Methods for separation efficiency determination from particle-resolved optical analysis

Assessing the efficiency of recycling processes requires quantitative data from complex material streams. Such data enable life cycle analysis, process intensification and the implementation of sustainable material cycles. This study investigates how data from an optical monitoring system under industrial operating conditions can be converted into quantitative measures of separation performance. Three consumer-grade RGB cameras were installed above transport belts before and after the manual sorting of copper-containing particles from a steel scrap fraction. The particle-discrete dataset from a simple image processing workflow was used to calculate grade efficiency curves, both for a one-dimensional analysis based on particle size and for a two-dimensional analysis using convexity as an additional shape factor. The conventional analysis indicated substantial removal of free copper-containing particles, while a no-sorting reference campaign revealed systematic deviations caused by particle rearrangement and overlap on the downstream belt. The multidimensional analysis showed that particle shape contains additional information on the separation outcome. These results demonstrate that the main value of low-cost optical monitoring does not lie in particle detection itself, but in the quantitative process analysis that can be derived from particle-resolved data. The proposed framework provides a basis for process monitoring, comparison of separation features, and identification of particle classes associated with uncertain or inefficient separation.