Aggregate Particles Shift From Angular To Stable Rounded Shapes Under Progressive Abrasion

A comprehensive morphological investigation examining crushed rock aggregate has established that progressive mechanical milling drives particles along a systematic evolutionary pathway, transitioning sharp, highly angular fragments into smoother, compact geometries that reach structural equilibrium. By tracking 39,600 crushed rock particles across eleven abrasion intervals ranging from zero to 2,000 revolutions within a Los Angeles abrasion drum, researchers quantified fourteen distinct geometric shape descriptors using digital image processing. The resulting multi-stage analysis revealed that the most intense morphological restructuring takes place between 300 and 1,000 revolutions, after which particle boundaries stabilize and geometric modification diminishes markedly despite continued mechanical interaction. This empirical evaluation clarifies how industrial mineral aggregates degrade under repeated physical impact and friction, providing civil engineers and geomaterial scientists with a rigorous quantitative basis for assessing aggregate durability and optimizing concrete mixture performance.

Crushed stone aggregate represents an indispensable foundational component of hydraulic concrete, asphalt pavements, and granular roadway sub-bases, often constituting 60 to 80 percent of the total volume of concrete mixtures. Because of this substantial volumetric presence, aggregate geometry directly governs the operational performance, workability, and economics of construction materials in both fresh and hardened states. In fresh concrete, particle angularity and surface texture heavily influence rheological behavior, finishability, bleeding, and internal segregation. Conversely, in hardened structural elements, aggregate morphology dictates compressive strength, elastic stiffness, drying shrinkage, permeability, and long-term durability. However, geometric irregularity can simultaneously generate localized stress concentrations under load, initiating microcracks that propagate through the cementitious matrix and alter macroscopic mechanical performance.

Despite the recognized engineering importance of particle morphology, evaluating aggregate geometry has traditionally relied on empirical manual indices or static two-dimensional characterizations that assess particles under a single, static material state. In practical applications such as mineral milling, ballast degradation, and mechanical handling, aggregates are continually exposed to dynamic mechanical wear, localized fracture, impact, and surface polishing. Prior research had demonstrated that progressive milling progressively alters particle geometry and gradation, but the stage-wise sensitivity and mathematical redundancy among widely used geometric descriptors remained largely unresolved. Furthermore, it remained uncertain whether aggregate shape continually shifts indefinitely under mechanical stress or asymptotically approaches a stable geometric geometry. The research team sought to resolve these questions by subjecting quarry-derived aggregates to controlled mechanical abrasion and systematically tracking their morphometric evolution through high-resolution imaging and multivariate statistics.

The experimental program utilized crushed gneiss aggregate acquired from local commercial quarries operating in the Northern Province of Sri Lanka. Sieve analysis was conducted to isolate coarse aggregate fragments strictly within the size range of 5 to 30 millimeters. To simulate the progressive impact, abrasion, and friction typical of heavy handling and operational wear, the material was divided into eleven discrete sample batches and processed in a standard Los Angeles abrasion drum in accordance with ASTM C131. Mechanical degradation was applied across designated intervals consisting of 0, 50, 100, 150, 200, 300, 500, 750, 1,000, 1,500, and 2,000 revolutions. Following each milling cycle, the aggregate material was resieved using 30-millimeter passing and 5-millimeter retained sieves to ensure that subsequent evaluations maintained consistent particle size boundaries. The retrieved aggregate particles were subsequently washed thoroughly to eliminate adhering powder and air-dried under controlled ambient laboratory conditions for 24 hours prior to optical characterization.

Accurate geometric quantification required developing an optical imaging protocol capable of isolating individual particle boundaries without optical distortions. Representative aggregate specimens from each abrasion stage were coated with a uniform matte black paint to maximize visual contrast against a pristine white testing surface. The coated stones were arranged on an A4 sheet in a spaced pattern designed to eliminate physical overlap. Preliminary optical testing revealed that overhead illumination generated peripheral contact shadows around irregular stone margins, which compromised edge boundary detection. To overcome this limitation, the authors introduced an underlying illumination system that eradicated boundary shadows and facilitated sharp, high-contrast silhouette capture. For each abrasion group, twelve independent image sets were acquired, yielding dimensional data for approximately 39,600 individual rock fragments.

Using these primary measurements, the investigators computed fourteen geometric shape descriptors compiled from established civil engineering and powder technology literature, categorizing them into functional morphological groups. Overall geometry was represented by form descriptors including circularity, roundness, form factor, and shape factor. Dimensional elongation and spatial anisotropy were evaluated through proportion descriptors comprising aspect ratio, Feret aspect ratio, SF-flatness, sphericity, and the Barksdale index. Surface irregularity and boundary indentations were monitored using boundary descriptors such as solidity, convexity, and extent. In addition, fullness ratio was monitored as an index of area compactness, while eccentricity evaluated ellipse-based geometric elongation. Because aggregate particles are inherently three-dimensional objects, the researchers noted that projecting them into two-dimensional silhouette profiles reports area fractions rather than volumetric mass, representing an intrinsic operational boundary of planar digital image processing.

