An Optimized Model for Clustering Rock Joints using Metaheuristic Algorithms

Abstract In the context of open-pit mining, numerical modeling and analysis are recognized as rigorous methodologies for investigating slope stability and characterizing slope mechanical behavior. However, initial considerations must be made, such as solution and model construction time, since numerical modeling of large-scale mines may take weeks or months. Engineers and consultants can address these challenges either by using supercomputers or by applying innovative procedures. Because of the unaffordability and high costs of supercomputers, mines often prefer innovative solutions. In this study, 376 joint data are collected from 19 benches of the Sungun copper mine. The joints are identified and clustered using the differential evolutionary fuzzy algorithm. The clustering resulted in three major joint sets, represented by average dip and dip direction values of 49/312, 66/095, and 57/210 for the first, second, and third sets, respectively. After clustering, numerical modeling of the Sungun pit is performed in 3DEC software. In the first case, 95 randomly selected joints are applied to their benches, yielding a factor of safety (FoS) of 1.61. In the second case, only three representative joints from the clustering output are applied to the model, resulting in a FoS of 1.59. The results showed that reducing slope joint data into representative classes with soft computing methods provides outcomes with less than 2% error (1.24%) compared to the full dataset. The result gives the user the ability and power to provide numerical modeling by soft computing techniques while maintaining the accuracy of the model results, giving fewer inputs to the model, thereby optimizing problem-solving and model-constructing time.