Mechanisms of Fouled Railway Ballast Deterioration Under Freeze–Thaw and Cyclic Loading: Implications for Sustainable Maintenance in Seasonal Frozen Regions

Abstract

Maintaining ballast performance in seasonal frozen regions is essential for resilient and sustainable railway infrastructure because freeze–thaw-driven fouling can shorten service life and increase maintenance-related material consumption. To investigate the deterioration mechanisms of fouled railway ballast in seasonal frozen regions, freeze–thaw cycle tests and cyclic loading model tests were conducted in sequence using a custom low-temperature geotechnical system. The test results processed by Origin software indicate that unfrozen water migrates toward the freezing front under temperature gradients and forms ice lenses during freezing. During thawing, meltwater is retained above the underlying frozen soil. Repeated freeze–thaw cycles therefore promote progressive water accumulation in the upper soil layers, eventually forming a clay layer with high water content. Under cyclic loading, interlayer thickening exhibited clear moisture thresholds relative to the clay liquid limit (LL = 24%). Below the LL (18–24%), ballast penetration and fines migration were limited and thickness increased slowly. Above the LL, rapid strength loss accelerated penetration and upward transport. At an initial water content of 32%, fines migration surpassed the ballast surface and the ballast became fully fouled, meaning that the fouled interlayer thickness equaled the full 100 mm ballast-layer thickness. Fouling severity increased sharply with moisture: the void contaminant index exceeded the maintenance criterion (VCI > 40%) at 28% water content and evolved into severe mud pumping at higher concentrations. Excess pore water pressure developed stratification with depth, maintaining an upward hydraulic gradient near the interface and yielding a net water loss of 2.24–6.91% in the upper fine-grained layer. These quantified thresholds and mechanistic insights provide actionable trigger points for condition-based maintenance and climate-adaptive design, helping extend track-bed service life and reduce resource-intensive ballast renewal in seasonal frozen regions.

1. Introduction

Ballasted tracks are widely used in railway lines crossing seasonally frozen soil regions of China [1]. The ballast layer, consisting of granular crushed stones, distributes sleeper loads, provides lateral and longitudinal resistance, and offers drainage that helps mitigate frost and water-related deterioration [2,3]. With increasing service time, ballast inevitably degrades through abrasion, breakage, and fouling. In seasonal frost environments, repeated freeze–thaw cycles followed by train loading can intensify the upward migration of fine-grained subgrade soils into ballast voids, promoting fouling and the development of a dense, moisture-sensitive fouled ballast interlayer at the ballast-subgrade interface [4]. As schematically shown in Figure 1, this interlayer impairs drainage continuity and stiffness transfer, accelerates differential settlement, and increases maintenance demand, thereby threatening track performance and safety [5,6,7]. From a sustainability perspective, such deterioration increases the frequency of tamping and ballast renewal, raises material and energy consumption, and generates additional waste (fouled ballast), motivating a clearer, mechanism-based identification of deterioration drivers and practical trigger conditions.

Extensive studies under non-freezing conditions have established the existence and mechanical relevance of this interface zone and clarified key drivers of its development [8,9,10]. Field investigations have identified a mixed transition zone between ballast and subgrade in in-service railways [11], and laboratory/model studies have shown that cyclic train loads, particularly under wet or near-saturated conditions, can generate excess pore water pressure in fine-grained subgrades, reduce effective stress, and drive ballast embedment and ballast–subgrade mixing, thereby forming and thickening an interlayer [12,13]. Mud pumping is frequently observed alongside ballast fouling, and a developed interlayer can aggravate pumping severity [14]. To support characterization, indices such as ballast penetration metrics and contamination descriptors have been proposed [15], and subsequent testing has repeatedly confirmed that moisture state governs strength loss and permanent deformation once the interlayer approaches saturation [16]. However, most of these frameworks were developed for temperate environments and do not explicitly incorporate freeze–thaw cycle-induced changes in moisture supply, drainage boundary conditions, and soil fabric.

In cold regions, the freeze–thaw cycle fundamentally alters the hydro-mechanical context that controls fines migration and interlayer evolution. Freezing can drive moisture redistribution toward the freezing front via cryosuction, increasing near-surface water availability for subsequent thaw weakening [17]. During thawing, meltwater accumulation and drainage restriction can enhance pore pressure build-up under cyclic loading, promoting fines expulsion and pumping-type migration [18,19]. Model tests indicate that a shallow water table combined with freeze–thaw conditions can markedly increase pore water pressure and trigger upward fines migration; transient frozen layers may impede drainage during thaw and worsen instability [20]. Freeze–thaw cycles can also modify pore structure and increase the susceptibility of fine-grained subgrade soils to cyclic deterioration [21], while coupled hydro-thermal-mechanical analyses suggest that water–fines migration may evolve into a self-reinforcing process that progressively reduces effective stress and facilitates mud intrusion [22]. Frost heave may further force fines into coarse voids, preconditioning intrusion and mixing at the interface [23]. Despite these advances, the literature still lacks a coherent experimental linkage between freeze–thaw-induced high-moisture conditioning and the subsequent interlayer development under cyclic loading, and the controlling role of hydraulic-gradient evolution in driving water/fines migration and ballast embedment has not been quantified in a systematic manner.

