摘要
为了准确地获得球磨机的制粉出力,提出了用灰熵关联分析法分析球磨机制粉出力的影响因素,选择对球磨机制粉出力影响较大的几个因素作为辅助变量建立模型,以最优拉丁超立方法选取的数据作为训练样本,基于BP神经网络建立球磨机出力软测量模型,并选出几组样本验证所建模型的可靠度.结果表明:所建立的球磨机出力软测量模型预测输出误差较小,有较强的泛化能力,具有很好的测量性能.
To accurately measure the ball mill output, factors influencing the mill output were studied via grey entropy correlation analysis, after which a soft sensing model was built up based on BP neural network by taking above influencing factors as auxiliary variables and using the data chosen by optimal Latin hypercube design as training samples. The model reliability was verified with several groups of samples. Results show that the model established for soft sensing of ball mill output has a small prediction error, strong generalization ability, indicating good forecasting performance of the model.
出处
《动力工程学报》
CAS
CSCD
北大核心
2015年第11期901-905,933,共6页
Journal of Chinese Society of Power Engineering
基金
"十二五"国家科技支撑计划资助项目(2013BAF02B11)
关键词
神经网络
球磨机
软测量
最优拉丁超立方法
neural network
ball mill
soft sensing
optimal Latin hypercube design
