Improving the identification from relay feedback experiments

Abstract

This paper proposes a new method for systems identification from relay feedback test. As a difference with most of the existing methods, which are based on the describing function technique, our proposal considers a more precise approach that allows to improve the estimation results. Furthermore, the method has also the advantage of being easy to implement, which makes it to stand out among other methods that require introducing additional elements on the loop or extend the experiment duration. The validity of the proposed method has been proved using a batch of models that represent the most common dynamic behaviors of actual industrial processes. The study reveals that with the new proposal, using data from a simple relay experiment, the error in the estimation for some kind of systems can be considerably reduced with respect to previous approaches.