Insight into Subgrade Long-Term Uplift Characteristic of Swelling Red-Bed Controlled by Temperature and Cyclic Load：From Laboratory Experiments Open Access

In order to ensure that railways can operate stably during their service life, the long-term deformation of subgrade needs to meet strict control standards [1-5]. Railway subgrades in China and other countries like Japan and the United States face the challenge of deformation, which seriously affects their operational safety. Among the various types of deformation, subgrade uplift is the greatest threat due to the expansive rock mass [6, 7]. Expansive geotechnical materials are a common cause of subgrade deformation in high-speed railways (HSRs), and it is important to study the swelling characteristics of expansive material and the swelling law under long-term cyclic action to control the deformation of subgrades in HSRs. Zhong [8] reported that the deformation of subgrades is due to the swelling and disintegration of the subgrade material. Dai [9, 10] agrees that the cause of inhomogeneous bulging in subgrades is the infiltration of rainfall through the weak structural surfaces to the deep rock mass. By means of numerical simulation, the swelling law of the subgrade under long-term rainwater infiltration conditions was fitted. Wang [11] concluded that the subgrade uplift is caused by long-term rainwater infiltration and established a model of subgrade uplift by FLAC and PFC. These scholars all highlight the importance of subgrade material swelling, whereas two factors that strongly affect the swelling are rarely studied: initial water content and cyclic load.

The red beds, as important expansive lithologies, are widely distributed in Southwest China, accounting for about one-third of the total red-bed area in China [12, 13]. The mudstones and siltstones in the red beds usually contain clay minerals such as montmorillonite, illite, and kaolinite [9]. They are often red or lime green in color due to the ratio of divalent to trivalent iron. The abundance of montmorillonite is the main reason for the swelling of red-bed rocks. Numerous studies have shown that the red layer in Southwest China has significant swelling characteristics, and the initial water content exerts a controlling role on the swelling characteristics. Huang [5] reported that the vertical pressure acting on the subgrade shows an inversely proportional relationship with swelling. Previous studies have conducted a series of water-absorption and swelling experiments on red-bed mudstones of the Jurassic Shaximiao Formation and Badong Formation in Southwestern China [14-17]. The experimental results show that the swelling characteristics of the red-bed mudstone from across formations are significantly different. The swelling characteristics of the Shaximiao Formation mudstone and siltstone vary, and the swelling and deformation is influenced by multiple reasons such as clay mineral content, initial water content, and intrinsic structural properties.

Heat treatment is commonly used to control the initial water content of the samples. The swelling potential of clay-rich minerals increases with decreasing initial water content [18], making low-water content samples critical for accurate testing. Chinese standard [19] specifies heating to 105–110°C (above water’s boiling point) to evaporate pore water and maximize swelling rates. In contrast, the International Society for Rock Mechanics testing methods omit thermal pretreatment for swelling measurements. Initial water content critically regulates swelling release processes. Studies indicate that ballastless HSR tracks require strict subgrade flatness tolerances: <4 mm uplift and <15 mm settlement [1, 2]. Subgrade deformation compromising train operations has been documented in China, France, the United States, and Spain; the subgrade deformation threatening the normal operation of trains has occurred [20]. Previous findings have shown that weakly swelling subgrades can threaten the safety of railway operations. Existing research identifies three primary uplift mechanisms: (1) Swelling by soluble inorganic salts through water rock interaction (common in expansive subgrades or saline groundwater areas); (2) Ground stress from tectonic activity and softrock creep (prevalent in tectonically active soft rock regions); (3) Frost heave from winter temperatures extremes. Most subgrade uplift involves multiple contributing factors. Swelling behavior in red-bed mudstones and fissile rocks is governed by initial water content, expansive mineralogy, depositional history, and microstructure [21], with release mechanisms depending on clay content, sedimentary genesis, hydration state, and environmental conditions [7, 12]. While previous studies focused on subgrade material mechanics under high-speed train vibration, long-term swelling responses to cyclic loading remain poorly characterized [22, 23].

In the central Sichuan Basin’s red beds traversed by HSR lines, the Neijiangbei station in eastern Sichuan Province experienced postconstruction subgrade uplift, drawing widespread concern [9-11]. This study investigates swelling mechanisms through experiments evaluating temperature-controlled initial moisture content and cyclic loading effects, supplemented by microstructural analysis. Our findings advance understanding of swelling evolution in heat-treated rocks and quantify cyclic loading impacts on siltstone swelling behavior.