Accordingly, three related questions remain unresolved for seasonally frozen ballasted tracks: (1) how freeze–thaw-induced moisture accumulation (including the formation of a high-water-content clay layer) establishes boundary conditions that promote rapid interlayer development; (2) how excess pore water pressure and vertical hydraulic gradients evolve under cyclic loading and govern migration pathways; and (3) how these hydraulic processes interact with ballast embedment to produce stepwise thickening of the fouled ballast interlayer and trigger the transition toward mud pumping.

Table 1. Comparison of representative studies and the present study.

Study	Primary Focus	Freeze–Thaw	Cyclic Loading	Key Limitation
Trinh et al. [11]	Transition identification	No	In-service	Evolution mechanism not quantified
Duong et al. [12,13]	Interlayer formation and fines migration	No	Yes	No freeze–thaw moisture preconditioning
Ding et al. [14]	Mud pumping–fouling interaction	No	Yes	Limited linkage to hydraulic gradients and staged thickening
Chang et al. [20]	Hydrothermal effects on pumping behavior	Yes	Related	Limited interlayer thickness quantification and trigger definition
Wang [22]	Water–soil migration modelling	Yes	Implicit	Lacks systematic experimental validation (thickness/indices)
Do et al. [23]	Frost-heave-induced fines intrusion	Yes	Not central	No pore pressure gradient evidence under cyclic loading
This study	Freeze–thaw, cyclic-loading model tests, Indices	Yes	Yes	Integrates gradients, pathways, thickness evolution, and practical triggers

contrasts representative previous studies with the present study and highlights the specific gaps addressed herein.

To address these gaps, a two-stage experimental program was designed to reproduce the field evolution of ballast–subgrade deterioration in seasonally frozen regions. Freeze–thaw tests were conducted to investigate the formation of a high-water-content clay layer at the ballast–subgrade interface, followed by cyclic-loading model tests under controlled initial water contents to examine the initiation and development of fouled ballast interlayers. The remainder of this paper is organized as follows: Section 2 introduces the materials and experimental methods. Section 3 reports the freeze–thaw test results, focusing on temperature variation and moisture redistribution. Section 4 presents the cyclic-loading test results, including interface observations, excess pore water pressure evolution, hydraulic-gradient response, water content changes, and a quantitative evaluation of ballast fouling. Section 5 discusses the deterioration mechanism and practical engineering implications.

2. Experimental Program

2.1. Low Temperature Geotechnical Testing System

A custom low temperature geotechnical testing system was used for both one dimensional freeze–thaw tests on fine-grained soils and cyclic loading model tests of subgrade. The test system consists of a freeze–thaw test chamber, a model cylinder, and a data acquisition system. As shown in Figure 2, the freeze–thaw test chamber integrates a low-temperature control system and a loading system. The temperature control system includes an internal low-temperature chamber and temperature-controlled bath, with a temperature range of −40 °C to +60 °C. The loading system consists of a servo motor, a loading rod, and an upper loading plate, and is capable of applying dynamic loads such as sine waves, with a load capacity of 20 kN and a maximum frequency of 5 Hz. The transparent acrylic model cylinder has an inner diameter of 200 mm, a height of 310 mm, and a wall thickness of 20 mm. Its base incorporates a water supply channel, while ventilation holes on the side walls allow for sensor placement. The data acquisition system comprises a CR350 data logger, an 8-channel data acquisition module, and sensors for measuring water, temperature, and pore water pressure. Additionally, the Mariotte bottle supplies water through the base of the specimen under a constant head.

2.2. Materials

2.2.1. Subgrade Soil (Low Liquid-Limit Silty Clay)

Ballast fouling is primarily induced by the material of low liquid limit silt, low liquid limit silty clay, or clay. In this study, low liquid limit clay was selected to simulate the subgrade material. Its physical and mechanical parameters are listed in

Table 2. Basic physical parameters of silty clay.

Specific Gravity Gs	Liquid Limit LL (%)	Plastic Limit PL (%)	Permeability Coefficient ksat,0 (m/s)	Maximum Dry Density ρd,max (g/cm3)	Optimum Water Content wopt (%)
2.79	24.17	14.06	3.49 × 10−10	1.73	17.9

, and the particle size distribution curve of this silty clay is shown in Figure 3.

