基于近红外光谱建立的模型预测反渗透复合膜生产线水相中间苯二胺、三乙胺和十二烷基苯磺酸钠等3种组分。以近红外光谱仪分析样品，得到的原始红外光谱图用Savitziky-Golay法平滑-二阶微分法预处理后，选择间苯二胺、三乙胺和十二烷基苯磺酸钠的PLS因子数为3，4，3，用偏最小二乘法（PLS）分别在1 630~1 699 nm、1 699~1 733 nm和1 662~1 690 nm波长区间内建模，3种水相组分模型的相关系数分别为0.999 3，0.986 3，0.999 8，校正均方根误差（RSMEC）分别为0.119，0.239，0.095。该模型用于预测集样品（混合标准溶液系列）的分析，并对3种水相组分的预测值和已知值进行线性拟合，所得相关系数大于0.980 0，预测均方根误差（RMSEP）分别为0.144，0.169和0.114。

Models based on near-infrared spectroscopy were used for predicting 3 components, including m-phenylenediamine, triethylamine and sodium dodecyl benzene sulfonate in water phase from the reverse osmosis composite membrane production line. The samples were analyzed with near infrared spectrometer. The raw near-infrared spectroscopy was pretreated by Savitziky-Golay smoothing-second order differential method, and used to build models of m-phenylenediamine, triethylamine and sodium dodecyl benzene sulfonate by partial least squares method (PLS) at wavelength ranges of 1 630-1 699 nm, 1 699-1 733 nm and 1 662-1 690 nm, respectively. The PLS factor numbers were chosen as 3, 4, 3 for m-phenylenediamine, triethylamine and sodium dodecyl benzene sulfonate, respectively. Correlation coefficients of the models of 3 water phase components were 0.999 3, 0.986 3, 0.999 8, and corrected root mean square errors (RSMEC) were 0.119, 0.239 and 0.095, respectively. The proposed models were used for the analysis of the prediction set samples (mixed standard solution series), and predicted values and known values were linearly fitted, and correlation coefficients of 3 water phase components obtained were greater than 0.980 0, and root mean square errors (RMSEP) of the prediction were 0.144, 0.169 and 0.114, respectively.