Status and prospects of frozen soil studies using CT technology

ShiJie Chen , ShuPing Zhao, Wei Ma, QianTao Zhu, LiLi Xing

State Key Laboratory of Frozen Soil Engineering, Cold and Arid Regions Environmental Engineering Research Institute, Chinese Academy of Sciences, Lanzhou, Gansu 730000, China

1 Introduction

Frozen soil is a complex and unique type of soil formed by soil particles, unfrozen water, ice, and gases.Very complex changes are produced in its internal structure from the effects of an exterior load or temperature gradient, involving the processes of phased change, mass transfer, and fissure formation. Because of experimental conditions and technical level constraints, there are still many unresolved issues pertaining to frozen soil research.In recent years, computed tomography (CT) technology has been widely applied to geotechnical and related fields.For frozen soil experiments, CT scanning is an effective method for nondestructive CT real-time monitoring and for quantitative description of internal structures.

A CT device was first designed in 1969 by Sir Godfrey Newbold Hounsfield (1973) from England. CT technology brought about a revolution in medical imaging after being made public during the 32nd Congress of the British Institute of Radiology in 1972. In 1990, the CT device was applied to the geotechnical field by the Key State Laboratory of Frozen Soil Engineering(KSLPE), Cold and Arid Regions Environmental and Engineering Research Institute (CAREERI), Chinese Academy of Sciences. A series of representative research findings were subsequently obtained with the use of a CT scanner, together with the deployment of specialized loading equipment. One such study conducted by Yanget al.(1998) was on the initial damage to rocks, which led to a variable formula for measuring rock damage represented in CT numbers and a study on the mechanism of damage propagation in sandstone under uniaxial com-pression. Renet al. (2000), Ren and Ge (2003), and Geet al. (2004) used CT technology to perform uniaxial and triaxial experiments on coal rock, fractured rock mass,and other research objects under a variety of load conditions, thereby tracking the changes and expansion patterns of the internal mesoscopic damages within rock masses. Yanget al. (1997), Wuet al. (2000), and Dinget al. (2003a) also utilized CT technology to carry out real-time and dynamic observations of the evolution and expansion of rock fractures. Subsequently, various agencies introduced the use of CT equipment for conducting geotechnical mesostructure tests (Dinget al., 2003b; Yinet al., 2006; Jianget al., 2011; Wanget al., 2012). For example, Wang and Feng (2009) utilized a high-precision micro-CT testing system to carry out micro-CT scans of five types of rock specimens. Using the absorption coefficient of the rays as the target, they were able to study the density distribution patterns of rock particles. In 2010,the Key State Laboratory of Frozen Soil Engineering further introduced and developed the Philips high-precision multi-voltage three-dimensional (3D) CT scanner and image processing workstation, which has a higher spatial and density resolution.

Summarizing the research findings from the application of CT technology in the geotechnical field over the past two decades, it was found that there were many more studies on rock and ordinary non-frozen soil than on frozen soil. One reason for this is that in order to carry out CT scanning of frozen soil, it is necessary to first research and develop specialized supplementary equipment that has precise temperature control and can comply with the operational requirements of the CT scanner. Another reason is the relatively complex character of the frozen soil itself. Regarding the use of test results from CT scans to explain the internal structural changes within frozen soil, there are still many technical problems that have not yet to be clearly understood, which affect their use and application.

With the above information as background, this paper begins by introducing the features of CT scanning. This is followed by a summary of the application status of CT technology in frozen soil research, including existing research findings and the identification of persistent problems and their possible solutions. Through this study,it is hoped that a basis can be established for the further application of CT technology in frozen soil research.

2 Features of CT scanning

2.1 Basic principles of CT scanner

A beam emitted by an X-ray tube is used to scan the selected surface from multiple directions. The amount of X-ray radiation that passes through the scanned surface is measured using a detector. The data are converted to a digital format using an analog/digital converter (ADC)and sent to a computer for storage and calculation. The X-ray absorption value per unit volume of the scanned surface is then obtained and sent via the computer to a digital/analog converter (DAC). The CT image is reconstructed through reverse projection and displayed on a monitor (Figure 1). The image is then clinically transferred onto a film. Hence, the CT image is actually the image calculated by the computer.

