Diagnosability of faults using finite-state automaton model

Abstract

This paper analyzes the diagnosability of faults in a plant modelled by Finite State Automaton Tables (FATs). The state-space is partitioned by choosing a set of boundaries for each state variable. Given the set of ordinary differential equations of the plant, it is possible to generate a Finite-State Automaton Table (FAT) representation of the plant that describes possible discrete state transitions under normal and under fault conditions. The FAT serves as input to a fault detection and isolation algorithm. In this paper, we extend our earlier work to propose a definition of fault diagnosability and analyze the nondiagnosable circumstance. The proposition is illustrated with an example.