Measuring the Visible Particles for Automated Online Particle Size Distribution Estimation


Optimisation and control of comminution, pelletization, and agglomeration processes is a complex task with large potential for gains in productivity, energy efficiency, and particle size distribution quality. One factor in realising these gains is the requirement for fully-automated online particle size distribution measurement. Moreover, online particle size distribution measurement that is based on the particles on the visible surface and can avoid or mitigate substantial errors that result from under-sizing overlapped particles and over-sizing areas of fines particles. Literature review indicates a body of substantially 2D photographic based particle size measurement systems that suffer from a large number of significant errors except in the simplest case of a monolayer of particles under good illumination with limited color variation in the material. In order to measure the visible particles on a pile there are a number of requirements including; avoiding particle delineation errors due to variation in material color and shadows, detect overlapping versus non- overlapping particles, and identify areas of fine particles as fines and not large rocks. Results are presented using a fully-automated online measurement system of crushed rock on conveyor belt using a 3D surface laser profile. The system provides continuous online measurement and results are shown on over four hours of production after the primary crusher. Size distribution results are calculated using volumetric estimation of the particles on the surface of the pile. The results show a strong capability to discriminate size variation especially in the 10% and 20% passing values. Furthermore results show that if areas of fines are misclassified as large rocks, and overlapped particles misclassified as small rocks, then a near complete loss of discrimination occurs for the P10 and P20 values which become largely constant values.

IMPC 2014

Thurley-IMPC2014-paperN30_final This paper is provided with the author's permission.