The natural ore distribution is unique, and the mining process entirely depends on it. Thus, every mining has its way of dressing ore due to the plan of industry. Therefore, in optimal and control energy systems, the relation between ore distribution, parameters of stages, and the final output are vital to understanding the entire dressing plant. In this way, the paper purpose of developing a learning-based inverse method to understand the relationship between the ore (gathered from a few different open pits) and the final recovery rate of minerals. The variational autoencoder’s exceptional property is suitable for the learning-based inverse method, and the low dimensional space in the encoding and decoding process connects to the first input ore and the final outputs regarding the daily plan. When the first input ore is determined corresponding to the planned recovery, we use the low dimensional space to express the stages’ appropriate parameters. The milling stage is the most crucial stage of the plant, and for the validation propose of the method, the real experiment that investigated the learning result of the selected ball milling stage. Finally, the predictive-based control system was considered based on generating a variational autoencoder-based learning feature.

天然矿石分布独特，开采过程完全依赖于此。因此，由于工业计划，每个矿山都有其选矿方式。因此，在优化和控制能源系统中，矿石分布、阶段参数和最终产量之间的关系对于了解整个选矿厂至关重要。通过这种方式，本文的目的是开发一种基于学习的逆方法来了解矿石（从几个不同的露天矿坑中收集的）与矿物最终回收率之间的关系。变分自编码器的特殊性质适用于基于学习的逆方法，编码和解码过程中的低维空间连接到关于每日计划的第一个输入矿石和最终输出。当第一个输入矿石确定对应计划采收率时，我们使用低维空间来表示阶段的适当参数。球磨阶段是工厂最关键的阶段，为了验证该方法的建议，实际实验调查了所选球磨阶段的学习结果。最后，基于预测的控制系统被认为是基于生成基于变分自编码器的学习特征。