A Structural Approach to Sensor Placement based on Symbolic Compilation of the Model

In the paper we address the problem of computing the Minimal Additional Sensor Sets (MASS) that guarantee a desired level of diagnostic discrimination for a system. The main contribution of this paper is the extension and the adaptation of techniques based on the symbolic compilation of qualitative system models to a structural approach suitable for the computation of MASS for component-oriented models consisting of sets of numerical equations. In this respect, the paper can be viewed as a bridge across the AI approaches to model-based sensor placement and the Fault Detection and Isolation approaches developed by the Automatic Control community. We show that the resulting method exploits the symbolic compilation techniques not only as a way to provide computational savings (including some theoretical guarantees on the computational complexity), but it also exhibits interesting new features, most notably the handling of multiple faults.