Saturday, November 21, 2026
Science
No Result
View All Result
  • Login
  • HOME
  • SCIENCE NEWS
  • CONTACT US
  • HOME
  • SCIENCE NEWS
  • CONTACT US
No Result
View All Result
Scienmag
No Result
View All Result
Home Science News Technology and Engineering

Sandstone Texture Outperforms Mineral Composition in Predicting Rock Strength Using Gradient Boosting

November 21, 2026
in Technology and Engineering
Denise Maddox
By Denise Maddox Scienmag Editorial Profile - Mechanical Engineering
Reading Time: 6 mins read
0
Sandstone Texture Outperforms Mineral Composition in Predicting Rock Strength Using Gradient Boosting

Sandstone Texture Outperforms Mineral Composition in Predicting Rock Strength Using Gradient Boosting

65
SHARES
587
VIEWS
Share on FacebookShare on Twitter
ADVERTISEMENT

A comparative engineering study investigating intact rock mechanics across four distinct geological formations in Iran has demonstrated that spatial fabric geometry and textural characteristics provide far more reliable predictions of sandstone mechanical properties than mineralogical composition alone. Researchers Davood Fereidooni and Matloob Hejazifar evaluated twenty representative sandstone blocks systematically sampled from the Caspian, Padeha, Fajan, and Hezardarreh formations, testing whether an aggregate composition index or a multi-parameter spatial-fabric metric better governs intact rock strength. Applying machine learning algorithms in Orange 3.39.0 software, the authors determined that models relying on the quantitative Texture Coefficient consistently surpassed those based on the Mineralogy Coefficient across five standardized mechanical properties: Schmidt rebound hardness, point load index, Brazilian tensile strength, uniaxial compressive strength, and elasticity modulus. The predictive models developed with Gradient Boosting delivered the highest overall statistical accuracy, while Support Vector Machine algorithms proved unstable when predicting uniaxial compressive strength.

Understanding what governs the mechanical strength and deformability of intact rocks remains an essential objective in geotechnical engineering, foundation design, and underground excavation. Direct mechanical testing on rock cores requires specialized laboratory apparatus, extensive specimen machining, and significant financial resources, presenting practical hurdles when rapid geomechanical site characterizations are required. While early foundational research established that microscopic rock features influence physical response, historical literature frequently assessed mineralogical composition or textural indices in isolation. The authors designed this study to resolve critical knowledge gaps by conducting a simultaneous, side-by-side comparative weight analysis of an aggregate composition-based index, termed the Mineralogy Coefficient, against a multi-parameter spatial-fabric index, known as the Texture Coefficient. This rigorous experimental setup allowed the researchers to clarify whether raw mineral abundance or the intricate spatial arrangement, grain interlocking, and boundary geometry of constituent particles exert primary control over sandstone fracture behavior.

To ensure diverse textural, depositional, and diagenetic variability, the research team collected twenty intact rock blocks from a uniform depth of approximately 30 centimeters across four contrasting geological formations in Iran. The weathered outer crusts of the rock masses were removed during field sampling to ensure that only unweathered, structurally intact interior material was extracted. The sampled units represent a broad geological spectrum: the Caspian Formation exhibits shallow marine sedimentation with early calcite cementation; the Padeha Formation reflects moderate burial diagenesis characterized by compacted textures under compressive stress regimes; the Fajan Formation represents continental red bed deposition under oxidizing environmental conditions; and the Hezardarreh Formation exemplifies shallow burial environments with minimal compaction. In accordance with International Society for Rock Mechanics guidelines, large lithologically uniform blocks measuring roughly 30 by 40 by 50 centimeters were transported to the laboratory to prepare standardized specimens, thereby minimizing inherent rock heterogeneity and directional anisotropy.

The laboratory phase integrated petrographic characterization, X-ray diffraction, scanning electron microscopy, and standardized geomechanical failure testing. From the twenty rock blocks, the researchers prepared 580 laboratory specimens adhering to International Society for Rock Mechanics and ASTM standards. In addition, Schmidt rebound hammer testing was performed directly on the large block faces. Petrographic thin sections, polished sections, and powdered samples passing the number 200 sieve were prepared to evaluate mineral abundances, cement fabrics, and elemental distributions across each lithology.

