Extraction of Pre-Blast Rock Fracture Information From 3D Point Cloud of Post- Blast Fragmentation

Abstract

Muck piles, obtained by mining blasting operations, contain a huge amount of rock fragments. Surfaces of the post-blast rock fragments consist of pre-blast (in-situ) fractures and post-blast (blasting induced) fractures. As the remaining parts of in-situ joints, pre-blast fractures include clues of joint information inside rock masses. By extracting the tremendous amount of the clues sealed in muck piles, a better knowledge of the inner structures of rock masses can be realized, which is very meaningful to rock mass characterization, blasting effect evaluation and further blasting design. In order to extract the clues and collect in-situ joint information, differentiating pre-blast fractures from post-blast fractures is the first and vital step, which is also the focus of the research in this thesis. We investigated topographic differences between pre-blast fractures and post-blast fractures by studying the point cloud data of individual rock fragments. 3D point cloud data of the fragment fracture surfaces were generated by photogrammetry. Geometrical information of the fracture surfaces was extracted from the point cloud data and analyzed using three methods, referred to as the RMS method, the area ratio method and the Fourier transform analysis method. The following steps were taken to investigate the fracture surfaces: (1) 33 post-blast rock fragment samples were collected from the University of Arizona San Xavier Mine. 100 real fractures were selected from the 33 samples. 3D point cloud models of the real fractures were generated via photogrammetry. By observing and comparing the surfaces of the 100 real fractures, they were classified into three groups: pre-blast fractures (23 fractures), post-blast fractures (61 fractures), and fractures containing rock bridges (16 fractures). Typical surface topographic patterns (Flat, U-shape, Sine-shape, Saw-shape, Trapezoid-shape, and Battlement-shape) were utilized to develop the simplified fracture models. The functional formulas and 3D point clouds of the simplified fracture models were then generated. (2) The raw data of the real fractures were edited in point cloud processing software before further analysis. The data processing work mainly includes fracture segmenting, fracture projecting and profile extracting. (3) The application of the RMS, area ratio and Fourier transform analysis methods were presented as a sensitivity analysis on the simplified fracture model data and validated against the real fracture data. The simplified fracture model data were utilized for a sensitivity analysis of the RMS and area ratio methods. The real fracture data were tested by all three methods and the test results were compared with the classification results. The results show that: (1) RMS is sensitive to surface amplitude change. If a fracture surface has greater undulation, it tends to have higher RMS value. Pre-blast fractures have a much gentler undulating surface than post-blast fractures. (2) Area ratio is sensitive to surface wavenumber change. If a fracture surface has denser undulation (shorter wavelength), it tends to have higher area ratio value. Pre-blast fractures have much sparser undulating surfaces than post-blast fractures. (3) Investigating geometrical features using the Fast Fourier Transform (FFT) considers both peak wavenumbers and amplitudes. Pre-blast fractures and post-blast fractures have very similar undulation patterns (range of wavelengths), but undulation intensity (amplitude) of pre-blast fractures is much gentler at the same wavelength. The wavenumber peak trendline of rock bridges has a relatively round turning angle, compared to the trendlines for pre-blast and post-blast fractures.