|M.Sc Student||Weisbord Amit|
|Subject||Predictor Based Control for Processes with Right Half|
|Department||Department of Mechanical Engineering||Supervisor||Professor Emeritus Zalman Palmor|
|Full Thesis text|
In feedback control theory, it is known that the presence of right half plain (RHP) zeros in a plant limits the achievable performance of its control. The bandwidth of the closed loop is restricted and the allowed control gain in the low frequencies is small. Systems with RHP zeros constitute one class of non-minimum phase (NMP) systems. A possible way to deal with the control of such plants is to apply optimal design methods like and . However, to the best of our knowledge, no widely accepted specific structures for the control of NMP systems resulted from those methods and to date there is no well-established control design methodology for NMP systems with RHP zeros.
Another class of NMP systems is systems with time delays or dead times. A well known control configuration for the latter is the dead time compensator (DTC) which is popular in the process industries. The DTC removes the delay from the characteristic equation allowing higher gains to be used and simplifies the design. The first and simplest DTC to be introduced was Smith predictor (SP). Several modifications of the SP for unstable processes have been suggested over the years, among which are the modified and filtered SP. In each of the schemes an additional component, termed here the modifier and filter respectively, is added to the predictor.
In this work the possibility of adopting the DTC configuration to systems with RHP zeros was examined. Two schemes for unstable plants with RHP zeros are considered. The modified predictor scheme (MPS), and the filtered predictor scheme (FPS). Assumptions representing feasibility conditions as well as control design requirements on the predictor are presented for both schemes. A theorem translating the requirements to the strong stabilizability condition of an auxiliary plant is presented for the MPS. Procedures for coping with the conditions and assumptions using interpolations are suggested for both schemes. Two controller design procedures are presented for both stable and unstable NMP processes. The procedures are demonstrated on examples taken from the literature and their advantages and drawbacks are discussed in comparison to the specific designs. It is shown that the new schemes considerably simplify reference tracking control design, particularly for unstable NMP processes. It is also shown that the schemes excite modes in the disturbance rejection CL transmission which complicate disturbance rejection design.