The Brazilian Test was modeled with FLAC to investigate the effect of loading angle on stress distribution and failure mechanisms in homogeneous and heterogeneous rock samples. The model was first calibrated to data of a Standard Brazilian test of a PMMA sample, found in the literature. Then several numerical Brazilian tests were performed using virtual materials with different deformability. The numerical results indicated that the loading angle (i.e. contact length between the jaw and the disc) is not constant in the Standard Brazilian test and changes depending on the amount of applied load and the deformability of the disc. This means that the contact length will be larger for materials with higher tensile strength and lower Young’s modulus. The contact length affects the distribution of both tensile and compressive stresses in the disc. Therefore, it can influence the failure mechanism of the disc. If the contact length is low, probability of shear failure increases close to the loading region, and if it is high, tensile failure is more probable at the center of the disc.
Furthermore, Uniaxial Tensile (UT), Standard Brazilian (SB) and Brazilian tests with constant loading angle of 30° (CLAB) were simulated. Heterogeneities were introduced as inclusions with their tensile strength and Young's modulus randomly selected from a normal distribution with different standard deviations. The analysis indicated that as the standard deviation of the Young's modulus and tensile strength distribution increases, the recorded peak load and peak tensile strength decreases for the three tensile test configurations. However, UT is much more sensitive to the variability of tensile strength than of Young's modulus. In contrast, the SB and CLAB tests indicated more sensitivity to Young's modulus variability.