Abstract. For sorting scrap metal materials, the ultimate goal is to separate some scrap metal from other scrap metal materials completely. In this paper, Laser-Induced Breakdown Spectroscopy (LIBS) is used to take sample data from cooked aluminum, aluminum, magnesium, stainless steel, iron, copper, and zinc 7 metal samples seven scrap metal samplings with principal component analysis methods of data reduction. And BP neural network is used to the classification of aluminum and non-aluminum metals, Experiment results show that the classification accuracy up to 93.40% . These results provide a method for sorting scrap metal and reference data with BP neural network and LIBS technique, which make online scrap metal sorting get better result.