The Role of Fabric, Stress History, Mineralogy, and Particle Morphology on the Triaxial Behavior of Nontextbook Geomaterials

Mining is an important economic activity worldwide, and extracting minerals from ore demands comminution. Therefore, ore processing includes simple crushing and screening methods (to obtain the coarser iron products) as well as more-sophisticated processes (to upgrade the ore quality and recover iron from smaller particles). All these processes result in large volumes of tailings worldwide. The iron ore tailings studied herein were collected from a beneficiation plant at Quadrilátero Ferrífero in the central region of Minas Gerais in Brazil. This region is responsible for approximately 65% of all Brazilian iron ore production. The tailings were obtained at the exit plant after the filtering processing. The particle-size distribution is presented in Fig. 1 following ASTM D7928 (ASTM 2021a). The studied tailings were composed predominantly of fine sand (60%) and silt (32%) fractions. The tailings were classified as silty sand (SM) according to the Unified Soil Classification System (ASTM 2017), and were nonplastic.

The industrial processing of the tailings provides some specific characteristics to these materials. Generally, the tailings are more angular and rougher than natural sands (e.g., Yang et al. 2019). The particle angularity and roughness also were observed in the present tailings using scanning electron microscopy (SEM) analysis for different particle sizes (Fig. 2). The beneficiation processes are not totally efficient, and the tailings had some residual iron contents. Energy-dispersive X-ray spectroscopy (EDS) analysis revealed that the iron ore tailings were composed mostly of iron oxide (28%) and quartz (64%). This high iron content affected the specific gravity ($G_s=2.92$), which was evaluated in accordance with ASTM D854 (ASTM 2014). The compaction characteristics were assessed using the standard effort ($w_{\mathrm{opt}}=12.8\%$, and ${\gamma }_{d\max }=18.7\mathrm{kN}/{\mathrm{m}}^3$) in accordance with ASTM D698 (ASTM 2021b).

The specimens were molded using moist tamping and static compaction, following the undercompaction method (Ladd 1978; Frost and Park 2003). Two molding conditions were adopted, one to obtain dense samples considering 100% of the compaction degree ($e_0\approx 0.59$) and the other to obtain loose samples considering 85% of the compaction degree ($e_0\approx 0.80$). The necessary dry tailings weight was mixed with water to achieve the optimum moisture content of 12.8%. The specimens were molded in cylinders 50 mm in diameter and 100 mm in height. Three layers were used in this process, and the tops of the first and second layers were scarified to guarantee the adherence of the subsequent layer.

Conventional isotropically consolidated drained (CID) triaxial tests were performed to assess the mechanical response of the iron ore tailings over a broad range of initial effective confining pressures ($p_0^{\prime }=50–\mathrm{4,000}\mathrm{kPa}$) and considering different stress histories. The specimens were named according to their stress history: overconsolidated (OC) refers to specimens that experienced some overconsolidation, whereas specimens molded in compression paths before the convergence to normal compression line (NCL) are referred to dense (D) or loose (L). Table 1 summarizes the 19 drained triaxial tests. The tests are identified using the notation $X\_Y\mathrm{kPa}\_Z$, where $X$ is the void ratio at molding, $Y$ is the initial confining pressure (50, 125, 250, 500, 1,000, 2,000, or 4,000 kPa), and $Z$ indicates the stress history (OC, D, or L).

Two extra tests were performed after defining the critical state line (CSL) over the initial range of effective confining pressures. One isotropically consolidated undrained (CIU) triaxial test with $e_0=0.80$ was conducted to check the CSL intercept adopted, and one isotropic compression test initially with $e_0=0.60$ was conducted up to the highest pressures achievable within the apparatus considered (4,000 kPa) to verify the influence of breakage during isotropic compression.

The recommendations of ASTM D7181 (ASTM 2020b) and ASTM D4767 (ASTM 2020a) were followed. Thus, the tests were performed on fully saturated specimens that were subjected to a saturation process composed of carbon dioxide percolation, water percolation, and application of back-pressure increments (maintaining a mean effective stress constant of 20 kPa). The Skempton $B$ values were measured after the saturation was finished, and were greater than 0.98. The consolidation phase was conducted by increasing the chamber pressure at a constant rate up to the desired ${\sigma }_3^{\prime }$ value. The rate varied from 2 to $5\mathrm{kPa}/\min$ and was adequate, because no excess pore pressure was generated during this procedure. All specimens were maintained under the desired confining pressure for at least 30 min, or until the volumetric strains ceased, to avoid the effects of incomplete consolidation. After the consolidation was terminated, shearing was started at a rate of $2.0\mathrm{mm}/\mathrm{h}$. The postshearing void ratio was determined using the moisture content of the specimen at the end of the test. The initial specific volumes were obtained following the procedure described by Shipton and Coop (2015). The procedure consists of obtaining the initial specific volumes in different forms in the most independent way possible. The average precision for specific volume determination obtained was 0.005.