Figure 1 Schematic of basic setup and data flow of CT system

2.2 Definition and function of CT numbers

The CT numbers are defined as follows:whereHis the CT number of a specific pixel;μrmis the coefficient of X-ray absorption;μSDis the coefficient of X-ray absorption of a standard substance; andkis the indexing factor.

The standard equation for the medical use of CT ma-chines is using water as the standard substance andk= 1000,i.e.,μSD=μw(μwis the X-ray absorption coefficient of water),the following formula is derived:

The unit of measure for CT numbers is the Hounsfield unit (HU). Hence, the CT value of water is 0 HU, while that of air is 1,000 HU.

Applying formulae(1)and(2), the X-ray absorption rate of a particular point can be expressed in terms of CT numbers and reflected in various degrees of grayness on a CT image.The brighter portions on the image indicate areas of the scanned object that have higher absorption coefficients,whereas the darker portions indicate areas with lower absorption coefficients. Essentially, a CT image is a two-dimensional (2D) distribution diagram of the X-ray absorption coefficients of a specific scanned surface of the object being measured.

The X-ray absorption coefficientμrmof the test specimen can be expressed as follows:

whereμmis the absorption coefficient of a particular substance.

Formula(4)is obtained by substituting formula(3)into formula(1):

Formula(5)is obtained by substituting formula(3)into formula(2):

Formulae(4)and(5)indicate that the density is inversely proportional to the absorption coefficient but directly proportional to the CT number. A CT image is a representation of the density diagram of a particular scanned surface. In a case whereμm, the absorption coefficient of the substance being tested, is known, changes in the measured CT numbers can be used to quantitatively describe density changes in the substance when undergoing tests.

3 Application of CT scanning technology in frozen soil research

3.1 Supplementary equipment needed for CT scanning of frozen soil

While the medical CT scanner was undergoing long-term development, many scholars introduced the technology to conduct research in the field of material science and observe the meso-structural changes within test specimens. For example, Torranceet al. (2008) employed CT scanning technology to observe the internal structural changes when a test specimen underwent freezing, moisture migration, and the formation of an ice lens.

The earliest application of CT scanning technology in frozen soil research was by CAREERI, Chinese Academy of Sciences, in 1990, when it set up a related workstation. The experimental station was made up of a GE 8800 CT scanner from the United States and loading equipment that was developed in-house.

In 1998, CAREERI acquired the more advanced Siemens SOMATOM Plus spiral CT scanner from Germany and developed a high-pressure triaxial CT experimental loading device. Simulating the natural environmental conditions of various soils, researchers carried out dynamic monitoring of the meso-structural changes occurring within the soils during experiments, thereby laying the foundation for further theoretical analysis and engineering practice. The experimental system comprises a hydraulic loading subsystem, dedicated triaxial testbed, and data processing subsystem. With this system, it is not necessary to suspend the scanning process for loading and unloading when carrying out cross-sectional scanning. This facilitates the capturing of real-time CT images of any part of the cross-section under different stress levels, thereby realizing the true meaning of real-time dynamic scanning. This system can be used for the real-time scanning of soil samples and rocks when undergoing uniaxial or triaxial compression failure. Fatigue and creep failures under the effects of cyclical loading can also be captured.

Later, experimental research involving the CT scanning of frozen soil was conducted by various scholars, including Linget al. (2003) and Liuet al. (2005). However, in these experiments, mechanical tests were first conducted using a compression testing machine. Only after a certain stage were the test specimens placed in the CT scanner for scanning and subsequent analysis.

In 2007, Zhenget al. (2009a, 2011), Zhao (2008), and Fanget al. (2012) made improvements to the triaxial apparatus of the CT scanning system that allowed the temperature to be accurately controlled within a range of ±1 °C, which fulfilled the temperature control requirements for frozen soil experiments. With these improvements, the scanning system evolved from the original loading device with no temperature control to the current form or with temperature control to triaxial loading, allowing mechanical tests and CT scans on frozen soil to be carried out concurrently.

