Search

[Display omitted] • A new method is proposed for sizing aerosols without refractive index. • The LSIF is used to characterize unconstrained PSD and refractive index. • Deep learning is adopted to retrieve unconstrained PSD efficiently (about 100 ms). • An integrated and cost-effective prototype is designed and verified. Existing light-based aerosol size analyzers suffer from measurement errors or even fail to measure particle size distribution (PSD) when the refractive index is unknown. To overcome this limitation, we introduced a new aerosol sizing method without the prior refractive index using the light scattering intensity field (LSIF) and deep learning. The LSIF is the intensity distribution of the light scattered in all directions around aerosol particles, which can be used to simultaneously characterize the PSD and the refractive index. A PSD retrieval approach based on deep learning was developed to efficiently retrieve (about 100 ms) the unconstrained PSD (141 independent channels). Our proof-of-concept prototype was tested on various aerosol samples with different particle sizes and refractive indices, where the relative standard deviation (RSD) of the particle size was 1.94% in the test of monodisperse aerosols and the Kullback-Leibler divergence of PSD was 0.13 in the test of polydisperse aerosols. [ABSTRACT FROM AUTHOR]

• Light-scattering angular spectrum is proposed for high-resolution particle sizing. • A compacted and lens-free optical system is designed for LSAS sampling. • The particle size distribution is inversed with high accuracy (D KL < 0.05). • The sensor is low-cost and portable for in-situ aerosol contamination measurement. In the study of aerosols, great concern has been focused on the particle size distribution (PSD) as a crucial factor of the physical and chemical properties of the aerosols. There are strong and growing demands for on-line and in-situ aerosol analysis technologies, but existing portable aerosol sensors can hardly provide high-resolution and accurate PSD measurement with limited volume and cost. To address this challenge, we introduce the concept of a new optical aerosol sensing method using light-scattering angular spectrum (LSAS), which describes the angular dependence of the scattering light intensity. The key novelty is a compacted LSAS sampling module for collecting the scattering light up to the observing angles. According to the wide-range LSAS sampled by our sensor, high-resolution PSD is predicted by non-negative regularization inversion algorithm. To validate the performance of our sensor, monodisperse di-ethyl-hexyl-sebacate aerosols (DEHS) samples of different sizes and smokes of smoldering cotton rope were tested. The maximum relative error (RE) of LSAS was smaller than 12.02%, while the Kullback-Leibler Divergence (D KL) between inversed and reference PSDs were smaller than 0.05. Without using any particulate pre-separation module and optical lens, we build a novel low-cost and miniaturized sensor only composed of a mirror, a linear charge coupled device (CCD) and a semiconductor laser, which has significant potential for high-resolution aerosol sizing in routine field measurements outside a laboratory. [ABSTRACT FROM AUTHOR]