a transformation schema to generate

a transformation schema to generate classifier structures thatcan be matched against existing reference model structures, thusproviding the experts a better understanding of the implicationsof adding new knowledge and detectors to the reference model.Second, we use real flight data to validate the new reference modelstructure bydetermining the improvementsin diagnostic accuracyand timeliness of isolation using well-defined metrics. Our overallapproach shows promise for targeted fault analysis that may leadto faster detection, and, therefore, avoidance of adverse eventssuch as an engine shutdown during flight. However, the task ofstudying and refining large, centralized reference models foraircraft systems is complex, especially for quantifying diagnosticaccuracy and false alarm rates across multiple fault modes. Wewill address this larger task along with detection of previouslyundetected faults (anomaly detection) in future work.Index Terms—Aviation safety, classification algorithms, diag-nosis, knowledge engineering, tree augmented Bayesian networks(TANs).