A Marble-Specific Rock Mass Classification Framework Integrating Empirical Parameters and Machine Learning Validation

Accurate assessment of rock mass quality in marble quarries remains challenging because conventional empirical classification systems are largely strength-dominated and insufficiently sensitive to discontinuity-controlled block instability. This study proposes a quarry-specific empirical framework, termed the Marble Rock System (MRS), designed to explicitly capture structural, hydro-mechanical, and alteration-driven controls governing bench-scale stability in dimension stone marble quarries. The primary objective was to develop and validate an empirically grounded classification system using machine learning as an independent diagnostic tool rather than as a black-box predictor.A comprehensive geomechanical database comprising ۸۵ quarry-scale records was developed from three active marble quarries in southern Rajasthan, India. Six physically interpretable parameters intact strength, weathering or serpentinization, joint frequency, joint surface condition, groundwater influence, and block stability were incorporated into the MRS framework. Supervised machine learning models, including artificial neural networks, support vector machines, and linear regression, were trained to predict independently derived factors of safety for validation. Model performance was evaluated using coefficient of determination, root mean square error, cross-validation, and classification metrics.Results show that MRS-based models achieved consistently higher predictive accuracy, improved class separability, and more stable generalization than models trained using conventional Rock Mass Rating inputs. Sensitivity analysis revealed that block stability and joint characteristics dominate stability prediction, while intact strength plays a secondary role. These findings confirm that marble quarry slope behaviour is primarily discontinuity-controlled. The proposed MRS provides a physically interpretable, empirically validated framework for quarry-scale stability assessment and offers a robust alternative to conventional classification systems for operational decision-making.

کلیدواژه ها:

Rock mass classification ، Block stability ، Machine learning validation ، Factor of Safety ، Discontinuity-controlled slopes