论文部分内容阅读
对二维相关光谱选择波数结合偏最小二乘判别分析检测水中杀螟硫磷残留进行了可行性研究。先用原始近红外光谱构建的二维相关同步光谱选择了(5365~7800)cm~(-1)来建立偏最小二乘判别模型。总共241个样品中,随机选择了161个纯水及含杀螟硫磷的水溶液样品作为训练集,另外80个样品作为验证集。在(5365~7800)cm~(-1)内,前处理方法为数据中心化时,训练集中的2个1μg/mL的水溶液样品被误判;前处理方法为标准正态变量转换或多元散射校正(含数据中心化)时,预测集中都有2个纯水样品被误判:总正确率都为99.2%。能检测出的水中的杀螟硫磷的最低浓度为1μg/mL。建立了一种快速判别水中杀螟硫磷残留的方法。
Feasibility studies on the determination of fenitrothion in water by two-dimensional correlation spectral selection wave number combined with partial least-squares discriminant analysis. The partial least squares discriminant model was established by using the two-dimensional correlation spectra constructed by near infrared spectroscopy (5365 ~ 7800) cm ~ (-1). Of a total of 241 samples, 161 pure water and fenitrothion-containing aqueous samples were randomly selected as the training set, and another 80 samples were used as validation sets. In the range of (5365-7800) cm ~ (-1), the two 1μg / mL aqueous samples in the training set were misjudged when the preprocessing method was data centripetal; the pretreatment methods were standard normal transformation or multiple scattering In the calibration (including data centering), two pure water samples were misjudged in the prediction set: the total correct rate was 99.2%. The lowest concentration of fenitrothion in water that can be detected is 1 μg / mL. A rapid method for the determination of fenitofos in water was established.