The proposed research seeks to provide a comprehensive framework for how to collect and assess degradation data over the product lifecycle from design to usage phases. The proposed framework will enhance both data-driven modeling and physics-based modeling techniques outlined in previous prognostic research. By using the coupled model scheme, the requirements for data quality and complete product design information can be reduced. The scheme can be validated on a critical rotary machinery component, such as a bearing, as well as for more complex components and structures, such an electrical vehicle battery.
The proposed work on developing a combined data driven and physics driven model for fault prognostication has the potential to yield significant cost reduction in manufacturing systems and processes. The work is supported by the Industry Advisory Board as well as individual industry members of the center and has the potential to extend the centers portfolio. The center has actively involved minority and female students in its research through Research Experiences for Teachers and Undergraduates (RET and REU) projects and will continue to do so through the proposed effort.
(PHM) research. The research efforts have focused on developing new modeling methodology and algorithms to discover new knowledge that includes the physics-based, systematic sensor selection process and virtual sensors that ensure the correct and accurate prognostics of the system’s health. Intellectual Merits By using the coupled model scheme, the requirements for data quality and complete product design information can be reduced to a minimum. The scheme can be validated on a critical rotary machinery component, such as a bearing, as well as for more complex components and structures, such an electrical vehicle battery. Thus, this research can be a cornerstone toward the goal of rapid, low-cost prognostics and maintenance. Two novel modeling methods and algorithms have been developed and validated: (1) Dual Extended Kalman Filter model (DEKF); and (2) Support Regression Vector-Particle Filter model (SVR-PF). The Dual Extended Kalman Filter (DEKF) model has been developed for SOC estimation by using two cooperating extended Kalman filters, where the first one is responsible for estimating the SOC while the second one estimates the cell parameters indicating the level of cell deterioration due to aging. The dual filter combination is capable of tuning Kalman gains and providing accurate estimates even when the dynamics of the parameters change as the cell ages (e.g., inner resistance, capacity). By comparing with the experimental data, the results of the DEKF method have shown an efficient SOC estimation with quick convergence and robust estimation of parameter changes in the long run. The SVR-PF model has been developed with novel degradation parameters introduced to determine battery health in real time. Moreover, new approach of RUL prediction has been investigated, which is able to provide the RUL value and update the RUL probability distribution to the latest time step. Results for both methods have shown that the SVR-PF has good monitoring and prediction capability. We use lithium-ion batteries as our primary testbed to develop, implement and validate the methods. Li-ion batteries have been widely used in clean vehicle and aerospace applications, which require very accurate performance assessment in real time and early warning of unacceptable capacity fade. We also extend the model to Nickel-hydrogen (NiH2) battery which has less known physical mechanism of degradation and capacity loss. The prediction model has shown to reduce the false alarm rate of the existing battery online monitoring system and improved accuracy of state-of-health prediction. Broader Impacts This research has addressed a significant question raised by the IMS Center’s industrialmembers and research partners. For many manufacturing companies, some systems are from thirdparties, of which the physics-based model is difficult to realize; for other situations, the data might not beavailable. The development and implementation of prognostics methodology can take the best advantageof any information from the physics knowledge and data either collected from the practical usage orgenerated by the virtual coupled models. The developed coupled model techniques can be applied to the various practical PHM applications, such as electrical vehicle batteries and renewable energy windturbine applications. As inspired by relevant research in other fields such as data-mining, biostatistics, and signalprocessing, the outcomes of this research could potentially impact those fields as well. During this project, IMS center has actively involved minority and female graduate students and has provided Research Experiences for Undergraduates (REU) projects.
