Advanced FDEM modelling of rock fracture process and applications in rock cutting

Throughout the years, rock fracture mechanism has been studied extensively experimentally, analytically and numerically due to its crucial importance for the field of civil, geotechnical and mining engineering as well as other fields, such as geothermal, hydraulic, oil and gas engineering, in which rock fracture plays an important role. However, in some cases such as dynamic rock fracturing and rock failure progressive process, the underlying mechanism of crack initiation and propagation has not been understood yet therefore it is worthwhile for further study. Recently, numerical methods have been increasingly applied to analyse the fracture process of rocks. The hybrid finite‚ÄövÑv™discrete element method (FDEM) is a widely used numerical technique in engineering applications involving material fracture, which, however, is computationally expensive and needs further development, especially when rock fracture progressive process is modelled. At the same time, some important fundamental components of FDEM such as the cohesive zone model (CZM) have been ignored in many recent publications in the field of rock fracture mechanics. This study aims to further develop a sequential hybrid FDEM code and then parallelize it with the use of a general‚ÄövÑv™purpose graphics processing unit (GPGPU) using compute unified device architecture (CUDA) C/C++. However, because the current contact detection algorithm in the sequential code is not suitable for GPGPU parallelization, a different contact detection algorithm is implemented in the GPGPU‚ÄövÑv™parallelized FDEM. Furthermore, a number of new features are implemented in the FDEM code, including the local damping, contact damping, contact friction, adaptive contact detection activation and mass scaling, and verified by simulating some simple tests. After that, the speed‚ÄövÑv™up performance of the GPGPU‚ÄövÑv™parallelized FDEM is discussed in terms of its performance on various GPGPU accelerators, which reveals that the GPGPU‚ÄövÑv™parallelized FDEM achieves a relative speed-up time of 128.6 and 284 times for 2D and 3D simulations, respectively, in comparison with the original sequential code. Furthermore, a number of simulations with both quasi‚ÄövÑv™static and dynamic loading conditions are conducted using the GPGPU‚ÄövÑv™parallelized FDEM, and the results obtained are compared qualitatively and/or quantitatively with published results from theoretical analysis and/or physical experiment to calibrate and validate the implementations. Based on the results obtained, the code can realistically model rock fracture under quasi static and dynamic under 2D and 3D loading condition. Moreover, it is found that the local orientation of elements affects resulted fracture mode in FDEM simulations, and the mixed mode I-II fracturing is the dominant failure mechanism in the FDEM simulations when unstructured meshes are used. Finally, a series of rock scratch tests with various cutting velocities and cutting depths are modelled using the GPGPU‚ÄövÑv™parallelized FDEM with calibrated input parameters and an appropriate value for local damping coefficient. On the basis of the results modelled, the failure mechanism, cutting force, chipping morphology and effect of various factors on them are discussed. Accumulation of fragments in front of the cutting tool is modelled successfully for the first time using FDEM by introducing a concept of local damping into the framework of FDEM. Changes of the fragment sizes and morphologies with increase in cutting depth can be fully captured like those reported in literatures on rock cutting experiments. Moreover, compared with recent numerical studies on rock cutting process, this study could model the change in average cutting force within ductile-brittle transition area similar to reported trends from experimental observations. Thus, the results obtained show that the GPGPU-parallelized FDEM is capable of producing quantitative predictions in addition to a qualitative representation of the rock failure process in rock scratch test. It is concluded that GPGPU-parallelized FDEM could be a powerful tool to further study rock cutting and improve cutting efficiencies.