What question did this study set out to answer?

The goal is to evaluate various methods for estimating uniaxial compressive strength (UCS) in rocks, focusing on both traditional and AI techniques.

February 14, 2026Open Access

Uniaxial compressive strength prediction in rocks: a comprehensive review from empirical equations to AI methods

Key Points

The goal is to evaluate various methods for estimating uniaxial compressive strength (UCS) in rocks, focusing on both traditional and AI techniques.
Reviewed empirical correlations and traditional methods for UCS estimation.
Highlighted machine learning approaches like artificial neural networks and adaptive neuro-fuzzy inference systems.
Assessed predictive accuracy based on rock types and conditions.
Many traditional methods demonstrated utility but varied in accuracy depending on specific rock conditions.
AI techniques achieved R2 values above 0.95, reflecting high predictive performance.
Integration of empirical methods with AI offers the most effective UCS prediction strategy.

Abstract

Abstract Uniaxial compressive strength (UCS) is one of the most fundamental parameters in rock mechanics, widely used in the design and stability assessment of geotechnical and mining structures. However, its direct determination requires high-quality samples, sophisticated laboratory facilities, and significant time and cost, which often limit its applicability in practice. As a result, a broad spectrum of indirect estimation techniques has been developed, ranging from simple empirical correlations to advanced artificial intelligence (AI) models. This review provides a comprehensive synthesis of the methods employed in UCS estimation, with a particular focus on both conventional index tests and machine learning approaches. Traditional methods such as the Schmidt rebound hammer (SRH), ultrasonic pulse velocity (UPV), point load test (PLT), and Brazilian tensile strength (BTS) have demonstrated considerable utility, though their predictive accuracy is highly dependent on lithology, rock anisotropy, and site-specific conditions. On the other hand, AI-based techniques, including artificial neural networks (ANN), adaptive neuro-fuzzy inference systems (ANFIS), and optimization-enhanced hybrid models, have achieved superior predictive performance by capturing nonlinear and multivariate relationships, often yielding coefficients of determination ( R 2 ) above 0.95. Despite their promise, AI methods require large and representative datasets, and issues of model interpretability and overfitting remain challenges. The comparison highlights that no single approach is universally applicable; rather, the integration of empirical knowledge with computational intelligence appears to be the most effective strategy. The study concludes that future research should prioritize the development of hybrid models and standardized open-access databases to enhance the accuracy, robustness, and practical applicability of UCS prediction in diverse geological settings.

Ask AI

Helpful

Bookmark

View Full Paper