The Qarhan Salt Lake area is situated in the hinterland of the Qinghai-Tibet Plateau in Qinghai Province. It belongs to the typical plateau continental climate. Due to the high-cold environmental conditions, rich salt-bearing host rocks, and strong weathering, the chlorine saline soil is well-developed in the Qarhan Salt Lake area, with an occurrence thickness of more than 20 m. The tested soil samples studied in this experiment were taken from the construction site of the K119 + 600 section of the Sebei-Qarhan Highway (G215), which is one of the first highways built in the Qarhan Salt Lake area (refer to Figures 1, 2). The maximum particle size of the original soil does not exceed 40.0 mm, and the non-uniformity coefficient (C _u) and curvature coefficient (C _c) are 23.3 and 1.58, respectively.

In order to ensure consistency between the tested soil sample and subgrade filling, a triaxial sample was prepared following the Test Methods of Soils for Highway Engineering (Ministry of Communications of the People’s Republic of China, 2020). A standard cylinder with a diameter of 50 mm and a height of 100 mm was utilized for this purpose. The maximum particle size of the sample was 10 mm. To prevent the influence of oversized particles on the test results, groups of particles with equal mass (particle size range 5–10 mm) were utilized to substitute for oversized particles (larger than 10 mm) in proportion. This not only ensures the skeletal structure of coarse particles in the soil but also maintains the content of other particle groups unchanged, preserving the continuity of the original soil grading. The particle grading curves of the soil samples before and after replacement are illustrated in Figure 3. According to the laboratory tests, the maximum dry density and optimal moisture content of the soil sample are 2.17 g/cm³ and 6.8%, respectively, with a salt content of 3.0%. It belongs to the chloride type of moderately saline soil.

The equipment used in this dynamic triaxial test is the temperature-controlled bidirectional vibration triaxial test system (DYNTTS) produced by GDS company in the United Kingdom. It mainly includes a power drive device, back pressure and confining pressure controller, temperature control system, and data acquisition device, as shown in Figure 4. The main technical indicators and parameters are as follows: the axial load range is 0–60 kN, the confining pressure control range is 0–20 MPa, the loading frequency range is 0–10 Hz, the axial displacement limit is ±100 mm, and the temperature range is N.T.P. (Normal Temperature and Pressure) to −30°C. The temperature is controlled by a refrigeration compression circulator, and the freeze-thaw cycle process of saline soil is achieved by continuously regulating the temperature of the frozen liquid within the pressure chamber through it.

The computed tomography (CT) is a highly accurate non-destructive testing technique. Its principle is to reflect the internal structure of an object by comparing the intensity before and after X-ray incidence, without damaging or changing the tested object. Generally, the higher the material density, the stronger the absorption ability of X-rays, and the greater the brightness displayed on CT images (Zhu et al., 2015; Hu et al., 2018). The CT scanning tests were performed using the YXLON FF85 CT, an innovative and versatile high-resolution CT system, as depicted in Figure 5. The sample scanning accuracy used in this detection test is 39 μm.

This research area features a typical plateau continental climate, with an altitude ranging from 2,680 to 2,800 m. The average temperature from January to April and September to December is below zero, with a historical minimum temperature of −36°C. The temperature from May to August ranges from 5°C to 15°C, with the highest temperature reaching 25°C. The temperature at night in summer ranges from −3°C to −5°C, with a significant difference in temperature between day and night. To simulate the freeze-thaw cycle caused by large temperature differences in the context of global warming and humidifying climate trends in high-altitude salt lake regions, this experiment controlled the minimum temperature at −25°C and the maximum temperature at 25°C, with a temperature reduction and heating rate of 0.1°C per minute. A significant body of literature suggests that for specific types of soil, when the freeze-thaw cycle surpasses 10 times, the trend of soil strength attenuation decreases significantly and stabilizes (Wang et al., 2017; Wang et al., 2020; Fan et al., 2021; Xu et al., 2021). Due to the excessively long freeze-thaw cycle test, conducting each cycle test from 0 to 10 times will span different seasons, leading to significant temperature variations in the indoor test environment and causing substantial errors in the test results. Therefore, to ensure the comprehensiveness of the test plan and the timeliness of the test results, the numbers of freeze-thaw cycles in this test have been set to 0, 1, 3, 5, and 7, based on the initial stage test results and conclusions drawn from numerous references, and the freezing and melting processes each continued for 12 h, simulating one freeze-thaw cycle.

Considering the impact on the natural ecological environment of the salt lake area, the entire route of the Sebei-Qarhan Highway is designed based on the principle of low embankment and gentle slope, with a height range of 1.0–1.5 m. When the volume-weight of the fill is 20 kN/m³, the soil confining pressure ranges from 10 to 30 kPa. Therefore, the confining pressure in the dynamic triaxial test is 20 kPa. Considering the limitations posed by complex geological and construction conditions, the compaction degree of CGSS is controlled at 0.92. As shown in Figure 6, the sinusoidal loading waveform is applied to the unidirectional pulse and continuous stress of the soil within the subgrade under vehicle load. Here, σ _c represents the confining pressure, σ _s denotes the static deviator stress, and σ _d indicates the cyclic stress amplitude. In order to simulate the long-term effect of vehicle load on the subgrade soil, the test is completed when the loading number reaches 10,000 times or the axial strain reaches 12%, with a load frequency of 1.0 Hz and a dynamic stress amplitude of 20–200 kPa, as listed in Table 1.