2.2.2. Ballast Simulant (Graded Crushed Stone)

Granite crushed stone served as the ballast simulant. The material consists mainly of angular particles, with a particle density of 2.69 and a maximum dry density of 2.07 g/cm³. Given equipment size limitations, the maximum particle size was restricted to 20 mm, which is less than one-fifth of the specimen diameter. Figure 4 compares the gradation curve of the in-situ ballast against the upper and lower limits specified in the Railway Crushed Stone Ballast standard. To accurately simulate the mechanical behavior of the in-situ ballast, its particle size distribution was modified using a gradation scaling method [24]. Previous studies have demonstrated that the analogous material prepared using this method can reasonably represent the original ballast in terms of its mechanical properties [25]. The ballast particles in different size fractions are also illustrated.

2.3. Specimen Preparation and Testing Procedures

2.3.1. One-Dimensional Freeze–Thaw Test

In accordance with the literature [25], fine-grained soil column specimens were prepared with a dry density of 1.6 g/cm³ and a water content of 18%. The silty clay was first sieved through a 0.075 mm mesh. Water was added and mixed thoroughly to achieve the target water content, after which the soil was sealed for 24 h to ensure uniformity. During specimen preparation, the soil was compacted in eight layers using the layered compaction method. During compaction, temperature sensors were installed at seven vertical heights of 25, 45, 65, 85, 105, 125, and 145 mm from the bottom, while water and pore water pressure sensors were placed at four heights of 25, 65, 105, and 145 mm, respectively. The sensor layout is shown in Figure 5. The specimen had a height of 160 mm and a diameter of 200 mm, and the corresponding compaction degree was 92.5%.

One-dimensional freeze–thaw tests were conducted to investigate the water and heat migration mechanism in fine-grained subgrade soils, focusing on the effect of top freezing temperature. The test conditions are presented in

Table 3. Overview of one-dimensional freeze–thaw test conditions.

Test No.	Top Freezing Temperature (°C)	Top Melting Temperature (°C)	Environment Temperature (°C)	Bottom Temperature (°C)	Freeze–Thaw Cycles
1	−5	+10	+2	+2	3
2	−10	+10	+2	+2	3
3	−15	+10	+2	+2	3

. A top-down unidirectional freezing mode was applied, with continuous water supply at the specimen bottom during the test. Temperature boundary conditions were set as follows: during freezing, the lateral and bottom temperatures were maintained at +2 °C, while target top surface temperatures of −5 °C, −10 °C, and −15 °C were imposed. To account for an observed near 3 °C difference between the bath setting and the actual soil surface temperature, the circulating bath was set to −8 °C, −13 °C, and −18 °C, respectively. During thawing, the top temperature was raised to +10 °C, with sides and bottom kept at +2 °C. Each test condition underwent three complete freeze–thaw cycles.

Prior to testing, the upper loading plate was first slowly lowered until contact with the specimen surface. The model cylinder was connected to the Mariotte bottle, with the water level maintained 10 mm above the base of the specimen to simulate an external water supply. Subsequently, the ambient environment, upper plate, and base were all set to a temperature of 2 °C. Once the internal temperature of the specimen had uniformly stabilized at 2 °C, the freezing phase was started by lowering the upper plate to the target sub-zero temperature. Freezing continued until the internal temperature gradient of specimen reached a steady state. The thawing phase commenced by increasing the upper plate temperature to 10 °C, with the other boundaries held at 2 °C. Thawing was completed when all soil layers exceeded 2 °C. This freeze–thaw cycle was repeated three times, and the test concluded after the third thawing phase.

2.3.2. Ballast–Fine Grained Soil Dynamic Loading Test

The specimen is a 260 mm cylinder with a diameter of 200 mm, consisting of a 100 mm ballast layer at the top and a 160 mm fine-grained soil layer below. The specimen diameter exceeds five times the maximum ballast particle size to minimize boundary effects. The fine-grained soil was prepared at a dry density of 1.6 g/cm³, consistent with the freeze–thaw tests. During preparation, silty clay was mixed to the target water content and then sealed for 24 h. The fine-grained soil was compacted in six layers to account for sensor installation and water variations, with layer thicknesses of 25, 40, 15, 25, 40, and 15 mm from the bottom upward. Water and pore water pressure sensors were installed at heights of 25, 65, 105, and 145 mm above the base, as shown in Figure 6. The ballast layer was placed in two compaction layers following the specified gradation to reach a relative density of 70%. Following the completion of filling, the upper loading plate was placed on the top surface of the ballast layer and carefully leveled.

To investigate the effects of fine-grained water content on fouled ballast interlayer formation under long-term train loading, a cyclic loading model test for subgrade was designed, the conditions for which are summarized in

Table 4. Overview of ballast–fine-grained soil dynamic loading test conditions.

