Research On Quantitative Analysis And Pollution Source Identification Method Of Polycyclic Aromatic Hydrocarbons In Atmospheric Deposition Particles Based On Raman Spectroscopy And Chemometrics

Polycyclic aromatic hydrocarbons(PAHs)are the main organic pollutants in atmospheric particulate matter,which are teratogenic and carcinogenic.Therefore,rapid and accurate analysis of the content and pollution sources of PAHs in atmospheric particulate matter is of great significance to the on-site analysis and risk assessment of atmospheric pollution.Taking atmospheric deposition particles as the research object,the quantitative and discrimination analysis method of PAHs is studied based on Raman spectroscopy and chemometrics in this work.The effects of different pretreatment methods and variable selection methods on the performance of the model were investigated.Furthermore,a fast and accurate method for PAHs content analysis and pollution source discrimination in atmospheric deposition particles was constructed based on the combination of Raman spectroscopy and chemometrics,of which will provide new ideas and methods for precise discrimination and comprehensive prevention and control of compound atmospheric pollution sources.The main research contents are as follows:Taking atmospheric deposition particles as the research object,a rapid quantitative analysis method for fluoranthene and anthracene in atmospheric deposition particles was established based on the combination of Raman spectroscopy and partial least squares(PLS).First,the effects of different spectral preprocessing methods(multivariate scattering correction,standard normal variation,first-order derivation and Savitzky-Golay smoothing)and different variable selection methods(particle swarm optimization and genetic algorithm)on the prediction performance of PLS calibration model were investigated.Then,a PLS calibration model was constructed based on the optimal input variables.It shown that PLS calibration model constructed based on standard normal variation combined with particle swarm optimization had excellent prediction ability(fluoranthene:R~2=0.9427,RMSE=2.0511 mg/g;anthracene:R~2=0.9180,RMSE=2.4159 mg/g).Compared with full spectrum,appropriate variable selection methods can help improve the prediction ability of model.Taking the ratio analysis of PAHs in atmospheric deposition particles as the research purpose,a method for analyzing the ratio of fluoranthene/(fluoranthene+pyrene)in atmospheric deposition particles was established based on combination of Raman spectroscopy and random forest(RF).First,the effect of different preprocessing methods on the prediction results of RF calibration model was investigated,and variable importance threshold was further selected and optimized.Then,the RF and PLS calibration models were established based on the optimal preprocessing method and variable importance threshold.The results shown that,compared with PLS,the RF calibration model had better prediction ability(R~2=0.9146,RMSE=0.0799).Therefore,RF combined with variable importance selection can significantly improve the prediction performance of the model,which is a feasible method of PAHs ratio analysis in atmospheric deposition particles.Taking atmospheric deposition particles as the research object,a method for discriminating pollution sources of PAHs in atmospheric deposition particles was constructed based on Raman spectroscopy combined with RF.First,unsupervised discrimination methods(principal component analysis and K-means method)of pollution sources were established.Further,the supervised RF discrimination model was established,and the effect of different preprocessing methods on the prediction performance of RF discrimination model was explored.Finally,the supervised RF,PLS and extreme learning machine discrimination models were constructed under the optimized preprocessing method and model parameters.It shown that the unsupervised discrimination methods were difficult to discriminate samples of different pollution sources,and the results of supervised discrimination analysis were more excellent,among which the prediction ability of RF discrimination model was the best(accuracy=0.9333).Therefore,the combination of Raman spectroscopy and RF based on the characteristic ratio method is a potential method for discriminating the pollution sources of PAHs in atmospheric deposition particles,which is expected to realize the rapid online discrimination of atmospheric pollution sources.