In 2010, the KSLPE of CAREERI acquired a Philips Brilliance 16-slice spiral CT scanner from the Netherlands,which has a higher resolution and allows for more accurate measurements than the earlier generations of CT scanners.With a spatial resolution of 0.208 mm and density resolution of 0.3%, this scanner allows high-precision multi-voltage 3D CT scans to be made. It is also configured with a digital imaging and communications in medicine (DICOM) stand-ard image processing workstation, which forms the technical basis for the rectification of image artifacts and the distribution measurements of multiphase materials.

3.2 Frozen soil research findings

3.2.1 Using CT numbers to analyze internal changes

From a CT image, the mean and variance of the CT numbers in the area of interest can be measured. Statistical methods can also be used to determine the effects of meso-structural and macroscopic density changes in specific parts of the test specimens. Through dynamic monitoring of uniaxial creep processes in frozen soil and an analysis of CT images and numbers, Wuet al. (1996) were able to establish the link between changes in the frozen soil mesostructure and its stress-strain relationship, which facilitated an exploratory analysis of the deformation characteristics of frozen soil. Puet al. (1995) applied mathematical analysis and experimental regression to derive mathematical equations for CT experiments on frozen soil. They linked the CT numbers of scanned images to related parameters in the frozen soil, thus establishing the basis for the use of CT scans in the quantitative analysis of frozen soil, after using different methods to separately freeze two specimens of silty clay.

Zheng (2009) and Zhenget al. (2009b) carried out nondestructive CT scans to determine changes in CT numbers produced by the freezing process. They concluded that axial freezing has a smaller impact on changes in the internal moisture content of the frozen soil specimens than radial freezing does.

Zhaoet al. (2010) used the improved triaxial apparatus and carried out dynamic CT scanning of frozen Lanzhou loess that was subjected to uniaxial compression under different temperature conditions. The CT numbers were plotted against the strain curves of the specimens under different temperature conditions and the findings were discussed accordingly.

3.2.2 Using CT numbers to define extent of damage

The internal structure and density of frozen soil change when they are subjected to the effects of external loading,which also causes CT numbers at the cross-section to change.Based on Yang and Zhang (1998), the formula for defining the extent of damage is as follows:

Formula(7)is obtained by substituting formula(4)into formula(6).

Assuming that the components of the cross-section of the specimen do not undergo any change during the mechanical tests,i.e.,μm0=μmi, formula(8)is obtained as:

Formula(9)is obtained after taking into account the impact of the instrument’s resolution:

wherem02is the spatial resolution of the CT scanner;H0is the initial CT number of the frozen soil specimen; andHiis the CT number after the specimen has undergone damage for a period of time equal toi.

It should be noted that the CT number represents the absorption value after the X-ray has passed through the specimen and is subjected to absorption in the process.When CT numbers are used to define the extent of the damage, the scale of the CT numbers must first be ascertained,i.e., the value of the indexing factorskandμSD, as stated in formula(1). The value ofkis related to the condition of the CT scanner and set when the CT scan is carried out.µSDis ascertained according to the standard substance adopted. HU is defined whenk= 1,000 andμSD=μw. Under this scenario, the CT number of water is 0, while that of air is -1,000. For industrial CT use, the value ofkis commonly set at 5,000, with aluminum (density of 2.8 g/cm3) as the standard object (Wanget al., 2010),i.e.,μSD=μAl(μAlis the coefficient of X-ray absorption of aluminum).Under this scenario, the CT number of aluminum is 0, while that of air is -5,000. For the medical CT scanner currently used in KSLPE, the value ofkis 1,024, and the CT number coverage ranges from -1,024 to 3,071. A phantom is adopted as the standard object,i.e.,μSD=μw. Hence, the CT number of water is 0, while that of air is -1,024.

Now, formula(9)changes in form to become formula(10):

Recognizing the connection between CT numbers and density, Liuet al. (2002) derived a formula to calculate frozen soil damage by applying the concept of additional damage to the frozen soil and using CT numbers as the variable to determine the meso-damages in rocks:

The incremental damage to the density is defined by Dinget al. (2000) as:

Formulae(11)and(12)are in basically corresponding situations.