The confining pressure and the back-pressure were monitored independently using electronic pressure transducers. Both were applied through a pressure controller by a servomotor system. The load was registered by a load cell. The axial external displacements were measured using a linear deformation transducer. Internally, two Hall effect sensors were used to evaluate the axial displacements, and one was employed to measure the radial displacements (Clayton and Khatrush 1986). The internal displacement measurements were taken throughout all the triaxial testing phases.

The use of internal measurements with the resolutions depicted in Table 2 allowed for the determination of stiffness from very small load increments. For the CID tests, the volume change was evaluated by measuring the flow volume of water entering and leaving the specimen using piston-type volume change devices. Table 2 presents the data acquisition and control system’s specifications, and Table 3 presents the devices’ measurement resolution.

Fig. 3 presents the compression paths of the two specimens with different initial specific volumes (${\nu }_0$) that achieved the highest confining pressure (4,000 kPa). The compression curves from different $v_0$ values converged very slowly, so even at the maximum stress level achieved, a unique NCL cannot be identified. Thus, for the usual and even for special tailings storage facilities in which the pressures would be at most in the range 2,000–10,000 kPa, the material would not be in a normally consolidated state.

Fig. 3 also shows the unloading response of the dense specimen considering an overconsolidation ratio of 4. The slope of the unloading–reloading line is similar to the slope of the virgin consolidation line, which demonstrates the predominance of elastic strains during isotropic compression for dense specimens at this pressure range. This compression path also indicates that particle breakage had not yet become the main deformation mechanism, because the plastic strains on unloading still were low.

This section presents the stress–strain behavior and the volume change response of the studied tailings in different initial states (loose, dense, and overconsolidated) for different stress levels. Only the tests sheared with 50, 1,000, and 2,000 kPa of confining pressure are presented for brevity. The overall characteristics of the tests are presented in Table 1.

The stress history influence on the iron ore tailings’ mechanical response was evaluated at two stress levels by comparing overconsolidated specimens with nonoverconsolidated specimens. Fig. 4 presents the stress–axial strain–volumetric strain response for the specimens sheared under 50 kPa confining pressure. The overconsolidation did not significantly affect the stress–strain response, as would be expected for granular materials. This occurred because the strength mobilization in these materials is due to friction between grains, without chemical or cementing bonding’s contributions. The peak strength and the initial stiffness were very similar for dense or overconsolidated specimens. Even the void ratio at the beginning of shearing was similar, indicating the small effect decurrent of the loading and unloading process. The main difference between the tests sheared under 50 kPa confining stress was that the initial volumetric compression was suppressed in the overconsolidated specimen because it already had experienced these strains during the isotropic compression up to a 4- times higher load (200 kPa).

The stress history influence also was evaluated at higher pressures. For the tests sheared with $p_0^{\prime }=\mathrm{1,000}\mathrm{kPa}$ shown in Fig. 5(a), the overconsolidation mainly affected the volumetric response. However, the overconsolidation ratio (the ratio of $p_{\max }^{\prime }$ to $p^{\prime }$ at the beginning of shear) did not influence the maximum strength and volumetric strains. Both overconsolidated specimens had higher dilative volumetric strains and smaller compression than the dense specimen. Because the void ratios at the beginning of shearing were very similar for OC and D specimens, the differences in volumetric behavior can be associated with changes in the particles arrangement after experiencing higher confining pressures (2,000 or 4,000 kPa). In the tests sheared with $p_0^{\prime }=\mathrm{2,000}\mathrm{kPa}$ [Fig. 5(b)], in addition to the changes in volumetric response, the main difference was that the overconsolidated ($\mathrm{OCR}=2$) specimen had a slightly lower peak strength than the dense specimens sheared under the same confinement level. For all stress levels, the results of the ultimate strength are difficult to compare due to the very marked strain localization occurring for these specimens.