The study area is situated in Neijiang, Sichuan Province, China. The investigated stratigraphic unit was deposited during the middle to late Jurassic and belongs to the Shaximiao Fm (⁠$J2s1$⁠). Sample lithologies include purple-red mudstones and green siltstones exhibiting horizontal or low-angle bedding, with locations illustrated in Figures 1(a) and 1(b). Mudstone and siltstone specimens were collected from a 10-m-deep quarry (Figure 1(c)) adjacent to a tunnel near Neijiangbei station, corresponding stratigraphically to the station’s foundation layer. Mudstone’s interbedded structure and weakly cemented lamellar joints hindered cylindrical sample preparation, limiting cyclic loading tests predominantly to siltstones. The experimental samples comprised:

45 cylindrical siltstone samples (100 mm × ϕ50 mm, ±1 mm)

75 pie-shape mudstone samples (20 mm × ϕ50 mm, ±1 mm)

75 pie-shape siltstone samples (100 mm × ϕ50 mm, ±1 mm)

All samples were obtained via air-cooled drilling (Figures 1(d)–1(f)).

Bulk densities measure 2.67 g/cm³ (mudstone) and 2.94 g/cm³ (siltstone). XRD analysis yields the following volumetric compositions:

Mudstone: Quartz (39.9%), feldspar (5.6%), plagioclase (20.2%), calcite (5.9%), hematite (2.1%), mica (1.6%), and clay minerals (24.7%: illite 21%, kaolinite 4.04%, chlorites 4.04% and illite—montmorillonite interstratified 70.92%).

Siltstone: Quartz (38.4%), feldspar (6.2%), plagioclase (35%), mica (1.2%), and clay minerals (19.2%: illite 35%, kaolinite 6.4%, chlorites 6.4% and illite—montmorillonite interstratified 64.3%). The contents of clay minerals are similar in two types of rocks. Both lithologies share comparable clay mineral contents. The purple-brick-red mudstones exhibit poor cementation and marked disintegration susceptibility, with microfractures observed on specimen surfaces. In contrast, lime-green siltstones demonstrate competent cementation and structural homogeneity.

In this study, swelling experiments are performed on mudstone and siltstone samples, including: (1) swelling of mudstone and siltstone under different initial water content conditions, which was controlled by heat treatment at different temperatures. (2) swelling of siltstone samples after cyclic loading by simulating the effect of train operation (Figure 2). Specific sample numbers are shown in Tables 1 and 2.

The experimental procedure is as follows:

One hundred and fifty samples, including 75 mudstones and 75 siltstones, were divided into four sets: set N and S for preliminary testing, set NS and SS for the main study, and set NCG and SCG as control groups (CGs) for mudstone and siltstone, respectively. We chose seven levels of heat treatment at 50, 70, 90, 100, 110, 130, and 150°C and used five samples for replicate tests at each temperature level. All samples and levels were divided into three groups: low-temperature (50–70°C, LT), medium-temperature (90–110°C, MT), and high-temperature (130–150°C, HT). Each sample was first dried at different heat treatment temperatures and quickly removed from the oven every 5 minutes to be tested for mass and then subjected to swelling experiments. Each experimental run consisted of three steps: (1) heat treatment, (2) scanning microscopy observation, and (3) swelling experiment. The recording time intervals for all experimental runs were 600 seconds. We stopped the tests when the test data were when the change of mass is less than 0.01 g. The CG is without heat treatment.

Forty-five siltstone samples were divided into two sets: set SN (nine samples) for triaxis test in order to obtain average critical stress at different circumferential pressures; set SC (thirty-six samples) for cyclic loading triaxial test. The microstructure of the destroyed samples was observed after each triaxis and cyclic loading experiment. For siltstones that remained intact after cyclic loading experiments, their swelling rate was measured subsequently.

The compositional changes in the clay-rich minerals of rocks in high-temperature environment can be traced by using thermos-gravimetric analysis (TGA). The TGA results show that montmorillonite and illite typically yield three inflections in TGA curves due to dihydroxylation at temperatures of 126, 650, and 906°C. Mineral composition changes at 650 and 906°C [24, 25]. As a result, we set the heat treatment temperatures at 50, 70, 90, 100, 110, 130, and 150°C. Heat treatment is stopped until the weight changes less than 0.01 g after three consecutive tests. The initial water content and water loss are calculated using Equations (1) and (2).

where $ω$ is the water content of the sample; $mw$ is the mass loss after 150°C heat treatment; $ms$ is the average mass of the sample after 150°C heat treatment; $m0$ is the initial average mass of each group; and $mt$ is the average mass of each group at the end of the experiment.

