柴油中硫氮污染物的快速分离与光谱机器学习定量检测OA

Sulfur-and nitrogen-containing pollutants in diesel fuel not only reduce fuel quality and combustion efficiency,but also lead to the emission of harmful gases,such as sulfur oxides and nitrogen oxides,posing serious threats to environmental safety and human health.Conventional analytical techniques,including GC-MS,can provide highly accurate analyses.However,they are generally limited by bulky,expensive instrumentation,complex sample pretreatment,lengthy analysis cycles,and high operational requirements,which make them unsuitable for on-site rapid screening and large-scale routine analysis.Therefore,the development of a rapid,low-cost,portable,and reliable analytical strategy for diesel pollutants is of great practical significance.In this work,a rapid detection method for typical sulfur-and nitrogen-containing pollutants in diesel based on thin layer chromatography-surface enhanced Raman scattering(TLC-SERS)combined with machine learning was developed.A GF254 silica gel plate was used as the stationary phase,while n-hexane/petroleum ether in a volume ratio of 1∶6 served as the developing solvent to achieve rapid chromatographic separation of aniline(AN)and dibenzothiophene(DBT)within 3 min.8.1×10-8 mol/L gold nanoparticles(Au NPs)colloids solution were employed as the SERS enhancement substrate to improve SERS signal intensity and detection sensitivity.SEM and UV-Vis were utilized to characterize the micromorphology and optical absorption properties of the as-prepared Au NPs colloids,and Nano Measurer software was adopted to analyze the particle size distribution.The synthesized Au NPs colloids exhibit good dispersibility and uniform particle size distribution,which contribute to stable electromagnetic"hot spots"and enhanced SERS signal intensity.Experimental conditions,including Au NPs concentration,laser power,detection time,and developing solvent composition,were systematically optimized.The proposed TLC-SERS system shows excellent selectivity and anti-interference capability toward structurally similar sulfur-and nitrogen-containing compounds.TLC separation effectively reduces interference from the complex diesel matrix and enables reliable identification of target pollutants.In addition,the prepared SERS substrate demonstrates good uniformity and reproducibility,with a relative standard deviation of 3.7%for the characteristic peak intensity of 4-mercaptobenzoic acid.To further improve quantitative analysis capability,a partial least squares regression model(PLSR)was established using SERS spectral data collected from different concentration levels.The PLSR model exhibits excellent prediction performance,with determination coefficients higher than 0.99 and low prediction errors for both AN and DBT.The recoveries of AN and DBT in spiked diesel samples reach 104.85%and 102.71%,respectively,indicating satisfactory analytical accuracy in complex fuel matrices.The qualitative detection limit of the proposed method is as low as 10×10-6(w).Compared with conventional laboratory-based analytical instruments,the developed TLC-SERS method possesses obvious advantages in portability,rapid analysis,low reagent consumption,and reduced operational cost,which make it highly suitable for on-site diesel quality monitoring and preliminary large-scale screening.This work provides a promising analytical strategy for rapid fuel pollutant detection and demonstrates the application potential of TLC-SERS combined with machine learning in petroleum analysis and environmental monitoring.