| Abstract Scope |
Cu impurities in steel scrap, mainly presented as motors and wires, could induce surface hot shortness during hot working recycled steel with high Cu content, limiting the efficiency of recycling. Impurity removal methods have been developed, including physical separation and chemical treatment, to overcome such detrimental effects. For physical separation, optical recognition was explored as a type of sensor-based sorting method, apart from regular shredding and magnetic separating, considering either the Cu metal in the motor rotors with reddish-brown color and the Cu wire with colorful insulation, or the potential difference of shape geometry between Cu impurities and Fe shreds. To further improve the optical detection, image classification through machine learning was adopted to optimize the recognizing process of shredded scrap collected from auto shredders. The results show that better recognition of Cu impurities could be achieved, resulting in a reduction of Cu content. |