基于范数优化极限学习机的矿浆浓度预测  

A prediction algorithm for pulp concentration using norm-optimized extreme learning machine

作　　者：王欢[1] 姜昌伟[2,3] 徐鑫[1] 孙为平[1] 鲁鹏云 张德政[2,3] 

机构地区：[1]鞍钢集团矿业公司,辽宁鞍山114001 [2]北京科技大学计算机与通信工程学院,北京100083 [3]材料领域知识工程北京市重点实验室,北京100083

出　　处：《中国矿业》2016年第8期112-116,共5页China Mining Magazine

基　　金：中央高校基本科研业务费专项资金资助项目资助(编号:FRF-BD-15-013A)

摘　　要：选矿过程中的矿浆浓度是一个重要的生产工艺参数,一般可以通过预测矿浆浓度来提高生产效率。由于矿浆浓度和其他的生产工艺参数往往非线性相关,这给矿浆浓度的预测带来了很大困难。本文针对此问题,基于极限学习机这一面向神经网络的新颖学习算法,提出了一种矿浆浓度预测新算法。首先,使用相空间重构方法对矿浆浓度数据进行预处理,从一维转换到多维。然后,使用基于L2范数的极限学习机算法(ELM-L2)建立时序预测模型,实现预测功能。围绕来自于某矿厂的真实生产数据进行了实验验证,结果显示,针对大规模的数据样本集,所设计的算法与传统神经网络预测算法相比,训练时间大约减少了30%,而预测精度大约提高了48%。实验结果表明了所设计预测算法的有效性。Pulp concentration as one of the most important production parameters plays an important role in the ore production. Generally, the production efficiency can be improved by a prediction for pulp concentration. Since there are some nonlinear relationships between the pulp concentration and other production parameters, it imposes very challenging obstacles to address this issue of prediction. A novel prediction method is proposed in this paper through the use of extreme learning machine （ELM） that is an effective learning algorithm developed for neural network. Firstly, the pulp concentration data is preprocessed by the phase space reconstruction method, and the time series prediction model is adjusted from one dimension to multiple dimensions. Secondly, an improved ELM algorithm using L2 norm （ELM-L2） is developed to implement the prediction. The experiments are conducted with a real-world production data set from a mine. Compared with the traditional prediction method using neural network, the proposed approach can reduce the training time by 30% and improve the prediction accuracy by 48% for a large-scale data set. The experimental results show the effectiveness of the proposed algorithm.

关 键 词：极限学习机 相空间重构 矿浆浓度 预测 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术] TD9[矿业工程—选矿]