The Research Of Online Monitoring Method For Ammonia & Nitrogen

Water is an important survival material, but with development of economic, water pollution which results in the serious consequences of cyanobacteria and red tide eutrophication are spreading and intensifying continually during the eutrophication. Nitrogen is one of the important nutrients which induces the serious consequences of cyanobacteria and red tide. So measurement of ammonia and nitrogen has great practical significance. It is not only difficult to achieve high-frequency time, short cycle and real-time but also has heavy workload by conventional manual sampling and monitoring. Therefore the online automatic monitoring of ammonia and nitrogen is essential.In order to determine the ammonia and nitrogen, first of all, various forms of ammonia and nitrogen are oxide into nitrate, and then the absorption spectrum of water sample is obtained by using ultraviolet spectrophotometer. The UV absorption spectrum of measuring nitrate and disturbing ferric ion, chromic ion are greatly overlapped. Wavelet transform is used to analyze the UV absorption spectrum of the mixture solution of nitrate, ferric ion and chromic ion in terms of the standard solution of nitrate. The eigen wavelet coefficients correlated with the content of nitrate are obtained and the nitrate content can be detected. The experimental results show that it is linear (r = 1.0008) during the nitrate content and its UV absorption spectrum and the average recovery rate reaches 100.17% according to the linear regression equation established by eigen wavelet coefficients insofar as 0.5-2.5mg/L. But this method is not suitable to solve the non-linear relationship between content and absorbency.The UV absorption spectrum of measuring nitrate and disturbing ferric ion, chromic ion are greatly overlapped and the relationship is non-linear between nitrate content and absorbency. To solve this problem, 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 to high-dimension space by kernel function, and then the SVM regression model is built to measure the nitrate content. The experiment results show that the maximum relative error of the nitrate content is 3.2% and the average recovery rate of the nitrate content is 100.9% insofar as 0.5-10 mg/L. Compared with traditional methods, wavelet transforms and SVM are simpler and don't requires 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 wastewater.