It is possible to associate the overconsolidated specimens’ lower peak strength (compared with that of the dense specimens) and distinct volumetric response with particle morphology and arrangement changes during isotropic compression and shearing. In the overconsolidated tests, the grain arrangement already had been modified to support higher stress than those experienced at a given moment, and the higher pressures can change the particles’ roughness and shape, which facilitates their relative movement. These differences could have reduced the interlocking contribution to the behavior of overconsolidated specimens.

Figs. 4 and 5 also plot the tests performed in initially loose conditions. A fully contractive volumetric response and a strain-hardening behavior are verified. The deviatoric stress after the peak of dense and overconsolidated specimens tended to the maximum strength of the loose specimen. However, there was no full convergence due to postpeak strain localization in the dense and overconsolidated specimens.

The critical state line (CSL) in the $q$-$p^{\prime }$-plane was estimated considering the stability in the deviatoric stress and the volumetric strains. Moreover, only tests performed with loose specimens were considered in the fitting due to the strain localization in dense specimens. Fig. 6 shows the CSL obtained. The fitting resulted in an $M$ value of 1.29, corresponding to a critical state friction angle of approximately 32.1°. Fig. 6 also demonstrates the suitability of the estimated CSL by analyzing the normalized results concerning the stress ratio ($\eta =q/p^{\prime }$) versus shear strain.

The loose, dense, and overconsolidated specimens delimited the CSL, making it possible to define the critical state locus in the $\nu$, $\ln p^{\prime }$-plane (Fig. 7). However, the CSL was determined using the endpoints of the tests that were closer to the critical state condition. Dense and overconsolidated specimens were not used in this adjustment because they had not yet achieved constant volume and stress states and presented significant strain localization. In addition, a loose specimen was tested under undrained conditions at low stresses to verify the CSL intercept ($\mathrm{\Gamma }$) adopted. As the specimen was looser than $\mathrm{\Gamma }$ at the beggining of shearing, it must liquefy ($p^{\prime }$ and $q$ tending to zero), as was observed. The curved CSL fitted were obtained using a power law to include all the stress ranges testedwhere $\mathrm{\Gamma }$ = curve intercept; $A$ = initial curve slope; $B$ is an exponent that controls the curvature; and $p_{\mathrm{ref}}^{\prime }$ was chosen as 100 kPa. The values adopted were $\mathrm{\Gamma }=1.75$, $A=0.02$, and $B=0.58$.

Pure isotropic hardening usually is assumed for geomaterials, although soils follow a conjunction of kinematic and isotropic hardening (Gajo and Muir Wood 1999; Jardine 1992; Kuwano and Jardine 2007). Kinematic hardening is the translation of the yield surfaces in the stress space without necessarily increasing its size and maintaining a constant shape (Mróz 1967). A kinematic hardening approach can model the typical stiffness degradation presented by soils. This is achieved by introducing a very small elastic region that moves with the current stress state (Gajo and Muir Wood 1999). Additionally, this approach can be extended to the concept of multiple surfaces and different degrees of yielding up to large yielding, at which abrupt changes in stress–strain response can be noticed (Jardine 1992).

Adopting kinematic hardening also allows for retaining information regarding the soil’s recent stress history. Atkinson et al. (1990) defined the recent stress history as the most recent loading followed by the material to reach the current state and the time required to do so, or a sudden change in the stress path direction. They characterized the recent stress history of soil using the angle of rotation $\theta$ required to follow the new stress path, with clockwise rotation taken as positive.

The effects of recent stress history on soil behavior have been investigated extensively for clays (e.g., Atkinson et al. 1990; Jardine 1992; Stallebrass and Taylor 1997; Baudet and Stallebrass 2004). Other authors (e.g., Kuwano and Jardine 2007; Hong et al. 2017) have verified the influence of recent stress history on sands. In studying clays, Stallebrass and Taylor (1997) found equal stiffness degradation curves for stress paths with a recent stress history of 90° and $-90°$. These angles of stress path rotation were achieved through isotropic compression followed by shearing at a constant $p^{\prime }$ triaxial compression ($-90°$) or extension (90°) stress path. However, the latter recent stress history also could be achieved by isotropically overconsolidating the specimen, then returning to the original pressure and shearing at a constant $p^{\prime }$ triaxial compression path. Hong et al. (2017) performed this procedure for Toyoura sand and achieved results equivalent to those of Stallebrass and Taylor (1997). Fig. 8 presents the recent stress history and stiffness degradation from Hong et al. (2017) for Toyoura sand [Figs. 8(a and b)] and the results obtained herein for iron ore tailings under low stress levels [Figs. 8(c and d)].