In order to understand the changes of water content under heat treatment, we use the relative water loss (RWL) to describe the state of initial water content before swelling experiment, as shown in Equation (3)

where RWL is relative water loss which is based on the weight loss of heat treatment at 150°C; $mi$ is the final mass loss at different temperatures, i = 50, 70, 90, 100, 110, and 130.

Thirty-six siltstone samples were selected to perform cyclic sinusoidal loading experiment under different frequencies, amplitudes, and circumferential pressure (Table 3). The cyclic load amplitudes were determined by the product of critical stress measured by triaxial test and CSR [26]. The experiment was ended after 5000 cycles or if the rock sample was damaged. The cyclic loading schematic is shown in Figure 3.

The microstructure of the samples was studied using a ZEISS EVO10 scanning electron microscope (SEM) equipped with a Leica EM ACE 200 28-nm golden ion sputter (sample pretreatment), a JSM-IT 500 tungsten lamp, and an OXFORD X-Max 80 energy dispersive spectrometer. The microstructure of the sample is affected by the stress state of the sampling section, so the observation surfaces are selected from the tension surface of the Brazilian splitting test, which is a laboratory test conducted in rock mechanics to indirectly determine the tensile strength of rocks. According to the mechanics of materials, the destructive surface of this test is the tensile surface. The ion sputter was used to create a conductive layer on poorly conductive rock samples.

The heat-treated and cyclic loaded rock samples were placed in a desiccator and cooled at room temperature. Oedometer was used for measuring the swelling rate every 600 seconds. The swelling rate was calculated by Equation (4). When the change of swelling rate is less than 0.01% for three consecutive records, we stop the experiment.

where $ε$ is swelling rate, %; $Ht$ is the sample height change during the test, mm; and $H0$ is the height at beginning of sample, mm.

Since water evaporates when the temperature exceeds 100°C, the boiling point is used as a discontinuity point to segment the water loss pattern of the sample. Figure 4 illustrates the changes in water loss of the two types of rock after heat treatment. The water content decreases linearly with temperature increase when the temperature is higher or lower than the boiling point. However, for the mudstones, the decrease slope of the water content is significantly different at two temperature ranges, in which the slope of line A is higher that of line B. For the siltstones, the decrease slope of water content is stable and also shows a linear relationship regardless of whether the temperature is higher or lower than the boiling point (line C).

Overall, the general trend is that as heat treatment temperature increases, the water loss increases for both samples. The effect of temperature on water loss is greater for the mudstone than for the siltstone. As shown in Figure 5, the swelling rate of mudstone and siltstone increases with increasing heat treatment temperature, but this pattern is not obvious for siltstone. The swelling rate and duration of the mudstone are much greater than that of siltstone at each heat treatment temperature. A linear fit of the results of the swelling experiment after heat treatment is line D and line E in Figure 5. Notably, both samples show anomalous swelling rate fluctuations at 90–110°C (blue area in Figure 5). The swelling rate of mudstone and siltstone ranges from 34.86% to 39.76% and 1.81% to 2.53%, respectively.

Strong weathering of the Shaximiao Fm mudstone poses difficulties in preparing triaxial test samples. Since the thickness of the siltstone in the study area is much greater than that of the mudstone, this study focuses on the swelling of siltstone. Based on the average value, the swelling of the siltstone samples after cyclic loading is significantly greater than that of the natural sample group. The swelling rate ranges of siltstone after cyclic loading from 2.23% to 5.46%. The maximum value occurs in the group with the load condition of “0.5 Hz-90%-1 MPa,” representing a 3.06-fold increase. The smallest value occurs in the group with the condition of “0.5 Hz-80%-5 MPa,” which is almost equivalent to the CG (average swelling rate in CG group is 1.87%).

