Research On Key Issues Of Water Quality Ultraviolet-Visible Spectroscopy Pretreatment

Water health is related to human life and health.With the development of economy,the situation of water pollution is not optimistic.The monitoring and management of water environment is an important means of water resources protection,and the research of water quality detection technology is of great significance.Traditional chemical methods have the problems of long detection cycle,expensive equipment,complicated operation,need to use chemical reagents,and easy to cause secondary pollution.Direct ultraviolet-visible spectroscopy has the advantages of no secondary pollution,fast analysis speed,and in-situ measurement.In recent years,it has become a research hotspot in the field of water quality testing.However,the water quality ultraviolet-visible spectrum detection system is susceptible to noise interference from the instrument itself and the external environment,and the measured spectrum data has a large number of system and stray light noise problems.At the same time,the collected spectral data is redundant and inconvenient for quantitative analysis of water quality,so it is necessary to preprocess the water quality spectral data to provide guarantee for subsequent water quality early warning and analysis.This dissertation has carried out research work on two key technical issues of water quality spectrum data preprocessing.The main research contents are as follows:(1)Design of water quality detection system based on ultraviolet-visible spectroscopy.Using Ocean Optics DH2000 deuterium tungsten halogen lamp combined light source and Japanese Hamamatsu TM series C10082 CAH spectrometer to build a water quality testing experimental platform,equipped with different concentrations of potassium hydrogen phthalate standard solutions,collecting water samples of sewage treatment plant outlets,drainage ditch,factory wastewater and other water were collected,and spectrum experiments were performed on the water samples to obtain the spectral data of the water samples.(2)Research on the denoising of water quality spectrum data by using genetic algorithm to optimize the wavelet threshold.Aiming at the water quality detection system of ultraviolet-visible spectroscopy which is susceptible to noise interference from the instrument itself and the external environment,based on the analysis of the noise source,a de-noising method applying genetic algorithm to wavelet threshold optimization is proposed.Comparative experiments were carried out with other traditional methods,and the method was also applied to the actual water sample spectral processing process.The experimental results show that the denoising method of optimizing wavelet threshold by genetic algorithm not only suppresses the noise in the spectral data,but also improves The system accuracy is improved,and a new solution is provided for the denoising treatment of water quality spectrum by ultraviolet-visible spectroscopy.(3)Research on dimensionality reduction of water quality ultraviolet-visible spectrum data.In view of the high-dimensionality of the data collected by the ultraviolet-visible spectroscopy detection system,the included spectral information is redundant and inconvenient for analysis.Principal component analysis is used to reduce the dimensionality of the collected spectral data,which can retain most of the useful information while reducing the dimensionality.The important information of the data is extracted,the data structure and computational complexity are simplified.(4)Research on classification of water quality ultraviolet-visible spectrum data.The classification of the spectral data of existing water samples is helpful for the rapid identification of contaminated water bodies,and is of great significance for subsequent water resources management and governance.Based on the poor classification effect of a single model on water quality spectral data,this research uses principal component analysis and K-nearest neighbor classification to establish a water quality discrimination model,conducts classification experiments on actual water samples,and compares with BP neural network and random forest algorithm.The experimental results show that,compared with other methods,principal component analysis combined with K-nearest neighbor classification has higher classification accuracy,can quickly and accurately identify the target water body,and has a low misjudgment rate and short calculation time.Aiming at the two key technical problems of water quality UV-visible spectrum pretreatment,this paper proposes a genetic algorithm to optimize wavelet threshold denoising and principal component analysis combined with K-nearest neighbor classification and discrimination.It has important guiding significance for the data denoising and classification model of spectrometers,and provides a new solution for the data preprocessing of the spectroscopy online detection system.