Complex optics for medical, optical and mechatronic applications, as well as sensor components, can be manufactured best and most cost-effectively using hot forming processes. Two main processes are used here: isothermal precision molding, which offers high molding accuracy, and non-isothermal glass forming, which scores with short cycle times.
Both isothermal precision molding and non-isothermal glass forming have many advantages and strengths, but the product quality of both processes is extremely dependent on the process conditions: For example, at very high process temperatures, the glass material sticks to the mold, which damages the product and the mold. At lower temperatures, the wear of the forming tool is lower, but the risk of glass breakage due to residual stresses in the glass is significantly higher.
Simulation models make a valuable contribution to improving the stability of the manufacturing processes and the results throughout. The relationships between force and temperature during forming are known qualitatively, but there are no closed analytical models for a quantitative solution - especially for extreme and critical process situations. This gap is closed by the research project "hyPro".
The aim of the research project "hyPro – Integration of hybrid intelligence in the process control of glass forming production plants" is to optimize the process of non-isothermal hot forming by recording and processing all important influencing parameters with the aid of modeling software. This software is directly integrated into the machine control system and offers machine users alternative processing strategies in the event of a deviation from the norm during production.
In the project, the research project team will combine data-driven machine learning (ML) models and process models based on expert knowledge. For the ML models, an artificial intelligence (AI) developed during the project will be trained with process data obtained during the forming of glass components using machine data, sensor data and quality measurements.
The "expert" models will be designed using data from expert knowledge, already known simulations and various physical models, such as surface energy models. Since the extensive knowledge of the researchers at the Fraunhofer IPT and their partners about process sequences and the laws of physics is systematically used in this project to form the process models and is incorporated into the simulation software, the amount of data for training the ML model is significantly reduced.
During the last stage of the project, the researchers will combine the ML model and the expert model to create a hybrid model in the form of modeling software. This software will be integrated directly into the process control of a project partner's forming plant. There, it automatically identifies quality-relevant process situations and offers the machine operators appropriate adjustments to the production parameters. The software is thus able to optimize production processes in non-isothermal forming.
The research project "hyPro - Integration of hybrid intelligence in the process control of glass forming production plants" is funded by the German Federal Ministry of Education and Research (BMBF) as part of the KI4KMU funding program.
Funding code: 01IS22053D