In order to clearly describe the mechanism of long-term performance degradation of the CGSS subgrade under the coupling effect of climate change and traffic loading, the typical samples of triaxial tests were analyzed by selecting the CT images from both transversal and longitudinal sections, as shown in Figure 7. The changes in micro parameters of the internal soil structure under freeze-thaw cycles and cyclic loads are primarily studied. In consideration of the integrity of soil structure, samples subjected to dynamic stress amplitudes of 30 kPa, 60 kPa, and 90 kPa are being analyzed.

The results indicate that the sample undergoes a significant initial strain after the first cyclic loading (N = 1), and the consistency of this strain is low, which is much smaller than the strain generated by subsequent cyclic loading (N > 1), as shown in Figure 8. Therefore, when analyzing the cumulative strain law of CGSS, the development of strain is divided into two stages based on the loading number: first, the initial strain (εⁱⁿⁱ ) generated under the first load; second, the regular cumulative strain (ε^reg ) generated by subsequent loads. Thus, the cumulative plastic strain can be expressed as the sum of ε i r r and ε r e g , as follows: ε p = ε i r r + ε r e g (1)

Figure 9 shows the variation of initial strain with dynamic stress amplitude for different freeze-thaw cycles (F). Previous studies have shown that freeze-thaw cycles have a significant impact on soil structure (Liu et al., 2016; Lu et al., 2016; Wang et al., 2020; Xu et al., 2021; Yang et al., 2021). From the figure, it can be seen that the initial strain increases non-linearly with dynamic stress. After the first freeze-thaw cycle, the initial strain increases by 1.16–1.33 times. When the number of freeze-thaw cycles is 5 and 7, the difference in initial strain between the two is very small. Using a power function to fit the initial strain under different freeze-thaw cycles, the parameters “a” and “b” exhibit a linear relationship with F (Figure 10), as depicted in fitting Eq. 2. ε i r r = a × σ d b a = 0.8840 + 0.04774 F ;   b = 0.2994 − 0.00326 F (2)

Figure 11 illustrates the variation of cumulative strain with dynamic stress amplitude for different freeze-thaw cycles (F). It can be seen that both dynamic stress amplitude and temperature (freeze-thaw cycle) have a significant impact on the cumulative plastic strain (Zhang et al., 2016). The cumulative strain curve of the failure mode sharply increases within 100 loading cycles, leading to rapid failure. When the freeze-thaw cycle ranges from 0 to 1, the critical dynamic stress for sample failure is between 150 kPa and 200 kPa. When the freeze-thaw cycle ranges from 3 to 7, the critical dynamic stress ranges from 90 kPa to 120 kPa. When the number of freeze-thaw cycles is 5 and 7, the cumulative deformation curves under the same dynamic stress amplitude are very close, which is consistent with the experimental results of the initial strain. Meanwhile, the critical dynamic stress varies greatly under different freeze-thaw cycles. When the freeze-thaw cycle in the initial phase (0–1 times), The critical dynamic stress range for specimen failure is 150–200 kPa. When the freeze-thaw cycle reaches 3–7 times, the critical dynamic stress range for specimen failure is 90–120 kPa.

Figure 12 illustrates the variation of cumulative strain of CGSS with different freeze-thaw cycles under semi-logarithmic coordinates. It can be observed that the stable cumulative strain curve under different freeze-thaw cycles exhibits a three-stage characteristic. The cumulative strain at different stages is fitted using a logarithmic function, as shown in Eq. 3. ε r e g = f F , σ d , N = k 1 σ d l g N , 0 ≤ N ≤ 20 k 3 σ d + k 2 σ d l g N , 1000 ≤ N k 1 = 0.00371 + 0.00266 F − 0.00039 F 2 k 2 = 0.00134 + 0.00126 F − 0.00023 F 2 k 3 = 0.00925 + 0.00273 F − 0.00023 F 2 (3)

Based on the triaxial cyclic test of compacted loess conducted by Wang (2019), it was observed that the cumulative strain under long-term, low-intensity cyclic loading shows an increasing trend when plotted on a semi-logarithmic coordinate system. This observation implies a significant difference in the evolution of strength properties between fine-grained and coarse-grained soils when exposed to long-term cyclic traffic loads. This difference can be attributed to the varying mechanical properties resulting from the distinct soil structures. In saline soil, the freeze-thaw cycle process significantly alters the original internal soil structure. Therefore, the subgrade soils endure not only numerous cyclic loads that are below the shear strength but also the detrimental effects of freeze-thaw cycles, ultimately leading to significant deformations in the subgrade (Zhou et al., 2016; Xu et al., 2020; Xu et al., 2021).

