大气气溶胶飞行时间质谱仪在对气溶胶粒子的测量过程中会产生大量包含单 粒子的化学成分和粒径信息的数据。介绍了DBSCAN对三种混合气溶胶单粒子质谱数据进行 聚类分析的研究,同以往的质谱分析方法相比,DBSCAN利用类的高密度连通性,可 以快速发现各种形状的类,更有利于质谱数据的分析。实验结果表明,DBSCAN算法 可以成功地对这三类物质进行分类。

Aerosol laser time-of-flight mass spectrometer(ALTOFMS) generates large amounts of data during monitoring atmospheric aerosol particles, which can be used to determine the chemical composition and diameter of individual aerosol particle in real-time. A density based spatial clustering of application with noise, DBSCAN algorithm, is used to classify mixed mass spectral of aerosol of NaCl, CaCl and FeCl with an ALTOFMS. Compared with previous algorithms, DBSCAN can identify clusters with arbitrary shapes and sizes, and it produces clusters without overlap and the continua between different types. Experimental results show that DBSCAN can classify the data into three clusters. The information can provide insight into particle sources, chemical transformations between particles and gases and distribution among or within particles.