中温煤沥青甲苯不溶物等4项指标的近红外预测模型的建立

Establishment of Near-Infrared Prediction Models for Toluene Insoluble Substance and Other 3 Indexes of Medium Temperature Coal Pitch

采用仪器所配的积分球分析模块采集了中温煤沥青样品的近红外漫反射光谱图,采用一阶导数(SG1D)、标准正态变化(SNV)等方法对所得的原始光谱进行了预处理。分别采用GB/T2292-2018、GB/T 2293-2019、GB/T 2294-2019、GB/T 8727-2008测定了样品的甲苯不溶物(TI)含量、喹啉不溶物(QI)含量、软化点和结焦值等4项指标。以4项指标的光谱数据为自变量,测定值为因变量,采用偏最小二乘法(PLS)分别建立了测定值与光谱数据的预测模型。结果表明:4项指标的预测值和测定值的相关性较好,决定系数(R2)分别为0.82,0.79,0.71,0.96,校正均方根误差(RMSEC)分别为0.48%,0.35%,1.11℃,0.35%,预测均方根误差(RMSEP)分别为0.59%,0.45%,1.10℃,0.52%,模型具有较好的预测能力,可满足煤沥青制成过程质量控制的需求。

Near-infrared diffuse reflectance spectra of medium temperature coal pitch samples were collected with integrating sphere analysis mode.SG1D,SNV and their combined methods were applied for pretreatments of the original spectra.Determined results of 4 indexes including TI content,QI content,softening point and coking value were obtained by GB/T 2292-2018,GB/T 2293-2019,GB/T 2294-2019 and GB/T 8727-2008.With values of NIR spectrum as independent variable and determined values as dependent variable,prediction models of 4 indexes were established by PLS.As found by results that the correlation between the predicted values and determined values was good,giving values of R2 of 0.82,0.79,0.71,0.96,and values of RMSEC found were0.48%,0.35%,1.11℃,0.35%,with values of RMSEP of 0.59%,0.45%,1.10℃,0.52%,respectively.The model had a good predictive ability and could be applied in the quality control of the manufacturing process of coal pitch.