Performances of data-driven prognostics approaches are closely dependent on form, and trend of extracted features. Indeed, features that clearly reflect the machine degradation, should lead to accurate prognostics, which is the global objective of the paper. This paper contributes a new approach for features extraction / selection: the extraction is based on trigonometric functions and cumulative transformation, and the selection is performed by evaluating feature fitness using monotonicity and trendability characteristics. The proposition is applied to time-frequency analysis of non-stationary signals using Discrete Wavelet Transform. The main idea is to map raw vibration data into monotonic features with early trends, which can be easily predicted. To show that, selected features are used to build a model with a data-driven approach namely, the Summation Wavelet-Extreme Learning Machine, that enables a good balance between model accuracy and complexity. For validation and generalization purpose, vibration data from two real applications of Prognostics and Health Management challenges are used: 1) cutting tools from Computer Numerical Control (CNC) machine (2010), and 2) bearings from platform PRONOSTIA (2012). Performances of the proposed approach are thoroughly compared with the classical approach by performing: feature fitness analysis, cutting tool wear "estimation" and bearings "long-term predictions" tasks, which validates the proposition.
Proton Exchange Membrane Fuel Cells (PEMFC) suffer from a limited lifespan, which impedes their uses at a large scale. From this point of view, prognostics appears to be a promising activity since the estimation of the Remaining Useful Life (RUL) before a failure occurs allows deciding from mitigation actions at the right time when needed. Prognostics is however not a trivial task: 1) underlying degradation mechanisms cannot be easily measured and modeled, 2) health prediction must be performed with a long enough time horizon to allow reaction. The aim of this paper is to face these problems by proposing a prognostics framework that enables avoiding assumptions on the PEMFC behavior, while ensuring good accuracy on RUL estimates. Developments are based on a particle filtering approach that enables including non-observable states (degradation through time) into physical models. RUL estimates are obtained by considering successive probability distributions of degrading states. The method is applied on 2 data sets, where 3 models of the voltage drop are tested to compare predictions. Results are obtained with an accuracy of 90 hours around the real RUL value (for a 1000 hours lifespan), clearly showing the significance of the proposed approach.
Fuel Cell systems (FC) represent a promising alternative energy source. However, even if this technology is close to being competitive, it is not ready for large scale industrial deployment: FC still must be optimized, particularly by increasing their limited lifespan. This involves a better understanding of wearing processes and requires emulating the behavior of the whole system. Furthermore, a new area of science and technology emerges: Prognostics and Health Management (PHM) appears to be of great interest to face the problems of health assessment and life prediction of FCs. According to this, the aim of this paper is to present the current state of the art on PHM of FCs, more precisely of Proton-Exchange Membrane Fuel Cells (PEMFC) stack. PHM discipline is described in order to depict the processing layers that allow early deviations detection, avoiding faults, deciding mitigation actions, and thereby increasing the useful life of FCs. On this basis, a taxonomy of existing works on PHM of PEMFC is given, highlighting open problems to be addressed. The whole enables getting a better understanding of remaining challenging issues in this area.
Integrating prognostics to a real application requires a certain maturity level and for this reason there is a lack of success stories about development of a complete Prognostics and Health Management system. In fact, the maturity of prognostics is closely linked to data and domain specific entities like modeling. Basically, prognostics task aims at predicting the degradation of engineering assets. However, practically it is not possible to precisely predict the impending failure, which requires a thorough understanding to encounter different sources of uncertainty that affect prognostics. Therefore, different aspects crucial to the prognostics framework, i.e., from monitoring data to remaining useful life of equipment need to be addressed. To this aim, the paper contributes to state of the art and taxonomy of prognostics approaches and their application perspectives. In addition, factors for prognostics approach selection are identified, and new case studies from component-system level are discussed. Moreover, open challenges toward maturity of the prognostics under uncertainty are highlighted and scheme for an efficient prognostics approach is presented. Finally, the existing challenges for verification and validation of prognostics at different technology readiness levels are discussed with respect to open challenges.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.