Qualitative Event-Based Diagnosis: Case Study on the Second International Diagnostic Competition
Matthew J. Daigle, Indranil Roychoudhury · 2010
We describe a diagnosis algorithm entered into the Second International Diagnostic Competition. We focus on the first diagnostic problem of the in-dustrial track of the competition in which a diag-nosis algorithm must detect, isolate, and identify faults in an electrical power distribution testbed and provide corresponding recovery recommen-dations. The diagnosis algorithm embodies a model-based approach, centered around quali-tative event-based fault isolation. Faults pro-duce deviations in measured values from model-predicted values. The sequence of these de-viations is matched to those predicted by the model in order to isolate faults. We augment this approach with model-based fault identifica-tion, which determines fault parameters and helps to further isolate faults. We describe the diag-nosis approach, provide diagnosis results from running the algorithm on provided example sce-narios, and discuss the issues faced, and lessons learned, from implementing the approach. 1