Partially Decentralized Control And Its Application In Quadruple Tank Process

Interaction is very common in industrial processes. Nevertheless, decentralized control is preferred in practice due to the fact that decentralized control is easy to implement, to tune and to maintain. However, if the interaction is strong in a multivariable system, a decentralized control may not achieve satisfactory performance. In this case, a centralized control has to be used to achieve better performance. But a centralized control is difficult to implement and maintain, so an alternative control structure, partially decentralized control (PDC), is a better choice. PDC can obtain comparable dynamic performance with the centralized control with a much simpler structure.In this thesis, a method to determine the structure of PDC is first discussed. The method is based on Gramian controllability and observability matrices. Then the expansion method is discussed for the design of PDC. Internal model control (IMC) and dynamic model control (DMC) methods are adopted to deal with the controller design problem of non-square systems that arises in the expansion design.Then the quadruple-tank process (QTP) is discussed. QTP is made of four interconnected water tanks and is a typical multivariable laboratory equipment with strong interaction. Moreover, the zeros of the process can be varied in the experiment. PDCs are designed for the minimum and the nonminimum phase QTP. Simulation results show that PDC can achieve close performance as the sophisticated multivariable controller; however, it has a simpler structure and is easier to implement and optimize.Finally, PDC is designed for a laboratory quadruple-tank equipment. Experiment results show that the designed PDC achieves good control performance.