Frost Heaving and Induced Pressure of Unsaturated Interfacial Zone between Gravel Ballast and Subgrade

Abstract

Most existing railroads are composed of gravel ballast. One of the major issues with gravel ballast is frost damage in cold regions. Gravelly soils are known to be not prone to frost action due to their low water retention capacity and high hydraulic conductivity. However, reports indicated continued frost damages resulting from the mixed zone between gravel ballast and subgrade. This study evaluated the frost heaving and induced pressure of gravel ballast–subgrade soil mixtures via 1D soil column testing in a cold chamber. Gravel ballast and subgrade soil were collected from the railroad in situ. Various mixing ratios and degrees of saturation were used as factors affecting the frost experiments. The mixtures were placed in the cold chamber, and vertical displacements and pressures were measured. Overall evaluations showed that gravelly soils are not a geomaterial prone to frost damage; however, the frost potential of gravel ballast increases as the degree of saturation and the mixing portion of the subgrade soil increase. Therefore, the interfacial zone between gravel ballast and subgrade soil, especially where possible mixing with low drainage exists, needs cautions of potential frost damage.

1. Introduction

A ballast in railroad tracks is a component that fixes ties (or sleepers) in a specific position and distributes the loads transmitted from the ties to the subgrade. Ballasts are classified into two types: gravel ballast and concrete ballast (or slab ballast). Minor types such as sand, Moorum, slag, etc. are not discussed in this study [1]. The typical design of gravel and concrete ballasts and their pros and cons of each are summarized in

Table 1. Characteristics of gravel and concrete ballasts.

Type	Gravel	Concrete
Typical design
Proportion	High	Low
Construction cost	Low	High
Track irregularity	High	Low
Noise during train	Low	High
Train vibration	High	Low
Maintenance	Frequent	Less
Repair	Easy	Difficult
Ventilation in tunnel	Need	Not need
Mud pumping	Exist	Not exist

. A gravel ballast is a ballast constructed by pouring gravels, while concrete ballast is made of prefabricated concrete blocks and cement mortar. At the initial era of railroad construction, gravel ballast was the major format in railroad construction due to its easy and economical constructability [2]. The stability of the gravel-ballasted track is secured by friction between the gravels, and the impact and vibration are absorbed through the elasticity of the ballast. However, concrete ballasts became popular in the world due to the frequent maintenance of gravel ballasts due to changes in track geometry, material deterioration, etc. [3,4]. Concrete ballast is a more advanced form than a gravel ballast due to its better engineering properties, maintenance, etc. despite the expensive initial construction cost [5]. Advanced railroad countries (France, Germany, Japan, China, etc.) are using concrete ballasts for new or expanded high-speed railroads. In South Korea, concrete ballasts have been used since 2005.

Although recently constructed concrete ballasts are superior, the problem is that most existing railroads are still gravel ballasts. One of the problems with gravel ballasts is the frost heaving in cold regions. In South Korea, Lee et al. [6] reported that the number of sites showing frost heaving was 129 with a −4 km distance in 2009, while Han et al. [7] stated 190 with a −8 km distance in 2011. To countermeasure the frost heaving on gravel ballasts, several approaches are attempted, such as the replacement of gravels, using calcium chloride, improving drainage, etc. Those maintenance costs are enormous. The frost potential of gravels is known to be low due to the high drainage capacity. However, it depends on the situation. For example, the subgrade, which is susceptible to freezing and thawing, can be mixed with gravel ballast during train operation, and then the mixture becomes vulnerable to frost damage. Most sites with frost heaving found mud pumping in summer. Mud pumping is the upward movement of subgrade soils to the gravel ballast surface. Subgrade soils during the rainy season are liquefied and easily move to the surface through gravel ballast. The subgrade soil–gravel ballast mixture brings a decrease in the void ratio, and correspondingly hydraulic conductivity, and eventually poor drainage, which turns out that the mixture is susceptible to frost heaving [8,9]. Therefore, the interfacial zone between the gravel ballast and the subgrade is the place where the frost action develops. The interfacial zone is generally unsaturated, as the gravels have high permeability [10]. There are some studies that observe the action of frost on gravel soils with fines [11,12,13,14]; however, few studies were conducted to evaluate the action of frost on gravel ballast–subgrade soil mixtures under unsaturated conditions.

