基于支持向量机测定水质中硝酸盐含量

Determination of Nitrate in Water Based on Support Vector Machine

针对水质中被测物硝酸盐与干扰物铁离子、铬离子紫外吸收谱线严重重叠的问题,将支持向量机应用于混合重叠光谱分析中,利用核函数将重叠的光谱数据进行高维空间变换后求得SVM回归模型,实现硝酸盐含量的光谱检测。实验结果表明:硝酸盐浓度在0.5—10mg/L范围内,所测得含量的最大相对误差为3.2%;平均回收率为100.9%,本法与传统的方法相比,要求条件较低,不需要物理或化学的分离,分析速度快,有望应用于在线监测污水中硝酸盐的含量。

The UV absorption spectrum of measured nitrate and disturbing ferric ion,chromic ion are greatly overlapped. A method is proposed with support vector machine （SVM） to the analysis of overlapped absorption spectrum in this paper. The overlapped absorption spectrum is transformed in high-dimension space by kernel function,and then the SVM regression model is built to measure the nitrate content. The maximum relative error of nitrate content is 3.2% and the average recovery of nitrate content is 100. 9% in the range of 0. 5-10mg/L. Compared with traditional methods, this method is simple and requires no physical or chemical separation with high analysis speed and accuracy. This method is expected to be applied to the online monitoring of the nitrate content of waste water.