Microscopic thin-section investigations, supported by X-ray diffraction and energy-dispersive X-ray spectroscopy, confirmed pronounced mineralogical and fabric diversity among the four sandstone groups. Samples from the Caspian Formation consisted of silicate grains including quartz, chert, and glauconite alongside bryozoan fossil fragments, with grains constituting 60 percent of the rock fabric embedded in 40 percent sparry calcite cement. Padeha sandstones displayed approximately 70 percent framework grains dominated by quartz, chert, and muscovite held within 30 percent schistose cement featuring minor iron oxides. Conversely, Fajan sandstones exhibited 60 percent calcite and 35 percent iron oxides with scattered angular quartz grains and calcite-filled veins, while Hezardarreh sandstones comprised roughly equal proportions of coarse grains and weathered sparry calcite matrix, featuring quartz, plagioclase, chert, biotite, and zeolite. Elemental mapping through scanning electron microscopy revealed that across all samples, quartz grains possessed higher intrinsic mechanical strength than the surrounding matrix, meaning that mechanical stress initially induces microcracks within the weaker cementitious matrix before coalescing into catastrophic macro-scale fractures.

To convert qualitative petrographic observations into computational predictors, the researchers quantified mineralogy using the Mineralogy Coefficient and texture using the Texture Coefficient. The Mineralogy Coefficient was calculated from the proportional abundance of composing minerals weighted by their individual Mohs hardness and specific gravity, yielding values between 0.097 in Hezardarreh calclithite sandstones and 0.147 in Padeha phyllarenite sandstones. The Texture Coefficient was quantified through JMicroVision 1.27 image analysis of thin sections, computing grain packing area weighting, grain aspect ratios, form factors, and orientation angle factors. Texture Coefficient values ranged from 0.750 in Hezardarreh sandstones to 0.931 in Padeha specimens. While the Mineralogy Coefficient captures the density and hardness of the mineral assemblage, it remains inherently blind to packing proximity and grain contacts. In contrast, the Texture Coefficient mathematically integrates grain interlocking, boundary tortuosity, and spatial alignment, effectively accounting for the geometric barriers that microcracks encounter during compressive and tensile loading.

In standard linear regression modeling, the Texture Coefficient proved distinctly superior to the Mineralogy Coefficient in explaining geomechanical performance. Empirical equations utilizing the Texture Coefficient yielded correlation coefficients ranging from 0.92 to 0.95 and coefficients of determination between 0.84 and 0.90 across all five mechanical parameters, supported by high F-statistics, Significance F values as low as 1.88 times 10 to the power of negative 10, and tight 95 percent confidence intervals. Conversely, regression models based solely on the Mineralogy Coefficient achieved correlation coefficients of only 0.69 to 0.75 and coefficients of determination spanning 0.47 to 0.57. Elasticity modulus exhibited the strongest linear relationship with both predictors, reaching an R² of 0.90 with the Texture Coefficient and 0.57 with the Mineralogy Coefficient, whereas Schmidt hardness demonstrated the weakest association with the Mineralogy Coefficient at an R² of 0.47. In 1:1 cross-plots comparing experimental measurements against model estimations, predictions generated from the Texture Coefficient aligned closely along the 45-degree parity line, whereas estimations derived from the Mineralogy Coefficient showed noticeable dispersion.

Data preprocessing utilized Min-Max scaling and Z-score standardization alongside an unsupervised Local Outlier Factor algorithm to mitigate scale distortion across the twenty instances. Models were evaluated using an 80:20 stratified train-test split, 5-fold cross-validation during hyperparameter tuning, and repeated Monte Carlo cross-validation across 100 iterations. Error assessment incorporated root mean square error, mean absolute error, mean absolute percentage error, and coefficient of determination. Gradient Boosting emerged as the most accurate and robust architecture across all property targets, achieving an R² of up to 0.96 and root mean square errors between 0.13 and 5.31. In particular, Gradient Boosting predictions for Brazilian tensile strength achieved an R² of 0.96 with a root mean square error of 0.13 for the Texture Coefficient variant and 0.14 for the Mineralogy Coefficient variant.

In sharp contrast to the stable performance of Gradient Boosting, Support Vector Machine regressors displayed severe algorithmic instability when predicting uniaxial compressive strength, yielding negative R² values of -0.18 for the Mineralogy Coefficient model and -0.06 for the Texture Coefficient model. These negative values indicate that the Support Vector Machine performed worse than a horizontal baseline fit to the target variable’s mean. The authors attributed this breakdown to several structural factors: uniaxial compressive strength displayed the widest numerical span in the dataset, ranging from 55.46 to 104.16 megapascals, creating substantial variance that challenged the algorithm’s epsilon-insensitive loss function. Furthermore, extreme values—such as Hezardarreh samples below 70 megapascals and Padeha samples exceeding 100 megapascals—acted as leverage points during support vector selection, while mapping features into high-dimensional space via the Radial Basis Function kernel exacerbated small-sample overfitting across twenty instances. Nevertheless, Support Vector Machines achieved high accuracy when predicting point load index and Brazilian tensile strength, reaching R² values between 0.90 and 0.96.

