The growing demand for aluminium worldwide makes aluminium recycling critical to realising a circular economy and increasing the sustainability of our world. One effective way to improve the impact of aluminium recycling is to develop cost-efficient automated sorting technologies for obtaining pre-defined high-quality aluminium scrap products, thus reducing undesirable downcycling and increasing environmental/economic benefits. In this work, an innovative facility, which includes singulation, sensor scanning, and ejection, is optimised for the automated sorting of aluminium scraps. The sorting facility is computationally studied by a virtual experiment model based on the discrete element method. The model considers particle-scale dynamics of complex-shaped scraps and mimics the automated operation of the facility. Based on virtual experiment modelling, the flow of scrap is optimized by computation, with the feasible operation of the sorting facility being proposed. Accordingly, the sorting facility has been built and model predictions are confirmed in actual operation.