基于RBF神经网络和RS理论的磨矿分级系统软测量模型

Model of soft measurement of grinding and classification system based on RBF neural network and RS theory

为了提高硫化镍选矿过程的效率并改善选矿产品的质量,运用RS理论研究了某选矿厂磨矿工艺多维数据的属性约简.在建立相应RBF神经网络预测模型基础上,给出了表征磨矿生产过程内在规律的最小知识表达,并基于该模型对选矿生产指标进行了预测.结果表明:磨矿工艺数据可以进行浓缩,生产过程经验操作能够找到相应的理论依据,从而加深了对生产工艺过程内在规律的认识;应用软测量技术获取了球磨机和旋流器内部状态主要关键参数,该模型分析过程相对简单,网络学习训练时间少、学习精度高;仿真结果表明估计值与分析值拟合良好.

In order to increase the efficiency of mineral processing and improve the quality of dressing products on nickel sulfide,attributes reduction of multidimensional data in certain ore plant was studied by using rough set theory.Based on the relevant RBF neural network predicting model,least knowledge expressions was given to express inherent law of ore grinding.The metallurgical performance was built by applying the model.The results show that attributes reduction of multidimensional data is feasible on grinding and classification system.The model is helpful for learning inherent law of mineral-processing,and theoretical basis of experimential operating is proved.Key parameters of ball mill and hydro cyclone were obtained by soft measurement technology.The analysis course is brief,the time of network learning and training is short,and learning precision is high.Simulation shows that estimating value is very close to analysis value.