Optimal non-parametric tuning of PID controllers based on classification of shapes of oscillations in modified relay feedback test

Abstract

In this work, tuning rules of the PID controller have been developed by categorizing a system's response into distinct classes. The classes are formed using the shapes of the test oscillations induced by the system under the Modified Relay Feedback Test (MRFT) produced by specific system models. It is proposed that a physical system can be categorized into one of the proposed classes and thus the tuning rules for a particular class can apply to any kind of system from this class. The idea of producing tuning rules that are based on the shape of the oscillations induced in the loop containing the process comes from the observations that oscillatory responses of physical systems reveal just a few different shapes depending on system dynamics. For applying the developed optimal tuning rules for an arbitrary system, first, certain system characteristics are determined using a priori knowledge of the class model. Then the system's response with the application of the MRFT is examined to classify the oscillation waveform/shape. In this work, such classification is carried out using a cross-correlation algorithm. Finally, a class tuning rules are applied.