Abstract
Health assessment of electric motors is a research topic of high relevance in the area of structural mechanics. In the early days, the health state of an electric motor was mainly determined by empirical knowledge. But this paradigm is shifting to advanced methods of predicting the health of single components of an electric motor using its physical simulation models from the design phase. However, the process of creating the models to become usable during operation is laborious and in many cases no simulation or even 3D-CAD models from the design phase are available. This article focuses on a combination of a physics-based and data-driven estimation of the motor health, especially for motors where no information from the design phase is available. In particular, the advancements of the development of the hybrid fusion method moSAIc are presented. moSAIc allows to transfer the knowledge inherent in physical degradation models of motors to unknown derivatives. The experiments show that the accuracy and robustness of moSAIc is significantly better compared to results of earlier stages.
Zusammenfassung
Die Zustandsschätzung von Elektromotoren ist ein aktuell sehr relevantes Forschungsthema im Bereich der Strukturmechanik. Zur Zeit wird der mechanische Zustand eines Elektormotors hauptsächlich durch empirisches Wissen aus Normen oder von Experten bestimmt. Dieses Paradigma verschiebt sich aktuell zu fortgeschrittenen Methoden, die in der Lage sind, den Zustand einzelner Komponenten des Motors durch physikalische Simulationsmodelle aus der Entwurfsphase in Kombination mit Messungen zu bestimmen. Ein Simulationsmodell aus der Entwurfsphase betriebsparallel einsetzbar zu machen, erfordert sehr viel Aufwand, da teilweise keine Simulations- oder 3D-CAD-Modelle zur Verfügung stehen oder mathematische Ordnungsreduktionen notwendig sind. Dieser Artikel beschäftigt sich mit einer Kombination von physikalisch-basierten und datengetriebenen Methoden zur Zustandsschätzung von Elektromotoren, insbesondere für Motoren von denen keine Informationen aus der Entwurfsphase verfügbar sind. Speziell wird in dem Artikel der Fortschritt der hybriden Fusionsmethode moSAIc präsentiert, welche es erlaubt, Wissen aus bekannten physikalischen Degradationsmodellen auf unbekannte Motorderivate zu übertragen. Die Experimente veranschaulichen, dass die Genauigkeit und Robustheit von moSAIc signifikant besser ist im Vergleich zu vorherigen Ergebnissen.
About the authors
Christoph Bergs started in October 2016 as an external PhD student at the Institute of Industrial Information Technology of the Karlsruhe Institute of Technology and at Siemens AG. His research interests lie in the field of automation, modeling and simulation of dynamic systems with a special focus on their industrial application.
Mohamed Khalil started in February 2017 as an external PhD student at the Chair of Structural Analysis at TUM. His research is focused on incorporating simulation models in predictive maintenance activities, and in enhancing AI models.
Marcel Hildebrandt started as PhD student at Siemens AG and LMU in April 2017. His research is focused on machine learning with graph-structured data and industrial applications of AI.
Michael Heizmann is professor at the Institute of Industrial Information Technology at the Karlsruhe Institute of Technology. His research areas include machine vision, image processing, image and information fusion, measurement technology and their applications in industry.
Roland Wüchner is deputy head at the Chair of Structural Analysis at TUM. His main research is dedicated to finite element methods in non-linear structural mechanics, isogeometric analysis, wind engineering, coupled systems and fluid-structure interaction.
Kai-Uwe Bletzinger is the professor at the Chair of Structural Analysis at TUM. His research focus includes structural and CFD optimization, coupled simulation problems, computational wind engineering, computational mechanics for shells and membranes.
Volker Tresp is professor in informatics at the LMU and Distinguished Research Scientist at Siemens AG. His current research interests focus on Statistical Relational Learning, which combines machine learning with relational data models and first-order logic and enables machine learning in knowledge bases.
