基于模糊聚类算法的大气粒子激光电离质谱数据分析

实验室自行研制了一台大气气溶胶飞行时间激光质谱仪(ATOFLMS)，它可以在线地对气溶胶单粒子进行物理和化学特性分析，利用双束连续激光对单个粒子的空气动力学粒径进行测量，并通过飞行时间完成单粒子化学成分的检测。该仪器在运行过程中将产生海量的实验数据，对这些数据的快速、自动处理并提取有价值的信息是整机系统的关键之一。文章介绍模糊聚类算法FCM(fuzzy c-means)在大气气溶胶单粒子聚类分析中的成功运用。利用该算法对连续24 h采集的室内空气气溶胶单粒子质谱数据进行了聚类分析，在得到的5个聚类结果中包含了无机的海盐粒子、矿物质粒子以及其他的三种二次气溶胶成分粒子类型。在对室内空气气溶胶粒子的粒径进行实时检测的结果表明室内可吸入颗粒物以细粒子为主，其中大于1 μm的粒子所占比重较小。小于1 μm的粒子均占95％以上， 在0.4～0.8 μm之间的粒子占据主要部分。

Data Analysis of Laser Desorption/Ionization Mass Spectrum of Atmospheric Aerosol Particles Using Fuzzy Clustering Algorithms

On-line measurement of size and composition of single particle using an aerosol time-of-flight Laser mass spectrometry (ATOFLMS) had been designed in our lab.Each particle’s aerodynamic diameter is determined by measuring the delay time between two continuous-wave lasers, A Nd∶YAG laser desorbs and ionizes molecules from the particle, and the time-of-flight mass spectrometer collects a mass spectrum of the generated ions.Then the composition of single particle is obtained.ATOFLMS generates large amount of data during the process period.How to process these data and extract valuable information is one of the key problems for the ATOFLMS.In this paper, the fuzzy clustering used to classify large numbers of mass spectral of air indoor by an ATOFLMS.Each revised spectrum is converted to a normalized 300-point vector, each point representing one mass unit.Then the positive ion mass spectra of a single particle are described as 300-dimensional data vectors using the ion masses as dimensions and the ion signal peak areas as values.The data vectors of all particles measured are written into a classification matrix.Each spectrum’s data was stored as one row in this matrix.The Fuzzy c-means algorithm is an iterative method starting the calculation with random class centers to find a substructure in the data.The procedure works in such a way that finally similar objects (particle spectra) have a minimum distance between their corresponding data vectors, on the one hand, and to the center of a cluster, on the other hand.So the aim of the iteration is to find local minima in the N-dimensional space where N is the number of evaluated peak masses.The particle data used in this study were collected over a period one day in Hefei.During the campaign, inorganic salts, mineral particles, and carbonaceous particles, with varying degrees of secondary components, were identified.The detection results of particle size exhibit that aerosol is predominanantly in the form of fine particles, and the particles whose diameter larger than 1 μm are scare.The particles whose diameter less than 1 μm are make up of 95% of the total particles, and these particles are major distributed in 0.4-0.8 μm.