为了达到压缩计算量以快速检测柴油硫含量的目的,该文分别利用回归系数分析法(RCA)、连续投影算法(SPA)和遗传算法(GA)对柴油全谱选取特征波长,作为 PLS 和 MLR 线性及LS-SVM非线性预测模型的输入变量,对比分析结果表明:RCA、SPA、GA算法得到波长个数分别为7、8、16,PLS、MLR线性模型的结果优于LS-SVM非线性模型.三种变量输入模型中,精度最高的是SPA-PLS模型,其RPD为1.99,RMSEP为0.0281,r2pre为0.91,SPA算法可以作为建立柴油硫含量快速预测模型的特征波长选取方法.%In order to compress computation to measure diesel sulfur content,the Regression Coefficient Analysis (RCA) and Successive Projections Algorithm(SPA) and Genetic Algorithm(GA) method were used to select fea-ture wavelength of diesel full spectrum,as PLS and MLR LS-SVM of linear and nonlinear prediction model of input variables. The result showed that RCA,SPA,GA algorithm to obtain the wavelength number were 7,8,16. PLS, MLR model was better than the LS-SVM model. In the three variable input models,the most accuracy was the SPA-PLS model. Its' RPD,RMSEP and r2prewere 1.99,0.0281 and 0.91.
展开▼
