A particle packing algorithm for pellet design with a predetermined size distribution

In this paper a particle packing algorithm is proposed which is to be used to predict the behaviour of pellets in the blast furnace on a first principal basis. Pellets consist of particles of various mineralogical composition and the structure in which they pack together to form a pellet is dependant on the size distribution of the particles and the pellet porosity. This packing structure can result in isolated volumes within the pellet where the local composition deviates from the overall average composition. This can result in, for example, melt formation at lower temperatures than expected, which will have a detrimental effect on the pellet strength. These local compositions result from the contacts between particles of different minerals and can thus be quantified by the coordination number of the particles. By using a validated coordination number model, which is unique for a particle packing algorithm, virtual pellets were created for a range of particle size distributions and porosities. The algorithm used the Monte Carlo method combined with the simulated annealing minimisation algorithm to solve the pellet simulations. The objective function is a combination of two functions, one describing the deviation from the target coordination number of the particles and the other the average fraction of overlapping volume of the particles per contact. In this way a realistic pellet structure was maintained while at the same time controlling the coordination number of the particles.