Hybrid Methods for PHM
Abstract
Utilization of Condition Based Maintenance (CBM) within industry and the defense community is increasing as data becomes more readily available and methods of analysis and machine learning improve. CBM is progressing from early implementations of Reliability Centered Maintenance (RCM), focused on improving preventative maintenance schedules, toward robust Prognostics Health Monitoring (PHM) capabilities. Development of a-priori PHM models can be time consuming and expensive; therefore, data collection capability is often deployed with a basic level of PHM capability. This research will develop hybrid PHM models that integrate knowledge-based models and physics-based models, focused on gear failures and a second TBD subsystem.The hybrid gear PHM research objectives are to: 1) empirically investigate gradual progression of damage in spur gears; 2) develop in-line methods for assessing damage levels using oil andvibration monitoring; 3) empirically investigate fatigue cracking; 4) develop crack-propagation data-driven models; and 5) develop diagnostics capability to differentiate between wear and fatigue failure. The research approach includes collection of oil and vibration data on multiple long-term mildly accelerated tests, image and material science based analysis of gear contact surfaces, analysis of metal particles in the oil, development of data driven models for damage assessment, fusion of sensor data to enhance predictive maintenance models, and comparison of wear vs. fatigue failure indicators. The anticipated outcomes are: 1) fault detection for tooth breakage and onset of wear; 2) diagnostics between wear and fatigue-based tooth breakage; 3) wear damage assessment utilizing vibration and oil monitoring sensors; 4) tooth breakage assessment and prognostics; and 5) fully documented datasets, equipped with data analyses and modeling tools for the gear research community.The second subsystem PHM research objectives are to: 1) develop approaches for integration of system performance, fault detection, diagnostics, and propagation models; 2) develop physics-based models associated with system operations; 3) create datasets to enable data-driven models and physics-based models; 4) identify existing physics of failure models; 5) characterize the physics of failure, damage assessment, and prognostics models; 6) develop approaches for integrating reparability considerations into RCM analysis and PHM algorithms; and 7) evaluate system level PHM models and algorithms on a low cost GPU based processor. The research approach includes development of a failure modes analysis as part of an RCM approach, augmented by functional and reliability block diagrams; development of a data collection plan for parametrization of the physics-based models and development of data driven models, as well as collection of seeded fault data for parameterizing physics-of-failure models and data-driven damage assessment and remaining useful lifetime models. The anticipated outcomes are: 1) hybrid PHM methods that include considerations for protecting reparability of components; 2) design considerations for implementation of system levelPHM in a GPU based processor; and 3) a fully-documented dataset, accompanied by data analysis and model development tools for the research community.As CBM+ has been mandated by the DoD, this research has direct impacts to the DoD community. The approaches researched and developed in this project are aimed at providing methods to develop PHM capability over time. Models may be developed and integrated as faults occur and the models mature. Initial capability based on physics based models or reliability data may be integrated with data driven approaches to further enhance and predictive outcomes.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jan 12, 2023
- Source ID
- N000142312035
Entities
People
- Michael Thurston
Organizations
- Office of Naval Research
- Rochester Institute of Technology
- United States Navy