Frost occurs when the following conditions are coincident: temperature is below 0 °C; there is the presence of water from groundwater or undrained water; and the soils can retain pore water [15]. Frost action has been studied in terms of frost heaving and frost heaving-induced pressure (frost heaving pressure) [16]. Frost heaving is highly dependent on the soil size. Soils more than 0.1 mm (e.g., more than sand) are unlikely to experience frost heaving due to high permeability. Soils with grain sizes of 0.05–0.1 mm (e.g., fine sands) start heaving, and those with grain sizes of 0.002–0.05 mm (e.g., silt) present the most heaving. Soils less than 0.002 mm (clayey soils) are too impermeable to form ice lenses, which is the fundamental cause of frost heaving [17,18,19]. Frost heaving pressure occurs under a certain confinement during the volumetric expansion of water (−9% of the initial volume of water). The frost heaving pressure and confinement are proportional; therefore, the higher the confinement, the higher the frost heaving pressure, and vice versa. Frost heaving and frost heaving pressure are inversely related. Therefore, the frost action of soils involves internal factors such as grain size distribution, soil structure, specific surface, and chemical properties of pore fluid as well as external factors such as confinement, source of water, and temperature.

This study focuses on the frost action of unsaturated gravel ballast mixed with subgrade soil. Gravel ballast and subgrade soil were collected in situ from a railroad. Gravel ballast–subgrade soil mixtures with various mixing ratios were prepared. A cold chamber was used to simulate the freezing conditions. Frost heaving was measured without confinement, whereas the frost heaving pressure was evaluated under confined conditions. Based on the results, a critical mixing ratio showing the threshold frost action is discussed. By comparing the results with data from the literature, the level of frost action on gravel ballast–subgrade soil mixtures is evaluated.

2. Materials and Methods

2.1. Gravel Ballast and Subgrade Soil

Gravel ballast and subgrade soil were collected in situ from a railroad site in Cheolwon, South Korea. Because the collected gravel ballast had a biased grain distribution during sampling, the distribution of the gravel ballast in situ was modified following KRCS A015 07 [20]. As the specification provides ranges, averaged values were chosen. The collected samples were sieved at 125 mm, 63.5 mm, 25.4 mm, 9.52 mm, 2.5 mm, and 0.15 mm [21] (Figure 1), and the gravel ballast was reconstituted in the following ratios: 63.5–125 mm (30%), 25.4–63.5 mm (23%), 9.52–25.4 mm (3%), 2.5–9.52 mm (14%), and 0.15–2.5 mm (20%). Subgrade soil was used with the original composition. Sieve analysis was conducted for the subgrade soil. The grain-size distributions of the gravel ballast and subgrade soil are shown in Figure 2. Based on a united soil classification system (USCS) [22], the ballast soils consist of −40% cobble (for soils retained on a 76.2 mm sieve) and −60% GP (for soils passing a 76.2 mm sieve, poorly graded gravel). The subgrade soil was classified as poorly graded sand (SP). The specific gravities of the ballast soil and subgrade soil were 2.67 and 2.70, respectively [23]. The subgrade soil had a maximum dry unit weight of 19.8 kN/m³ and an optimum water content of 10.3% [24].

