The aim is to develop an AI-supported assistance system for in-process control (KIKA-IPK), which enables more resource-efficient process and material configuration through self-learning image feature correlations. The machine operator's experience is used to model visual quality features and process properties through machine learning.
The intended project result is an assistance system that maps optical quality characteristics and process variables in an AI model to enable resource-efficient configuration of the process parameters. Application partners provide optical image characteristics and process data for additive manufacturing processes via a cloud interface. The AI model is demonstrated for additive metal deposition welding and the drop-on-demand process for personalized medication printing under real conditions.
The solution approach systematically recognizes cognitive human abilities in the event of deviations in additive manufacturing, correlates these with process properties and initiates measures in real time via the machine control. The innovative aspect lies in the machine recognition of new quality characteristics, their correlation with process properties and the adjustment of the process control variables.
Specific project goals of DiHeSys
Preparation of a specification to replace the wet chemical investigation, interface, system specification, EFS, role specification, metrics, planning DoE.
DoE and implementation, data generation and data provision, quality and process metrics with thresholds and continuous user feedback.
Integration into the AI workflow: real-time acquisition, data synchronization, data quality, image and process data stream integration into the cloud environment.
AI learning process: role-specific presentation of results of metrics, user feedback, measures and evaluations.
AI evaluation: Determination of thresholds and specific instructions for process adjustment.
Integration and system test: Validation of active ingredient content, optimization, adaptation and provision of “FLEXDOSE ® Printer” for validation on GP ® .
Contact person:
Jochen Lutz MA, Product Owner at DiHeSys Digital Health Systems GmbH
Project partners:
Comentários