ABSTRACT

Coal content in gangue is an important production index, and a commercial method to detect it is yet to be devised. The prediction of coal and gangue particle volumes is crucial. The shape clustering method is adopted to automatically classify coal or gangue particles based on their shapes to build volume models that can adapt to different shapes. Subsequently, volume models for coal and gangue particles of different shapes are established. Without shape clustering, the average relative errors of the volume model for gangue are 13.41%, 12.87%, 11.42%, and 9.12% for particles sizes of 25–13 mm, 50–25 mm, 100–50 mm, and >100 mm, respectively, whereas they are 12.54%, 11.82%, 10.36%, and 7.69%, respectively, after shape clustering. Without shape clustering, the average relative errors of the volume model for coal are 9.97%, 11.10%, 12.44%, and 11.06%, respectively, whereas they are 9.08%, 8.98%, 11.53%, and 8.27% after shape clustering. The reduction in error indicates the effectiveness of the proposed volume prediction.

摘要

煤矸石中的含煤量是一项重要的生产指标，目前尚无商业化检测方法。煤和煤矸石颗粒体积的预测至关重要。采用形状聚类方法，根据形状自动对煤或煤矸石颗粒进行分类，构建适应不同形状的体积模型。随后，建立了不同形状的煤和煤矸石颗粒的体积模型。在没有形状聚类的情况下，对于 25–13 mm、50–25 mm、100–50 mm 和 >100 mm 的粒径，煤矸石体积模型的平均相对误差分别为 13.41%、12.87%、11.42% 和 9.12% ，而它们在形状聚类后分别为 12.54%、11.82%、10.36% 和 7.69%。在没有形状聚类的情况下，煤体体积模型的平均相对误差分别为 9.97%、11.10%、12.44% 和 11。06%，而形状聚类后分别为 9.08%、8.98%、11.53% 和 8.27%。误差的减少表明所提出的体积预测的有效性。