水体有机污染物浓度检测中的紫外光谱分析方法

Ultraviolet Spectrum Analysis Methods for Detecting the Concentration of Organic Pollutants in Water

水体中的大多数有机污染物在紫外区域有较强的吸收,因此可利用紫外吸光度检测水体中的有机污染物浓度。在检测过程中,通过平滑、导数、标准正态变量变换等光谱预处理后,采用主元回归、偏最小二乘、支持向量机等方法建立回归模型,并由该模型依据待测样本的紫外光谱数据计算出有机污染物浓度。为了探究不同的预处理方法、建模方法在有机污染物浓度检测中的特点及内在联系,本文对一组来自污水厂进出水的光谱数据采用不同的预处理和建模方法进行实验研究,研究结果表明:当获得的原始数据较好时,可以直接进行建模,进行光谱预处理反而会使模型效果下降;由于本实验中污水的样本数小于光谱数据点数,所以最小二乘支持向量机更适合于本组实验样本。

The absorption rate of ultraviolet could be used to measure the concentration of organic pollutant, as most of the organic pollutant has stronger absorption rate in ultraviolet region in water. In the present paper, principal component regression （PCR）, partial least squares （PLS）, and support vector machine （SVM） were respectively used to model a regression model after the spectrum preprocessing, such as smoothing, derivation, standard normal variate transformation （SNV）, etc. Then, the concentration of organic pollutant could be measured via the ultraviolet spectrum and the regression model. In the experiments, a group of water samples from the wastewater treatment process were used to verify the effects of the various preprocessing and modeling approaches. The results showed that for the good spectrum data, direct modeling without the spectrum pretreatment could be used since the pretreatment would worsen the results. LSSVM approach is more applicable in the case of small-size samples.