The swelling pattern of siltstone after cyclic loading is similar to that of mudstone and siltstone after heat treatment and can also be divided into three stages. The first stage is the rapid swelling stage, during which the swelling rate increases rapidly and reaches its maximum. This stage is within 6 hours after the test begins. In the second stage, the slow swelling phase, the swelling rate continues to increase but decelerates gradually, ending 24 hours after the test begins. In the final stable stage, the swelling behavior of siltstone samples ceases and the curve remains stable. Although discontinuous nodes in the swelling stage of each sample vary slightly, the general trends are consistent. Figure 6 shows the swelling of siltstone under different frequencies, circumferential pressure, and CSR conditions. Swelling rates were positively correlated with CSRs (R² > 0.89) and negatively correlated with circumferential pressure (R² > 0.54), while no significant correlation with loading frequency was observed.

Figures 7(a)–7(d) show results for mudstone samples treated at different temperatures, while siltstone samples at different temperatures are show in Figures 7(e)–7(h). The microstructure of the mudstone sample exhibits layered distribution, with few natural fractures in the original sample (Figure 7(a)). Detachment traces of detachment of embedded particles can be found in samples treated at different temperatures (Figure 7(b)).

As the heat treatment temperature increases, the orderly layered structure of mudstone is disrupted, and the area of the planar structure decreases. SEM images reveal a pattern of intensified fragmentation. After heat treatment, siltstone sample develops small cracks, which may be related to the drying shrinkage of clay minerals [7]. However, the planar structure in Figure 7(d) is similar to that in Figure 7(a), indicating that liquid gasification at higher temperatures has minimal impact on the microstructure and preserves the original physical structure. The microstructure of the siltstone sample shows a particle stacking structure, with many clay mineral particles and nonclay particles stacked in a planar shape. The original sample (Figure 7(e)) has many natural fractures, but the fractures are not fully connected. Crystals of chlorite can be found in the fractures. As shown in Figures 7(f)–7(h), it is evident that the heat treatment process causes damage to the microstructure, not only increasing the number of cracks that rupture between particles but also increasing the number of cracks inside the particles.

In the cyclic load test, only five samples failed with similar failure mode, which were “0.5 Hz-95%-1 MPa,” “0.5 Hz-95%-2 MPa,” “1 Hz-95%-1 MPa,” “1 Hz-90%-1 MPa,” and “1.5 Hz-95%-2 MPa.” Using the sample of “1.5 Hz-95%-2 MPa” as the representative sample, the failure surface was observed via SEM (Figure 8). Particles on the failure surface of normal triaxial test samples are uniformly arranged, with no obvious large cracks; interparticle cementation remains intact, and cohesive cracks exist between particles (Figure 8(a)). Particle diameters range from 100 µm to 200 µm. In contrast, failure surface of the cyclically loaded samples exhibits not only coarse cracks but also disrupted interparticle cementation, forming scattered particles with a diameter below 100 µm. Agglomerations of small particles in Figure 8(c) show diameters below 10 µm. These phenomena indicate that after cyclic loading, although the external shape of the siltstone does not change, there are significant changes in the microstructure, including but not limited to the breakage of cementation and the formation of coherent cracks between and within large particles.

Through normal loading and cyclic loading tests, we observed that the failure surface was enriched with shed particles of different sizes, exhibiting distinct physical forms (Figure 9). These particles were mounted on conductive adhesive tape for SEM observation (Figures 9(b) and 9(c)). Psephicity was quantified using Image Pro Plus 6.0 to quantify the particle roundness. Psephicity values (Figure 9(a)) and long/short axis frequency ratios (Figures 9(d) and 9(e)) were higher for cyclically loaded samples. Particle psephicity from failure surfaces was primarily distributed between 1 and 2. After cyclic loading, particles exhibited more spherical morphologies with reduced sharp angles. The proportion of particles with psephicity values from 1 to 1.5 under cyclic loading was twice that under normal loading, indicating ellipsoidal morphologies. Major and minor axes of the particles after cyclic loading showed similar dimensions (Figures 9(d) and 9(e)).

It is well known that changes in the water content of rock are affected by multiple complex factors such as temperature, humidity, rainfall, groundwater table fluctuations, and surface runoff. The swelling characteristics are comprehensively by initial water content, expansive mineral, depositional history, and microstructure [27, 28]. Heat treatment is a widely used method to control rock initial water content. To obtain maximum swelling potential, Chinese testing standards typically adopt a “complete drying – to - complete wetting” protocol (mass stabilization after drying → volume stabilization after drying wetting) to determine maximum swelling rates [19]. Conventional engineering practical assumes that the water in the rock can be completely removed via high-temperature treatment. However, advances in unsaturated soil mechanics and rock mechanics have revealed complex water retention mechanisms in geomaterials [29, 30]. Given that heat treatment remains a standard method for controlling the initial water content in laboratory tests, we performed swelling experiments on heat-treatment samples to evaluate the accuracy of conventional swelling test protocols.