The actual mixed zone between the gravel ballast and subgrade soil was not easy to sample, including various uncertainties. Therefore, artificial mixture was used to simulate soils mixed at the interfacial zone between the ballast and subgrade based on the porosity of gravel ballast. This study focuses on the frost action of gravel ballast as a major material mixed with subgrade as a sub-material; therefore, the subgrade soil was added to the gravel ballast soil with a mixing ratio M (e.g., mass_subgrade/mass_total × 100%) of 0%, 5%, 10%, 15%, 20%, and 25%. M = 25% indicates a mixture consisting of 25% subgrade soil and 75% gravel ballast based on mass percentage. The dry unit weight (γ_d) and porosity (n) of the ballast–subgrade mixtures were measured using a mold with a diameter of 28 cm and height of 23 cm. Once the dry gravel ballast and subgrade soil were mixed, the mixture was placed into the mold in five layers with vibrations. Then, the steady-state dry unit weight of the mixtures was evaluated. Rather than compaction, only vibrating energy was applied to simulate the mixing due to the dynamic energy during train operation. The measured γ_d and corresponding n values are shown in Figure 3. The values of γ_d and n show an inverse relationship. The trends in Figure 3 showed a complex relationship. When M increases, the compaction efficiency of the mixture increases as the subgrade soils are finer than the gravel ballast. However, the mass of the gravel ballast decreases as M increases. The coefficients of uniformity and curvature rapidly change with respect to Figure 2. Therefore, γ_d–M seems to exhibit an extraordinary relationship.

2.2. Frost Experiments

A mold measuring 28 cm in diameter and 23 cm in height was used for the frost experiments. The mixtures were prepared and placed into a mold following the sample preparation method used to measure the dry unit weight of the mixtures. Thermocouples were inserted at specimen heights of 30% and 70% to monitor the temperature change in the mixture (Figure 4). As the mixtures are gravelly soils, most of the added water (15 °C) fills the void of the mixture from the bottom of the experimental mold rather than being uniformly distributed throughout the entire specimen. A scour pad was placed on the surface of the mixture. A specific volume of water was prepared based on the target degree of saturation (S). Half of the prepared water was sprayed over the mixture surface at an infiltration rate of 0.001 cm/s to simulate uniformly wet conditions of the mixture. The remaining water was percolated along the inner wall of the mold to prevent the mixtures from forming excessive preferential hydraulic paths and washing.

The preliminary tests showed no frost heaving when S was less than 55%. This is because when the volume of the pore fluid is low, water merely freezes and migrates to unsaturated pores, causing no increase in volume [25]. In addition, during freezing, water theoretically expands by −9% compared to its initial volume. Therefore, unsaturated conditions of S = 70%, 80%, and 90% were selected in this study to examine the frost action of gravel ballast–subgrade soil mixtures. The saturated condition (i.e., S = 100%) was omitted because gravel ballast is rarely saturated.

After preparing the specimens, they were transferred to a cold chamber and kept at −25 °C for 65 h under undrained conditions. For the frost heaving test, a plate was placed on the surface of the mixture, and a displacement transducer (PY2 LVDT, Gefran, Provaglio d’Iseo, Italy) was used to measure the upward displacement at the middle of the plate, assuming 1D heaving (Figure 4a). For the frost heaving pressure test, the mixture was placed in the same manner as the frost heaving setup; however, the plate on the mixture surface was confined by a load cell connected to a rigid reference frame (Figure 4b). The change in pressure was monitored during the freezing phase. The assumptions on the confinement were intended to measure the maximum frost heaving and maximum frost heaving pressure.

The experimental details are summarized in

Table 2. Details of frost experiments conducted in this study.

Test Type	M [%]	S [%]	Test Name
Frost heaving	0	70/80/90	h-M0-S70/80/90
	5		h-M5-S70/80/90
	10		h-M10-S70/80/90
	15		h-M15-S70/80/90
	20		h-M20-S70/80/90
	25		h-M25-S70/80/90
Frost heaving pressure	0	70/90	p-M0-S70/90
	5		p-M5-S70/90
	15		p-M15-S70/90
	25		p-M25-S70/90

. For instance, ‘h-M10-S70’ indicates a frost heaving test for a mixture with a mixing ratio of 10% (10% w/w subgrade soil and 90% w/w gravel ballast) and a degree of saturation of 70%. For the frost heaving tests, M = 0%, 5%, 10%, 15%, 20%, and 25% with S = 70%, 80%, and 90% were used. For the frost heaving pressure tests, reduced cases such as M = 0%, 5%, 15%, and 25% with S = 70% and 90% were used because of limitations with respect to the laboratory conditions.