Comprehensive sensitivity analyses utilizing three distinct ranking frameworks—an aggregated scoring system, univariate regression, and the ReliefF algorithm—further confirmed the governing role of rock texture. In the multi-metric scoring system aggregating all error metrics across a 40-point scale, the point load index model based on the Texture Coefficient achieved the highest individual score of 34, while the total group score for Texture Coefficient models reached 143 compared to 77 for Mineralogy Coefficient variants. ReliefF analysis similarly identified the Texture Coefficient combinations for point load index and Brazilian tensile strength as the top-ranked predictors, each scoring 0.584, while the Mineralogy Coefficient model for Schmidt hardness ranked lowest at 0.530. Mechanistically, high Texture Coefficient values reflect an interlocking, tortuous grain framework that forces microscopic crack tips to divert along irregular paths, absorbing considerable fracture energy. Conversely, low Texture Coefficient fabrics dominated by soft calcite cement provide planar, low-energy cleavage pathways that accelerate structural failure under load.

The study authors emphasized critical boundary conditions that govern the interpretation and field application of their predictive models. Because the empirical relationships were established using twenty rock samples from four Iranian formations, the findings represent localized geomechanical insights rather than universally calibrated engineering formulas. The current dataset excluded highly cemented quartz arenites or quartzites, where mineralogical quartz weld contacts might override textural geometry, as well as highly fractured, weathered, or intensely porous lithologies. Bootstrapping with 1,000 resamples demonstrated narrow confidence intervals for Gradient Boosting metrics, confirming model stability within the tested suite, while combining Mineralogy and Texture coefficients into multivariate models improved uniaxial compressive strength predictions by 12 to 15 percent. Fereidooni and Hejazifar concluded that while texture-based machine learning offers a powerful preliminary screening tool to complement traditional testing when specimen retrieval is difficult, practical deployment across broader engineering environments requires further validation on larger and more diverse global lithological datasets.

Subject of Research: Petrographic and machine learning prediction of sandstone mechanical properties

Article Title: A comparative study of the mineralogy and texture effects on the sandstone mechanical properties using some machine learning algorithms

Article References: Fereidooni, D., & Hejazifar, M. (2026). A comparative study of the mineralogy and texture effects on the sandstone mechanical properties using some machine learning algorithms. International Journal of Geo-Engineering, 17(1), Article 19. https://doi.org/10.1186/s40703-026-00276-6

Image Credits: AI Generated

DOI: 10.1186/s40703-026-00276-6

Keywords: sandstone geomechanics, texture coefficient, mineralogy coefficient, gradient boosting, support vector machine, uniaxial compressive strength, brazilian tensile strength, point load index, schmidt rebound hardness, elasticity modulus, petrographic analysis, x-ray diffraction, scanning electron microscopy, rock mechanics, orange data mining, catboost engine, iranian geology, diagenetic fabric

Cite Scienmag News

Denise Maddox. (November 21, 2026). Sandstone Texture Outperforms Mineral Composition in Predicting Rock Strength Using Gradient Boosting. Scienmag. https://scienmag.com/sandstone-texture-outperforms-mineral-composition-in-predicting-rock-strength-using-gradient-boosting/

Denise Maddox. "Sandstone Texture Outperforms Mineral Composition in Predicting Rock Strength Using Gradient Boosting." Scienmag, 21 November 2026, https://scienmag.com/sandstone-texture-outperforms-mineral-composition-in-predicting-rock-strength-using-gradient-boosting/. Accessed 21 November 2026.