Although the stress paths followed for iron ore tailings were not the same as those in Hong et al. (2017), the overall response with low influence of recent stress history on stiffness was the same. This could be explained by the specimens having not yet achieved full yielding. Coop and Lee (1993) argued that, in sands, the full yielding is delayed until particle breakage dominates and the compression paths in the $\nu$, $\ln p^{\prime }$-plane steepen to join the NCL, which had not yet occurred in Fig. 3. Furthermore, these observations agree with the general trend observed in quartz sands, in which for a typical range of pressures they have the same sensitivity to pressure during the loading or unloading processes (He et al. 2019; Lo Presti et al. 1993). This analysis is possible because of the proximity of void ratio values of the specimens, which has a great influence on stiffness.

Fig. 9 depicts the stiffness degradation curves for specimens sheared at 2,000 kPa of confining pressure. The results for the specimens sheared at 1,000 kPa indicated a similar response and are omitted for brevity. The stiffness of tailings was slightly lower for the overconsolidated specimens than for the specimen isotropic compressed to 2,000 kPa and subsequently sheared. This effect, along with the changes in stress–strain response (reduction in maximum mobilized strength and volumetric strains) observed for the overconsolidated specimens at high pressures [Figs. 5(a and b)], can be related to the reduction of the interlocking contribution on these specimens.

The state boundary surface (SBS) of a material usually is assumed to have a constant shape and only to increase in size. Thus, it is necessary to adopt a normalization technique in order to represent the SBS. These techniques take the pressure at some reference state as the normalizing pressure. Fig. 10 presents possible choices for reference states and the corresponding normalization. In this paper, the reference state selected was the critical state, considering a CSL in the compression plane in the form of Eq. (1). By adopting the pressure at critical state ($p_{\mathrm{cs}}^{\prime }$) as the normalizing parameter, it was possible to plot the state paths followed by the specimens tested (Fig. 11). In addition, the tests were normalized further with respect to the deviatoric stresses by considering the critical state strength, $M$, so that the critical state in this plane would plot at (1,1).

Li and Coop (2019) first investigated the SBS of tailings. They found surfaces that changed from clay shape to sand shape as the fines content was reduced and the tailings became coarser. The SBS of clays usually have an elliptical shape with the apex at the critical state (Roscoe et al. 1958), whereas those for sands have an apex before reaching the critical state (Chandler 1985; Coop 1990). Fig. 12 presents the SBS for clays and sands to illustrate how changes in SBS shape influence the materials’ flow rule. When the SBS apex coincides with the critical state, an associated flow rule (the strain increases in the same direction as the stress increases) can represent the plastic strains occurring. However, when this apex does not coincide with the critical state, the material necessarily follows a nonassociated flow rule (the plastic strain increases in a direction independent of the stress increase direction). The iron ore tailings studied herein had an apex before reaching the critical state, indicating the existence of nonassociated flow for this material (Fig. 11). This also can be demonstrated by superimposing the plastic strain vectors on the SBS and verifying their nonorthogonality (Fig. 13). The plastic strains were obtained by considering an elastic model, determining the elastic strains for each stress increase, and subtracting them from the total strain increase (Kuwano and Jardine 2007).

Another feature of the SBS is its intercept with the normalized mean effective pressure axis. For an initially isotropic compressed specimen, the value at which its state path intercepts the abscissas axis indicates the spacing ratio (the ratio of the $p^{\prime }$ on the CSL to that on the NCL for the same specific volume). However, Li and Coop (2019) stated that this normalization could not work well for the curved part of CSL because the value of $p_{cs}^{\prime }$ for liquefiable states would approach zero, giving a very large apparent SBS. Coop (1990) also demonstrated that the behavior of some nontextbook materials (carbonate sands) reduces to the classical critical state framework (straight CSL and NCL in the $\nu$, $\ln p^{\prime }$-plane) when sufficient pressures are achieved. Thus, Li and Coop (2019) only regarded as the true SBS the surface defined in the region where the CSL is approximately straight and parallel to the NCL. In turn, Consoli et al. (2023) demonstrated for two iron ore tailings from different stages of treatment the achievement of linear critical state lines and normal compression lines at high pressures with a spacing ratio of 2.71.

