High-precision glass optics are used as a key component of many new products in numerous markets. The sophisticated and complex optics are best manufactured using hot molding. Manufacturing using the non-isothermal molding process is significantly more cost-effective than isothermal precision molding.
However, large-scale manufacturing using the non-isothermal molding process faces the challenge that optimum process control is subject to unknown influences that are difficult to detect with current technology. For example, a non-optimal temperature distribution can quickly lead to mold defects and defects in the glass optics.
The use of automated machine learning systems (AutoML) in combination with hybrid artificial intelligence (AI) models offers the possibility to analyze process data in an automated and user-friendly way. Temperature distribution and flow behavior - and thus also the quality of the optics - can be optimized in this way.
The aim of the research project "OptiMassKi - Hybrid AI for process optimization in the series production of complex optics" is to optimize processes in hot molding by predicting glass flow behavior using hybrid artificial intelligence.
In the project, the partners are developing a hybrid model consisting of a "black box" and a "white box." The black box is trained with sensor data and the white box is formed from suitable rheological models, expert knowledge or simulations.
The researchers develop various functionalities in the form of modules, distinguishing between the configuration for creating the hybrid model and the actual use. Both the creation process and the use of the hybrid model are implemented automatically in a user-friendly software tool.
Fraunhofer Institute for Production Technology IPT, Aachen