AUTOMATIC TUNING OF PID CONTROLLERS USING MULTISINE SIGNALS

Abstract

It is useful to study the relative performance of PID and PI control so that the know ledge can be made used of by auto-tuners. A quantitative measure of the relative performance of PID and PI control is also useful in the design of a knowledge based feedback controller. In this thesis, a comparison study has been performed to find out how much improvement in performance can an optimal PID controller give compared to an optimal PI controller. The optimal controllers are designed by using formulae which minimize some common performance indexes. Set-point response and load disturbance response are considered. The present frequency domain based automatic tuning of PID controller uses mostly steps or pulses as perturbation signals to identify the process. This contains frequencies across a wide spectrum and will likely result in a waste of energy. Also, in the on-line methods based on the tracking of several points on the Nyquist curve of a process, identification can only be performed when the user makes set-point changes which may not be often in practice. In this thesis, a novel approach of automatic tuning using multisine signal as the perturbation signal has been developed. This approach is efficient in the sense that all the signal power can be made concentrated exactly on the frequencies of concern. Also, this approach eliminates the need to introduce any set-point changes in the identification process since the multisine signal can simply be added to the control signal.