Investigation of Condition Indicators, Operational Conditions and Gear Health State Using Data Mining Techniques

Damage progression tests were performed in the NASA Glenn Spiral Bevel Gear Fatigue Rig. Six gear sets with varying levels of tooth damage were tested, and vibration-based gear condition indicators, amount of debris generated, and oil temperatures were measured. The damage state was documented with photographs taken at inspection intervals throughout the test and was quantified with a numerical continuous damage factor. Condition indicator performance was first assessed with traditional methods, and then data-mining methods in the form of a clustering analysis were applied to the operational data and condition indicator data. This analysis was then fed into a decision tree model to predict the gear damage state. Results indicate gear health state can be determined with minimum knowledge of the dataset.