Skip to content
Licensed Unlicensed Requires Authentication Published by De Gruyter (O) May 27, 2021

Probabilistic model-based fault diagnosis for the cavities of the European XFEL

Probabilistische, modellbasierte Fehlerdiagnose für die Kavitäten des European XFEL
  • Ayla Nawaz

    Ayla Nawaz received her M. Sc. in Electrical Engineering from the Hamburg University of Technology in 2015. During her master she also started working on parameter estimation projects for the Deutsches Elektronen-Synchrotron (DESY), Hamburg. Since 2016 she is working as a research associate for DESY and the University of Lübeck, interested in fault diagnosis, probabilistic modeling and the handling of large data sets.

    EMAIL logo
    , Christian Herzog né Hoffmann

    Christian Herzog, né Hoffmann, studied mechatronics at Hamburg University of Technology (TUHH) from 2008–2011 and received his Ph. D. in 2015 studying nonlinear and distributed control with the Institute of Control Systems, TUHH. Since 2015 he is a tenured researcher at the Institute for Electrical Engineering in Medicine at the University of Lübeck and active in probabilistic algorithms in systems and control. He is also an active researcher in the domains of engineering ethics, participative approaches, co-creation as well as responsible research and innovation.

    , Jan Graßhoff

    Jan Graßhoff received the B. Sc. and M. Sc. degrees in computer science from Universität zu Lübeck, Germany, in 2014 and 2016, respectively, where he is currently pursuing the Ph. D. degree. For the master’s degree thesis, he was with the Research Unit, Dräger Medical, Germany, where he was working on parameter and state estimation problems. Since 2016, he has been working as a Research Associate with the Institute for Electrical Engineering, University of Lübeck. His research interests include probabilistic signal processing and estimation problems in biomedical applications.

    , Sven Pfeiffer

    Sven Pfeiffer studied electrical engineering at the Hamburg University of Technology (TUHH) from 2004–2010. He received his Ph. D. in 2014 on model-based feedback and feedforward control methods for linear accelerators. Since then he is with the Deutsches Elektronen-Synchrotron, where he is part of the low-level radio frequency group and is responsible for the design, development, and implementation of control schemes for the European XFEL and other DESY facilities.

    , Gerwald Lichtenberg

    Gerwald Lichtenberg is a Professor of Physics and Control Engineering at the Faculty of Life Sciences at the HAW Hamburg. His research areas are in the fields of model-based and learning control as well as fault diagnosis of complex systems such as local energy networks, building systems, or particle accelerators. The methodological focus of his work is on tensor decomposition methods and multilinear models.

    and Philipp Rostalski

    Philipp Rostalski is a Professor for Electrical Engineering in Medicine and head of the corresponding institute at the University of Lübeck. His research is focused on model- and learning-based methods in signal processing and control for safety critical systems. The primary application domains are biomedical and autonomous systems.

Abstract

The European X-ray Free Electron Laser (EuXFEL) is a complex system with many interconnected components and sensor measurements. We use factor graphs to systematically design a probabilistic fault diagnosis method for its cavity system. This approach is expandable to further subsystems and considers uncertainties from measurements and modeling. After representing a model of the cavity system in the factor graph framework, we infer marginal distributions, e. g., of the fault classes using tabulated message-passing definitions. The emerging fault diagnosis method consists of an unscented Kalman filter-based residual generator and an evaluation of the residuals using a Gaussian mixture model. We include message-passing definitions for the training of the Gaussian Mixture Model from noisy data using the expectation-maximization algorithm.

Zusammenfassung

Der European X-ray Free Electron Laser (EuXFEL) ist ein komplexes System mit einer großen Anzahl zusammenhängender Komponenten und Messungen. In dieser Arbeit verwenden wir Faktorgraphen zum Entwurf probabilistischer Fehlerdiagnosemethoden für die Resonatoren des EuXFEL. Dieser Ansatz ist leicht auf weitere Systeme erweiterbar und Unsicherheiten in den Messungen oder der Modellierungen werden berücksichtigt. Ein Resonatormodell wird als Faktorgraph definiert und Randverteilungen werden mit Hilfe tabellarischer Berechnungsvorschriften für Nachrichten berechnet. Die entstehende Fehlerdiagnosemethode besteht aus dem Unscented Kalman-Filter zur Generierung von Residuen und einer Auswertung der Residuen mithilfe einer Gaußschen Mischverteilung. Wir fügen Definitionen für das Training der Gaußschen Mischverteilung aus verrauschten Daten unter Verwendung des Expectation-Maximization Algorithmus hinzu.

