Data-Based Fault Diagnosis via Dissipativity-Shaping

Abstract

In this letter, a novel data-based fault diagnosis framework for the linear process is developed using dissipativity shaping method. First, the dissipativity for finite-horizon trajectory data is learnt and shaped such that it is valid for <bold xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">all but one</b> certain fault (to be identified). Then, using the <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">weaving response lemma</i> , the data-based dissipativity for fault diagnosis purposes is extended over an arbitrarily long horizon. Next, from the available input-output trajectories, the corresponding fault-selective coefficient matrices of the dissipativity condition of the process are shaped using semi-definite programming, which are further fed into the online diagnosis algorithm. By checking the dissipation rate against the threshold with process input-output trajectories, the fault can be diagnosed. Finally, the presented method is illustrated through an example of a heat exchanger.