Inversion of particle-size distribution with improved conjugate gradient algorithm

Abstract

In order to ameliorate convergence of the algorithm to invert the particle-size distribution (PSD) from laser diffraction data, an improved conjugate gradient algorithm (ICGA) is proposed. This method is independent of any given a priori information of the particle-size distribution. In the algorithm, each objective function is constructed according to an equation of the system of equations. Then iterations are carried out continuously between objective functions by choosing conjugate gradient directions, and thus the objective functions are tied up. An iteration step-adjusting parameter is introduced, which depends on the row index vectors of the matrix equation. Two narrowly distributed particulate-certified reference materials, their mixture, and a widely distributed particle plate are used as samples to verify the algorithm. Experimental results show that the ICGA is sufficiently convergent and that the convergence points are stable. The presented method can be used to invert unimodal and multimodal PSD with high precision.