Closed loop manufacturing of precision components in small series based on machine learning using high-frequency fine-granular process data (PräziLoop)

Current manufacturing processes for high-precision components are often associated with high investment, maintenance and personnel costs. In addition, long measuring and reworking times lead to productivity losses.

As there is often little training data available, the use of AI poses a problem, particularly in small and medium-sized companies. The PräziLoop project is investigating the extent to which this problem can be solved with current algorithms and methods for the specific case of grinding within a closed production cycle. The aim is to develop a model that requires very little or no training data.

By integrating the AI solution directly into the machine control system, the machine can independently make adjustments after each manufactured part and improve the quality of the components for the following parts. The classification of parts into good parts, suspect parts and reject parts enables targeted reworking or sorting out of faulty parts. This also significantly reduces the risk of defective products being delivered to the customer.

A central concern of the project is also to ensure usability, transparency and trust in the AI results in order to promote acceptance in industrial practice.