Optical Diffraction Inversion Algorithm Research Of Atmospheric Particulate

With the improvement of requirements for environment protection,pollution discharge standard is more and more strict all around the world,for technical support,major research institutions and foreign companies have also developed a variety of pollutants corresponding monitoring technology and equipment,including fixed pollution source emission,especially the key fine/ultrafine particles.This Has become the current focus.This paper mainly studies how to accurately measure smoke particle size and particle size distribution,people attaches great importance to the studies of particle measurement,this has become a important branch in modern modern surveying,there are a lot of particle size distribution measurement methods,from the perspective of the historical developing process of particle size distribution measurement,such as screening method,sedimentation method,imaging method and microscope method,electric induction method and laser scattering method,etc.Among them,the laser scattering method is widely used for the measuring speed,good repeatability and suitable for online advantages of non-contact measurement,based on the Fraunhofer diffraction pattern,we do the simulation on matlab for the scattering of particle swarm,get the diffraction intensity function to help us understand the diffraction pattern of particle swarm,and as a guide for our optical device selection.After completing the forward simulation,we do the inversion simulation based on the diffraction intensity.Inversion method includes no model inversion and the model inversion,no model inversion includes the machine learning regularization: TSVD algorithm and the Tikhonov algorithm,and we use the L-curve algorithm for the determination of regularization parameters;the model inversion is R-R distribution model.In the end,we get the conclusion: Model inversion in unimodal R-R distribution inversion effect is good,but the response time is too long when there are many peaks,and it's difficult to determine actual distribution model;No model inversion,the response time is too long in Chihine method,and the Tikhonov regularization inversion algorithm is best,and it's most suitable for our project.