Open Access

ARTICLE

Machine Learning-Based Prediction of Rock Fracture under Uniaxial Loading Using Infrared Radiation

Naseer Muhammad Khan^1,2, Liqiang Ma^3,*, Majid Khan⁴, Sajjad Hussain⁵, Waleed Inqiad⁶, Tariq Feroze², Danial Jahed Armaghani^7,*

1 Department of Mining and Explosives Engineering, Missouri University of Science and Technology, 324 McNutt Hall, 1400N. Bishop Ave., Rolla, MO, USA
2 Sustainable Advanced Geomechanical Engineering, National University of Sciences and Technology, Risalpur, Pakistan
3 Xinjiang Key Laboratory of Coal-Bearing Resources Exploration and Exploitation, Xinjiang Institute of Engineering, Urumqi, China
4 School of Civil and Resources Engineering, University of Science and Technology Beijing, China
5 Department of Mining Engineering, University of Engineering & Technology, Peshawar, Pakistan
6 Department of Civil Engineering, College of Engineering & Physical Sciences, Aston University, Aston St, Birmingham, UK
7 School of Civil and Environmental Engineering, University of Technology Sydney (UTS), Ultimo, NSW, Australia

* Corresponding Authors: Liqiang Ma. Email: ; Danial Jahed Armaghani. Email:

(This article belongs to the Special Issue: Computational Intelligent Systems for Solving Complex Engineering Problems: Principles and Applications-III)

Computer Modeling in Engineering & Sciences 2026, 147(3), 18 https://doi.org/10.32604/cmes.2026.081660

Received 06 March 2026; Accepted 27 May 2026; Issue published 30 June 2026

Abstract

Rock fracture behavior under stress is vital for risk evaluation in underground engineering excavation because the presence of water can significantly increase the extent of cracks and fractures in rock, leading to structural damage. This can result in catastrophic failures, including rock bursts, coal bursts, and water inrush. Hence, reliable prediction of rock damage and fracture processes is still lacking, which, in turn, enables the safe and efficient conduct of engineering projects in rock-mass environments. Thus, this study examines both dry and saturated sandstone samples under loading using Infrared Radiation (IR), Acoustic Emission (AE) monitoring, and Particle Flow Computation (PFC) techniques to effectively evaluate the fracture process in rocks under loading. Additionally, seven different artificial intelligence techniques, such as Gene Expression Programming (GEP), Gradient Boost Regression (GBR), Extreme Gradient Boosting (XGB), Adaptive Boosting (AdaBoost), Light Gradient Boosting Machine (LGBM), Categorical Boosting (CatBoost), were employed along with Explainable Machine Learning (XML) to predict the rock damage and fracture process. These models helped in the development of early warning signals to prevent catastrophic accidents. Both the experimental and simulation results have shown that the fracture density measured in terms of PFC and AE cumulative energy is significant in the saturated conditions compared to the dry conditions. Also, stress levels of 0.72 and 0.75 were found to be the warning signs in both dry and saturated conditions, based on the IR index (Average Infrared Radiation Temperature, AIRT) and AE characteristics. The comparison showed that the prediction accuracy of the XGB algorithm was the highest, followed by GBR, CatBoost, LGBM, GEP, and AdaBoost. However, GEP expressed its output in the form of an empirical equation owing to its grey-box nature, and thus, the law of fracture estimation in the form of an empirical equation was developed. The XML methods were added in order to enhance the interpretability of the high-performing, but black-box, XGB model. Such methods, along with a user-friendly Graphical User Interface (GUI), improved the model transparency and facilitated the integration of data-driven decision-making. XML and GUI tools may be instrumental in improving the safety measures adopted in coal mines and tunnels by reducing the risks and increasing operational safety.

---

Keywords

---