Development, statistical validation, and machine learning prediction of bench stability for a marble-specific rock mass classification system with analytical hierarchy process-derived parameter weighting: a case study for Rajasthan Quarries, India

This study develops and evaluates the Marble Rock System (MRS), a preliminary, site-specific geomechanical classification framework for dimension-stone marble quarries. MRS addresses three deficiencies of general-purpose systems in the quarry context: absence of a mineralogical alteration parameter, insufficient structural weighting relative to intact strength, and inadequate kinematic block stability encoding. The scope is restricted to three Rajasthan quarries representing dolomitic, massive, and serpentinized marble lithologies. Six geomechanical parameters (M1: intact Uniaxial Compressive Strength (UCS); M2: weathering and serpentinization; M3: Rock Quality Designation and joint frequency; M4: joint spacing and surface condition; M5: groundwater condition; M6: block stability disposition) were weighted using the Analytical Hierarchy Process (AHP; through pairwise comparisons by five experienced practitioners (consistency ratio CR = 0.002 < 0.10). A database of 45 primary field records from three quarries was assembled,105 additional noise-augmented records were generated exclusively for machine learning training. Statistical correlations were computed on primary records only. Four supervised classifiers (Random Forest [RF], Extra Trees [ET], Support Vector Machine [SVM], Artificial Neural Network [ANN]) were evaluated under stratified cross-validation with SMOTE class balancing, sensitivity analysis (primary versus augmented), and confusion matrix with per-class metrics. MRS correlates with RMR (R 2 = 0.569; p < 0.001) and with deformability modulus exponentially (R 2 = 0.908; p < 0.001). On primary records only ( n = 45; 5-fold CV), RF achieved 0.977 ± 0.039 accuracy; on augmented data ( n = 150; 10-fold CV), 0.980 ± 0.031. The small performance gap confirms augmentation did not bias results. However, ML results are exploratory given the binary class distribution (Classes B and C only). Model performance is not generalizable to the full five-class MRS range. Joint parameters M3 and M4 are the dominant predictors, consistent with AHP weights and field observations. MRS is the first marble quarry classification index to derive parameter weights through formal multi-criteria analysis (AHP) with documented consistency checking, to exclude non-engineering aesthetic parameters, and to integrate machine learning as an operational scoring support tool (not a validation mechanism). A geological column, stereonet analysis, sensitivity analysis separating primary from augmented ML performance, and a seven-priority validation roadmap are contributed.

Authors

Institutions

Publication Details

Journal
Journal of Engineering and Applied Science
Published
2026-11-18
DOI
https://doi.org/10.1186/s44147-026-01308-9
Primary Topic
Rock Mechanics and Modeling
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Development, statistical validation, and machine learning prediction of bench stability for a marble-specific rock mass classification system with analytical hierarchy process-derived parameter weighting: a case study for Rajasthan Quarries, India

Anup Kumar Tripathi, Akhil Avchar, Mritunjay Kumar, Shambhavi Sinha
2 links
Journal of Engineering and Applied Science
Rock Mechanics and Modeling
article

Development, statistical validation, and machine learning prediction of bench stability for a marble-specific rock mass classification system with analytical hierarchy process-derived parameter weighting: a case study for Rajasthan Quarries, India

Anup Kumar Tripathi, Akhil Avchar, Mritunjay Kumar, Shambhavi Sinha
article en

Abstract

This study develops and evaluates the Marble Rock System (MRS), a preliminary, site-specific geomechanical classification framework for dimension-stone marble quarries. MRS addresses three deficiencies of general-purpose systems in the quarry context: absence of a mineralogical alteration parameter, insufficient structural weighting relative to intact strength, and inadequate kinematic block stability encoding. The scope is restricted to three Rajasthan quarries representing dolomitic, massive, and serpentinized marble lithologies. Six geomechanical parameters (M1: intact Uniaxial Compressive Strength (UCS); M2: weathering and serpentinization; M3: Rock Quality Designation and joint frequency; M4: joint spacing and surface condition; M5: groundwater condition; M6: block stability disposition) were weighted using the Analytical Hierarchy Process (AHP; through pairwise comparisons by five experienced practitioners (consistency ratio CR = 0.002 < 0.10). A database of 45 primary field records from three quarries was assembled,105 additional noise-augmented records were generated exclusively for machine learning training. Statistical correlations were computed on primary records only. Four supervised classifiers (Random Forest [RF], Extra Trees [ET], Support Vector Machine [SVM], Artificial Neural Network [ANN]) were evaluated under stratified cross-validation with SMOTE class balancing, sensitivity analysis (primary versus augmented), and confusion matrix with per-class metrics. MRS correlates with RMR (R 2 = 0.569; p < 0.001) and with deformability modulus exponentially (R 2 = 0.908; p < 0.001). On primary records only ( n = 45; 5-fold CV), RF achieved 0.977 ± 0.039 accuracy; on augmented data ( n = 150; 10-fold CV), 0.980 ± 0.031. The small performance gap confirms augmentation did not bias results. However, ML results are exploratory given the binary class distribution (Classes B and C only). Model performance is not generalizable to the full five-class MRS range. Joint parameters M3 and M4 are the dominant predictors, consistent with AHP weights and field observations. MRS is the first marble quarry classification index to derive parameter weights through formal multi-criteria analysis (AHP) with documented consistency checking, to exclude non-engineering aesthetic parameters, and to integrate machine learning as an operational scoring support tool (not a validation mechanism). A geological column, stereonet analysis, sensitivity analysis separating primary from augmented ML performance, and a seven-priority validation roadmap are contributed.

Journal of Engineering and Applied ScienceVol. 73(1)
National Institute of Technology Karnataka (IN)
Rock Mechanics and Modeling
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon