摘要
为了快速检测煤炭的灰分和硫分,以屯兰选煤厂0.42、0.177、0.125 mm三个粒度等级煤样为研究对象,采集了300个煤样的近红外光谱图,利用主成分回归(PCR)算法,建立了基于学生氏残差剔除异常样品后的定量回归模型,并与标准值进行了对比分析。分析结果表明:粒度等级0.42 mm煤样的灰分模型效果较好,校正集相关系数和均方根误差分别为0.927 0和0.039 9,预测集相关系数和均方根误差分别为0.911 1和0.043 9,模型的稳定性较高;粒度等级0.177 mm煤样的硫分回归建模效果较好,校正集和预测集的相关系数均达到0.96以上,校正集均方根误差和预测集均方根误差分别为0.019 5和0.016 7,模型具有较强的预测能力。该研究为煤质内部成分的快速检测提供了有效的分析方法。
For rapid determination of coal internal ash and sulfur contents, test is made with coal samples of 3 sizes （0.42 mm, 0. 177 mm and 0. 125 mm） . On the basis of the near-infrared spectrums of 300 collected coal samples, a student＇s residual-based quantitative regression model is established after abnormal samples have been rejected, using the principle component regression algorithm （PCR） . A comparison with the standard values indicates that a satisfactory modeling result can be expected for the determination of the ash of the 0.42 mm size coal and the sulfur of the 0. 177 mm coal. In the former case, the calibrated correlation coefficient and root mean square are respectively 0. 927 0 and 0. 039 9 as against the predicted figures of 0. 911 1 and 0. 043 9 while in the latter case, the calibrated correlation coefficient reaches over 0.96 with a root mean square error of 0. 019 5 as against the predicted figure of 0. 016 7. The model was a high stability and predicative ability. It provides an effective means for analysis and rap- id determination of coal internal properties.
出处
《选煤技术》
CAS
2016年第4期15-18,共4页
Coal Preparation Technology
关键词
煤质分析
近红外光谱
主成分回归
粒度
灰分
硫分
analysis of coal property
near-infrared spectrum
principle regression analysis
size
ash
sulfur