The principles of CT scanning imaging and the CGSS samples after the triaxial test are shown in Figures 13, 14. After passing through the material, variances in X-ray intensity are produced due to differences in attenuation, which are then captured at the receiving terminal. Then, the changes in radiation intensity are converted into CT value variations through detectors, computers, etc., and represented in grayscale values on the image. The attenuation law they proposed satisfies Eq. 4 (Yao et al., 2021). I = I 0 ⁡ exp − u m ρ X u = u m ρ (4)

Where I and I ₀ are the light intensity before and after X-ray irradiation; u _m is the unit mass absorption coefficient of the tested sample, ρ is the density of the tested sample; X is the length of ray penetration; u is the attenuation coefficient of a line passing through a tested sample of 1 cm.

Due to the differences in X-ray absorption among different materials, the CT value represents the material’s ability to absorb X-rays, which can determine the density status of the material. Hounsfield proposed the following definition for the CT value (Hounsfield, 1973): H ρ = u − u m u w × 1000 (5) Where H _ρ is the CT value of the scanned sample; u _w is the absorption coefficient of water (u _w = u _w, H _ρ=0, i.e. u _w can be used as the CT standard value).

Figure 15 displays scanning images of CGSS samples under various sections and test conditions, illustrating the soil structure’s condition under different freeze-thaw cycles and cyclic loads. Figures 15A,B show the CT images without undergoing freeze-thaw cycles (F = 0) and with dynamic stress amplitudes of 30 kPa and 60 kPa. After 10,000 loading cycles, numerous pores and small internal cavities appear on various sections of the CGSS samples, with no cracks, and the initial damage is not readily apparent. When the number of freeze-thaw cycles is 1, and the dynamic stress amplitude remains at 60 kPa, the number of pores in the section increases, and local transverse cracks appear. When the number of freeze-thaw cycles increases to 3 and 5, the cracks in the central area of sample sections develop significantly. Additionally, when the dynamic stress amplitude increases to 90 kPa, the sample exhibits significant lateral deformation, with the central part showing a drum shape. When the number of freeze-thaw cycles reaches 7, the crack width rapidly expands, forming a Y-shaped crack throughout the sample. When the number of freeze-thaw cycles is 5 and 7, the damaged area is very similar, which aligns with the findings of the dynamic triaxial test.

The CT value can reflect the average density of geotechnical materials, and its variation reflects the damage to the internal structure. As the degree of internal defect development increases, the average CT value decreases. The CT values of the central (cross-sectional) and axial (vertical-sectional) surfaces of samples under different working conditions were obtained from the scanning test. From Figure 16, it can be seen that there is a significant difference in CT values among different scanning sections under various working conditions. The CT value without freeze-thaw cycles is around 120 Hounsfield units (Hu), and the CT value is highest when the dynamic stress amplitude is 60 kPa.

This indicates that without freeze-thaw cycles, the soil continues to be compacted under dynamic loads, and the compaction deformation between soil particles is much greater than the lateral deformation. When the number of freeze-thaw cycles is 1, the CT value rapidly decreases. With a dynamic stress amplitude of 30 kPa, the CT value decreases by 18.55%, and with a dynamic stress amplitude of 90 kPa, the CT value decreases by 29.27%. It can be observed that the CT value of the cross-section is typically sensitive to freeze-thaw cycles. This sensitivity suggests that the freeze-thaw cycle process can notably decrease the soil’s compactness, resulting in deformation and failure under dynamic loads. Under the same number of freeze-thaw cycles, when the amplitude of dynamic stress increases from 60 kPa to 90 kPa, the CT value of the cross-section decreases by 16.35%–21.84%, and the CT value of the vertical section decreases by 17.86%–26.47%. Under the same dynamic stress amplitude, as the dynamic stress amplitude increases, the CT value on the scanning section continues to decrease. The degree of decrease in the cross-section is greater than that in the vertical section. When the number of freeze-thaw cycles is 5 and 7, the CT values of the transverse and vertical sections are relatively close, which is consistent with the results of the cyclic dynamic triaxial test. This indicates that during the dynamic triaxial test, the CT value on the vertical section of the sample is more sensitive to the increase in dynamic load, while the CT value on the cross-section is more sensitive to the early freeze-thaw effect.

To quantify the coupling effect of freeze-thaw cycles and dynamic load on CGSS performance, the CT damage variable D can be used for analysis, as shown in Eq. 6 (Hu et al., 2018). D = 1 m 0 2 ρ 0 − ρ ρ 0 = 1 m 0 2 ∆ ρ ρ 0 (6) Where m ₀ is the spatial resolution; Δρ is the density change value of the scanning layer; ρ ₀ is the initial density value of the material.

According to Table 2, under the same freeze-thaw cycle, the dynamic stress amplitude increases from 60 kPa to 90 kPa, and the damage variable value D increases by 9.3%–27.3%. When the dynamic stress amplitude is 60 kPa, the number of freeze-thaw cycles F increases from 0 to 7, and the damage variable value D increases by 1.02 times. This indicates that cyclic loading continuously enhances the expansion and connectivity of cracks between soil particles after freeze-thaw cycles, which is also the main reason for the possible cracking of the subgrade during long-term service.