The water in the rock sample equilibrates dynamically with ambient humidity. Under constant environmental conditions, water molecules in the sample are exchanged with water molecules in the air by Brownian motion. Temperature elevation increases air saturation capacity while reducing relative humidity, inducing sequential water loss: (1) hydrogen-bonded water, (2) cation-adsorbed water on hydrophilic mineral surfaces, until equilibrium is reestablished [29]. As shown in Figures 4 and 5, a linear relationship between the initial water content and swelling rate of the mudstone and siltstone was established (Figure 10). The siltstone and mudstone contain 19.2–24.7% clay minerals. When mineral content effect is excluded, microstructural integrity primarily governs swelling behavior.

Clay-dominated geomaterials exhibit swelling through hydration-induced expansion of layered minerals (e.g. montmorillonite, illite). Under low initial water content, cations hydration (Na⁺ and Ca²⁺) promotes enhanced double-layer repulsion and interlayer expansion, while discontinuous pore water distribution generates high matrix suction that amplifies swelling potential [29]. Thus, lower initial water content correlates with higher hydration capacity and dimensional changes. Thermal treatment at varying temperatures was employed to modulate swelling potential. Given that ambient temperature controls initial water content, we selected temperatures optimizing moisture remove without mineralogical alteration. SEM and swelling experiment results validate this approach although temperatures 70–150°C exceed natural conditions they simulate field-equivalent moisture states. Swelling potential was quantified using empirical relationships (Figure 10). Despite low R² values (mudstone, R² = 0.50; slitstone, R² = 0.73), a linear correlation emerges confirming that pretest thermal conditioning controls initial water content and subsequently biases the results of the swelling experience.

Cyclic loading induces microstructural damage in geomaterials, generating a microcrack network and reducing the interparticle cementation [31], thereby altering rock swelling behavior. Mechanistically:

1. Enhanced crack networks facilitate water infiltration, exposing clay minerals to hydration and reactivating secondary swelling in incompletely hydrated minerals;

2. Cementation degradation compromises structural integrity, amplifying swelling deformation postlocal stress relaxation.

Pore structure reorganization during cyclic loading drives anisotropic swelling. This swelling behavior emerges from synergistic damage accumulation and water migration, manifesting macroscopically as continuous deformation. Figure 11(a) shows a conceptual diagram of the microstructure of the siltstone samples as observed via SEM. Cyclic load causes the particles in the sample to be squeezed, resulting in displacement tendencies. As the loading progresses, random failures occur on the weak surfaces within the sample, including interparticle fracturing and intraparticle damage (Figure 7(a)). These failures accumulate until penetrating cracks form (Figures 8(b) and 8(c)). During cyclic loading, particles experience compressive stresses, but individual loading cycles fail to generate sufficiently weak through-fractures. Microscopically, intergranular damage predominates, which does not affect the macroscopic structure but compromises intergranular cementation. Irregular natural pores and cracks (Figure 7(e)) contain mineral filling and mineral cementation. Before cyclic loading, the swelling potential of some clay minerals was fully released, while others remained constrained by the surrounding nonexpansive minerals (Figures 8(a) and 11(b)). After cyclic loading, particles are crushing and bond destruction occurs (Figures 8 and 11(c)), indicating that cyclic loading reduces constraints on expansive minerals, thereby enhancement of the swellability.

Traffic loading affects the subgrade in two ways. (1) Continuous train-induced cyclic loading compacts the subgrade [22, 23]. (2) It alters the microphysical properties of the subgrade. Cyclic load usually rarely causes macroscopic structural damage, but induces microscopic damage accumulation, leading to mineral bonding failure between particles. Previous studies demonstrate that particle-bonded geomaterials undergo damage under cyclic loading, modifying mechanical properties. Compared to laboratory-reshaped samples, undisturbed samples retain natural spatial characteristics and particle bonding that connects internal particles. Cementation by nonexpansive minerals inhibits swelling release. Statistical results reveal that cyclic loading modifies the internal physical structure of the siltstone through external dynamics. Although no sample failure occurred experimentally, these structural changes explain postcyclic loading swelling rate variations. Cyclic loading induces loose granulation of nondetached particles (Figure 8(c)). In this study, cyclic loading simulates repeated train traffic loads. Swelling experiment confirms that cyclic load damages the subgrade structure, facilitating swelling release of swelling during water fluctuations. This aligns with the continuous subgrade uplift of the Neijiangbei station after operation, suggesting a strong connection between the HSR operation and subgrade uplift.

