Identification and control of relay-stabilizable processes

Several methods of identification of the frequency response of stable processes have been developed. However, for unstable processes, these methods are not applicable. Although obtaining Nyquist data is even more critical for the design of robust controllers for unstable systems. In this thesis, a sinusoidal-signal based relay method is proposed which can be used to identify several points on the Nyquist plot of an unstable process. The method is referred to as RFS (Relay Feedback with Sinusoidal input). The procedure involves injecting a mixed sinusoidal signal into the feedback loop stabilized by a standard relay. Fourier analysis techniques are then used to estimate Nyquist points corresponding to various frequencies. Information obtained from the RFS method can then be used to design robust controllers using tuning rules which consider both the minimum and maximum gains needed to stabilize the process as well as the shape of the Nyquist plot surrounding (-1,0).; The method is demonstrated using two simulation examples. The first example shows how the results of RFS can be used to design a robust controller for a high order process with two unstable poles. The second example deals with the identification of unstable process with one unstable pole and a significant delay.