3. Results

3.1. Frost Heaving

3.1.1. Heaving Behavior

Figure 5 shows the frost heaving for gravel ballast–subgrade soil mixtures with varying mixing ratios and degrees of saturation. Despite the presence of pure gravelly soils, h-M0-S70/80/90 exhibited certain levels of frost heaving (h) in the presence of pore water (Figure 5a). When S = 70%, −500 min was required to reach h at steady state, whereas −100 min and −150 min were required for S = 80% and 90%, respectively. The ultimate h (h_ult, h selected at 20 h) of h-M0-S70 and S80 were similar (−0.3 mm), but that of S90 showed a higher value (−0.75 mm). The behavior of h over time for h-M0-S70 was different from that for h-M0-S80/90. h-M0-S70 shrank in the initial phase and gradually expanded to h_ult. However, h showed a rapid increase in the case of h-M0-S80/90. Water contracts at 4 °C and expands below 4 °C [26]. Therefore, it is natural to observe that h decreases in the initial phase and then increases. However, it seems that the initial freezing mechanism of the gravel ballast differs from that of S. When S is lower, frost shrinkage occurs, followed by frost heaving, whereas frost heaving occurs immediately when S is higher. This implies that the formation of ice lenses in the pore network occurs easily and rapidly when S is high, resulting in prompt frost heaving.

When M = 5%, the frost behavior of the mixtures was similar to that when M = 0% (Figure 5b). h-M5-S70 exhibited frost shrinkage in the initial phase; however, later on, it frost-heaved. The amount of frost shrinkage (approximately −0.5 mm) was higher than the ultimate frost heaving (approximately 0 mm). The amount of h_ult were also analogous to that for M = 0%.

The frost behavior for M = 10% was slightly different from those for M = 0% and 5% (Figure 5c). All cases exhibited frost shrinkage during the initial phase, regardless of S. Frost shrinkage was maximized at −200 min; thereafter, it gradually increased. H_ult was observed at −400–600 min. In this case, 10% of the subgrade soils were mixed with 90% of the gravel ballast. It is inferred that the migration and release of pore fluid in subgrade soil induces frost shrinkage during initial freezing [19]. Eventually, the mixtures were heaved owing to the expansion of ice lenses from the agglomerated pore fluid. Similar behaviors were observed when M = 15% and 25% (Figure 5d,e, respectively). At M = 25%, the mixtures exhibited the highest h_ult (Figure 5f). Interestingly, the frost shrinkage of h-M25-S70 was the lowest in the M range of 0–25%. The effect of the migrating pore fluid was maximized for the h-M25-S70 specimen.

3.1.2. Summary of Frost Heaving Tests

The frost heaving tests showed various observations in terms of time and heaving amount with respect to the mixing ratios. The time to reach an h_ult was defined as t_ult. The heaving ratio (η) is defined as h_ult over the initial specimen height (h₀) in a percentage, and the total heaving ratio (Δη) is defined as the absolute heaving amount (h_ult – h_min) over h₀ in a percentage, as shown in Equations (1) and (2), respectively:

$$\eta =\frac{h_{\mathrm{ult}}}{h_0}\times 100 [\%]$$

(1)

$$\Delta \eta =\frac{h_{\mathrm{ult}}-h_{\min }}{h_0}\times 100 [\%]$$

(2)

where h_min indicates the maximum frost shrinkage. If there is no frost shrinkage (h_min = 0), Δη is the same as η.

The results are shown in Figure 6. When the mixtures had a low M and high S (e.g., h-M0-S90 and h-M5-S90), h_ult was observed at a relatively earlier freezing duration (e.g., −2 h) (Figure 6a). Except for these the two cases, t_ult was generally measured within the range of 8–10 h. These observations suggest that at a high S, the pure gravel ballast exhibits quicker frost heaving. As S is lowered, the stabilization of frost heaving becomes slower regardless of M. Therefore, it is concluded that a proper drainage system at the gravel ballast–subgrade interface can prevent the occurrence of frost heaving.

