With the growing sophistication and complexity of man-made systems, it is becoming increasingly difficult to detect and isolate system faults. We propose, in this work, a solution to the failure diagnosis challenge. Our approach makes use of discrete event system theory and introduces structures called diagnosing supervisors. These structures generate the sequences of events necessary for fault detection and isolation. As well, these supervisors satisfy other control objectives.
We first explain the idea of diagnosability and then provide the theoretical foundations
for diagnosing supervisors. Two software programs have been developed for this thesis.
We supply tutorial guides for their use, as well as details of this algorithms. Next,
examples are given which illustrate our fault diagnosis approach. Following the
examples, the results of a simulation, which show the accuracy and reliability of our
method, are provided. Conclusions are drawn and future research areas are recommended.