Predictions of the Hardgrove grindability index, a predictor of the crushing and pulverization propensity of coal, have been made using both regression and neural network techniques. All techniques suffer from shortcomings. In general, input parameters must be selected based on a sound knowledge of coal chemistry and petrology, with avoidance of redundant parameters, avoidance of closure in the data sets that add to 100% (individually the proximate and ultimate analyses, petrology, and (approximately) major oxides), and a constrained coal rank and provenance setting. Predictions based on a specific set of coals are not necessarily translatable to different ranks or maceral suites. In general, for high volatile bituminous coals, combinations of coal rank (vitrinite reflectance or volatile matter), liptinite content, and ash percentage produce the best predictions.

已经使用回归和神经网络技术对 Hardgrove 可磨性指数进行了预测，该指数是煤的压碎和粉化倾向的预测指标。所有技术都有缺点。一般来说，输入参数的选择必须基于对煤化学和岩石学的充分了解，避免冗余参数，避免数据集中增加到 100%（分别是近似和最终分析、岩石学和（大约） ) 主要氧化物)，以及受限制的煤等级和来源设置。基于一组特定煤的预测不一定可转换为不同的等级或煤层组。一般而言，对于高挥发性烟煤，煤阶（镜质反射率或挥发分）、脂铁质含量、