Aerosols Cluster Methods Study And Its Application Based On Single Particle Aerosol Mass Spectra Information

Atmospheric aerosol have an important impact on human health, visibility,global climate and environment changes, thus it was gotten more and more attentionby people. Traditional aerosol analytical methods are primary off-line analyticalmethods and bulk analysis, such as scanning electron microscopy and IC and so on.However, these methods need to pre-collect aerosol particles, which will take toomuch labor power and time. Moreover, aerosols would condensation; undergoreaction, gas to solid transition during the process of sampling, storing, andconservation, which will make the results deviate the original constitution condition.Single particle aerosol mass spectrometry (SPAMS), as a new-styled aerosolanalytical tool, has high time and spatial resolution, which can detect the aerosolparticles from the single particle level. The instrument can produce large amounts ofdata (hundreds of mass spectra per minute) in the process of sampling. Relying onmanual processing is not only slow, heavy workload, prone to error, but greatlydiminishing the advantage of real-time measurement of the instrument, thus it isessential to develop automatic, efficient methods of data analysis.This thesis is primary divided into three parts: principal and characterization ofSPAMS, aerosols cluster methods study, aerosols source apportionment:(1) Principle and function of Single particle aerosol mass spectrometry.Structures of the single particle aerosol mass spectrometry were described,including aerosol inlet system, two-laser sizing system, laserdesorption/ionization system and TOFMS system. Besides, size calibrationmethods, data acquisition system and data processing methods were alsointroduced.(2) Aerosols cluster methods research. Based on MATLAB language hierarchicalclustering、 density-based clustering and partitioning clustering includingART-2a、DBCAM and k-means were written, and were compared through the data obtained by SPAMS. DBCAM algorithm was written, and then the clustertheory、cluster speed and cluster results of ART-2a and DBCAM were compared.Furthermore, the improved method of ART-2a through square root of particles’eigenvectors was demonstrated, the experiments show that this method canimprove the precision greatly. This is the third chapter of the thesis, which iscompleted alone(3) Aerosols source apportionment. This chapter describes the sourceapportionment experimental methods and experimental results. Theexperimental analysis automobile exhaust particles, biomass burning particlesand dust particles including their size distribution and chemical compositioninformation, and automobile exhaust and dust particles were analyzed by PMF.The experimental results show that there are significantly differences of thediesel and petrol vehicle emissions in the particle size of the particles and theircomposition, the composition of the different biomass combustion are alsodifferent, different sources of dust particle component correspond to significantdifferences. The aerosol source apportionment experiment not only can have acertain understanding of the information on specific emission sources ofaerosol particles, and can provide some theoretical support for the analysis ofenvironmental pollution. This is the third chapter of the thesis, which iscompleted alone.