Therefore, the scale of CT numbers must first be clearly defined when quantitative CT number calculations are being used to define frozen soil damage. Otherwise, the calculated value of the damage variable will not be accurate.

3.2.3 Using CT images for direct analysis

When a sample is first subjected to the effects of external loading or a temperature gradient, the changes to its internal structure at the initial stage are very small. Given the limitations of the human eye in identifying various shades of gray,it is generally difficult to notice any corresponding changes that show up on the CT images. As the effects continue to be exerted over time, internal structure change of the sample become more apparent, including the appearance of new meso-fractures, perforation of smaller pores to form larger pores, moisture migration, and the phase change of ice water.These changes show up through the redistribution of grayscale tones in the CT image. The trend of the change to the internal structure of the specimen can be determined by comparing the new CT image and the original image and analyzing the differences. Hence, through the analysis of CT images to understand the internal structural characteristics of the specimen in a nondestructive state, combined with changes in CT numbers of different portions of the specimen,it is possible to analyze the evolution of its internal structure.

Maet al. (1997) conducted triaxial compression creep tests on Lanzhou loess (150 mm height, 61.8 mm diameter,and 27% moisture content) at a temperature of -5 °C. CT scans were performed at the same time. By applying statistical methods to CT numbers of various scanned surfaces, they analyzed the distribution characteristics of the CT numbers at the full cross-sections and the central and peripheral regions,as indicated on the CT images of the frozen soil samples.They then obtained the CT bimodal curve for the structural changes within the frozen soil when it underwent the triaxial creep. Specifically, for the purpose of scanning, a specific area or segment of interest was first identified and fixed. The changes in CT numbers of the identified surface during the various experimental stages were documented and then compared. Through an analysis of changes in CT numbers,the evolution of the changes to the internal structure of the specimen was established.

Based on these findings, they believed that during the creep process, both structural weakening and strengthening phenomena exist. The effects of confining pressures inhibited the proliferation and expansion of the frozen soil specimen to a certain extent, leading to structural strengthening. As a result, the structural strengthening phenomenon was dominant as frozen soil underwent the stages of unstable and steady creep. With the further accumulation of inelastic deformation,if structural strengthening continued to dominate, the deformation would follow the characteristics of absorption.However, if the weakening effects overcame that of the confining pressures, non-absorption creep would develop, which would eventually lead to frozen soil failure.

4 Existing issues and possible solutions when using CT scanning technology for frozen soil research

4.1 Obtaining high-quality CT images

A high-quality CT image should have the following characteristics: (1) a high degree of separability, including high spatial resolution and low noise; (2) dependability, including linearity of the CT numbers and amount of image artifacts; and (3) an appropriate scanning dosage.

The CT images obtained from the scanning of rocks differ from those obtained via medical CT scanning in two main ways:

1) Calibration issues

The objects of geotechnical scanning are very different from the human body. Because the water content in an adult human body amounts to approximate 65% of the total body weight, a phantom is applied as the standard object, and air calibration is performed before scanning. The resulting CT numbers obtained from medical CT images have a very small margin of error. Furthermore, the diameter of the phantom (200 mm in the case of the Philips Brilliance 16-slice spiral CT scanner) is similar to that of a normal adult human torso. Hence, the CT numbers obtained from the scanning of human bodies are more accurate and precise than those from geotechnical scanning.

Another factor is the conditions under which medical scanning is carried out compared to geotechnical scanning.For the former, a low-dose scanning program is normally set for the various portions of the human body. For the latter, the scanning of many specimens can only be done with the assistance of specialized loading equipment, which makes it necessary to increase the scanning dosage in order to obtain CT images of a higher quality.

2) Quality of geotechnical CT images

During geotechnical experiments, the CT numbers obtained from the scanning of rock and soil specimens contain a certain degree of inaccuracy. This is because there is a greater degree of difference between the compositions and dimensions of the geotechnical specimens compared to the standard phantom. The actual manifestation is in terms of the image noise of a uniform specimen; that is, the standard deviation of the CT numbers for the area of interest located at the center of the image tends to be greater.