About the authors

Ayla Nawaz

Ayla Nawaz received her M. Sc. in Electrical Engineering from the Hamburg University of Technology in 2015. During her master she also started working on parameter estimation projects for the Deutsches Elektronen-Synchrotron (DESY), Hamburg. Since 2016 she is working as a research associate for DESY and the University of Lübeck, interested in fault diagnosis, probabilistic modeling and the handling of large data sets.

Christian Herzog né Hoffmann

Christian Herzog, né Hoffmann, studied mechatronics at Hamburg University of Technology (TUHH) from 2008–2011 and received his Ph. D. in 2015 studying nonlinear and distributed control with the Institute of Control Systems, TUHH. Since 2015 he is a tenured researcher at the Institute for Electrical Engineering in Medicine at the University of Lübeck and active in probabilistic algorithms in systems and control. He is also an active researcher in the domains of engineering ethics, participative approaches, co-creation as well as responsible research and innovation.

Jan Graßhoff

Jan Graßhoff received the B. Sc. and M. Sc. degrees in computer science from Universität zu Lübeck, Germany, in 2014 and 2016, respectively, where he is currently pursuing the Ph. D. degree. For the master’s degree thesis, he was with the Research Unit, Dräger Medical, Germany, where he was working on parameter and state estimation problems. Since 2016, he has been working as a Research Associate with the Institute for Electrical Engineering, University of Lübeck. His research interests include probabilistic signal processing and estimation problems in biomedical applications.

Sven Pfeiffer

Sven Pfeiffer studied electrical engineering at the Hamburg University of Technology (TUHH) from 2004–2010. He received his Ph. D. in 2014 on model-based feedback and feedforward control methods for linear accelerators. Since then he is with the Deutsches Elektronen-Synchrotron, where he is part of the low-level radio frequency group and is responsible for the design, development, and implementation of control schemes for the European XFEL and other DESY facilities.

Gerwald Lichtenberg

Gerwald Lichtenberg is a Professor of Physics and Control Engineering at the Faculty of Life Sciences at the HAW Hamburg. His research areas are in the fields of model-based and learning control as well as fault diagnosis of complex systems such as local energy networks, building systems, or particle accelerators. The methodological focus of his work is on tensor decomposition methods and multilinear models.

Philipp Rostalski

Philipp Rostalski is a Professor for Electrical Engineering in Medicine and head of the corresponding institute at the University of Lübeck. His research is focused on model- and learning-based methods in signal processing and control for safety critical systems. The primary application domains are biomedical and autonomous systems.

References

1. Baoping Cai, Lei Huang and Min Xie. “Bayesian networks in fault diagnosis.” IEEE Transactions on Industrial Informatics 13.5 (2017), 2227—2240.10.1142/11021Search in Google Scholar

2. Ali Abdollahi. “Optimization and Bayesian Inference in Model-Based Decision Making.” Doctoral dissertation. University of Connecticut, 2017.Search in Google Scholar

3. David M. Blei. “Build, compute, critique, repeat: Data analysis with latent variable models.” Annual Review of Statistics and Its Application 1 (2014), 203–232.10.1146/annurev-statistics-022513-115657Search in Google Scholar

4. Marco Cox, Thijs van de Laar and Bert de Vries. “A factor graph approach to automated design of Bayesian signal processing algorithms.” International Journal of Approximate Reasoning 104 (2019), 185–204.10.1016/j.ijar.2018.11.002Search in Google Scholar

5. Daphne Kollerand Nir Friedman. Probabilistic Graphical Models: Principles and Techniques. MIT Press, 2009.Search in Google Scholar

6. Hans-Andrea Loeliger et al.“The factor graph approach to model-based signal processing.” Proceedings of the IEEE 95.6 (2007), 1295–1322.10.1109/JPROC.2007.896497Search in Google Scholar

7. K.P. Murphy. Machine Learning: A Probabilistic Perspective. Adaptive Computation and Machine Learning Series. MIT Press, 2012. ISBN: 9780262018029.Search in Google Scholar

8. Gregory Provan. “A graphical framework for stochastic model-based diagnosis.” In: 3rd Conference on Control and Fault-Tolerant Systems (SysTol), 2016. IEEE, 2016, 566–571.10.1109/SYSTOL.2016.7739809Search in Google Scholar