The symbol η represents the general trends of frost heaving on gravel ballast–subgrade soil mixtures (Figure 6b). Except for a few cases, the general trends of η exhibited an increase as S and M increased. When S = 90%, η was approximately 300–900% and 140–230% higher than those for S = 70% and 80%, respectively. The trends of Δη were generally similar to those of η; however, the level of Δη for S = 70% increased because of frost shrinkage (Figure 6c). Therefore, it is concluded that the frost heaving for S = 70% is lower than that for S = 80% and 90%; however, the total fluctuation may have to be considered during freezing.

3.2. Frost Heaving-Induced Pressure

3.2.1. Heaving Pressure Behavior

The variation of the frost heaving-induced pressure (p) with times for a different S and M is shown in Figure 7. Under conditions of confinement, the frost heaving was less than 0.1 mm; therefore, the upward movement can be neglected. When M = 0% (i.e., pure gravel ballast), no pressure was induced under confinement. As shown in Figure 5a, frost heaving was observed under free confinement. Figure 7a shows that the expansion of water is dissipated within the pore space for the pure gravel ballast. Therefore, the frost heaving pressure can be neglected under a sufficient overburden pressure for pure gravel ballast. However, the pressure increased as M increased. At M = 5%, p increases from t − 300 min (Figure 7b). As M increased, the trends in p were similar each other, but the level of p increased (Figure 7c,d). A small negative pressure was induced in the initial phase owing to frost shrinkage, but it is negligible. Considered against the irregular increase in h with time, as shown in Figure 5, the increase in p showed almost linear. This would indicate that, under confinement, the formation of ice lenses occurred while migrating the pore spaces of the mixtures, resulting in a moderate variation in the frost heaving pressure.

The minimum value in the p (−2.5 kPa) was observed at t = 28.5 min for p-M25-S70 (Figure 7d). The position of the minimum value corresponds to the position of the largest frost shrinkage for h-M25-S70 (Figure 5f). Frost shrinkage was also affected by the behavior of frost heaving pressure.

3.2.2. Summary of Frost Heaving Pressure Tests

The ultimate frost heaving pressure (maximum p, defined as p_ult) increased as M increased. The p_ult is located in a t range of 500–700 min. The p_ult values are summarized in Figure 8. For S = 70%, p_ult were 0.1, 3.1, 4.5, and 7.6 kPa for M = 0, 5, 15, and 25%, respectively. For S = 90%, a relatively linear increase was observed at p_ult = 0.3, 4.5, 9.4, and 16.7 kPa for M = 0, 5, 15, and 25%, respectively. The p_ult for S = 90% is approximately two times larger than that for S = 70%. Overall, the frost heaving pressure of pure gravel ballast can be neglected under constrained conditions; however, the pressure becomes perceptible as subgrade soil is mixed in the gravel ballast.

4. Discussion

Level of Frost Heaving in Different Soil Types According to Degree of Saturation

The frost heaving and pressure of the gravel ballast–subgrade soil mixtures were evaluated through laboratory tests. However, these results need to be compared with those for other types of soil to clarify the degree of frost action on gravelly soils. Several previous studies have evaluated frost heaving in various soil types. For a comprehensive evaluation, typical soil types were classified as gravel (>4.75 mm), sand (4.75–0.075 mm), fine (<0.075 mm), silt (0.075–0.02 mm), and clay (<0.02 mm) [22]. The type of soil that forms more than 50% of the total weight of a soil mixture was considered as the major soil type, and the other soil types as minor soil types. Accordingly, five soil types were artificially categorized in this study: (a) gravel with sand, (b) gravel with fine, (c) sand with silt, (d) silt with clay, and (e) clay with silt. For example, gravel with sand consisted of more than 50% gravelly soils and less than 50% of sandy soils. Gravel with sand included the gravel ballast–subgrade soil mixtures of this study. Silt with clay mainly consisted of silty soils with minor clayey soils. Data from the literature were not ready to use [13,14,25,27,28,29,30]; therefore, several assumptions were made while summarizing the literature. For instance, when the water content (w) was given, the degree of saturation (S) was calculated using the specific gravity (G_s) and void ratio (e) (e.g., Se = wG_s). When e was not available, e was derived using the dry unit weight (γ_d) (e.g., γ_d = [G_s∙γ_w]/[1 + e], where γ_w is the unit weight of water, 1 g/cm³) [31]. The summarized data are shown in Figure 9 in terms of the frost heaving ratio η. The data on gravel with fine in this study were compared with previous data on this soil mixture [12,13,14]. Some studies performed frost tests on subgrade soils (sand with silt) [25,27,30]. Wan et al. [29] analyzed the frost-heaving behavior of the clayey silt area in northeastern China (silt with clay), while others focused on that of clayey soils, including kaolinite and montmorillonite (clay with silt) [17,28,32,33].

