Fuzzy Model Optimization Using Genetic Algorithm for Aircraft Engine Diagnosis

All Papers

An accurate and up-to-date diagnostic model is critical for economic aircraft engine operation. However, for many commercial airline fleets, monitoring and diagnosing engine faults is often left to human operators due to lack of effective modeling. Individuals must manually interpret engine performance parameters and this results in inconsistent evaluation and the potential for error. A recent work has been performed to capture the knowledge of diagnostic engineers and apply fuzzy logic to engine fault classification. The developed fuzzy model is capable of capturing various engine fault signatures based on expert experience, and it provides accurate diagnosis by assessing the similarity between those fault signatures and observed trends in sensor data. The major drawback with this approach is the manual knowledge extraction and model maintenance process. Manual model tuning is not only labor intensive, but it also brings another source of inconsistency to the overall system performance. In this paper, we will present a hybrid approach to augment the existing expert knowledge-based diagnostic model with automatic learning capability using a genetic algorithm. The proposed approach not only allows for automatic model tuning, but also enables the diagnostic model to adapt throughout the engine service life as operating conditions change.