Lehrstuhl für Technologien und Management der Digitalen Transformation
Herr Univ.-Prof. Dr.-Ing. Tobias Meisen
Industrial Transfer Learning
In the field of production, methods of machine learning, deep learning and artificial intelligence offer great potential for developing innovative solutions for optimization or automation. A central challenge here is to ensure a sufficiently large data basis - whether through experiments on a real machine or through simulations. In addition, the learning models must always be trained on new valuable observations as soon as the production process changes fundamentally (e.g. manufacturing a new product, changing the material). The research area industrial transfer learning investigates how these challenges can be mastered in the future. For this purpose, innovative processes and technologies are being developed based on transfer learning and related continual learning strategies. The goal is that an AI uses its knowledge from previous similar problems to solve a new problem, thus increasing its learning efficiency. With regard to the production area, industrial transfer learning is defined as follows:
In the field of production, industrial transfer learning refers to machine learning and artificial intelligence methods that make use of source data from different production process domains (e.g. simulation) or process variations (e.g. different product) with the goal to create robust, accurate and data efficient models for a certain target task.
The focus of our research lies on methods based on artificial neural networks. On the one hand, approaches for the continual learning of neural networks across different tasks and variations are investigated with the goal that the networks can use their knowledge from previous variations without forgetting them (Continual Learning, Incremental Learning, Lifelong Learning). On the other hand, approaches are investigated to train neural networks first in simulated environments and then to transfer them to complex real environments (Simulation to Reality Transfer).
Selected relevant publications
Industrial Transfer Learning: Boosting Machine Learning in Production
2019 IEEE 17th International Conference on Industrial Informatics (INDIN) , page 274--279.
Transfer-Learning: Bridging the Gap between Real and Simulation Data for Machine Learning in Injection MoldingProcedia CIRP, 72:185--190