Statistical evaluation of the extracted dataset began with distribution and association analyses conducted across all milling stages using JMP Pro and MATLAB. Most shape metrics exhibited distinctly non-normal, skewed distributions across the aggregate population, reflecting the inherent heterogeneity of quarry-crushed stones. Compactness and boundary descriptors such as solidity, convexity, and fullness ratio displayed narrow distributions heavily concentrated near their upper mathematical limits, showing that particles maintained generally compact cores throughout mechanical processing. Conversely, circularity, roundness, sphericity, and aspect ratio displayed broad dispersions during early abrasion stages. As milling progressed, the central tendencies of circularity, roundness, and sphericity increased steadily, accompanied by a marked narrowing of their interquartile ranges. Meanwhile, aspect ratio, Feret aspect ratio, SF-flatness, and eccentricity declined progressively, confirming that mechanical abrasion systematically eliminated elongated geometries, sharp corners, and planar irregularities to produce an increasingly uniform and homogeneous particle population.

Pairwise Pearson correlation analysis demonstrated substantial mathematical and geometric redundancy among several established descriptors. Circularity exhibited near-perfect positive correlations with sphericity (r = 0.999) and form factor (r = 1.000), while displaying an almost complete inverse correlation with shape factor (r = -0.994). This co-dependence demonstrated that these four parameters measure nearly identical geometric attributes, rendering the simultaneous tracking of all four mathematically redundant. Similarly, SF-flatness was perfectly negatively correlated with the Barksdale index (r = -1.000) and strongly positively correlated with roundness (r = 0.994), while aspect ratio demonstrated a strong inverse relationship with roundness (r = -0.963). In contrast, Feret aspect ratio exhibited comparatively weak correlations across the descriptor matrix, functioning as an independent metric of particle dimensional anisotropy. Descriptors of boundary regularity and compactness also formed a coherent cluster, with solidity correlating positively with fullness ratio (r = 0.73) and convexity (r = 0.62).

To evaluate which parameters most effectively distinguish morphological change, the researchers conducted analysis of variance and calculated the omega-squared effect size (ω²) for each descriptor across the milling stages. The effect-size ranking identified convexity as the single most sensitive shape metric to mechanical abrasion, yielding an effect size of ω² = 0.35. This pronounced sensitivity aligns with physical observations showing that initial degradation is dominated by the rapid chipping and fracturing of sharp protrusions, concavities, and protruding edges. Circularity, form factor, sphericity, and shape factor also demonstrated robust sensitivity, exhibiting effect sizes ranging between ω² ≈ 0.27 and 0.29. Solidity displayed moderate responsiveness with an effect size of ω² ≈ 0.18. In contrast, Feret aspect ratio, extent, fullness ratio, roundness, and aspect ratio exhibited limited stage-wise sensitivity with effect sizes below 0.06, while the Barksdale index, SF-flatness, and eccentricity contributed negligibly to explaining abrasion progression.

Multivariate ordination using principal component analysis, accompanied by PERMANOVA and PERMDISP testing, confirmed that aggregate morphology advances along a coordinated, directional evolutionary trajectory. The first two principal components captured 78.4 percent of total morphological variation, with PC1 accounting for 55.4 percent and PC2 explaining 23.0 percent. Group centroid trajectories plotted within this multivariate space demonstrated that the greatest geometric restructuring takes place between 150 and 1,000 revolutions, reflecting combined edge chipping, localized fracture, and macro-scale smoothing. Beyond 1,000 revolutions, centroid displacement slowed substantially, confirming that particle morphology enters a stabilization phase where degradation transitions into fine surface wear, micro-abrasion, and polishing. Crucially, when PCA was repeated on residual descriptors after statistically eliminating particle area and perimeter influences, the directional trajectory remained identical, demonstrating that the observed evolution reflects true morphological restructuring rather than changes in particle size.

Pairwise PERMANOVA tests verified statistically significant separations among intermediate milling groups, whereas comparisons among late stages (1,500 to 2,000 revolutions) exhibited elevated p-values, confirming morphological stabilization. Homogeneity of multivariate dispersion was confirmed via PERMDISP, ensuring that observed group separations reflected genuine centroid shifts rather than within-group variability. Based on these combined statistical and physical findings, the researchers proposed a streamlined descriptor dominance framework. Rather than employing redundant metric suites or relying solely on unvalidated composite indices, practical aggregate characterization can be conducted using convexity to monitor boundary smoothing and circularity to quantify overall particle-form regularization. The authors emphasized that because specific wearing rates, transition boundaries, and stabilization thresholds were established using gneiss aggregates, these numerical benchmarks must be experimentally validated before direct application to lithologies possessing contrasting mineralogies, cleavage structures, or crushing strengths.

Subject of Research: Morphological evolution and geometric shape descriptor sensitivity of crushed rock aggregate under progressive mechanical abrasion

Article Title: Characterizing aggregate shape evolution during mechanical abrasion using digital image analysis and multivariate analysis of geometric shape descriptors

Article References: Janarth, A., Bandara Wijekoon, S. H., Kandeepan, R., & Subramaniam, D. N. (2026). Characterizing aggregate shape evolution during mechanical abrasion using digital image analysis and multivariate analysis of geometric shape descriptors. Discover Civil Engineering, 3(1), Article 231. https://doi.org/10.1007/s44290-026-00631-7

Image Credits: AI Generated

DOI: 10.1007/s44290-026-00631-7

Keywords: aggregate morphology, Los Angeles abrasion, digital image processing, geometric shape descriptors, gneiss aggregate, convexity, circularity, multivariate analysis, principal component analysis, PERMANOVA, aggregate degradation, concrete materials, pavement engineering, particle shape evolution, mechanical abrasion, form factor, shape factor, sphericity