Test No.	Cyclic Stress Amplitude ${\overline{\mathit{\sigma }}}_{\mathbf{cyc}}$ (kPa)	Upper Water Content of Fine-Grained Soil (%)	Bottom Water Content of Fine-Grained Soil (%)
A1	100	18%	18%
A2	100	20%	18%
A3	100	22%	18%
A4	100	24%	18%
A5	100	26%	18%
A6	100	28%	18%
A7	100	30%	18%
A8	100	32%	18%
A9	100	34%	18%

. The lower layer was maintained at the optimum water content of 18% across all test conditions. The upper layer water content was varied at 18%, 20%, 22%, 24%, 26%, 28%, 30%, 32%, and 34% to simulate different wetting conditions. The upper ballast layer of the test specimens remained consistent under different testing conditions.

An initial static axial stress of 15 kPa was applied to the specimen to simulate the self-weight load of the sleeper and train. With reference to related studies, the dynamic stress variation induced by heavy haul trains on the subgrade surface ranges approximately from 35 to 185 kPa [26]. Therefore, a cyclic load amplitude of 100 kPa was selected for the test. The load schematic is presented in Figure 7. To simplify the process of waveform decomposition and experimental analysis, a sinusoidal waveform was adopted for the cyclic loading [27]. The loading frequency is determined by the train operating speed v and vehicle length L, calculated using the formula f = v/L [28]. Based on typical parameters of heavy haul railways, where the train speed ranges from 50 to 80 km/h and the vehicle length is approximately 14 m [29], the frequency range of dynamic stress on the subgrade structure is approximately 0.99 to 1.59 Hz. Accordingly, a loading frequency of 2 Hz was selected for the test, with the total number of loading cycles set to 50,000.

3. Freeze–Thaw Tests: Temperature and Moisture Redistribution in Silty Clay

3.1. Temperature Evolution and Freezing-Front Development

Figure 8 illustrates the temperature changes over time at various heights in the specimen under top freezing temperatures of −5 °C, −10 °C, and −15 °C, respectively. Under all test conditions, the soil undergoes rapid freezing, slow freezing, and temperature stabilization in three distinct stages. Initially, the soil cools rapidly due to a high temperature gradient, which promotes efficient heat transfer. As temperatures near the freezing point, the release of latent heat from the ice-water phase change slows the cooling rate. Further cooling reduces the temperature gradient, which in turn decreases the cooling rate. A lower top freezing temperature, associated with a higher initial temperature gradient, leads to a greater final frozen depth and a lower steady state temperature at the same layer height [30]. Notably, more freeze–thaw cycles extend the time required to reach steady state temperatures. Under open water supply, external water and water from unfrozen zones migrate toward the freezing front during freezing. This increases the initial water content in subsequent cycles. The additional latent heat release upon freezing slows the freezing front advancement, thereby extending the time to reach thermal equilibrium. Moreover, under the same top freezing temperature, deeper soil layers exhibit a delayed thermal response compared to upper layers. During cooling, soil near the cold source freezes first. Heat transfer from deeper layers is weakened by the thermal resistance of soil, leading to lower steady state temperatures near the top. In the subsequent thawing, downward heat transfer is similarly resisted, causing more pronounced warming in shallow layers. Consequently, the temperature profile shifts with depth.

3.2. Moisture Accumulation and Formation of a High-Water-Content Clay Layer

Figure 9 shows the water content distribution along the height of fine-grained soil specimens following each freeze–thaw cycle under top freezing temperatures of −5 °C, −10 °C, and −15 °C, respectively. In Figure 9a, under the −5 °C condition, the water content in the lower layer A and the mid-upper layers C and E increased after the first freeze–thaw cycle compared with the initial state before testing, while the water content in the top layer G remained significant unchanged. Following the second cycle, the water content in layer G shows the highest level and continued to accumulate in subsequent cycles. As shown in Figure 9b, under the −10 °C freezing condition, by the end of the first cycle, the water content in the mid-lower layer C increased to the maximum, with water accumulation also observed in the top layer G. After the second cycle, the water content at different heights gradually increased, with layer C still exhibiting the highest value. By the end of the third cycle, the water content in layer G increased markedly and became the highest among all layers. In Figure 9c, at −15 °C, after the first and second cycles, the water content at the bottom layer A remained the highest, while the top layer G showed a gradual accumulation trend. By the end of the third cycle, the water content in layer G increased sharply to the highest level. In summary, under all tested freezing temperatures, a water enriched zone developed in the top layer G after multiple freeze–thaw cycles. The mechanism involves freezing suction from the advancing freezing front, which draws water from the unfrozen bottom and external supply toward the freezing zone, leading to ice accumulation and ice lens formation near the front. During thawing, the top ice melts first, but the underlying frozen soil and low permeability zones act as an aquitard that impedes downward meltwater percolation. Repeated cycles promote continuous water accumulation in the top layer, ultimately forming a high-water-content region that significantly exceeds both the initial state and the lower layers.

