July 12, 2021 (online)
Evolutionary Computation in Practice
In the Evolutionary Computation in Practice (ECiP) track, well-known speakers with outstanding reputation in academia and industry present background and insider information on how to establish reliable cooperation with industrial partners. They actually run companies or are involved in cooperation between academia and industry.
If you attend, you will learn multiple ways to extend EC practice beyond the approaches found in textbooks. Experts in real-world optimization with decades of experience share their approaches to creating successful projects for real-world clients. Some of what they do is based on sound project management principles, and some is specific to our type of optimization projects. If you are working in academia and are interested in managing industrial projects, you will receive valuable hints for your own research projects.
In 2021, ECiP will be online (virtual) for the first time, which is a great challenge for a track that relies on personal interactions. We will do our best to enable opportunities for establishing contacts among participants.
|Michael Affenzeller Andreas Beham||Prescriptive Analytics in Production: Practical Applications of Evolutionary AI|
|Tea Tušar||Roadblocks to Finding Optimal Tunnel Alignments with Evolutionary Algorithms|
|Niklas Körwer |
|AI-Based Control of Conveyor Applications Using a Digital Twin for Model Training|
|Hans Fleischmann||Model-based condition and process monitoring of automated assembly systems based on machine learning methods|
Updated: May, 26th 2021
Michael Affenzeller, Andreas Beham (University of Applied Sciences Upper Austria): Prescriptive Analytics in Production: Practical Applications of Evolutionary AI
Abstract. Prescriptive Analytics is an interdisciplinary topic in an interdisciplinary field, or put another way it is a synergistic hybridization of various methods and algorithms from statistics, computer science, artificial intelligence, mathematics and operations research. Its aim is to provide optimized recommendations for action in various application areas. In this way, knowledge gained in the digital world is brought back to the real world, providing better and more efficient procedures, designs and processes.
When controlling production and logistics processes, dynamic events within the planning horizon must be continuously taken into account. By combining heuristic algorithms and machine learning, optimization procedures can be developed which optimally assist process owners with customised and comprehensible recommendations for action. Besides developing a software environment for adaptive optimization procedures, other key milestones include forming new dynamic problem models and defining suitable benchmarks. Furthermore, machine learning methods are used as an integral part of solution processes to predict future events, develop potential future scenarios and automatically select and parameterize appropriate optimization algorithms. The performance of the developed methods is ultimately examined through simulation experiments and is evaluated under realistic conditions.
The presentation will cover theoretical aspects as well as real world examples demonstrating how the open source framework HeuristicLab can be used for modeling, optimization and machine learning tasks for concrete challenges in the domain of production, logistics and systems research.
CV. MICHAEL AFFENZELLER has published several papers, journal articles and books dealing with theoretical and practical aspects of evolutionary computation, genetic algorithms, and meta-heuristics in general. In 2001 he received his PhD in engineering sciences and in 2004 he received his habilitation in applied systems engineering, both from the Johannes Kepler University of Linz, Austria. Michael Affenzeller is professor for heuristic optimization and machine learning at the University of Applied Sciences Upper Austria, Campus Hagenberg, and head of the research group HEAL http://heal.heuristiclab.com/ . Since October 2014 he serves as the head of studies for the Master degree program Software Engineering and as vice-dean for R&D at the faculty of informatics, communications and media.
CV. ANDREAS BEHAM received his MSc in computer science in 2007 and his Ph.D. in engineering sciences in 2019, both from Johannes Kepler University Linz, Austria. He currently holds an assistant professor position at the R&D facility at University of Applied Sciences Upper Austria, Hagenberg Campus and is leading several funded research projects. He is co-architect of the open source software environment HeuristicLab. He has published more than 80 documents indexed by SCOPUS and applied evolutionary algorithms, metaheuristics, mathematical optimization, data analysis, and simulation-based optimization in industrial research projects. His research interests include applying dynamic optimization problems, algorithm selection, and simulation-based optimization and innovization approaches.
