Entwicklung der digitalen Sortieranalyse im Digital Waste Research Lab

This master's thesis concerns the development and testing of digital methods for the sorting analysis in waste management. Conventionally, the characterization of waste samples is primarily manual and laboratory-based, a process associated with significant time consumption and personnel costs. In the Digital Waste Research Lab (DWRL) at the Montanuniversity Leoben, an automated, sensor-based sorting analysis is therefore being developed to enhance efficiency, reproducibility and accuracy. The focus is particularly on near-infrared spectroscopy (NIR), which, when combined with mechanical conveyor units and algorithmic solutions, enables (semi-)automated and standardized characterization of non-hazardous mixed municipal waste. Initially, basic tests are carried out to determine the optimum material feed and singulation. Based on these findings, various handling methods are applied to enable multiple surface detection of representative waste samples. The resulting pixel-based proportions are then converted into masses or mass proportions by means of specific surface weights, in order to generate a reliable sorting analysis. Initial results show that adapted recording methods and suitable conversion factors enable accurate characterisation of waste streams with total deviations of less than 2,4% (RMSE) after only a few measurement cycles. Conversely, conversion with literature values attains moderate accuracies of more than 54 % (RMSE).