Fusion techniques for vibration and oil debris/quality in gearbox failure testing

Improved diagnosis is key towards the development of machinery prognostics. A conceptual approach to combine the oil and vibration feature data is reviewed and discussed. The methodology requires the correlation of failure mode phenomena with appropriate features in controlled transitional tests that reflect the evolution of machine failures. The Mechanical Diagnostics Test Bed described herein was designed and built to provide a means to acquire such transitional data for gearboxes. A series of tests with collected data from accelerometers, temperature, torque, speed, and oil debris measurements in cooperation with NRL, has been conducted. With in-situ inspection, the tests allowed the correlation of processed features with damage assessment as it occurred. Methods to fuse non-commensurate vibration and oil features are explored for the potential to provide better estimation of failure mode identification and severity. The identification of specific feature vectors to track damage and provide prognostic precursors are also discussed. Collecting the data and developing algorithms in this manner will help to accelerate advances in CBM.