References
1. Y. Lei, N. Li, L. Guo, N. Li, T. Yan, and J. Lin. Machinery health prognostics: A systematic review from data acquisition to RUL prediction, Mechanical Systems and Signal Processing (104): 799–834, 2018.10.1016/j.ymssp.2017.11.016Search in Google Scholar
2. M. Hildebrandt, M. Khalil, C. Bergs, V. Tresp, R. Wuechner, K.-U. Bletzinger, and M. Heizmann. Remaining useful life estimation for unknown fleet derivatives using a hybrid modeling approach, 2019 IEEE 17th International Conference on Industrial Informatics (INDIN), 2019.10.1109/INDIN41052.2019.8972200Search in Google Scholar
3. C. Bergs, M. Heizmann, and H. Held. Hybrid modeling approaches with a view to model output prediction for industrial applications, 2018 IEEE 16th International Conference on Industrial Informatics (INDIN), 2018.10.1109/INDIN.2018.8471964Search in Google Scholar
4. C. Okoh, R. Roy, J. Mehnen, and L. Redding. Overview of remaining useful life prediction techniques in through-life engineering services, Procedia CIRP (16): 158–163, 2014.10.1016/j.procir.2014.02.006Search in Google Scholar
5. E. Santecchia, A. Hamouda, F. Musharavati, E. Zal-nezhad, M. Cabibbo, M. El Mehtedi, and S. Spigarelli. A review on fatigue life prediction methods for metals, Advances in Materials Science and Engineering, 2016.10.1155/2016/9573524Search in Google Scholar
6. C. Heinrich, M. Khalil, K. Martynov, and U. Wever. Online remaining lifetime estimation for structures, Mechanical Systems and Signal Processing (119): 312–327, 2019.10.1016/j.ymssp.2018.09.028Search in Google Scholar
7. X.-S. Si, C.-H. Wang, W. Hu, and D.-H. Zhou. Remaining useful life estimation – a review on the statistical data-driven approaches, European Journal of Operational Research, 2011.10.1016/j.ejor.2010.11.018Search in Google Scholar
8. X. Li, Q. Ding, and J.-Q. Sun. Remaining useful life estimation in prognostics using deep convolution neural networks, Reliability Engineering & System Safety (172): 1–11, 2018.10.1016/j.ress.2017.11.021Search in Google Scholar
9. Z. Li, D. Wu, C. Hu, and J. Terpenny. An ensemble learning-based prognostic approach with degradation-dependent weights for remaining useful life prediction, Reliability Engineering & System Safety, 2017.10.1016/j.ress.2017.12.016Search in Google Scholar
10. R. K. Singleton, E. G. Strangas, and S. Aviyente. Extended kalman filtering for remaining-useful-life estimation of bearings, IEEE Transactions on Industrial Electronics, 62(3): 1781–1790, 2015.10.1109/TIE.2014.2336616Search in Google Scholar
11. W. Ahmad, S. A. Khan, M. Islam, and J.-M. Kim. A reliable technique for remaining useful life estimation of rolling element bearings using dynamic regression models, Reliability Engineering & System Safety, 2018.10.1016/j.ress.2018.02.003Search in Google Scholar
12. L. Liao and F. Koettig. A hybrid framework combining data-driven and model-based methods for system remaining useful life prediction, Applied Soft Computing (44): 191–199, 2016.10.1016/j.asoc.2016.03.013Search in Google Scholar
13. M. Seera, C. P. Lim, D. Ishak, and H. Singh. Offline and online fault detection and diagnosis of induction motors using a hybrid soft computing model, Applied Soft Computing (12): 4493–4507, 2013.10.1016/j.asoc.2013.08.002Search in Google Scholar
14. C. Hu, B. D. Youn, and P. Wang. Ensemble of data-driven prognostic algorithms with weight optimization and k-fold cross validation, 2010 AMSE International Design Engineering Technical Conferences and Computers and Information in Engineering Conference, 2010.10.1115/DETC2010-29182Search in Google Scholar
15. V. Tresp and M. Taniguchi. Combining estimators using non-constant weighting functions, Advances in neural information processing systems: 419–426, 1995.Search in Google Scholar
16. R. A. Jacobs, M. I. Jordan, S. J. Nowlan, and G. E.Hinton. Adaptive mixtures of local experts, Neural computation 3(1): 79–87, 1991.10.1162/neco.1991.3.1.79Search in Google Scholar PubMed
17. O. C. Zienkiewicz, J.Z. Zhu, and Robert L. Taylor. The Finite Element Method: its Basis and Fundamentals, Butterworth-Heinemann, 7th Edition, 2013.Search in Google Scholar
18. Y.-L. Lee. Fatigue testing and analysis: theory and practice, Butterworth-Heinemann, Volume 13, 2005.Search in Google Scholar
19. E. Haibach. Betriebsfestigkeit, Springer, VDI, 2005.Search in Google Scholar
20. S. Bach, A. Binder, G. Montavon, F. Klauschen, K.-R. Müller, and W. Samek. On pixel-wise explanations for non-linear classifier decisions by layer-wise relevance propagation. PloS one 10(7): 130–140, 2015.10.1371/journal.pone.0130140Search in Google Scholar PubMed PubMed Central
© 2019 Walter de Gruyter GmbH, Berlin/Boston
