RESUMO
Detection methods for microbiological aerosols based on single particle mass spectrometry (SPAMS) and a fluorescent aerosol particle sizer (FLAPS) have been developed progressively. However, they encounter interference and inefficiency issues. By merging FLAPS and SPAMS technologies, the majority of inorganic ambient aerosols may be eliminated by the FLAPS, thus resolving SPAMS' large data volume. SPAMS, on the other hand, may eliminate the secondary fluorescence interference that plagues the FLAPS. With the addition of the enhanced machine learning classifier, it is possible to extract microbial aerosol signals more precisely. In this work, a FLAPS-SPAMS instrument and a Random Forest classifier based on Kendall's correlation expansion training set approach were built. In addition to analyzing the outdoor microbial proportions, the interference components of non-microbial fluorescent particles were also examined. Results indicate that the fraction of outdoor microbial aerosols in fluorescent particles is 25.72% or roughly 2.57% of total particles. Traditional ART-2A algorithm and semi-empirical feature clustering approaches were used to identify the interference categories of abiotic fluorescent particles, which were mostly constituted of EC/OC, LPG/LNG exhaust, heavy metal organics, nicotine, vinylpyridine, polycyclic aromatic hydrocarbons (PAHs), and polymers, accounting for 68.51% of fluorescent particles.