Research On Aerosol Sensing Method Based On Fast Light Scattering Theory Algorithm

With the global industrialization development,aerosols,known as “haze” and“PM2.5”,have been affecting human life activities and health through different mechanisms.Recently,aerosols have again attracted much attention as one of the virus transmission routes.Studies have shown that the particle size and distribution characteristics of aerosols can estimate their residence time in the air,floating distance,the difficulty of entering the human body,and the location of deposition.Optical scattering is widely used in situ measurement of aerosols due to its non-invasive,non-destructive and high sensitivity.However,the most serious challenge of this method is that the classical Mie scattering theory model involves high order recurrence relations of Bessel,Hankel and Legendre functions,which has high computational complexity and is difficult to be applied to embedded computing chips.In addition,Mie scattering theory is a strict solution to Maxwell’s equations under the assumption of spherical particles,which means that it cannot be applied to non-spherical particles.Moreover,in Mie scattering theory,there are many parameters that affect the scattering process,and the serious coupling between the parameters leads to difficulties in retrieving the information of aerosol characteristic size and particle size distribution(PSD).To address the above issues,this thesis devotes to investigating the low-complexity light scattering model for aerosol sensing.The contents of this thesis are listed as follows:Firstly,aiming at the problem that the classical Mie scattering model has high computational complexity and cannot be used for non-spherical particles,the numerical characteristics of the Mie scattering model are analyzed,and the relationship between the scattered light intensity and the particle size characteristics is discussed.Then,a low-complexity Mie scattering algorithm(LCMSA)based on the “volume-surface area response” of particles is proposed,and the algorithm is extended to non-spherical particles.On this basis,the stability and reliability of the algorithm under different incident light wavelength,scattering angle,and particle refractive index are investigated.The calculation time of LCMSA on PC is 28 times faster than that of the classical Mie scattering model algorithm.The average error of the total scattered light intensity calculated by LCMSA on STM32H7 and TMS320C6416 T DSP for aerosol particles under different PSD is 1.50%.Secondly,aiming at the problem of large errors in traditional measurement methods of aerosol Sauter mean size,a low-error general calculation method of aerosol Sauter mean size based on LCMSA is proposed.In this method,the integral term in the Sauter mean size definition formula is directly related to the scattered light intensity through the correlation between the scattered light intensity and the particle volume-surface area in LCMSA.The proposed method not only reduces the error in the calculation of the Sauter mean size based on the traditional “three-stage” model in principle,but also extends the calculation of the Sauter mean size to non-spherical particles.In addition,the method of joint calculation of particle ovality and Sauter mean size for non-spherical particle is investigated.The spatial distribution characteristics of scattered light intensity of non-spherical particles under multiple scattering angles are decomposed into their corresponding ovality,and then the equivalent Sauter mean size is calculated by LCMSA under the parameters of non-spherical particles.The proposed method is used in typical non-spherical aerosol N-Heptane combustion smoke experiments.The results demonstrate that the ovality of smoke particles is calculated to be 10:1.The average error of multiple experimental measurements of Sauter mean size is only 11.97% which are 7.63% and12.36% lower than that of directly equating them to spherical particles and based on the traditional “three-stage” model,respectively.Thirdly,aiming at the problem that the traditional inversion methods of aerosol PSD cannot balance the accuracy and computational complexity,a model-free inversion method of aerosol PSD based on incident light wavelength scattering scanning is proposed.Then,LCMSA is introduced to expand the actual measured limited light intensity information,and thus an aerosol PSD inversion method with computational scanning instead of physical channel scanning is realized.The error of this method is compared with the modeled inversion method under various PSDs.The results show that the average inversion error of the proposed method using five incident lights under simple distribution is only 13.10%,which is only 0.53% higher than the modeled PSD inversion error when the models are correctly matched,and 79.95% lower than the inversion error when they are mismatched.For complex PSDs where it is difficult to pre-define the distribution model,the PSDs curves obtained by the inversion of this method have the same double-peaked shape as the real distribution.Based on the method,an optical scattering sensor is developed and tested on the aerosol integrated experimental platform for aerosols generated by smouldering cotton smoke and aerosol generator.The experimental results show that the PSD inversion results completely preserve the real distribution characteristics of the aerosol.