Denise Maddox. "Sandstone Texture Outperforms Mineral Composition in Predicting Rock Strength Using Gradient Boosting." Scienmag. November 21, 2026. https://scienmag.com/sandstone-texture-outperforms-mineral-composition-in-predicting-rock-strength-using-gradient-boosting/

Tags: brazilian tensile strengthcatboost enginediagenetic fabricelasticity modulusgradient boostingiranian geologymineralogy coefficientorange data miningpetrographic analysispoint load indexrock mechanicssandstone geomechanicsscanning electron microscopyschmidt rebound hardnesssupport vector machinetexture coefficientuniaxial compressive strengthX-ray diffraction
Share26Tweet16
Previous Post

Pediatric Invasive Meningococcal Disease Drives Lasting Medical Complications and Steep Healthcare Costs

Next Post

Harmonic Increment Sequence Enables Smooth Exponential Approximation of the Minimum Function

Related Posts

Fingertip pain threshold from millimeter waves reaches 41.9 degrees Celsius
Technology and Engineering

Fingertip pain threshold from millimeter waves reaches 41.9 degrees Celsius

November 21, 2026
ResNet Deep Learning Model Accurately Classifies Brain Stroke on CT Scans
Technology and Engineering

ResNet Deep Learning Model Accurately Classifies Brain Stroke on CT Scans

November 21, 2026
Rail Subsidies Yield Highest Cargo Volume Growth for Low Time Sensitivity Goods
Technology and Engineering

Rail Subsidies Yield Highest Cargo Volume Growth for Low Time Sensitivity Goods

November 21, 2026
Electrochemical Cycling Accelerates Lithium Diffusion in Amorphous NMC111 Thin Film Cathodes
Technology and Engineering

Electrochemical Cycling Accelerates Lithium Diffusion in Amorphous NMC111 Thin Film Cathodes

November 21, 2026
Model Predictive Control Improves Hydraulic Cylinder Position Tracking Over Standard PID
Technology and Engineering

Model Predictive Control Improves Hydraulic Cylinder Position Tracking Over Standard PID

November 21, 2026
Molecular Polarity Screening Yields Stable High Mobility Perovskite Transistors
Technology and Engineering

Molecular Polarity Screening Yields Stable High Mobility Perovskite Transistors

November 21, 2026
Next Post
Harmonic Increment Sequence Enables Smooth Exponential Approximation of the Minimum Function

Harmonic Increment Sequence Enables Smooth Exponential Approximation of the Minimum Function

  • Mothers who receive childcare support from maternal grandparents show more optimized

    Mothers who receive childcare support from maternal grandparents show more parental warmth, finds NTU Singapore study

    27656 shares
    Share 11059 Tweet 6912
  • University of Seville Breaks 120-Year-Old Mystery, Revises a Key Einstein Concept

    1061 shares
    Share 424 Tweet 265
  • Bee body mass, pathogens and local climate influence heat tolerance

    682 shares
    Share 273 Tweet 171
  • Researchers record first-ever images and data of a shark experiencing a boat strike

    546 shares
    Share 218 Tweet 137
  • Groundbreaking Clinical Trial Reveals Lubiprostone Enhances Kidney Function

    531 shares
    Share 212 Tweet 133
Science

Embark on a thrilling journey of discovery with Scienmag.com—your ultimate source for cutting-edge breakthroughs. Immerse yourself in a world where curiosity knows no limits and tomorrow’s possibilities become today’s reality!

RECENT NEWS

  • Early Biologic Therapy Yields Durable Inactive Disease in Systemic Juvenile Arthritis
  • Dynamic Mode Decomposition Enables Real-Time Motion Detection in Streaming Video Footage
  • Global Dairy Analysis Identifies Dominant Mastitis Pathogens and Rising Klebsiella Infections
  • Comorbid Anxiety in Bipolar Disorder Signals Lower Mood Stabilizer Response and Complexity

Categories

  • Agriculture
  • Anthropology
  • Archaeology
  • Athmospheric
  • Biology
  • Biotechnology
  • Blog
  • Bussines
  • Cancer
  • Chemistry
  • Climate
  • Earth Science
  • Editorial Policy
  • Marine
  • Mathematics
  • Medicine
  • Pediatry
  • Policy
  • Psychology & Psychiatry
  • Science Education
  • Science News
  • Social Science
  • Space
  • Technology and Engineering

Subscribe to Blog via Email

Enter your email address to subscribe to this blog and receive notifications of new posts by email.

Join 5,151 other subscribers

© 2026 Scienmag - Science Magazine

Welcome Back!

Login to your account below

Forgotten Password?

Retrieve your password

Please enter your username or email address to reset your password.

Log In
No Result
View All Result
  • HOME
  • SCIENCE NEWS
  • CONTACT US

© 2026 Scienmag - Science Magazine

Discover more from Science

Subscribe now to keep reading and get access to the full archive.

Continue reading