However, Poorooshasb (1989) defined the SBS as the surface in the state space that encloses all the possible states that a sample of a granular medium may assume. Therefore, the large state paths obtained in Fig. 11 also would be enclosed by the true SBS because the specimen existed in these states.

Poorooshasb (1989) proposed a different plane to plot the results and obtain a trace of the SBS [Fig. 14(a)]. Fig. 14(b) demonstrates the SBS obtained for iron ore tailings herein on the state parameter ($\psi =\nu -{\nu }_{\mathrm{cs}}$) versus stress ratio space proposed by Poorooshasb (1989). The indicated SBS is nonlinear, as was that found by Todisco and Coop (2016) for another material. Thus, there is a trend of stabilization with decreasing state parameter, not predicting infinite stress ratios, and consequently dilatancies, as the specimens move farther from the critical state at the dry side (lower state parameter values).

In granular cohesionless media, the effects of structure usually have been neglected. The terms structure and fabric usually are used indistinctly; however, this study adopted the definition of Mitchell and Soga (2005), in which structure is the combination of bonding (composition and interparticle forces) and fabric (particle arrangement). Tailings generally are considered to be unstructured materials due to their recent deposition history. However, the importance of their structure—specifically, their fabric—must be considered in some situations (Fourie et al. 2022). This implies that, at a given void ratio, some tailings can sustain higher stresses than could the same material without the effects of structure (Schnaid 2021). Therefore, it is proposed that the large apparent SBS (Fig. 11) obtained at the curved portion of the CSL (Fig. 7) may arise because of elements of fabric.

Accounting for fabric effects can be an arduous task, because it still is difficult to describe this property quantitatively (Collins et al. 2010; Coop 2015; Liu et al. 2021). However, the effects caused by fabric on the overall response can be accounted for indirectly, for example, by describing the evolution of the SBS (Baudet and Stallebrass 2004).

Baudet and Stallebrass (2004) presented a constitutive model for structured clays based on the sensitivity framework (Cotecchia and Chandler 2000). In their model, the structure allows the material to exist with higher void ratios than the equivalent reconstituted soils. The structure’s influence is represented by a large metastable SBS that decreases with the degree of destructuration until it reaches the intrinsic SBS or even a slightly larger SBS due to a persistent fabric element.

Following the development of Baudet and Stallebrass (2004), the large apparent SBS in Fig. 11 would be related to unstable elements of fabric that gradually would disappear according to the degree of destructuration. The SBS still did not converge to a unique surface even with pressures as high as 4,000 kPa (Fig. 11). Thus, the destructuration of the studied tailings is associated with pressures increasing to values much higher than those usually achieved in conventional laboratory testing.

Moreover, Baudet and Stallebrass (2004) represented structure using a horizontally amplified space in which the material can exist. This amplification originates from their structured samples having a NCL parallel to the intrinsic NCL, while sharing a unique CSL. However, for tailings, the structure allows them to exist in metastable states in which the CSL cannot be defined fully. This gives rise to the nonlinear CSL observed at low pressures and to the large apparent SBS observed (Fig. 15).

Although the state paths in Fig. 11 do not collapse into a unique SBS with the normalization adopted, they are enclosed by and have the same shape as the true SBS. Fig. 16 plots in the compression plane the peak points indicated in Fig. 11. These points define a locus of instability, similar to what Carrera et al. (2011) observed. This instability locus can be related to the break of the fabric that allowed the material to exist at these states. However, the instability does not take place in the tests conducted because the stress probe direction was in a strain-hardening direction (increasing $p^{\prime }$) rather than in a strain-softening direction (decreasing $p^{\prime }$) as in the case of undrained tests on loose samples (Sasitharan et al. 1994).

The usual linear approximations of the relationship between ${\eta }_{\max }$ and minimum plastic dilatancy ($D_{\min }^P$) do not fit well the experimental data obtained (Fig. 17). This was because the commonly adopted flow rules are a function of Degree 1 of the mobilized stress ratio. To illustrate how the degree of $\eta$ for the flow rule influences ${\eta }_{\max }$ versus $D_{\min }^P$, an adapted version of the flow rule for modified cam clay is adopted in this paper; the derivation is presented in the Appendix.

