Discrete combustion of micron-sized aluminum particle dust cloud in air was investigated by a three-dimensional simulation. In light of presence of radiation and conduction heat transfer mechanisms, simulations became more realistic. Particles were randomly dispersed within a rectangular control volume using two different methods. An average flame front speed was obtained at each particle concentration using both methods. The second method yielded more accurate average flame front velocities. At lower dust concentrations, the results were in a good agreement with experimental outcomes. In the control volume composed of particles distributed using the second method, flame front was approximated with a fitted surface obtained via genetic algorithm. The temperature profiles of four random particles were plotted to provide insight into preheating time order of unburned particles. Later, the effect of particle diameters on flame front velocities was demonstrated. In the end, the flammability lean limit, as a significant parameter in safety issues, was calculated for different particle diameters. (C) 2017 Elsevier Ltd. All rights reserved.