摘要
利用基于激光诱导击穿光谱技术(LIBS)的煤质快速检测装置进行实验研究,在传统的偏最小二乘法(PLS)模型基础上通过预先的谱线筛选,剔除部分波动过大不利于建模的噪声信息,建立基于谱线筛选的PLS模型。相对于传统的PLS模型,该基于谱线筛选的PLS模型在基本不改变定标样品误差的情况下,显著降低了预测样品的分析误差,其中预测样品的灰分预测结果平均误差从2.262%下降至1.972%,最大误差从5.093%下降至3.575%,模型的性能得以显著提高。
Experimental study was carried out by using laser-induced breakdown spectroscopy(LIBS)-based coal quality rapid detection device.Based on the traditional partial least squares(PLS)model,through the pre-screening of spectral lines,the noise information which was too fluctuated and not suitable for model establishment would be eliminated,and then the PLS model based on screening of spectral lines was established.Compared with the traditional PLS model,this PLS model based on screening of spectral lines reduced the analyzing errors of predicted sample obviously under the circumstance of maintaining the calibration sample error constant,the average error of predicted ash content of predicted sample decreased from 2.262%to 1.972%,the maximum error decreased from 5.093%to 3.575%,the performance of model was obviously improved.
作者
李春艳
茌方
刘翠茹
LI Chun-yan;CHI Fang;LIU Cui-ru(Nanjing Coal Quality Supervision and Inspection Company Ltd.,China Guodian Corporation,Nanjing 210031,China)
出处
《煤质技术》
2019年第1期37-39,共3页
Coal Quality Technology
关键词
煤质分析
灰分
激光诱导击穿光谱
偏最小二乘法
谱线筛选
模型
误差
coal quality analysis
ash
laser-induced breakdown spectroscopy
partial least squares method
spectral line screening
model
error
