Comparative Study on Decoupling Methods for Time-Delay Systems

Abstract

Decoupling of a multivariable system is much pronounced control problem in many indus- trial process. This has received much importance since last 20 years to develop decoupling controllers as an alternative to multi-loop PI controllers in order to achieve satisfactory set- point responses when there exist multiple time delays, non-minimum phase zeros and large uncertainties. In this thesis, static and dynamic decoupling control strategies are discussed. Static decouplers are designed for control of multi-variable processes by RGA interaction analysis to determine the nature of interaction for static decouplers applied to multi-loop con- trollers and integral mode only at low and high frequencies. Here it is shown that static decouplers applied to integral modes perform better for processes in which non-diagonal ele- ments decrease faster as frequency increases. This approach is sensitive to process changes and require detailed models. In dynamic decoupling, internal model control and inverted decoupling methods are dis- cussed here. Internal model control approach comprises both open-loop decoupling and closed- loop decoupling techniques. Internal model control is a centralized controller derived from the desired diagonal system matrix defined taking in consideration of time-delay compensation, non-minimum phase zeros and H2 optimal performance objective; which effectively perform both decoupling and controlling functions. In open-loop IMC decoupling, the controller is de- signed from open-loop equation of process assuming that the model is exactly matching with the process. In closed-loop IMC decoupling, only model of the system is considered and con- troller is derived from closed-loop equation of the overall system. The presence of RHP zeros is elaborately discussed in different cases like no RHP zeros, finite RHP zeros, infinite RHP and LHP zeros and infinite RHP but finite LHP zeros. Inverted decoupling approach utilizes inverse transfer matrix of the multivar