Skip to content
Licensed Unlicensed Requires Authentication Published by Oldenbourg Wissenschaftsverlag September 13, 2016

Advanced multi-sensory process data analysis and on-line evaluation by innovative human-machine-based process monitoring and control for yield optimization in polymer film industry

Multisensorische Prozessdatenanalyse und schritthaltende Auswertung mittels innovativer, Mensch-Maschine-basierter Prozessbeobachtung und -führung für die Ausbeuteoptimierung in der Polymerfilmindustrie
Michael Kohlert

Michael Kohlert received the Diploma in economic & electrical engineering at the Institute of Integrated Sensory Systems, TU Kaiserslautern, Germany in 2010, respectively. He made his PhD at the Institute of Integrated Sensory Systems, TU Kaiserslautern, Germany in 2015. He is responsible for the department of IT & Automation at the Mondi Gronau GmbH, a polymer film plant. His research interests include Advanced Automation/Manufacturing, Industrie 4.0, Multi-sensory systems, Human-Machine-Interface, Computational Intelligence, Machine Learning.

IT & Automation, Mondi Gronau GmbH, Gronau, Germany

EMAIL logo
and Andreas König

Andreas König received the Diploma in Electrical Engineering at TU Darmstadt in 1990 and the PhD at the Institute of Microelectronics Systems of the same university in 1995 in the field of intelligent vision system design for inspection tasks and of neural hardware. He then was with Fraunhofer IITB (now IOSB), in Karlsruhe continuing research in visual industrial inspection and aerial image analysis. He moved to TU Dresden, and from 1996 served as Assistant Prof. of Electronic Devices in the Electrical Engineering dept., and established a group in the field of integrated intelligent vision sensors, circuits, and systems design and application. In 2002, he was appointed deputy chair of the computer engineering chair, computer science dept. at TU Chemnitz. In 2003, he was appointed Prof. for Integrated Sensory Systems in the Electrical Engineering dept., TU Kaiserslautern. He established the Institute of Integrated Sensory Systems with research on heterogeneous integrated, intelligent multimodal sensor, circuit, and system design. Activities comprise reconfigurable or adaptive self-x systems, neural and evovable hardware, wireless, autonomous/self-sufficient sensors and sensor swarms or networks, CMOS and MEMS design and 3D integration, automation and synthesis activities, multidimensional signal processing and sensor fusion, and interactive large high-dimensional data analysis and visualization. He is author or coauthor of more than 120 technical publications in the field, he was chair/co-chair of int. conferences in the field, e.g., HIS 2007 and KES 2011 at TU Kaiserslautern, he is Senior Member of IEEE, founder and chair of the IEEE CIS German Chapter, member of MIR Labs, DHV, AHMT, the interest group on microelectronics for neural networks, and of several editorial boards.

Institute of Integrated Sensor Systems, TU Kaiserslautern, Kaiserslautern, Germany

From the journal tm - Technisches Messen

Abstract

High material waste in the order of more than 1000 millions of Euro/year in polymer film industry provides an economic as well as environmental incentive for manufacturing optimization in the polymer film industry. Advanced complex industry processes from microelectronics to pharmaceutical industries provide huge datasets (big data) from heterogeneous multi-sensory monitoring. Process optimization, energy efficiency, yield optimization by higher data analysis, e.g., as in microelectronics could be transfered to polymer fields. Inspired by Industry 4.0, e.g., big data method approaches, condition monitoring, recommendation or human-machine interaction should provide options to be introduced in this way. Analytical tools are available to support manufacturers in quality and yield optimization, for real-time support. As a research vehicle for the development of methods for efficient process interfacing, a particular polymer film process was investigated with focus on novelty, and anomaly detection. A process line with 160 sensory channels has been monitored for several months. 21.900 process datasets of normal condition samples consist of about 160 dimensions were investigated. Accuracies of 99% were achieved, and a first prototype of a condition monitoring GUI for process recommendation was conceived. The results now allow process problem prediction in advance of occurrence. In future work, a broadening of the approach to other production steps and lines as well as methodological improvements starting from the sensor level with a focus towards intelligent condition conitoring and self-x properties will be pursued.