4. Cyclic Loading Tests: Evolution of the Fouled Ballast Interlayer

4.1. Post-Test Observations: Interlayer Formation and Fines Migration

Figure 10 presents the post-test interface between the ballast layer and the fine-grained soil layer. For the specimen with an initial water content of 18% in the upper fine-grained soil, the interface showed only slight interlocking after loading, and no water accumulation was observed. When the initial water content of the upper layer ranged from 20% to 32%, fine-grained soil intruded into the ballast after loading, forming a distinct fouled interlayer. Concurrently, significant water accumulation was also evident near the upper interface, and this increased with the initial water content of the fine-grained soil. During testing, the specimen with 34% initial water content exhibited severe mud pumping, which prevented clear identification of the interface. Therefore, it is not included in the figure.

Figure 11 presents the particle migration in the specimen after testing. Under cyclic loading, the migration height of fine particles increased gradually with the rise in the initial water content of the upper soil layer. In the lower water content range of 18% to 24%, the increase was relatively gradual. When the water content reached 26%, however, the migration height showed a marked abrupt increase. As the water content increased to 30%, pore water pressure drove the fine particles within the subgrade to penetrate the ballast layer structure and surge to the surface, manifesting as pumping mud.

4.2. Excess Pore Water Pressure Response and Vertical Hydraulic Gradient

Figure 12 shows the excess pore water pressure against cyclic loading along the vertical direction of the specimen. The excess pore water pressure induced by cyclic loading exhibits obvious stage characteristics and spatial variation. In the initial loading stage, the excess pore water pressure in each soil layer accumulated rapidly to its peak. The upper-middle layer E reached the highest peak, followed by the top layer G, while the lower layers A and C showed the lowest values. This difference mainly stems from the distance between the soil layer and the interface as well as the initial water content. Although the pressure in layer G, adjacent to the interface, increased rapidly initially, its quick dissipation limited peak accumulation. In contrast, layer E, located farther from the interface, experienced marked accumulation because dissipation was hindered after the rapid rise, leading to the highest peak. The lower layers A and C, with lower initial water content and greater distance from the interface, exhibited a weaker pore water pressure response and thus the lowest peaks. As loading continued, the dissipation behavior of excess pore water pressure clearly depended on layer position. In layer G, near the interface, the pressure dissipated rapidly and stabilized at a low level. In layer E, however, dissipation after the peak was noticeably slower due to a longer drainage path. For the lower layers A and C, except for the 34% high water content condition where penetration of the ballast layer disrupted the upper structure and caused gradual dissipation in layer C, the excess pore water pressure largely stabilized after reaching its peak under other water content conditions. This stabilization was influenced by both the low water content and distance from the interface. By the end of loading, except under the 34% high water content condition, the residual pore water pressure at the top layer G was significantly lower than in layers E, A, and C. This demonstrates the noticeable impact of efficient drainage near the interface on the residual pore water pressure.

Water redistribution under cyclic loading is closely linked to the evolution of pore water pressure. Temporal changes in pore water pressure provide an effective indicator of whether water is being stored locally or transferred within the soil. During loading, a spatial pressure difference may develop, and the resulting pore water pressure gradient can generate a suction-like driving effect that promotes seepage between measurement points. To better interpret the migration mechanism, both pore water pressure and its vertical gradient are therefore evaluated. The pore water pressure gradient between two locations is computed as [31]:

$$i={\displaystyle \frac{u_{w1}-u_{w2}}{\Delta h}},$$

(1)

where i is the pore water pressure gradient (kPa/m); u_w1 and u_w2 are the pore water pressures at two locations (kPa); Δh is the distance between the two locations (m). With this sign convention, i > 0 indicates the pore water flows from u_w1 location to u_w2 location, i < 0 indicates the opposite direction, and i = 0 suggests no discernible pore water migration between the two points.

Figure 13 shows the pore water pressure gradient calculated using Equation (1) during cyclic loading along the vertical direction specimen. For all moisture contents, the pore pressure gradient between E and G increases sharply at the early stage of cyclic loading and then gradually decreases, while remaining relatively high. Throughout the test, the gradients between E and G are consistently positive, indicating sustained upward water migration under all test conditions; this effect is most pronounced at the initial loading stage. In contrast, the gradients between C and E initially become negative and decrease rapidly, and then gradually increase toward zero. Accordingly, water migrates downward at the beginning of loading, but this downward migration progressively diminishes. Over the entire loading process, the gradients between A and C remain close to zero, suggesting no significant water migration in the middle-to-lower soil layer. Notably, for the specimen with a high moisture content of 34%, a clear overall water migration develops in the later loading stage.