To address the aforementioned issues, two courses of action have to be taken:

1) Make adjustments to the scanning program by taking into account the actual conditions of the specimen to be scanned

Because the dimensions of geotechnical specimens are generally smaller, an infant phantom with a smaller diameter should be used when calibrating the phantom. Next, the conditions of the specialized loading equipment and the specimen itself should be considered when deciding on an appropriate and specific high-power scanning program.

2) Consider the implications of the X-ray theory

The mass absorption coefficient of a single element for X-rays of different wavelengths is as follows:

whereKis a constant;Zis the atomic number; andλis the X-ray wavelength.

The scanning voltage of a CT scanner,i.e., the voltage of the X-ray tube, is inversely proportional to the X-ray wavelength. Hence, modifying the scanning voltage will lead to a corresponding change in the X-ray wavelength. The same element, when scanned under different scanning voltages,will produce a plurality of mass absorption coefficients. In other words, for the same substance, X-ray scans with different energy levels will result in different CT numbers.

A multi-voltage CT scanner is able to carry out scanning using different voltages. Utilizing this feature, frozen soil scanning is performed using various scanning voltages to obtain images of higher quality. For example, the scanning of cement concrete was performed using 3 voltages: 90, 120,and 140 kV. It can be seen from figure 2 that the CT images obtained with the 120 and 140 kV scanning voltages have clearer boundaries. The internal fissures also show up more clearly.

Figure 2 CT images obtained under different scanning voltage conditions

The mean and standard deviation of the CT numbers obtained from the image scanned using the 140 kV scanning voltage were smaller than those found when using 120 kV(Table 1). For this specific specimen, using the 140 kV scanning voltage produced CT images of a higher quality and CT numbers that were less discrete. Therefore, during the process of scanning frozen soil with a multi-voltage CT scanner, the appropriate scanning voltage should be selected prior to each scan, depending on whether the target to be observed consists of soil particles or ice.

Table 1 Mean and standard deviation of CT numbers under different scanning voltage conditions

Furthermore, CT image quality is also affected by other factors, including the scanning tube current, thickness of the scanned specimen, and reconstruction matrix. In order to obtain high-quality CT images and numbers, appropriate scanning parameters must be selected based on the actual specimen. For example, a greater tube current should be selected when scanning specimens with larger dimensions.

4.2 Issues faced when using CT numbers for analysis

For a homogeneous substance (for example, metallic materials), the distribution diagram of the CT numbers can be directly converted to a density diagram based on the standard table of mass absorption coefficients,i.e., when

Formula(4)can be converted into formula(14):

A method of calibrating and measuring the density of materials is provided in the United States ASTM E1935-97(2008) standards. Specimens of two or more mate-rials are selected. Theμmof each material is checked using the related manuals, whileρis measured by weighing. The specimens are scanned using a particular voltage. This establishes the relationship between the CT numbers and the X-ray absorption coefficientμrm, which is related to the values ofaandbin formula(14). The distribution of the CT numbers can be measured by scanning a particular slice of the specimen. The distribution ofμrmcan then be obtained using formula(14). Subsequently, the density distribution of the slice zone of the specimen can be computed.

If the frozen soil specimen is treated as a whole, with a minimum of two sets of experiment results, the relationship between the overall density and CT numbers can be derived from the average density of the specimen and the average value of CT numbers for different slices of the specimen. The values ofaandbin formula(14)can also be obtained. Using this method, CT numbers of a particular slice of frozen soil can be converted into the density distribution. A CT image essentially represents the density diagram of a slice of frozen soil that has been scanned. However, this is still on a macroscopic scale and does not characterize the internal structural changes within the frozen soil.

In reality, frozen soil is a complex substance formed by soil particles, unfrozen water, ice, and gases. The masses of the soil particles, water, ice, and gases within the frozen soil are represented byms,mw,mi, andmg, respectively, whilemrepresents the total mass. This means that the overall mass absorption coefficient of the frozen soil specimen complies with the formula:

whereμsis the absorption coefficient of the soil particles;μwis the absorption coefficient of water;μiis the absorption coefficient of ice; andμgis the absorption coefficient of the gases.