Tea Tušar (Jožef Stefan Institute): Roadblocks to Finding Optimal Tunnel Alignments with Evolutionary Algorithms
Abstract. Solving real-world optimization problems presents many challenges. Some expected, others less so. This talk will explore a tunnel alignment optimization problem and the five challenges, or roadblocks, we encountered on our way to solving it. The first stems from the requirement that the objectives and constraints of our optimization problem are not fixed, but can be selected by the user. The second comes in the form of a relatively low number of function evaluations at our disposal. The third is the unexpected change of the problem definition at half way, which adds new restrictions to the tunnel alignment. The fourth arises from the difference between the test problems and those coming from the real world. The final, fifth roadblock is the consequence of optimization being the last link in the chain – the one in which hidden issues of the entire chain are revealed and time is running out.
CV. Tea Tušar is a research associate at the Department of Intelligent Systems of the Jožef Stefan Institute, and an assistant professor at the Jožef Stefan International Postgraduate School, both in Ljubljana, Slovenia. After receiving the PhD degree in Information and Communication Technologies from the Jožef Stefan International Postgraduate School for her work on visualizing solution sets in multiobjective optimization, she has completed a one-year postdoctoral fellowship at Inria Lille in France where she worked on benchmarking multiobjective optimizers. Her research interests include evolutionary algorithms for singleobjective and multiobjective optimization with emphasis on visualizing and benchmarking their results and applying them to real-world problems.
Niklas Körwer, Martin Bischoff (Siemens AG): AI-Based Control of Conveyor Applications Using a Digital Twin for Model Training
Abstract. Machine builders are faced with faster changing markets and increased customer demands everyday. With conventional engineering methods these demands cannot be met anymore which has lead to an increase in the application of simulation and artificial intelligence in key engineering disciplines. Instead of prototypes, digital twins are being created and machine learning algorithms are applied to various machines and processes to improve their performance or to ensure a consistent quality. In this presentation we will show how an artificial intelligence was trained using a digital twin of a real machine and how this AI is integrated into a conventional PLC (Programmable Logic Controller) to improve the machine throughput.
CV. Niklas Körwer is a sales specialist for motion control at Siemens AG in Cologne, Germany. He joined the Siemens AG in 2013 via a cooperative study program at the DHBW Mannheim and moved to Cologne in 2016 to work as a technical consultant for factory automation. Two years later he switched to a part-time position in order to continue his studies at the TH Cologne for a Master in Automation and IT which he completed last year.
During his master’s thesis he focused on the training of an artificial intelligence using a simulation of a parcel sorting machine. The thesis was supervised by Prof. Dr. Thomas Bartz-Beielstein (Institute IDE+A at TH Köln) and Dr. rer. nat. Martin Bischoff, who works as a research scientist at Siemens Technology in Munich, Germany.
Hans Fleischmann (Schaeffler Group): Model-based condition and process monitoring of automated assembly systems based on machine learning methods
Abstract. Within the scope of the fourth industrial revolution, Industry 4.0, the manufacturing industry is trying to optimize the traditional target figures of quality, costs and time as well as resource efficiency, flexibility, adaptability and resilience in volatile global markets. Cyber-physical production systems are used as intelligent overall systems to control production processes, machines and product quality . The technical complexity of automated assembly systems and their associated maintenance processes are rising due to the demands on their adaptability and the increasing automation. The scope and objective of this talk is to define and explain the model-based condition and process monitoring of automated assembly systems based on machine learning methods at the Schaeffler Group.
CV. Dr.-Ing. Hans Fleischmann is a scientist in the field of information and automation technology at the Schaeffler Group. Previously, he was a research assistant and doctoral student at the Institute for Factory Automation and Production Systems at the Friedrich-Alexander-Universität Erlangen-Nürnberg. His current research focuses on distributed condition monitoring systems, machine learning methods and communication standards used in cyber-physical production systems.
- Thomas Bartz-Beielstein, IDE+A, TH Köln, Germany
- Bogdan Filipič, Jožef Stefan Institute, Slovenia
- Sowmya Chandrasekaran, IDE+A, TH Köln, Germany
GECCO 2021 Webpage: https://gecco-2021.sigevo.org/HomePage