多输入多输出LSSVM磨机负荷软测量  

Soft Sensor Approach for Mill Load on Multi-input Multi-output LSSVM

作　　者：梁礼明[1] 肖盈丁 吴健[1] 

机构地区：[1]江西理工大学,江西赣州341000

出　　处：《煤矿机械》2015年第11期306-309,共4页Coal Mine Machinery

基　　金：国家自然科学基金项目(51365017);江西省自然科学基金项目(20132BAB203020);江西省教育厅科学技术研究项目(GJJ13430)

摘　　要：针对湿式磨机在磨矿过程中负荷(ML)难以测量的特点,提出了一种基于核函数选择的多输出LSSVM磨机负荷软测量方法。引入能量熵的概念,借助超球体表示和核函数本身蕴涵的度量特征,提出基于样本分布能量熵的支持向量机核函数选择方法;利用LSSVM建立模型对多输入多输出磨机负荷进行软测量,且将据此测量的结果与BP神经网络测量结果进行了比较。仿真结果表明,借助样本分布能量熵选择LSSVM核函数,有助于SVM泛化能力和磨机负荷软测量准确率的提高。The measurement of mill load （ML） in grinding process is difficult to realize, aiming at the problems ,a new soft sensor for mill load on multi-output I_SSVM based on kernel function selection was proposed. Introduced a concept of energy entropy, with the super sphere description and measurement features of kernel function, put forward method of selection kernel function based on energy entropy of sample distribution; then use LSSVM to establish multi-input multi-out model of the mill load measurement and compare the results with the results of BP neural network. Simulation results show that, with energy entropy of sample distribution to select the kernel function, the generalization ability of SVM and the accuracy of soft measurement of load can be improved.

关 键 词：磨机负荷 核函数 能量熵 最小二乘支持向量机 

分 类 号：TP27[自动化与计算机技术—检测技术与自动化装置]