Formulae(16)and(17)are obtained after substituting formula(15)into formula(14):

wherevsis the total volume of the soil particles;vwis the total volume of water;νiis the total volume of ice;νgis the total volume of the gases; andνis the total volume of the specimen.

The absorption coefficient of a homogeneous substance does not change as long as the conditions for the CT scan remain constant. Hence, there is a quantitative relationship between the CT numbers and density,i.e., formula(14).However, the situation is much more complicated for frozen soil. Even if the overall density of the specimen does not vary,the mass components (soil particles, water, ice, and gases)may be undergoing changes. In addition, because of the interactions between the various components, there are also some differences between the mass absorption coefficients of the components within the frozen soil and that of a substance which comprises only one of those components. From formulae(16)and(17), it can be seen that a quantitative relationship between the CT numbers and density does not exist at the mesoscopic scale.

A possible solution is to first conduct experiments to determine the individual mass absorption coefficient of each component (soil particles, water, ice, and gases), as well as the proportion of each component within the frozen soil.After scanning, statistical analysis methods can be used to determine the absorption coefficient of each component before the statistical relationship between the CT numbers of frozen soil and density is established. A large amount of work is involved in this solution. There is also a need to design reasonable experiments to obtain the various data. Thus, the biggest obstacle when applying CT scanning technology to frozen soil research is finding a way to establish a close relationship between the changes in CT numbers for the images and the changes undergone by the various components of the region of interest in the frozen soil specimen, and then, on that basis, reflecting the internal structural changes within the frozen soil and the changes in the mass transfer.

4.3 Issues faced when using CT images

With the continuous development of digital information technology, digital image processing technology has gradually developed into an independent discipline and been applied to various disciplines and fields.

The design of the CT scanner itself causes the quality of CT images to be constrained by various factors, including noise, beam hardening, repositioning, and resolution. The digital image processing of CT images is thus needed to address the following aspects:

1) Reduction in system errors

It includes controlling the noise in CT images, image uniformity, linearity of the CT numbers, and positioning accuracy.

2) Quality control of CT images

It involves two aspects. The first is the removal of CT artifacts, including artifacts caused by the beam, ring artifacts caused by differences in the detector channel, scattering artifacts, and those caused by the partial volume effect. The other aspect is the rectification of data distortions generated during the experiment, including correction to the dimensions of the specimen and the setting of the specimen’s edges and boundaries.

3) Prioritization of information

The useful information obtained from CT images should be enhanced while useless information should be weakened or suppressed. Image enhancement technologies (such as image smoothening, sharpening, and point operations) can be used to achieve the best display effects.

After the aforementioned issues have been addressed,computer analysis and processing of the information contained in the CT image can be carried out to extract the desired research data. Currently, the most commonly used method is to extract useful information for 3D reconstruction of the image.Intensive cross-section scanning is carried out next to obtain CT numbers of the different parts of the specimen.

After the threshold value of the CT numbers is set, numbers that exceed the threshold are excluded. For numbers that are lower than the threshold, the boundary of the area where these numbers are derived will be marked out to create a new 2D image of the subject specimen. One image (or several images combined) is then used to create a new 3D image to reflect the spatial distribution of the information. For example,some scholars have extracted information on fissures from CT images for 3D reconstruction. Dynamic CT images of 3D resampling can reveal the spatial location of fissures in various parts of the specimen in a vivid and intuitive manner, as well as illustrate the evolution process starting from fissure initiation to propagation and eventually percolation.

In the medical field, Bayet al. (1999) carried out 3D reconstructions of digital images. By tracking the changes in CT numbers before and after human bones were subjected to pressure, they recreated the displacement and deformation fields as the bones changed under pressure, analyzed the deformation details of various parts of the bones, and attempted to provide the basis for selection of suitable materials for making artificial bones.

Wanget al. (2003) employed CT technology to conduct quantitative research on damage to materials. CT images and virtual partitioning technology were used to obtain full images of the various sections, which were in turn used to reconstruct a 3D delaminated concrete structure. Quantified damage parameters were revealed in the process, as well as the potential application of these parameters to mechanics modeling. The findings were subsequently verified using an asphalt mixture.

