近红外光谱_偏最小二乘法测定三元溷合物中各组分的含量_张明锦.pdfVIP

第 24 卷 第 1期 重 庆 理 工大 学 学 报(自然科学) 2010年 1月 Vo.l 24 No. 1 Journal of Chongq ing Un iversity of Technology( N atural Science) Jan. 2010 - 1, 2 1 张明锦 , 杜一平 ( 1. , 200237; 2. , 810008) : 为了提高近红外光谱定 分析的预测精度和建模效率, 提出了一种基于峰纯度值的 变 优选方法, 结合偏最小二乘建模, 用于近红外光谱多组分同时测定对庚烷氯苯和甲苯三 组分混合体系进行分析, 只选择其中的 10 个波长变 建立校正模型, 与全谱范围的定 校正模 型比较结果表明该方法可使最终所得的模型中变 数大大减少, 可通过波长变 优选最小化 冗余信息, 提高预测精度及建模效率 : 近红外光谱; 偏最小二乘; 纯度值; 变 筛选 : O657 : A : 1674- 8425( 2010) 01- 0123- 04 Determ ination of Component Concentration in TernaryM ixture by Near Infrared Spectroscopy with Partial Least Square Re ression 1, 2 1 ZHANG M ing-jin , DU Y -i p ing ( 1. Research Centre for Analysis and M easu rem ent, East China Un iversity of Science and Technology, Shanghai 200237, Ch ina; 2. Departm ent of Chem istry, Q inghai Norm alUn iversity, Xining 810008, China) Abstract: In order to mi prove prediction accuracy andmodeling efficiency for quantitative calibration in near infrared ( N IR) spectroscopy, a novelw avelength variable selection method based on value of purity is proposed, wh ich is comb ined w ith partial least square ( PLS) for smi ultaneous determ ination of mu lticomponent w ith N IR. By analysis of the m ixture of heptane, chlorobenzene and methylbenzene, 10 variables are selected and used to bu ild quantitative calibration model. Compared w ith the model constructed w ith full spectral region, the proposed variable selection method can reduce the number of variables greatly an