Application of Machine Learning in Determination of Nitrate Nitrogen Based on Ultraviolet Spectrophotometry
LIU Si-xiang1, 2, FAN Wei-hua1, 2, GUO Hui1, ZHAO Hui1, JIN Qing-hui1, 2*
1. State Key Laboratory of Transducer Technology, Shanghai Institute of Microsystem and Information Technology, Chinese Academy of Sciences, Shanghai 200050, China
2. University of Chinese Academy of Sciences, Beijing 100049, China
Abstract:Ultraviolet spectrophotometry has been widely applied in determination of water quality parameters because of its advantagous properties compared to chemical method, such as high efficiency, easy operation and non-secondary pollutions. Nitrate nitrogen is one of major pollutants in waste water. The standard ultraviolet spectrophotometry used to determinate the concentration of nitrate nitrogen in water is firstly to determinate the absorbances at wavelength 220 and 275 nm, which is used to calibrate the former, followed by the plotting of standard curve. While the linear equation described by Lambert-Beer's law and the linearity in the superposition of absorbances of various substance, on which the standard ultraviolet spectrophotometry based, are not fitted well anymore with the increase of concentration. In addition, it was found to be difficult to construct absorption model of nitrate solution at wavelength 220 nm in experiment. To overcome the disadvantages in single-wavelength or double-wavalengths spectrophotometry, the absorbances at the wavelength that covered by the absorption peak are introduced into the construction of the model and to avoid the increase of the model complexification resulted by the introduction of more wavelengths, we run the principal components analysis on the original absorbances data. The data with dimensions compressed from 107 down to 4 after process construct the absorbance model using locally weighted linear regression. Good performance were achieved in both training samples set and test samples set using this model and it was able to deal with the non-linear relation between the absorbance and concentration thus raised the upper range limit concentrations of nitrate nitrogen up to hundreds mg·L-1 from 4 mg·L-1 defined in the standard method. Meanwhile the principle and procedure of this analytical method are suitable for the absorbance model construction of other solutions.
Key words:Ultraviolet spectrophotometry; Nitrate nitrogen; Principal component analysis; Locally weighted linear regression
刘思乡,范卫华,郭 慧,赵 辉,金庆辉. 机器学习在紫外法测定硝酸盐氮浓度中的应用[J]. 光谱学与光谱分析, 2017, 37(04): 1179-1182.
LIU Si-xiang, FAN Wei-hua, GUO Hui, ZHAO Hui, JIN Qing-hui. Application of Machine Learning in Determination of Nitrate Nitrogen Based on Ultraviolet Spectrophotometry. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2017, 37(04): 1179-1182.
[1] Wang L L, Liu X H, Shi X X, et al. Advanced Materials Research, 2013, 726-731: 1534.
[2] Agustsson J, Akermann O, Barry D A, et al. Environ. Sci. Process Impacts, 2014, 16(8): 1897.
[3] Drolc A, Janez V. Bioresour. Technol., 2010, 101(11): 4228.
[4] Huang Guoxin, Huang Yuangying, Hu Hongyan, et al. Chemosphere, 2015, 130: 8.
[5] Bro R, A K Smilde. Analytical Methods, 2014, 6(9): 2812.
[6] Douglas C M, Elizabeth A P, G. Geoffrey V. Introduction to Linear Regression Analysis, Fifth Edition. New Jersey: John Wiley & Sons, 2012. 190.
[7] Hou Dibo, Liu Shui, Zhang Jian, et al. Journal of Spectroscopy, 2014, 2014: 1.
[8] Wong T. Pattern Recognition, 2015, 48(9): 2839.