论文部分内容阅读
利用拉曼光谱检测技术,对甲醇柴油的甲醇含量和黏度进行定量检测研究。93个甲醇柴油样品作为被检测的对象,划分校正集(72个)和预测集(21个)。分析比较了光谱的不同预处理方法的全交互验证偏最小二乘(PLS)模型效果;然后以最优预处理方法得到的光谱数据为输入,结合连续投影算法(SPA)建立不同的回归校正模型,并进行比较分析。结果表明,甲醇含量的多元散射校正-偏最小二乘(MSC-PLS)模型预测效果最优,其校正集相关系数RC为0.9761,交互验证相关系数RCV为0.9551,校正集均方误差(RMSEC)为1.5089,交互验证均方误差(RMSECV)为2.0630;黏度的MSC-PLS模型预测效果也是最优的,RC为0.9794,RCV为0.9580,RMSEC为0.0907mPa·s,RMSECV为0.1292mPa·s。
The methanol content and viscosity of methanol-diesel oil were quantitatively studied by Raman spectroscopy. Ninety-three methanol-diesel samples were tested as calibration sets (72) and prediction sets (21). The effects of different pretreatment methods on PLS model were analyzed and compared. Based on the spectral data obtained from the optimal pretreatment method, different regression models were established by using the continuous projection algorithm (SPA) , And comparative analysis. The results showed that methanol-based multivariate scatter correction-partial least squares (MSC-PLS) model had the best prediction performance. The calibration set correlation coefficient RC was 0.9761, the correlation coefficient RCV was 0.9551, the RMSEC, (RMSECV) was 2.0630. The viscosity of the MSC-PLS model was also the best one with RC of 0.9794, RCV of 0.9580, RMSEC of 0.0907 mPa · s and RMSECV of 0.1292 mPa · s.