank 1 and tank 2 respectively. A decentralized PI gain synthesis using a Genetic algorithm with maximum sensitivity as an objective function is implemented. The proposed work also includes the design of a decentralized controller that uses a linear quadratic regulator and Regional pole-placement using LMI (D-stability). Lastly, by altering the plant model’s gain and time constant, the robustness of the proposed controller is assessed in terms of model uncertainty.