Population balance models (PBMs) are widely used to predict particle size variation in grinding. An accurate determination of particle breakage and selection functions in PBMs requires detailed knowledge on the stress causing particle breakage. This study proposed a combined experimental and numerical approach to determine the two functions. In the physical experiments, the breakage of individual iron ore particles in a Fritsch impact mill with different mill speeds was investigated and the product sizes were analysed. The simulations based on the discrete element method (DEM) were carried out under similar conditions to examine the dynamics of the particles inside the mill, such as number of impacts, impact velocity and impact angle. The simulation results showed that most of the particles experienced multiple impacts before they were ejected from the mill. The first impacts occurred at low impact energy which had no contribution to particle breakage. The second and third impacts, on the other hand, posed high impact energy on the particles and were the main cause of particle breakage. Based on the simulation results and the experimental data, two semi-empirical selection and breakage functions were determined. The accuracy of the equations was confirmed by applying them to predicting the size distribution of ground particles under different mill conditions. (C) 2017 Elsevier B.V. All rights reserved.