Zusammenfassung

Hohe Ausschussmengen von mehr als 1000 Millionen €/Jahr in der Kunststofffolienindustrie sind ein steigender Einflussfaktor und Optimierungsanreiz im globalen Wettbewerb, sowohl aus ökonomischen als auch aus Nachhaltigkeitaspekten. Die Verfahrenstechnik stößt an ihre Grenzen mit den bislang verwendeten Methoden der Prozessanalyse. Aus dem Feld der Halbleiterindustrie und Pharmazeutik werden aus diesem Grund Anwendungen übernommen, deren Einsatzgebiet sich mit der Analyse von großen Datenmengen an Maschinen beschäftigt. Beeinflusst vom Industrie-4.0-Charakter, wie Big-Data-Methoden, Echtzeit-Überwachung, Mensch-Maschine-Interaktion und Empfehlungswesen, können hier Anwendungspotenziale erkannt und umgesetzt werden. Für die Anwendung im polymeren Umfeld zur Ausschuss- und Kostenreduktion wurde solch ein Gebiet untersucht und implementiert. Der Extrusionsprozess wurde dazu analysiert mit Blick auf Neuheits- und Anomaliedetektion. Eine typische Maschine mit 160 sensorischen Merkmalen wurde dazu über mehrere Monate überwacht und ca. 21.900 Datensätze extrahiert, bestehend aus normalen und kritischen Prozesszuständen. Genauigkeiten von 99% und der Prototyp eines Mensch-Maschine-Interface in Form einer Stand-Alone-Applikation mit Empfehlungssystem wurden erreicht. Der Vorteil liegt nun in der Vorhersage von Prozesszuständen durch Echtzeit-Überwachung. In zukünftiger Arbeit wird sowohl eine Verbreiterung des Ansatzes auf andere Produktionsschritte und -linien als auch eine methodische Verbesserung von der Sensorebene aufwärts in Richtung Intelligent-Condition-Monitoring und Self-X-Eigenschaften verfolgt.

About the authors

Michael Kohlert

Michael Kohlert received the Diploma in economic & electrical engineering at the Institute of Integrated Sensory Systems, TU Kaiserslautern, Germany in 2010, respectively. He made his PhD at the Institute of Integrated Sensory Systems, TU Kaiserslautern, Germany in 2015. He is responsible for the department of IT & Automation at the Mondi Gronau GmbH, a polymer film plant. His research interests include Advanced Automation/Manufacturing, Industrie 4.0, Multi-sensory systems, Human-Machine-Interface, Computational Intelligence, Machine Learning.

IT & Automation, Mondi Gronau GmbH, Gronau, Germany

Andreas König

Andreas König received the Diploma in Electrical Engineering at TU Darmstadt in 1990 and the PhD at the Institute of Microelectronics Systems of the same university in 1995 in the field of intelligent vision system design for inspection tasks and of neural hardware. He then was with Fraunhofer IITB (now IOSB), in Karlsruhe continuing research in visual industrial inspection and aerial image analysis. He moved to TU Dresden, and from 1996 served as Assistant Prof. of Electronic Devices in the Electrical Engineering dept., and established a group in the field of integrated intelligent vision sensors, circuits, and systems design and application. In 2002, he was appointed deputy chair of the computer engineering chair, computer science dept. at TU Chemnitz. In 2003, he was appointed Prof. for Integrated Sensory Systems in the Electrical Engineering dept., TU Kaiserslautern. He established the Institute of Integrated Sensory Systems with research on heterogeneous integrated, intelligent multimodal sensor, circuit, and system design. Activities comprise reconfigurable or adaptive self-x systems, neural and evovable hardware, wireless, autonomous/self-sufficient sensors and sensor swarms or networks, CMOS and MEMS design and 3D integration, automation and synthesis activities, multidimensional signal processing and sensor fusion, and interactive large high-dimensional data analysis and visualization. He is author or coauthor of more than 120 technical publications in the field, he was chair/co-chair of int. conferences in the field, e.g., HIS 2007 and KES 2011 at TU Kaiserslautern, he is Senior Member of IEEE, founder and chair of the IEEE CIS German Chapter, member of MIR Labs, DHV, AHMT, the interest group on microelectronics for neural networks, and of several editorial boards.

Institute of Integrated Sensor Systems, TU Kaiserslautern, Kaiserslautern, Germany

Received: 2015-12-4
Revised: 2016-2-26
Accepted: 2016-3-5
Published Online: 2016-9-13
Published in Print: 2016-9-28

©2016 Walter de Gruyter Berlin/Boston

Downloaded on 9.12.2022 from https://www.degruyter.com/document/doi/10.1515/teme-2015-0120/html
Scroll Up Arrow