Representative and Reliable Modeling for Rock Materials: 2D vs 3D

Determination of strength and deformation parameters of rocks are crucially important for many engineering structures. In recent decades, numerical modelling and analysis have become a critical part of engineering projects due to significant advances in software technologies and computational infrastructure. As with the dilemmas brought by every new development, differences between 2D and 3D modelling approaches need to be considered. In nature, rocks are exposed to stresses from multi-dimensions. In order to create representative and reliable models to assess the behaviour of rock, this natural phenomenon must have been taken into account. On the other hand, in many studies on rock mechanics and modelling, 2D models are used. The reasons behind this choice can be listed as computational load, time, effort, and the results being in the preferred and reliable range. In addition, it is an inevitable fact that the importance of 3D modelling will increase since developing technologies and engineering structures begin to push the limits of the known engineering experience of engineers. In this study, 2D and 3D models were created by using Particle Flow Code (PFC) for a Castlegate sandstone sample to evaluate the response of 2D and 3D models under similar stress conditions. For this purpose, uniaxial compressive strength, triaxial compressive strength, and tensile strength tests were performed on both intact rock and gapped (circular in 2D and spherical in 3D) models to obtain data from different scenarios. Although both models provide similar test results, 3D models offer much more detail, especially in parameters such as crack initiation, propagation, and stress localization. Here, it can be said that the differences in the behaviours arise from the number of the balls in 3D models, which is more than 3 times in 2D, and the number of contacts which is more than 6 times, respectively. However, because of this resolution, the model response to stress conditions is closer to nature. The computational load for the 3D model is much higher than the 2D model because of the resolution. Even though 3D models have some drawbacks compared to 2D models in terms of computational load, time, and effort, it can be said that the data they provide is much more representative and reliable, especially in terms of model behaviour.