9. Massimo Altarelli et al.“The European X-ray free-electron laser technical design report.” In: DESY 97.2006 (2006).Search in Google Scholar

10. G. Geloni et al.“Coherence properties of the European XFEL.” New Journal of Physics 12.3 (Mar. 2010), 035021.10.1088/1367-2630/12/3/035021Search in Google Scholar

11. Thomas Schilcher. “Vector Sum Control of Pulsed Accelerating Fields in Lorentz Force Detuned Superconducting Cavities.” Doctoral dissertation. Universitat Hamburg, 1998.Search in Google Scholar

12. V. Ayvazyan et al.“Superconducting cavity quench detection and prevention for the European XFEL.” In: 16th International Conference on RF Superconductivity. DESY-2014-00617. HF-Technik Linearbeschleuniger, 2013.Search in Google Scholar

13. Frank R. Kschischang, Brendan J. Frey and H.-A. Loeliger. “Factor graphs and the sum-product algorithm.” IEEE Transactions on Information Theory 47.2 (2001), 498–519.10.1109/18.910572Search in Google Scholar

14. John Winn and Christopher M. Bishop. “Variational message passing.” Journal of Machine Learning Research 6.Apr (2005), 661–694.Search in Google Scholar

15. Thomas P. Minka. “Expectation propagation for approximate Bayesian inference.” In: Proceedings of the Seventeenth Conference on Uncertainty in Artificial Intelligence. Morgan Kaufmann Publishers Inc., 2001, 362–369.Search in Google Scholar

16. Justin Dauwels, Sascha Korl and Hans-Andrea Loeliger. “Particle methods as message passing.” In: 2006 IEEE International Symposium on Information Theory. IEEE, 2006, 2052–2056.10.1109/ISIT.2006.261910Search in Google Scholar

17. Christian Herzog, Eike Petersen, Philipp Rostalski, et al. “Iterative approximate nonlinear inference via Gaussian message passing on factor graphs.” IEEE Control Systems Letters, 2019.Search in Google Scholar

18. Rolf Isermann. Fault-Diagnosis Systems: An Introduction from Fault Detection to Fault Tolerance. Springer, Berlin and New York, 2006.10.1007/3-540-30368-5Search in Google Scholar

19. Onno Zoeter and Tom Heskes. “Change point problems in linear dynamical systems.” Journal of Machine Learning Research 6.Dec (2005), 19992026.Search in Google Scholar

20. Dorothea Kolossa et al.“Independent component analysis and time-frequency masking for speech recognition in multitalker conditions.” EURASIP Journal on Audio, Speech, and Music Processing 2010.1 (2010), 651420.10.1186/1687-4722-2010-651420Search in Google Scholar

21. Christopher M. Bishop. Pattern Recognition and Machine Learning. Springer Science & Business Media, 2006.Search in Google Scholar

22. Andrew R. Runnalls. “Kullback-Leibler approach to Gaussian mixture reduction.” IEEE Transactions on Aerospace and Electronic Systems 43.3 (2007), 989–999.10.1109/TAES.2007.4383588Search in Google Scholar

23. Arthur P. Dempster, Nan .M Laird and Donald B. Rubin. “Maximum likelihood from incomplete data via the EM algorithm.” Journal of the Royal Statistical Society. Series B (Methodological) 39.1 (1977), 1–38.10.1111/j.2517-6161.1977.tb01600.xSearch in Google Scholar

24. Alexey Ozerov, Mathieu Lagrange and Emmanuel Vincent. “Uncertainty-based learning of Gaussian mixture models from noisy data.” Computer Speech and Language 27.3 (2013), 874–894.10.1016/j.csl.2012.07.002Search in Google Scholar

25. J. Dauwels, S. Korl and H. Loeliger. “Expectation maximization as message passing.” In: Proceedings of International Symposium on Information Theory, 2005, 583–586.10.1109/ISIT.2005.1523402Search in Google Scholar

Received: 2020-04-23
Accepted: 2020-10-16
Published Online: 2021-05-27
Published in Print: 2021-06-25

© 2021 Walter de Gruyter GmbH, Berlin/Boston

Downloaded on 23.2.2025 from https://www.degruyter.com/document/doi/10.1515/auto-2020-0064/html
Scroll to top button