Use of neural networks in the analysis of particle size distribution by laser diffraction: tests with different particle systems

Abstract

The application of forward light scattering methods for estimating the particle size distribution (PSD) is usually limited by the occurrence of multiple scattering, which affects the angular distribution of light in highly concentrated suspensions, thus resulting in false calculations by the conventionally adopted algorithms. In this paper, a previously proposed neural network-based method is tested with different particle systems, in order to evaluate its applicability. In the first step of the study, experiments were carried out with solid–liquid suspensions having different characteristics of particle shape and size distribution, under varying solid concentrations. The experimental results, consisting of the angular distribution of light intensity, particle shape and suspension concentration, were used as input data in the fitting of neural network models (NN) that replaced the optical model to provide the PSD. The reference values of particle shape and PSD for the NN fitting were based on image analysis. Comparisons between the PSD values computed by the NN model and the reference values indicate that the method can be used in monitoring the PSD of particles with different shapes in highly concentrated suspensions, thus extending the range of application of forward laser diffraction to a number of systems with industrial interest.