4.3. Water Content Evolution in the Fine-Grained Layer During Loading

Figure 14 illustrates the variation of volumetric water content along the vertical direction of specimens with the number of cyclic loading cycles under different initial water content conditions. For the specimen with an initial water content of 18%, the water content in each soil layer increased slightly during early loading due to compaction, then stabilized. Overall, no significant water migration was observed. In contrast, for specimens with initial water contents of 22%, 26%, 30%, and 34%, the water content in the lower soil layers remained essentially unchanged throughout the loading. However, the water content in the upper fine-grained soil layers E and G decreased rapidly during initial loading before gradually stabilizing. Notably, a higher initial water content in these upper layers corresponded to a faster rate of decrease during early loading and a greater total water loss upon stabilization. Specifically, the stabilized water content in layer G of the 22% specimen decreased by 2.24% from its initial value. As the initial water content increased to 26%, 30%, and 34%, this reduction increased to 2.85%, 3.69%, and 6.91%, respectively. This water migration pattern agrees with findings reported by Han et al. [25]. The mechanism involves excess pore water pressure generated under cyclic loading, which dissipates primarily through the more permeable interface between fine-grained soil and ballast. This dissipation creates an upward hydraulic gradient that drives pore water upward into the overlying ballast layer, ultimately reducing the water content in the upper fine-grained soil.

4.4. Quantification of Interlayer Growth and Ballast Fouling

4.4.1. Interlayer Thickness Versus Initial Water Content

To analyze the evolution of the fouled ballast interlayer, post-loading specimens were sampled vertically in layers to measure the interlayer thickness. The relationship between the thickness and initial water content is plotted in Figure 15. The results indicate that as the initial water content increased, the migration height of fine particles, the penetration depth of ballast, and the thickness of the fouled interlayer all exhibited an increasing trend. For clarity, the thickness evolution can be categorized into three moisture regimes: (i) w₀ < w_L (18–24%), (ii) w_L < w₀ < 32% (26–30%), and (iii) w₀ ≥ 32%.

When the initial water content ranged from 18% to 24%, the fine-grained soil had relatively low water and higher stiffness, providing significant resistance to ballast penetration. Although the increased water content caused slight softening, its effect on the overall soil strength was limited. Therefore, the increases in fine particle migration height and fouled interlayer thickness were gradual. Once the initial water content reached 26%, exceeding the liquid limit, the soil structure began to break down and shear strength dropped sharply. This greatly reduced ballast penetration resistance and intensified penetration. Simultaneously, driven by the pore water pressure gradient, the upward migration of fine particles became more pronounced, causing significant increases in both interlayer thickness and migration height. Under these conditions (w_L < w₀ < 32%), the fine particle migration height remained below the top of the ballast layer; therefore, the fouled interlayer thickness was defined as the vertical difference between the upward migration height and the ballast penetration depth. When the initial water content further increased to 32%, the overlying fine-grained soil approached a fluid state due to excessive water and essentially lost its structure. Under cyclic loading, ballast particles penetrated extensively, severely disrupting the original soil–ballast interface. Meanwhile, fine particles were forced upward by rapid dissipation of excess pore water pressure and the squeezing action of penetrating ballast. The migration height eventually surpassed the top of the ballast layer, fouling the entire ballast layer. Accordingly, for w₀ ≥ 32%, the fouled interlayer thickness was taken as the full ballast-layer thickness, i.e., the vertical distance from the top to the bottom of the ballast layer. In summary, the apparent thickening of the fouled ballast interlayer transitions from a gradual growth below the liquid limit, to a rapid growth once w₀ exceeds w_L (due to strength attenuation and enhanced particle transport), and finally to full-depth ballast fouling at w₀ ≥ 32%. The formation mechanism can be described as a synergistic process of ballast penetration, upward migration of fine particles, and interlayer evolution.

4.4.2. Ballast Fouling Quantified by Void Contaminant Index (VCI)

The fine particle content within the fouled ballast interlayer significantly influences track geometry and the long-term dynamic stability of railway substructures. To assess the degree of ballast fouling, the Void Contaminant Index (VCI), proposed by Tennakoon et al. [2] and Indraratna et al. [15], is used. VCI is defined as the volume ratio of fouling (silty clay) to the voids within clean ballast (graded crushed stone).

$$\mathrm{VCI}={\displaystyle \frac{V_{\mathrm{f}}}{V_{\mathrm{vb}}}},$$

(2)

where $V_{\mathrm{f}}$ is the volume of fouling and $V_{\mathrm{vb}}$ is the volume of voids within the clean ballast.

By introducing relevant physical parameters, the VCI can be further expressed as:

$$\mathrm{VCI}={\displaystyle \frac{1+e_{\mathrm{f}}}{e_{\mathrm{b}}}}\times {\displaystyle \frac{G_{\mathrm{sb}}}{G_{\mathrm{sf}}}}\times {\displaystyle \frac{M_{\mathrm{f}}}{M_{\mathrm{b}}}}\times 100,$$

(3)

where $e_{\mathrm{b}}$, $G_{\mathrm{sb}}$, and $M_b$ represent the void ratio, specific gravity, and mass of the clean ballast, respectively, while $e_{\mathrm{f}}$, $G_{\mathrm{sf}}$, and $M_f$ denote the void ratio, specific gravity, and mass of the fouling (silty clay), respectively.

