Research On Fusion Processing Method Of Multi-parameter Data For Detecting Surface Water By Spectroscopy

For water quality testing,chemical methods are commonly used,but this method has the potential to cause secondary pollution;and the measurement cycle is too long.It has the shortcomings of low sensitivity and cannot be reused in biosensing.The measurement process of chromatography and mass spectrometry in physical methods is complicated,requires professionals to operate,and the equipment is relatively expensive.In recent years,spectroscopy has been widely used in water quality testing.It has the advantages of fast speed,no secondary pollution,fingerprint detection,and traceability of pollution.It has become a research hotspot in the field of water quality testing.The 24 parameter indicators in surface water contain most of the possible pollutants in production and life,which can effectively measure the degree of water pollution.Based on the distinction of surface water parameter indexes,some researches have been carried out on the measurement of certain specific parameters.The main research work of the thesis is as follows:1)Design of experimental system for detecting surface water by spectroscopy.The high-power deuterium halogen lamp combined light source combined with the PG2000pro spectrometer manufacturer is used to build the ultraviolet-visible spectroscopy water quality detection system;the ATR20202 power manufacturer is 600mW and the narrow linewidth laser with wavelength of 785nm is used.The spectrometer adopts the ATR3 0007-785 high-resolution optical fiber spectrometer to build a Raman spectroscopy water quality detection system.Configure the standard solution of different concentrations in the experiment to simulate surface water parameters and obtain spectral data.2)Research on the method of multi-source spectroscopy for parameter measurement.Aiming at the problem that UV-Vis spectroscopy is not accurate enough to predict water quality parameters,the method of multi-spectral data fusion of Raman spectroscopy and UV-Vis spectroscopy is adopted to improve the accuracy of measuring water quality parameters.In the experiment,the data fusion strategy is the combination of the importance analysis of projection variables and the partial least squares regression method,and the information of two different spectra is applied:Raman and ultraviolet-visible spectroscopy.Four different modelings were carried out for the low-concentration COD value of 5-100mg/L and the NO3-N of 0.1-20mg/L.The multi-source spectral model is a feature-level fusion model that uses the normalization method to remove dimensions and cascades into a new data set matrix.The feature-level fusion model is used to filter out the feature variables through the importance of projection variables.Finally,the processed matrix is combined with the partial least squares regression method.The results show that in the data-level fusion data,the COD R2=0.9668,RMSECV=1.6113,NO3-N R2=0.9633,RMSECV=0.5968.In the feature-level fusion,the COD’s R2=0.9766,RMSECV=1.2957;the NO3-N’s R2=1,RMSECV=0.24.The fused model has stronger monitoring capabilities and is closer to the real value.3)Research on the classification of water quality by multiple parameters of surface water.To solve the problem of imperfect water quality classification and evaluation,a research on high conflict water quality multi-parameter data fusion method based on interval evidence theory is proposed.The traditional D-S evidence theory is combined with interval numbers,and high conflict correction is adopted.Achieve a classification result that can be directly viewed.The experiment provides a basic probability distribution generation method based on interval similarity,and is adjusted with a coefficientα,which can keep the most original data while modifying conflicting data.After calculation,the water quality classifications I to V are 0.3516,0.5224,0.0913,0.0247,0.0099,the probability of uncertainty is 0,and the value with the greatest reliability is the current water quality classification.The integrated reliability can clearly understand the probability of each level of support,and can quickly determine the current level of water quality.Provides a new idea for water quality classification.The thesis discusses two methods of spectroscopic inspection of surface water.By studying the multi-source spectra of Raman spectroscopy and ultraviolet-visible spectroscopy,the importance of using the feature-level fusion method of data fusion to carry out the light source Screening to achieve a certain dimension retention to determine the parameter concentration.Through the decision-level fusion of multiple parameters,the probability classification of surface water quality is realized.