نوع مقاله : مقاله پژوهشی
نویسندگان
گروه مهندسی معدن، دانشکده مهندسی شهید نیکبخت، دانشگاه سیستان و بلوچستان، زاهدان، ایران
چکیده
کلیدواژهها
عنوان مقاله [English]
نویسندگان [English]
Summary
This study aims to develop a novel method for identifying and analyzing structural joint sets, crucial for rock mass stability and engineering projects like tunneling and mining. A hybrid, multi-stage approach was used, combining the K-means algorithm for initial clustering and Agglomerative Hierarchical Clustering (AHC) for analyzing complex relationships. The dataset includes orientation and dip measurements from 172 joint planes in the Lucho granite mass of Zahedan. The K-means algorithm improved initial clustering accuracy by reducing intra-cluster variance, while AHC formed six final clusters with high spatial coherence, enhancing inter-cluster variance and reducing intra-cluster dispersion. This method effectively eliminated noise and outliers, facilitating 3D spatial pattern analysis and revealing complex data relationships. The findings demonstrate the method's superiority over conventional approaches, providing valuable quantitative parameters for geological pattern interpretation, joint orientation analysis, and potential applications in geoscience and engineering fields.
This study presents a novel data mining approach that leverages clustering algorithms, including K-means and hierarchical clustering (AHC), to identify and classify fractures. This method offers advantages over conventional techniques by reducing user interpretation errors and enabling quantitative analysis. The research focuses on the Zahedan granite batholith, located in the Sistan and Baluchestan province, which features a complex geological structure with varying rock types such as diorite, granodiorite, and biotite granite. Joint structures are prevalent, posing challenges for geological and mining operations. A total of 172 geological structural joints were surveyed in the field, and their orientations were analyzed, revealing significant dispersion in 3D space. This data is critical for understanding stress patterns and improving mining operations.
This study combines K-means and AHC algorithms to provide a comprehensive approach for joint analysis in rock masses, demonstrated through a case study of the Luccio granite mass in Zahedan. The method successfully identified six distinct joint clusters with clear spatial and geometric characteristics. By applying K-means clustering and refining with AHC using cosine similarity, the clusters were optimized, resulting in minimal intra-cluster variance and high spatial correlation. This approach enhanced the accuracy and reliability of joint identification, especially by removing outliers. The spatial distribution analysis revealed diverse tectonic influences, with high-dip joints (J3, J4, J5, J6) posing risks to wall stability, while low-dip joints (J1, J2) required careful management. This methodology provides a more accurate understanding of joint distribution and serves as a foundation for advanced studies, numerical modeling, and geomechanical analysis. It is applicable in mining project planning, tunnel design, slope stability, and modeling rock mass behavior. Future developments, especially for larger datasets and underground structure simulations, will further improve the understanding of joint distribution patterns and geological structures in rock masses.
کلیدواژهها [English]
)