Prediction of rock mechanical parameters using gamma ray log data with regression and machine learning approaches
Mining of Mineral Deposits · 30.06.2026 · 20 (2) · 89-100
DOI: 10.33271/mining20.02.089 ↗Kurz gefasst
Fachartikel in der Zeitschrift Mining of Mineral Deposits, online erschienen am 30.06.2026. Verfasst von Andyono Broto Santoso und 5 weiteren Autor:innen.
Abstract (Originalsprache)
Purpose
This study aims to investigate the feasibility of using gamma ray (GR) wireline logs as a principal predictor for estimating key rock mechanical parameters, including internal friction angle (φ), cohesion (c), and uniaxial compressive strength (UCS), as a rapid and cost-effective alternative to conventional laboratory testing.
Methods
A dataset of 54 depth-matched records, comprising GR log values and laboratory-derived mechanical properties, was compiled from multiple boreholes in sedimentary formations. Predictive models were developed using linear regression, Random Forest (RF), Extreme Gradient Boosting (XGBoost), and Artificial Neural Networks (ANN). Data preprocessing included median imputation, feature scaling, and an 80/20 train-test split. Model performance was evaluated using the coefficient of determination (R2), mean squared error (MSE), and root mean squared error (RMSE). SHapley Additive exPlanations (SHAP) values were used to improve model interpretability.
Findings
Correlation analysis revealed strong positive associations between GR and UCS (r = 0.852), cohesion (r = 0.799), and friction angle (r = 0.785). Among machine learning models, XGBoost delivered the best performance for cohesion (R2 = 0.6764, RMSE = 5.09 kPa) and UCS (R2 = 0.4427, RMSE = 145.42 MPa), while RF was optimal for friction angle (R2 = 0.4955, RMSE = 2.35°). ANN consistently underperformed relative to tree-based models. Linear regression provided competitive baseline performance, indicating a predominantly linear relationship between GR logs and mechanical properties at wireline resolution. Originality. This work systematically positions GR logs as the principal predictor of rock mechanical properties – an approach rarely adopted in prior research that predominantly relied on sonic or density logs. The study provides the first systematic empirical comparison of linear and nonlinear models using GR as the primary input feature, establishing practical regression equations directly applicable to field deployment. Practical implications. The derived regression equations and validated modeling framework enable engineers to perform rapid geomechanical assessments using widely available GR log data, thereby reducing dependence on costly, time-consuming laboratory testing. This approach is particularly valuable during early exploration stages and in mining contexts where core sample availability is limited.
Zitieren
APA
Santoso, A. B., Sujatono, S., Farhan, M. A., Syaputra, R., Rahmad, B., & Dwinagara, B. (2026). Prediction of rock mechanical parameters using gamma ray log data with regression and machine learning approaches. Mining of Mineral Deposits, 20(2), 89-100. https://doi.org/10.33271/mining20.02.089
Vancouver
Santoso AB, Sujatono S, Farhan MA, Syaputra R, Rahmad B, Dwinagara B. Prediction of rock mechanical parameters using gamma ray log data with regression and machine learning approaches. Mining of Mineral Deposits. 2026;20(2):89-100. doi:10.33271/mining20.02.089
BibTeX
@article{Santoso2026,
author = {Santoso, Andyono Broto and Sujatono, Supandi and Farhan, Muhammad Alfiza and Syaputra, Rizky and Rahmad, Basuki and Dwinagara, Barlian},
title = {Prediction of rock mechanical parameters using gamma ray log data with regression and machine learning approaches},
journal = {Mining of Mineral Deposits},
year = {2026},
volume = {20},
number = {2},
pages = {89--100},
doi = {10.33271/mining20.02.089}
}
EUVIMED