Engineering for Smart Manufacturing (E4SM)

1 Background

A basic idea of Industry 4.0 is to utilize data of all areas of a production process in an integrated way. The acquired extensive data enables completely new applications for machine learning and artificial intelligence techniques, which can be utilized to achieve considerable increases in efficiency by using assistance systems.

Since several years, novel machine-learning-based approaches have been applied successfully in various contexts to solve very specific tasks (e.g. person recognition in video data, self-adapting strategies in games), often surpassing classical problem-solving strategies and even humans in their accuracy and efficiency. However, so far mainly very restricted problems are addressed. In the context of Industry 4.0, however, the full potential will only be unleashed if machine-learning-based methods can be realized in such a way that they incorporate data from different areas of the production process into the learning processes. Furthermore, questions of IT security, safety, reliability and predictability need to be addressed prior to use in larger industrial contexts.

2 Objective

The goal of the five-year project "Engineering for Smart Manufacturing" (E4SM) at the TU Ilmenau, which has been funded by the Carl Zeiss Foundation with three million euros since May 1, 2019, is to research innovative methods for the development and operation of assistance systems based on machine learning for intelligent manufacturing in industrial application scenarios. In the context of Industry 4.0, the demands and specifics of manufacturing and assembly processes of small and medium-sized enterprises (SMEs) will be particularly considered. The goal is to make the use of machine learning and assistance systems for SMEs more predictable and controllable, thus lowering the entry barriers for the use of these technologies.

3 Scope of examination

In coordination with the associated partners, the Thuringian Center for Mechanical Engineering (Thüringer Zentrum für Maschinenbau, ThZM) and the SME 4.0 Competence Center Ilmenau (Mittelstand 4.0-Kompetenzzentrum Ilmenau), the "Jigless Laser Beam Welding" and the "Multi-variant Assembly Processes" were identified as exemplary and also transferable and demonstrable application scenarios. In their context, the methods and techniques worked on in this project are to be researched, tested and demonstrated.

In laser beam welding, there is a process-specific relative displacement of the components to each other, which requires precise positioning and fixation. The required (clamping) devices represent a high cost factor. If, however, visually guided robot systems could actively hold the parts and simulate the clamping movement, a product-specific clamping device would be dispensable. For the first time, it would be possible to avoid expensive, time-consuming equipment and achieve a more flexible division of production processes.

In manufacturing process chains for highly variable small series there are always challenging, fine-motor sub-tasks that still cannot be performed sensibly and economically by robots and are therefore performed by humans. Therefore, smooth and effective processes require a close and coordinated collaboration between humans and assistance robots. In order to support the assembly process, we plan to use mobile robotic assembly assistants, which, based on initial knowledge and experience, can learn to interact with humans in the interaction process, become increasingly proactive and thus continuously improve their actions.

A unique feature of the E4SM project is the clear focus on integrated and holistic engineering methods for the application of learning-based assistance systems in manufacturing. Thus, the aim is to integrate developed partial solutions in the important core areas of collaborative assistance robotics, management and analysis of heterogeneous data sets from industrial manufacturing process chains as well as IT security and IT safety by means of a holistic (software) development process.

Michael Reichel (ari)

4 Project Consortium

In order to achieve the goals of the E4SM project, which is funded by the Carl Zeiss Foundation, the expertise from seven internationally recognized fields of study of the two faculties "Computer Science and Automation" (Informatik und Automatisierung) and "Mechanical Engineering" (Maschinenbau) at the TU Ilmenau will be bundled and combined in an interdisciplinary way with a newly established junior research group. Involved are the following groups:

  • Neuroinformatics and Cognitive Robotics
  • Telematics/Computer Networks
  • Software Engineering for Safety Critical Systems
  • Databases and Information Systems
  • System and Software Engineering
  • Production Engineering
  • Quality Assurance and Industrial Image Processing
  • Junior Research Group Machine Learning

Associated partners are the Thuringian Center for Mechanical Engineering (Thüringer Zentrum für Maschinenbau, ThZM) and the SME 4.0 Competence Center Ilmenau (Mittelstand 4.0-Kompetenzzentrum Ilmenau). In addition, the E4SM project is guided by a top-class company advisory board consisting of the Honda Research Institute Europe, Robert Bosch GmbH, LASO tech Systems GmbH, Metralabs GmbH, Henkel und Roth GmbH and TÜV Thüringen e.V..

5 Status

The project started on May 01, 2019 and was presented to the public on May 09, 2019. First prototypical implementations of the two application scenarios and first experiments were presented to the company advisory board in a workshop on June 18, 2020. In the application scenario " jigless laser beam welding" the influences of welding speed and sheet thickness on the weld seam quality were investigated for only partially fixed sheets. The welding processes were recorded multisensorically. The next step is to use machine learning to analyze the sensory data to determine whether the current welding process is promising or requires intervention. In the application scenario "variant rich assembly processes" experimental data of an assembly process were recorded. The next step is to analyze the actions of the assembler with machine learning so that a robot can recognize the individual processing steps. In the long term, this should enable the robot to act proactively. In addition, the first procedures for gripping workpieces by a robot arm were prototypically implemented.


Prof. Dr.-Ing. Horst-Michael Groß
Neuroinformatics and Cognitive Robotics Group

Prof. Dr.-Ing. Günter Schäfer
Telematics/Computer Networks Group

Dr.-Ing. Markus Eisenbach
Head of the E4SM Junior Research Group Machine Learning
Website Project E4SM