Resource-efficient and low-waste elastomer production through AI-based quality and process optimization – REVEAL

Manufacturing silicone seals through injection moulding involves high quality requirements, complex process influences and time-consuming post-processing steps. However, traditional quality control measures only provide limited insight into the underlying process parameters. Consequently, there is a lack of optimisation and defective parts incur considerable costs. This is where the REVEAL research project brings transparency and controllability to the process.

REVEAL aims to develop an AI-supported system for automated in-line quality control and data-based process optimisation in elastomer injection moulding. By digitising manufacturing parameters and using multimodal AI models, complex defect patterns can be detected and explained at an early stage. A virtual assistant will translate the AI results into transparent recommendations for the operators to take action on. This should significantly reduce waste and resource consumption, as demonstrated by the product portfolio of an industrial partner.

For TMDT, the project involves researching generalisable deep learning methods for the visual inspection of elastomer components. This is complemented by research into multimodal AI approaches that jointly evaluate image and process data, enabling robust fault detection and process control. Particular focus is given to explainable AI (XAI) methods that make the models comprehensible and prepare them for use in a virtual assistance system.