煤全水分近红外光谱检测的预处理方法研究  

Study of Pretreatment of Detecting Moisture of Coal with Near Infrared Spectroscopy

作　　者：叶晨晖[1] 陆辉山[1] 闫宏伟[1] 高强[1] 

机构地区：[1]中北大学机械工程与自动化学院,太原030051

出　　处：《煤炭技术》2014年第4期257-259,共3页Coal Technology

基　　金：国家自然基金(41201294);江苏省农产品物理加工重点实验室开放基金(JAPP2012-2);山西省青年科技基金(2009021019-3)

摘　　要：水分含量多少是煤质好坏的重要指标,在研究煤的基础理论及煤加工中具有特殊意义。本实验采用多元散射矫正(MSC)、平滑处理、微分处理等预处理方法,结合主成分回归法及偏最小二乘法分析了预处理方法对煤粉全水分模型的影响;发现多元散射矫正(MSC)处理结合偏最小二乘法获得的模型最佳,其校正模型R=0.988 1,RMSEC=0.501,预测结果 R=0.955,RMSEC=0.601;发现平滑处理也可以提高模型可靠性,但过度平滑会使模型可靠性变差;综合比较主成分回归法模型与偏最小二乘法模型,发现偏最小二乘法获得的模型要好于主成分回归获得的模型。Moisture content is an important indicator of how much coal is good or bad,has a special significance in the study of the basic theory of coal and coal processing.In this study,using multiple scatter correction（MSC）,smooth handling,pretreatment methods of differential treatment based on principal component regression and analysis of the impact of minimum two became law on pulverized coal pretreatment total moisture model;Multiple scatter correction found（MSC） treatment combined with the least squares method to obtain the best model,its calibration model R=0.9881,RMSEC=0.501,R=0.955,RMSEC =0.601;Smoothing can also be found to improve the reliability of the model,but excessive smoothing model reliability will deteriorate;Comprehensive comparison of principal component regression model and least squares model,found that the smallest model to get into law is better than the second principal component regression model obtained.

关 键 词：煤 近红外光谱 预处理 最小二乘法 主成分回归法 

分 类 号：TD94[矿业工程—选矿]