Determination of shear failure regions of rock joints based on point clouds and image segmentation

Abstract

To better understand the shear mechanical behaviour, the information value (IV) and the fractal theory (FT) models are employed to determine the shear failure regions of rock joints under the low effective normal stress conditions. First, high density point clouds of rock joints at the initial stage (pre-shearing) were collected, using a laser scanning technology and four topography parameters (slope angle, horizontal orientation, elevation difference and curvature) were produced from point clouds, to represent the mesostructure features. Second, direct shear testing were conducted at the end stage (post-shearing), and shear striae on the rock joints were measured to assess shear failure regions, based on image segmentation. Third, by combining morphological features with experimental observations, correlation analyses are performed, to investigate the influences of topography parameters on shear failure behaviors of rock joints, based on IV and normalized weight Wi. Similar findings are observed from the two algorithms, indicating that horizontal orientation has a larger impact on shear failure than the other three parameters. Furthermore, under the low normal loading, the shear failure of rock joints are more likely to occur in regions with a slope angle in the range of 30° to 40°, a horizontal orientation in the range of SE to SW (facing shear direction), an elevation difference in the range of 0.15 to 0.30 mm and a curvature in the range of −1000 to −200 mm−1 and 200 to 1000 mm−1. It is notable that the effect of normal loads level on shear failure regions should not be neglected. Eventually, a model is developed to predict the shear failure regions of rock joint and receiver operating characteristic (ROC) analyses show that shear failure determinations, based on the proposed model, match the actual situation.