Jump to ContentJump to Main Navigation
Show Summary Details
More options …

tm - Technisches Messen

Plattform für Methoden, Systeme und Anwendungen der Messtechnik

[TM - Technical Measurement: A Platform for Methods, Systems, and Applications of Measurement Technology
]

Editor-in-Chief: Puente León, Fernando / Zagar, Bernhard

12 Issues per year


IMPACT FACTOR 2017: 0.476

CiteScore 2017: 0.46

SCImago Journal Rank (SJR) 2017: 0.239
Source Normalized Impact per Paper (SNIP) 2017: 0.566

Online
ISSN
2196-7113
See all formats and pricing
More options …
Volume 81, Issue 5

Issues

Automated fault detection using deep belief networks for the quality inspection of electromotors

Automatische Fehlerdetektion mittels Deep Belief Netzwerken zur Qualitätskontrolle von Elektromotoren

Jianwen Sun / Reto Wyss / Alexander Steinecker / Philipp Glocker
Published Online: 2014-04-28 | DOI: https://doi.org/10.1515/teme-2014-1006

Abstract

Vibration inspection of electro-mechanical components and systems is an important tool for automated reliable online as well as post-process production quality assurance. Considering that bad electromotor samples are very rare in the production line, we propose a novel automated fault detection method named “Tilear”, based on Deep Belief Networks (DBNs) training only with good electromotor samples. Tilear consctructs an auto-encoder with DBNs, aiming to reconstruct the inputs as closely as possible. Tilear is structured in two parts: training and decision-making. During training, Tilear is trained only with informative features extracted from preprocessed vibration signals of good electromotors, which enables the trained Tilear only to know how to reconstruct good electromotor vibration signal features. In the decision-making part, comparing the recorded signal from test electromotor and the Tilear reconstructed signal, allows to measure how well a recording from a test electromotor matches the Tilear model learned from good electromotors. A reliable decision can be made.

Zusammenfassung

Die Analyse von Vibrationssignalen zur Fehelerdetektion elektromechanischer Komponenten und Systeme stellt ein wichtiges Werkzeug in zuverlässiger und automatischer Qualitätssicherung des Produktionsprozesses dar. Davon ausgehend, dass fehlerhafte Elektromotoren nur einen geringen Anteil einer Charge ausmachen, schlagen wir einen neuen Inspektionsansatz namens ,,Tilear” vor. Dieser Ansatz basiert auf einem Deep Belief Netzwerk (DBN), welches mit unterschiedlichen Signalmustern guter Elektromotoren trainiert wurde. Tilear generiert einen Auto-Encoder mittels DBNs mit dem Ziel, die Eingangssignale so genau wie möglich zu rekonstruieren. Tilear besteht aus zwei Teilen: (i) Training und (ii) Entscheidung. In der Trainingsphase wird Tilear nur mit Vibrationssignalen guter Motoren angelernt. Auf diese Weise kann Tilear ausschliesslich Signalmuster rekonstruieren, die sich einem guten Motor zuordnen lassen. In der nachfolgenden Entscheidungsphase wird ein aktuelles Vibrationsmuster mit der entsprechenden Rekonstruktion von Tilear verglichen. Auf diese Weise wird die Abweichung vom idealen, vorab gelernten Motorsignal quantifiziert und kann für eine Entscheidung in der Qualitätskontrolle verwendet werden.

Keywords: Electromotor; fault detection; deep belief networks; vibration signals; non-desctructive testing; online quality inspection

Schlagwörter: Elektromotor; Fehlerdetektion; Deep Belief Netzwerke; Vibrationssignal; zerstörungsfreie Prüfung; Echtzeit Qualitätskontrolle

About the article

Jianwen Sun

Jianwen Sun is a PhD student working at Microassembly & Robotics group, CSEM Alpnach. He is Also affiliated to Institute of Neuroinformatics, University/ETH Zürich, Switzerland. His main research area includes signal representation and processing, deep learning theory, and quality inpsection using machine learning techniques.

Microassembly & Robotics, Untere Gründlistrasse 1, CSEM Alpnach, 6005 Alpnach Dorf, Switzerland

Reto Wyss

Reto Wyss received his PhD degree from Institute of Neuroinformatics, University / ETH Zürich, Switzerland. His main research area includes invariant pattern recognition and navigation in autonomous agents. He is currently the CEO of ViDi Systems SA offering a full range of ground breaking vision tools and services.

ViDi Systems SA, Z. I. du Vivier 22, 1690 Villaz-St-Pierre, Switzerland

Alexander Steinecker

Alexander Steinecker is a business development manager at CSEM. He is a physicist and received the PhD degree (Dr. rer. nat.) in natural sciences from University of Bonn, Germany. For CSEM he is in charge of strategic positioning and industrializing CSEM's technology platforms in the microrobotics and packaging domain. From his former experience as R&D engineer he gained a profound expertise in delivery of advanced solutions in robotics and automation to industry.

Microassembly & Robotics, Untere Gründlistrasse 1, CSEM Alpnach, 6005 Alpnach Dorf, Switzerland

Philipp Glocker

Philipp Glocker is the section head of Microassembly & Robotics groupd consisting of 8 R&D Engineers, CSEM Alpnach. As an electrical engieer with an MBA degree, he received the Diploma degree in electrical engineering from ETH Zürich. He has a background in industrial R&D for automation and robotics, and has been dealing with all major production leaders like Siemens, Bosch, etc within his 25 years experience in the machinery industry. He has been working for CSEM Alpnach since 2007. He is also coordinating the national R&D initiatives.

Microassembly & Robotics, Untere Gründlistrasse 1, CSEM Alpnach, 6005 Alpnach Dorf, Switzerland


Accepted: 2013-11-14

Received: 2013-10-24

Published Online: 2014-04-28

Published in Print: 2014-05-28


Citation Information: tm - Technisches Messen, Volume 81, Issue 5, Pages 255–263, ISSN (Online) 2196-7113, ISSN (Print) 0171-8096, DOI: https://doi.org/10.1515/teme-2014-1006.

Export Citation

©2014 Walter de Gruyter Berlin/Boston.Get Permission

Citing Articles

Here you can find all Crossref-listed publications in which this article is cited. If you would like to receive automatic email messages as soon as this article is cited in other publications, simply activate the “Citation Alert” on the top of this page.

[2]
Elina Stoffel, Anton S. Becker, Moritz C. Wurnig, Magda Marcon, Soleen Ghafoor, Nicole Berger, and Andreas Boss
European Journal of Radiology Open, 2018, Volume 5, Page 165
[3]
Tim de Bruin, Kim Verbert, and Robert Babuska
IEEE Transactions on Neural Networks and Learning Systems, 2017, Volume 28, Number 3, Page 523

Comments (0)

Please log in or register to comment.
Log in