The dynamic fracture process of rock materials is of importance for several industries, such as the rock drilling process in geothermal and mining applications. Gaining knowledge and understanding of dynamic rock fracture through numerical simulations can enhance the rock drilling process, for example by optimising the drill bit geometry and drilling parameters. In order for a numerical simulation of rock fracture processes to be accurate, the model needs to be able to capture key aspects of rock materials. Generally, rock materials are said to be britle and heterogeneous. The heterogeneity is partly due to the varying mechanical properties of constituent minerals, and partly due to the varying sizes, shapes, and directions of these minerals. The main objective of this thesis is the development of a heterogeneous rock model to be used for dynamic drilling processes. In the first article in this thesis, a heterogeneous bonded particle model is developed. Here, the heterogeneity is introduced in two steps – a geometrical heterogeneity using statistically distributed grain shapes and sizes, and a mechanical heterogeneity by distributing bonding parameters using a Weibull distribution. The model is applied to the quasi-static Brazilian disc test and a parametric study is conducted on the heterogeneity index and intergranular cement strength. The results show that crack initiation and propagation are highly dependent on the degree of heterogeneity. In general, the model was found to replicate typical phenomena associated with britle heterogeneous materials, for example unpredictability of macroscopic strength and crack properties. In the second paper of this thesis, an extensive dynamic experimental characterization of two igneous rock materials – Kuru grey granite and Kuru black diorite – is conducted. Here, a Split-Hopkinson configuration together with high-speed photography and digital image correlation is utilized to obtain the compressive and indirect tensile behavior of the rock materials. By using a significantly high frame rate of 671,000 fps in the digital image correlation analysis, it is shown that the point in time for crack initiation in the Brazilian disc can be estimated. From this, it is shown that the main splititng crack in the Brazilian disc occurs at 70 and 77 % for the two rock materials. In the third paper of this thesis, the heterogeneous bonded particle model from the first paper is further developed and calibrated using the dynamic experimental data for Kuru black diorite from the second paper. In contrast to the first paper, where one Weibull distribution is used, three Weibull distributions are used here. The first distribution is used for assigning average bonding parameters of the grains, the second for the intragranular bonding parameters and the third for the bonding parameters of the intergranular cementing. First, a homogeneous bonded particle model, i.e., without heterogeneous grains and no statistical distribution of bonding parameters, is calibrated so that the average experimental results are replicated. Then, using this homogeneously calibrated model, the heterogeneous model is activated, and a parametric study is conducted on the heterogeneity index for the average grain properties and the intergranular cement strength. The results show that this modelling approach is able to capture key phenomena of dynamic rock fracture, such as stochastic crack initiation and propagation, as well as peak stress, overloading, strain rate and crack propagation time. In the fourth paper, the proposed heterogeneous bonded particle model from previous papers is validated using a laboratory rock drilling experiment. The rock material is dynamically characterized using the methodology from the second paper and the grain structure is obtained from a scan of the rock surface. Three of the constituent minerals are represented in the model in terms of their size, occurrence, and mechanical properties. Furthermore, the model is calibrated in both compression and tension, where both the peak stress values and fracture behavior are captured. The model is then used to simulate the laboratory rock drilling experiment, where crater depth, load and rock fragment sizes are compared with the experiments. The results show that the simulation is able to capture peak load values and the rock fragment sizes are similar to that of the experiments.