This flow rule is based on a new parameter $k$ related to the particle morphology evolution. de Bono and McDowell (2014) showed using discrete-element modeling with crushable or uncrushable particles that the change from a dilative to a compressive trend in granular materials occurs because of the balance between dilation caused by shearing and compression which originates from particle crushing. Based on this concept, the $p_0^{\prime }$ value at which the initial dense specimen begins to have contractive behavior (Fig. 7, 0.60_4000 kPa_D) is taken as the instant at which the two deformation mechanisms (dilation due to shearing and compression due to particle breakage) cancel out. Thus, $k$ is assumed to vary linearly from 0 (at which there is no influence of particle breakage) to 1 (compression due to breakage suppress dilatancy trend) from 0 kPa to the pressure at which the volumetric behavior trend changes (4,000 kPa in this case). Fig. 17 depicts the new flow rule derived in relation to the experimental data.

The modified flow rule adopted demonstrates the nonlinear nature of the relation between ${\eta }_{\max }$ and $D_{\min }^P$. These results reinforce the relationship between plastic strains, particle breakage, stress level, and fabric.

Particle breakage often is related to large volumetric deformations and very high pressures. However, there can be many types of breakage (Fig. 18). Fragmentation [Fig 18(b)] may require extremely high pressures originating from stress concentrations, whereas other types, such as chipping [Fig. 18(c)], could occur at lower stress levels. These mechanisms of particle breakage occurring at lower stress levels may not affect the soil particle-size distribution directly, but instead affect its particle morphology.

The shape of the particles determines how they arrange and interact, resulting in changes in the macroscopic response. Usually, particle shape is associated directly with interlocking effects, and more-angular particles provide increased strength and stiffness. However, particle shape may affect other fundamental properties, such as the internal friction angle (Ciantia et al. 2019) and yielding behavior (de Bono and McDowell 2020).

The iron oxide contents of the iron ore tailings studied were concentrated mainly in the smaller particles, for which the recovering processes are less efficient. Because particle strength is a combination of mineral strength and particle size (McDowell and Bolton 1998) and smaller particles usually have higher coordination numbers (and thus, less probability of stress concentrations), it is not expected that the smaller particles will break. Instead, the quartz minerals present in tailings result from industrial processes, which gives them characteristics such as high angularity, low sphericity, and high surface roughness. Thus, during loading, the smaller iron particles may act like the balls in a ball mill, degrading the larger quartz particles and changing their shape to more spherical and less rough.

The particle breakage patterns described in Fig. 18 are related directly to contact mechanics and particle morphology. Usually, particles with fewer contacts are more prone to fragmentation [Fig. 18(b)] (Artoni et al. 2019; Karatza et al. 2019; Todisco et al. 2017). Karatza et al. (2019) observed that particles with a higher coordination number usually present chipping and that less spherical particles present more breakage in general because of stress concentrations at the contacts.

Aspects of particle morphology for intact tailings were determined by analyzing SEM images aiming to confirm the mechanisms occurring. Image-based methods to determine for particle morphology have been used in soil mechanics to obtain information about the microscopic characteristics of particles and their arrangements (e.g., Bowman et al. 2001; Sun et al. 2019; Yang et al. 2019; Zheng and Hryciw 2015). The studied tailings were found to be angular [convexity $(C)\approx 0.85$] and elongated [aspect ratio $(\mathrm{AR})\approx 0.60$]. Approximately 250 particles was evaluated in the calculations, following the recommendations of Bowman et al. (2001) for the statistical significance of shape descriptors. These particle characteristics provide arrangements with more contacts than do spherical particles (Adesina et al. 2023; Fang et al. 2022; Karatza et al. 2019; Tian and Liu 2019). Thus, due to the nature of the tailings—elongated, angular, and containing fines—it is expected that they have more contacts and are more susceptible to breakage modes such as chipping. This was confirmed by the analysis of posttested samples.

The particle-size distributions of posttest samples were plotted by mass percentage rather than the usual cumulative distribution to investigate the changes in particles due to loading. Plotting data this way allows observing some breakage patterns and investigating the effects on specific particle sizes (Dong et al. 2024; Vilhar et al. 2013; Zhang and Baudet 2014). The changes in particle-size distribution were related mainly to the mean particle size (Fig. 19). In this range (0.1–0.2 mm), the most marked difference in the curves occurred; however, these differences are not observed directly in the distribution of finer particles ($<0.075\mathrm{mm}$). This indicates that the mechanism occurring is related more to changing particle morphology than to breaking particles and increasing fines content.