The values of η varied from zero to more than 10% (Figure 9). Gravel with sand exhibited low values of η in the range of less than 1%, whereas gravel with fine sand had a wider range of η (less than 2%). In fact, the η range of gravel with fine was closely located where those of gravel with sand are plotted. However, the frost heaving of gravel with sand was negligible when S was less than 55%, but that of gravel with fine sand was η − 1% when S was less than 40%. This observation indicates that additional fines make gravelly soils more prone to frost heaving than sandy soils. Therefore, gravel with fine can be regarded as having a higher frost potential than gravel with sand. Sand with silt and silt with clay were located in a similar range of η (e.g., approximately 0.5–2.5%). Silty soil is considered the most frost-susceptible soil owing to its moderate water retention capacity and hydraulic conductivity to form ice lenses [15]. The η of sand with silt and silt with clay were located at S = 30–100%. In addition, clay with silt showed an extremely high η compared to other types of soil. Kaolinite and typical clayey soils generally show η > 5% at S > 70%. Studies have indicated that ice lenses are usually formed by fissuring the clay body, causing excessive heaving of the clay above the ice lenses. One value of η by montmorillonite is located at η = 1.3 at S = 38% [17]. Montmorillonite has an exceptionally high water retention capacity, which results in low frost heaving [15]. However, except for such special cases, clay with silt had the highest frost heaving potential among the five groups.

Overall, gravelly soil itself has a low frost potential compared to other soil types; however, it may increases depending on the additives. The mixing of gravelly, sandy, and fine soils can occur at the interfacial zone, especially on subgrade and railroads. Frost action occurs repeatedly during the freezing and thawing processes. Insignificant frost heaving can be signified through multiple repetitions of freezing and thawing actions. Therefore, the complete segregation of two materials (e.g., membrane, geosynthetics, etc.), proper drainage, and frequent maintenance can help prevent frost damage to infrastructure.

5. Conclusions

Frost heaving and pressure tests were performed for gravel–sand mixtures with different mixing ratios to examine the frost action at the interfacial zone between the gravel ballast and subgrade. Frost heaving and induced pressure of the mixtures were measured over time in a cold chamber. Pure gravel ballast is considered to have exceptional frost resistance. However, the results of this study revealed that gravel ballast becomes susceptible to frost when finer soils are mixed in it. The specific findings are as follows:

• For pure gravel ballast, no frost heaving occurs when S is less than 55%. However, even pure gravel ballast experiences frost heaving as S increases. The induced pressure is dissipated within the soil network under fully confined conditions.
• The level of frost heaving increases as S and M increase. Depending on the mixture matrix, frost shrinkage is sometimes observed during the initial freezing phase. However, the mixture gradually frost-heaved. The initial frost shrinkage disappears as S increases. The level of frost heaving pressure under fully confined conditions also increases as S and M increase.
• The observed level of gravelly soils was lower than that of fine soils in this study compared to data reported previously in the literature. However, the frost potential increased as the proportion of frost-susceptible soil increased when there was sufficient water. Repeated freezing cycles can amplify the frost potential of gravelly soils. Therefore, a proper drainage system and prevention of fine intrusion are required to reduce frost damage in gravel ballast, particularly at the interfacial zone.