Fanet al. (2004) utilized the visual characteristics of the human eye and the color modes and image formats of digital computer images. By adopting a method that transforms grayscale to colors, they were able to provide pseudo color enhancement of grayscale CT images, thereby improving the resolution of the CT image significantly.

One way to use image processing methods for frozen soil scans is to extract from the CT images information related to the deformation and failure of the specimen, such as the equivalent fissure width and lateral deformation. By analyzing the evolutionary mechanism of the mesofissures, a fissure evolution model can be constructed. Another way is to extract useful information (such as changes in soil density and the formation of ice crystals) during the process of frost heaving for 3D reconstruction.

5 Conclusion

Between 1990, when the earliest geotechnical experimental workstation was established in China, and the introduction of the new-generation high-speed CT scanner and the development of related supplementary equipment in 2010,KSLPE of the Chinese Academy of Sciences employed CT scanning technology to carry out various geotechnical experiments and studies, achieving a wide range of research findings. In this paper, we have provided a general overview of the application of CT scanning technology in the field of frozen soil research, including its development history and various research findings. We have illustrated that this technology is an effective means to study the internal mesostructure evolution of frozen soil. With the use of supporting experimental equipment and digital image processing in post-production, we believe that CT scanning technology can play an even greater role in frozen soil research.

The authors are grateful to the editors and reviewer for constructive suggestions. This paper was financially supported by the National Natural Science Foundation of China (Nos. 41023003, 40971046, 41201181), the Foundation of State Key Laboratory of Frozen Soil Engineering (No. Y252J81001) and the Youth Foundation of Cold and Arid Regions Environmental Engineering Research Institute, Chinese Academy of Sciences (No.51Y251B91).

Bay BK, Smith TS, Fyhrie DP,et al., 1999. Digital volume correlation: three dimensional strain mapping using X-ray tomography. Experimental mechanics, 39: 217–226. DOI: 10.1007/BF02323555.

Ding WH, Wu YQ, Pu YB,et al., 2003a. Measurement of crack width in rock interior based on X-ray CT. Chinese Journal of Rock Mechanics and Engineering, 22(9): 1421–1425.

Ding WH, Wu YQ, Pu YB,et al., 2003b. X-ray CT approach on rock interior crack evolution under low strain rate. Chinese Journal of Rock Mechanics and Engineering, 22(11): 1793–1797.

Ding WH, Wu YQ, Pu YB,et al., 2000. New progress in real time detecting of geotechnical experiment by use of X ray computed tomography. Journal of Glaciology Geocryology, 22: 218–222.

Fan LM, Li N, Ding WH, 2004. Application of pseudo color enhancement method to geotechnical CT image analysis. Chinese Journal of Rock mechanics and engineering, 23(13): 2257–2261.

Fang LL, Qi JL, Ma W, 2012. Freeze thaw induced changes in soil structure and its relationship with variations in strength. Journal of Glaciology and Geocryology, 34: 435–440.

Ge XR, Ren JX, Pu YB,et al., 2004. Macro and Meso Testing Study on Damage Mechanics of Geotechnical Media. Science Press, Beijing.

Hounsfield GN, 1973. Computerized transverse axial scanning (tomography).Part I. Description of system. British Journal of Radiology, 46:1016–1022. DOI: 10.1259/0007-1285-46-552-1016.

Jiang YZ, Wang YL, Wang SJ, 2011. 3D reconstruction for CT images of expansive soil based on Matlab. Journal of Southwest University (Natural Science Edition), 33(3): 144–148.

Ling XZ, Xu XY, Qiu MG,et al., 2003. Study on CT scanning of Harbin frozen silty clay before and after dynamic triaxial test. Chinese Journal of Rock Mechanics and Engineering, 22(8): 1244–1249.

Liu ZL, Li HS, Zhu YL,et al., 2002. A distinguish model for initial and additional micro-damages on frozen soil. Journal of Glaciology and Geocryology, 24: 676–680.

Liu ZL, Zhang XP, Li HS, 2005. A damage constitutive model for frozen soils under uniaxial compression based on CT dynamic distinguishing. Rock and Soil Mechanics, 4: 542–546.

