基于BP人工神经网络的球磨机钢球配比预测模型

采用BP神经网络对实验室磷矿球磨机磨矿中的钢球配比与磨矿产品粒级分布的关系进行建模， 解决选矿厂磨机生产中钢球配比的计算问题. 建立的BP神经网络预测模型通过磨矿产品粒级分布来预测对应的球磨机内钢球配比，预测绝对误差控制在3%以内，但预测相对误差较大且不稳定，说明在钢球配比与磨矿产品粒级分布的关系建模中该建模方法具有一定研究价值，该模型进一步优化后可具有工业应用价值.

Prediction Model on Matching Steel Ball in Mill Based on BP Artificial Neural Network

The error back prorogation （BP） artificial neural network was applied to establish the prediction model about the relationship between the particle size distribution of grinding product and the proportion of matching steel balls of different sizes in the ball mill of phosphate ore in the laboratory, which can predict the proportion of different size balls in the ball mill through the particle size distribution of grinding product. The mean absolute percentage error of the prediction can be controlled in 3%, but the mean relative percentage error of prediction is unacceptable, which illustrates that the modeling method has some research values, but it should be studied in-depth to reduce the error of model for application in the factory.