9509975 Lafortune The research covered by this proposal describes: (i) the development of a comprehensive methodology for failure diagnosis of large-scale complex dynamic systems modeled as discrete event systems; (ii) application areas where the methodology will be applied and tested. This research will build on previous work of the PIs which so far has been restricted to passive diagnosis of live languages,i.e., to live languages where no control commands are issued specifically for the purpose of diagnosing failures. In passive diagnosis, the following issues will be addressed; (i) what is the most detailed diagnosis of failures that one can achieve with a given set of sensors? (ii) given the desired diagnosability requirements of a system, what is a minimum set of sensors such that these diagnosability requirements are achieved? (iii) how is diagnosability of non-live languages defined and under what conditions is a non-live language diagnosable? This research will also address the problem of active diagnosis, which is the situation where control commands are issued specifically for the purpose of diagnosing failures. In active diagnosis, the objective will be the systematic construction of supervisors such that the resulting closed-loop system is diagnosable within the allowable delay limits. Several application areas where problems of failure diagnosis arise have been identified. In collaboration with colleagues from industry, discrete event models for failure diagnosis in these applications areas will be developed. These models will guide the theoretical research and will test the methodology to realistic systems. ***
