KI-kognitionsunterstützendes Assistenzsystem zur Inprozesskontrolle in der Fertigung (KIKA-IPK); Teilprojekt: ML-basierte visuelle Prozesskalibrierung und Regelung

Abstract

The KIKA-IPK research project has developed scalable AI cloud services with assistance tools that independently learn how image features and process characteristics are related. The empirical knowledge of machine operators was modelled using machine learning methods. User feedback and anomaly detection enabled AI models to be trained even with comparatively little training data. The result was two use cases (3D metal printing and drug printing) in which assistance systems were developed to analyse optical quality features and process variables during production in order to enable more resource-efficient process parameters. This is particularly suitable for small series and single-item production. A total of four concrete results were achieved: the development of three AI-based assistance systems and an associated individual cloud platform for connecting AI service providers to industry partners.