Research On Quantitative Method Of Sulfate Based On Raman Spectroscopy

The "gypsum rain" phenomenon is a pollution phenomenon in which fine gypsum particles generated by a thermal power plant using a wet desulfurization system are discharged to the atmosphere through a tail flue.Studies have shown that the higher the content of fine grains in the droplets of the desulfurization slurry,the more solids are released into the atmosphere.Therefore,it is necessary to control the crystal size in the original slurry,which is decided by the supersaturation,and further reflect by the concentration of SO42-in the slurry.In recent years,due to its rapid and non-contact characteristics,Raman spectroscopy is has gradually become a measure means for monitoring and controlling the crystallization process in industrial crystallization.The quantitative theoretical basis of Raman spectroscopy provides the possibility of monitoring the real-time concentration of gypsum crystallization process in the desulfurization tower as a measurement tool.This thesis studies the application of Raman spectroscopy in establishing a quantitative model of sulfate solution,explores the effects of different spectral data pre-processing methods and different modeling methods on the model’s predictive ability,and finally selects the best pre-processing method and construction method.Model type and design experiments to explore the stability of the model.In this thesis,the Raman spectra of gypsum solids and solutions are measured,and the vibration modes of their Raman response are analyzed.The establishment of the sulfate quantitative model first needs to analyze the original Raman spectrum data,select the appropriate pretreatment method,and extract the effective information in the spectrum.Based on this,this thesis designed to set up 55 sets of correction sets and 33 sets of prediction sets in the experiment,using different preprocessing methods to reduce the noise of the original spectral data,smoothing the baseline,etc.,comparing a single preprocessing method and a variety of different preprocessing The processing effect of the combination method on the spectral data determines that the combination method of convolution smoothing,standard normal transformation and second derivative is the optimal pretreatment method.In this thesis,two methods of model establishment are selected-in the establishment of quantitative models:univariate linear regression and partial least squares.For the establishment of a univariate linear regression model,the method of single peak fitting and internal standard fitting is selected and compared.The results show that the internal standard method reduces the shadow noise information generated by laser power and environmental changes.thereby improving the predictive ability of the model.For partial least squares modeling,the optimal quantitative analysis model was determined from the analysis of the number of principal components,the choice of wavenumber range and the choice of pretreatment methods.In this thesis,an experiment was designed to consider the influence of temperature and the addition of impurity ions on the model’s predictability.Four different temperatures were set in the experiment,and the correlation between the spectral matrix and the two variables of concentration and temperature was investigated by partial least squares modeling.The results showed that there was no obvious linear relationship between the spectral intensity information and the temperature,which was verified With the verification of the set,the quantitative model can still predict well,indicating that the accuracy of the quantitative model is basically not affected by temperature.In addition,a control group experiment was set up to verify the influence of the introduction of two impurity ions on the model.The results show that the impurity ions increase the useless information in the spectral data,which is not conducive to the establishment of the model and has an impact on the predictive ability of the quantitative model built.