Granite crushed stone with a specific gravity of 2.69 was selected as the ballast material, consistent with values reported by Indraratna et al. [15]. Referring to the void ratio of compacted ballast documented by Duong et al. [12,13], the void ratio of the compacted ballast in this test was determined as 0.62. Substituting the parameters of the fine-grained soil and ballast into Equation (3) yielded the VCI of the fouled ballast interlayer at the end of the test. Figure 16 illustrates the variation of VCI with the initial water content of the overlying fine-grained soil. Between 18% and 22% water content, VCI increased gradually. However, at 24% water, VCI increased sharply. This phenomenon can be attributed to the disintegration of soil structure and marked reduction in shear strength as the fouling water content approached the liquid limit, which intensified fine particle migration. Notably, at 34% water, VCI increased drastically. Corresponding observations in Figure 11h,i, above, revealed clear mud pumping under cyclic loading, with ballast voids almost completely filled by fine particles. According to Indraratna et al. [32], track maintenance is required when VCI exceeds 40%. In this test, when the initial water content of the overlying soil layer reached 28%, the final VCI of the ballast layer under cyclic loading exceeded 40%, indicating severe fouling and the need for remediation. In practice, replacing heavily fouled ballast is recommended to ensure normal railway operation. Additionally, appropriately reducing train speed and freight volume during spring thaw can help mitigate mud pumping.

4.4.3. Changes in Ballast Gradation with Depth After Loading

Following testing, ballast was collected from three distinct layers labeled I, II and III from bottom to top. Layer I had a thickness of 40 mm, and Layers II and III were each 30 mm thick. Specimens from each layer were dried, weighed, washed to remove adhered mud, dried again, and sieved to determine particle size distribution. Figure 17 presents the post-test particle size distributions for all nine test groups along with the initial gradation for comparison. In Figure 17a–d, corresponding to initial water contents of 18–24% in the overlying fine-grained soil, the gradation curve of Layer I shifted upward and rightward relative to the initial curve, with the shift increasing with water content. This indicates the intrusion of fine particles into the voids of the ballast skeleton in Layer I, forming a fouled interlayer where fouling severity rose with water. In contrast, the curves of Layers II and III largely overlapped with the original, suggesting negligible upward migration of fines. In Figure 17e–h, representing water contents from 26% to 32%, the shift in the curve of Layer I stabilized, implying that its voids were nearly filled by fines. Further migration then predominantly affected Layer II, where the curve shifted markedly upward and rightward as water increased, reflecting rapid accumulation of fines. Layer III exhibited only minor curve shifts, indicating limited fine particle ingress. At 34% water, shown in Figure 17i, the curve of Layer I shifted downward compared to the 32% case, while curves for Layers II and III shifted upward, most notably in Layer III. This resulted from downward squeezing under cyclic loading and pore pressure driven migration, leading to substantial fine particle penetration through the ballast, a phenomenon referred to as pumping mud. Consequently, fines content increased sharply in Layer III, with partial migration from Layer I to the upper layers. In summary, high water concentrations promote significant upward migration of fine particles, progressively destabilizing the ballast layer and degrading both its gradation and mechanical integrity.

5. Discussion: Deterioration Mechanism and Engineering Implications

Based on the above experimental phenomena, this section divides the deterioration mechanisms of fouled railway ballast in seasonal frozen regions into four stages, as shown in Figure 18.

(1)

Stage 1: Water migration during freezing

In winter, unfrozen water in the lower part of the subgrade bed surface moves toward the upper freezing front due to the temperature gradient, forming ice lenses. This increases the ice content in the frozen zone, especially near the freezing front. At the same time, frost heave pushes fine-grained soil at the ballast-subgrade interface into the voids of the overlying ballast skeleton.

(2)

Stage 2: The development of a high-water-content clay layer during the thawing

In spring, the top soil of the subgrade surface thaws first, while the underlying layers remain frozen. This forms a temporary aquitard that traps meltwater in the upper part of the surface layer. After repeated freeze–thaw cycles, a high-water-content clay layer forms at the top of the subgrade surface.

(3)

Stage 3: Mud formation and intrusion

Under train loads and water softening, the fine-grained soil in the high-water-content clay layer mixes with water, forming a fluid mud. Driven by pore water pressure gradients induced by train loading, this mud is pumped into the ballast layer, where it settles and accumulates, filling voids between ballast particles.

