Fraunhofer-Gesellschaft Hybrid sensor for material identification and characterization
To record the data in waste containing plastic, the “Waste4Future” project partners developed a suitable hybrid sensor system that records the material properties, including the internal structures and external geometry of the bulk material. In particular, the research partners have addressed previously unsolved challenges in plastics sorting, such as colour-independent sorting (robust characterization even of soot-coloured, black plastics), aged plastics (enabling sorting according to different ageing states), identification of additives and fillers (determination of occurrence and, if necessary, proportions). By developing a multi-sensor system with AI-based data evaluation, considerable progress was made here, which was demonstrated using a sorting conveyor belt line as a demonstrator.
The consortium expanded previously existing sorting sensor solutions to include new sensor technologies (THz line scan camera, active thermographic sensors, line laser, phased array airborne ultrasound line scan camera). Using machine learning methods, a real-time capable algorithm was developed for the hybrid sensor that enables reliable material identification.
The focus was on developing a model that can recognize, classify and sort different plastics, including black ones. The THz sensor technology developed in the project proved to be particularly suitable for sorting black plastics. The models were trained with special test specimens and then the sorting of a) model specimens made of PE and PA in various sizes and b) the sorting of a real SLF fraction with various impurities were validated. The model showed a high accuracy with high resolution and classification rate, taking into account the special challenges of waste streams containing plastics (such as high material variability, small particle sizes).
A high-performance multiband chip was developed for the ThZ sensor as part of the project. Suitable fused processing of both bands will make it possible to achieve a higher distance resolution and significantly increase the information content about the samples under investigation. Significant improvements were also made to the lens and reflector, as well as to the interaction of the sensor technology with the architecture and assembly line speed of the demonstrator.