最小二乘支持向量机结合中红外光谱测定甲醇柴油甲醇含量

采用中红外光谱法对甲醇柴油的甲醇含量进行检测分析。首先,对采集到的原始光谱进行预处理(标准正则变换、多元散射校正、一阶微分、二阶微分、Savitzky-Goly平滑),采用偏最小二乘法和最小二乘支持向量机建立了甲醇柴油的甲醇含量预测模型,并比较了不同预处理方法对模型预测能力的影响。实验结果表明,LSSVM的建模效果最佳,其预测集相关系数R~2为0.981 8,预测均方误差RMSEP为1.3917%(体积比)。因此,中红外光谱技术可用于甲醇柴油中甲醇含量的快速检测,且可以达到很好的效果。

Methanol Content Determination in Methanol Diesel with Mid Infrared Spectroscopy Analysis Using Least Square Support Vector Machine

The mid-infrared spectroscopy was used to detect the content of methanol in the methanol diesel. Firstly, the collected original spectra were pre-processed by the standard normal transformation(SNV), multiple scatter correction (MSC), first derivative, second derivative and Savitzky-Goly smoothing methods. The quantitative model of methanol was established by partial least square method and least square support vector machine (SVM). The influence of different pre-processing methods on the model prediction capability was also investigated. The results show that LSSVM modeling is the optimal forecasting model with the prediction set correlation coefficient R² of 0.981 8 and the prediction mean square error (RMSEP) of 1.3917%(volume ratio). So, the mid infrared spectrum technology can be used for the rapid detection of methanol in the methanol diesel and achieve very good results.