Dai and Wang [9, 11] proposed that the reason for the long-term subgrade uplift at Neijiangbei station lies in continuous infiltration of water due to the rock fissures caused by excavation. Additionally, it has been suggested that the excavated section experienced basement rebound due to foundation rock creep under external ground stresses [32]. Notably, field monitoring data reveal that subgrade long-term uplift persisted after the operation began in December 2015 with no significant convergence observed. In regions with strong tectonic activity or deeply buried tunnels, subgrade deformation typically results from tectonic or gravitational stresses; however, the Shaximiao Fm mudstone and siltstone exhibited poor strength and high creep susceptibility. Some scholars attribute subgrade uplift to horizontal geostatic stresses [32], while other scholars propose that the Shaximiao Fm mudstone and siltstone’s natural water resistance allowed prolonged rainwater infiltration into expansive particles, causing delayed swelling. Zhong [8] conducted a long-term immersion experiment on Neijiangbei station which in Shaximiao Fm, showing that the red-bed rocks swelling correlated with soaking time, with 98% of the samples completed the stabilization after 145.2 hours [33]. Notably, the subgrade uplift of the Neijiangbei HSR station has two significant characteristics:

1. Preconstruction geological investigations detected no significant swelling in the Shaximiao Formation mudstone and siltstone.

2. Continuous uplift occurred within 2 years postoperation.

Figure 12 presents a rock-swelling-based mechanism for Neijiangbei’s subgrade uplift, differing from prior studies. Road cutting altered the original hydrological environment, inducing subgrade water content fluctuations (Figure 12(b)). These changes triggered mudstone/siltstone swelling, initiating early upward displacement. Moreover, cyclic loading accumulated internal damage of red-bed rocks, disrupting the original cementation mode between particles. Over time, operational vibration loads progressively activated swelling in subgrade materials (Figure 12(c)).

This study investigated red-bed subgrade uplift mechanisms through laboratory experiments. Although simulated cyclic loading conditions were simplified, real-world train-induced vibrations are far more complex, necessitating further research. Subgrade uplift arises from multifactorial interactions, with this study focusing on the combined effects of water content and cyclic loading.

This article investigates the swelling characteristics of red-bed mudstone and siltstone under the influence of heating and cyclic loading through a series of experiments. The mechanism for the subgrade uplift of Neijiangbei station is discussed from the perspective of rock swelling. The following conclusions are obtained:

1. The microstructure of mudstone is layered while that of siltstone is particle stacking. Temperature has more significant influences on the increase of microcracks of mudstone than that of siltstone. The initial water content of red-bed rocks is influenced by the heat treat temperature and negatively correlated with the swelling rate. The swelling rate of mudstone and siltstone ranges from 34.86% to 39.76% and 1.81% to 2.53%, respectively.

2. Cyclic load promotes the increase of particle fragments and cracks of siltstone. The shape of exfoliated particles after cyclic load approaches to be round than that after normal load. The accumulative damage causes further swelling release. The swelling rate range of siltstone after cyclic loading from is 2.23% to 5.46%, which maximum 3.06 times higher than that of the natural siltstone samples. Swelling rates were positively related to CSRs (R² > 0.89) and negatively related to circumferential pressure (R² > 0.54), while no significant correlation with cyclic loading frequency.

3. The reason of long-term subgrade uplift at Neijiangbei station are attributed to two factors: 1) the fluctuation of water content caused by environment changes leads to the swelling releasing of red-bed rocks; 2) the accumulative damage of microstructure of subgrade rocks caused by continuous cyclic loading induce further swelling during the high speed railyway operation.

The datasets generated and analyzed during the current study are available from the corresponding author upon reasonable request.

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this article.

This research is funded by the National Natural Science Foundation of China (42477200, XL), Chengdu Science and Technology Program (2022-YF05-00340-SN, SXL), and the Innovative Practice Bases of Geological Engineering and Surveying Engineering of Southwest Jiaotong University (YJG-2022-JD04). We thank Dr. Weizhen Fang from the Analytical and Testing Center of Southwest Jiaotong University and Dr. Siyuan Zhao from the Key Laboratory of Deep Earth Science and Engineering (Sichuan University) for their help.