(4)

Stage 4: The evolution of the fouled ballast interlayer

As temperatures rise in summer, water evaporates or drains from the ballast. The intruded mud within the ballast layer gradually settles, and consolidates. Ultimately, a dense, fine particle dominated interlayer forms within the ballast, particularly near the ballast-subgrade interface. This layer disrupts the original clean, porous, and freely draining structure of the ballast, leading to hardening, loss of elasticity, and impaired drainage. It becomes a root cause of secondary failures such as mud pumping.

Engineering implications: The results enable practical guidance for seasonal frozen regions. The deterioration is primarily moisture-controlled, with clear threshold behavior. Interlayer development accelerates when the upper fine-grained layer approaches or exceeds the liquid limit of about 24%, and mud pumping was observed at high water contents of about 30%. Routine monitoring should therefore focus on the moisture state of the fine-grained layer near the ballast–subgrade interface and the degree of ballast fouling. The void contamination index provides a useful maintenance metric, and in this study, it exceeded 40% when the upper-layer water content reached 28%, indicating the need for remediation. Excess pore water pressure and its vertical gradient can also be tracked as process indicators of upward seepage and pumping potential. Inspections and preventive maintenance should be prioritized before and during spring thaw, when meltwater retention and thaw weakening are most pronounced. Mitigation should emphasize drainage improvement, ballast cleaning or replacement where fouling is severe, and interface separation or reinforcement, such as geotextiles or geogrids, to limit ballast–subgrade mixing. Temporary operational measures during spring thaw may further reduce pumping risk when needed.

6. Conclusions

One dimensional freeze–thaw cycle tests and cyclic loading model tests were conducted in sequence to investigate the development of high-water-content clay layer and the evolution of fouled ballast interlayer in seasonal frozen regions. The main conclusions are as follows:

(1)

The temperature variation along the height of the fine-grained soil specimen showed pronounced nonlinear characteristics during freeze–thaw cycles. The shallow soil cooling rapidly than deeper soil. Under open water supply during freezing, the temperature gradient drove unfrozen water migration toward the freezing front, forming ice lenses. During thawing, meltwater was hindered by the low permeability of the underlying frozen layer and accumulated in the upper soil. As freeze–thaw cycles increased, the water content in the top layer increased significantly, eventually forming a clay layer with high water content.

(2)

Pore water pressure in the upper layer increased and dissipated rapidly, accompanied by a persistently positive pore pressure gradient. In the mid-upper layer, both the peak value and dissipation rate were strongly controlled by the initial water content. Meanwhile, the gradient evolves from negative toward zero as loading proceeded. In the lower layer, pore pressure accumulated gradually because of limited moisture and constrained drainage, and the gradient remained close to zero. Dissipation of excess pore pressure near the interface promoted upward moisture migration. Overall, the amount of moisture migration increased with initial water content.

(3)

Ballast penetration combines with the upward migration of fine particles to form a fouling ballast interlayer under cyclic loading. As the initial moisture content of the overlying fine-grained soil increases, the interlayer thickens gradually. Once the initial moisture content exceeds the liquid limit, the interlayer thickness increases markedly. At moisture contents above 30%, fine particles migrate further upward to the top of ballast, resulting in mud pumping. Overall, the evolution of the fouling interlayer follows a synergistic mechanism governed by ballast penetration, upward moisture movement, and fine particle accumulation.

(4)

The particle size distribution of the fouled ballast interlayer deteriorated considerably as initial water content increase. The fouling degree was quantified using the void contamination index, which increases with the initial water content of overlying fine-grained soil. When water content exceeded 28%, the index of the specimen surpassed 40%, reaching the engineering threshold requiring remedial measures. This index can provide a quantitative basis for identifying fouled ballast interlayers and guiding replacement decisions in practice.

(5)

The evolution mechanism of fouling in ballasted tracks within seasonal frozen regions can be summarized as follows. During winter freezing, water migrates upward and forms ice lenses near the freezing front. In spring thaw, meltwater is trapped above the frozen layer below. After multiple freeze–thaw cycles, a high-water-content clay layer forms in the upper soil, where fine-grained materials soften into mud. Under cyclic train loading, excess pore water pressure drives the mud upward. As temperatures rise in summer, evaporation and drainage allow the mud to settle and consolidate, eventually forming a dense fouled ballast interlayer near the ballast–subgrade interface.

(6)

The results suggest practical field indicators and triggers for seasonal frozen regions. Interlayer evolution accelerates when the upper fine-grained soil exceeds the liquid limit (24%), and mud pumping was observed under high water contents (≈30%). The ballast fouling degree can be evaluated using VCI. A VCI > 40% occurred at an upper-layer water content of 28%, indicating the need for remediation. Maintenance should be prioritized around spring thaw, and mitigation should emphasize drainage improvement, ballast cleaning or replacement, and interface separation or reinforcement to reduce fines intrusion and pumping risk.