Ma W, Wu ZW, Pu YB,et al., 1997. Monitoring the change of structures in frozen soil in triaxial creep process by CT. Journal of Glaciology and Geocryology, 19: 52–57.

Pu YB, Wu ZW, Ma W,et al., 1995. CT mathematial equation of CT experiment on frozen soil. Journal of Glaciology and Geocryology, 17:135–139.

Ren JX, Ge XR, Pu YB,et al., 2000. Real-time CT test on the meso-damage evolution of rock under uniaxial compression. China Civil Engineering Journal, 33(6): 99–104.

Ren JX, Ge XR, 2003. Whole process of unloading damage of fractured granite CT real-time testing. Progress in Natural Science, 13(3):275–280.

Torrance JK, Elliot T, Martin R,et al., 2008. X-ray computed tomography of frozen soil. Cold Regions Science and Technology, 53(1): 75–82. DOI:10.1016/j.coldregions.2007.04.010.

Wang J, Huang SH, Cai YF, 2010. Research on measuring ICT material density. Computer Engineering and Applications, 46(2): 203–205.

Wang LB, Frost JD, Voyiadjis GZ, 2003. Quantification of damage parameters using X-ray tomography images. Mechanics of Materials, 35(8):777–790. DOI: 10.1016/S0167-6636(02)00206-5.

Wang SJ, Chen ZH, Li X,et al., 2012. Pore-damage evolution and mechanical properties of remolded soil by CT triaxial test. Transactions of the Chinese Society of Agricultural Engineering, 28(7): 150–154.

Wang XH, Feng ZC, 2009. Study on distribution laws of rock density. Chinese Journal of Rock Mechanics and Engineering, 28(S2): 3484–3489.

Wu YQ, Ding WH, Pu YB,et al., 2000. Rock under compression CT dynamic observation of density damage increment. Progress in Natural Science, 10(9): 830–835.

Wu ZW, Ma W, Pu YB,et al., 1996. Monitoring the change of structures in frozen soil in uniaxial creep process by CT. Journal of Glaciology and Geocryology, 18: 306–311.

Yang GS, Xie DY, Zhang CQ,et al., 1998. Application of CT technique in the damage detection of rock materials. Journal of Experimental Mechanics, 4: 24–29.

Yang GS, Xie DY, Zhang CQ,et al., 1997. The discussion about damage constitutive relation of rock under uniaxial stress identified by CT. Rock and Soil Mechanics, 2: 29–34.

Yang GS, Zhang CQ, 1998b. Rock Damage and Its Detection. Shaanxi Science and Technology Press, Xi’an.

Yin XT, Dang FN, Ding WH,et al., 2006. Morphologic measurement of crack in CT images of rock and soil. Chinese Journal of Rock Mechanics and Engineering, 25(3): 539–544.

Zhao SP, Ma W, Zheng JF,et al., 2010. CT real-time monitoring on frozen Lanzhou loess at different temperatures and under uniaxial loading. Rock and Soil Mechanics, 31(S2): 92–97.

Zhao SP, 2008. Theoretical analysis for coupled problem of temperature,moisture and stress fields in frozen soil. Graduate school of the Chinese Academy of Sciences Dissertation for Ph.D., Cold and Arid Regions Environmental and Engineering Research Institute, Chinese Academy of Sciences.

Zheng JF, Ma W, Zhao SP,et al., 2011. Study of meso-damage changes of frozen Lanzhou loess under triaxial compression based on CT real-time monitoring. Journal of Glaciology and Geocryology, 33: 839–845.

Zheng JF, Zhao SP, Ma W,et al., 2009a. Application of CT testing technology in evaluating the uniformity of soil samples. Journal of Lanzhou University (Natural Sciences), 45(2): 20–25.

Zheng JF, Zhao SP, Ma W,et al., 2009b. Study of the effect of freezing on frozen soil specimen preparation. Journal of Glaciology and Geocryology,31: 130–138.

Zheng JF, 2009. Study on the mesostructure changes of frozen soil under force. Graduate school of the Chinese Academy of Sciences Dissertation for Ph.D., Cold and Arid Regions Environmental and Engineering Research Institute, Chinese Academy of Sciences, pp. 1–133.