Investigation of Voids Characteristics in an Asphalt Mixture Exposed to Salt Erosion Based on CT Images

The performance of an asphalt mixture will deteriorate under the condition of salt erosion, but there are different opinions on the mechanism of deterioration. Few studies have focused on the relation between the change of void characteristics and performance deterioration of an asphalt mixture exposed to salt erosion. To explore the relation between the air voids characteristics of an asphalt mixture and mechanical damage under salt erosion, the mechanical damage in an asphalt mixture was measured by splitting strength. The asphalt mixture specimens, immersion solutions, asphalt mortar, and aggregate were scanned with CT technology. To segment the voids, the Otsu method was used over asphalt mortar and solution range of CT values. A three-dimensional reconstruction of the CT image was performed with Mimics 20 software to calculate the asphalt mixture’s void characteristics. On this basis, the relationships between the change in void characteristics and splitting strength were analyzed. The results showed that the ideal calculated void fraction can be obtained by threshold segmentation of the image void/asphalt mortar interface with the local CT value Otsu method. Under the salt corrosion environment, the increase of open voids of an asphalt mixture is linearly correlated with the decrease of splitting strength, while salts’ crystallization in the open voids produces crystallization pressure, accelerating the volume growth of open voids. The early damage of an asphalt mixture suffered from the salt may be mainly physical damage. These results can provide a useful reference for the performance of damage research on asphalt mixtures in salt enrichment areas.


Introduction
In coastal areas with frequent marine fog and high salinity and humidity, inland saline lakes, saline soil areas, and northern areas where chlorine salts, such as road deicing salts and snow-melting agents, are used widely, salt erosion compromises a pavement's performance [1,2]. Most research to date largely has studied salt erosion's influence on asphalt mixture performance according to changes in splitting strength, rutting factor, and other mechanical properties before and after exposure to different salt erosion conditions (dry-wet cycle, freeze-thaw cycle, etc.) and different salts (NaCl, Mg 2 Cl, Ca 2 Cl, Na 2 SO 4 , etc.) [3][4][5]. Because the macro-mechanical test index is based on repeated tests,

Aggregate and corrosive salinity
All of the coarse and fine aggregates used in the test were made of basalt, and the filler was ground limestone ore powder. The mineral aggregate's gradation was AC-13, and each grade of aggregate's passing rate is shown in Table 2. The asphalt mixture's optimum asphalt-stone ratio is 4.78%, as determined by the standard Marshall test. Industrial salt was selected for the test. The technical indices are shown in Table 3.

Methods
The asphalt mixture specimens were prepared by the standard Marshall test method according to the "Test Rules for Asphalt and Asphalt Mixture in Highway Engineering" (JTG E-20-2011). The specimens' initial void fraction was measured by the surface drying method. Eleven groups of molded dry specimens were used to meet the specifications' requirements, and were labeled AC(0)~AC (10). There were four specimens in the blank group, and the remainder included 16. All immersed specimens had to be saturated with water before the test. The specific test steps are as follows: place the specimen in a box filled with clear water with the water surface 2 cm higher than the top surface of the specimen and place the box in the vacuum instrument at a negative pressure of 3.7 + 0.3 kPa for 20 min. After vacuum treatment, the salt water erosion test was carried out in the test environment shown in Table 4. The specific laboratory test environments were as follows: The salt immersion environment was as follows: the sample was immersed horizontally at 40 • C. During the test, the test box is covered to prevent test error attributable to water evaporation.
The drying and wetting cycle environment was a 24 h cycle (soaking for 12 h, drying for 12 h); the soaking condition were the same as above, while the drying condition entailed removing the sample horizontally and placing it in an oven set at 40 • C.
According to the "Test Rules for Asphalt and Asphalt Mixture in Highway Engineering" (JTG E20-2011), the freeze-thaw splitting test for an asphalt mixture was used to test the AC (0) specimens and other specimens' initial splitting strength at 7, 14, 21, and 28 days of age.

CT Scanning Test
The CT equipment used in this experiment was the Hitachi ECLOS.16 row spiral CT (HITACHI, Zun'yi, Guizhou Province, China), the CT test stand is shown in Figure 1, the scanning parameters of which are shown in Table 5.

Soak
AC(1) AC(2) AC(3) AC(4) AC (5) Dry-wet cycle AC(6) AC(7) AC(8) AC (9) AC (10) The specific laboratory test environments were as follows: The salt immersion environment was as follows: the sample was immersed horizontally at 40 °C. During the test, the test box is covered to prevent test error attributable to water evaporation.
The drying and wetting cycle environment was a 24 hour cycle (soaking for 12 hours, drying for 12 hours); the soaking condition were the same as above, while the drying condition entailed removing the sample horizontally and placing it in an oven set at 40 °C.
According to the "Test Rules for Asphalt and Asphalt Mixture in Highway Engineering" (JTG E20-2011), the freeze-thaw splitting test for an asphalt mixture was used to test the AC (0) specimens and other specimens' initial splitting strength at 7, 14, 21, and 28 days of age.

CT Scanning Test
The CT equipment used in this experiment was the Hitachi ECLOS.16 row spiral CT (HITACHI, Zun'yi, Guizhou Province, China), the CT test stand is shown in Figure 1, the scanning parameters of which are shown in Table 5.
The initial asphalt mortar, aggregate, distilled water, and a 20% salt solution, respectively, were put in small glass bottles (22 × 60 mm), and their corresponding CT values were obtained by CT scanning. The scanning sample is shown in Figure 2.  Four standard Marshall specimens were selected and referred to as AC1, AC2, AC3, and AC4. The test environment is shown in Table 6. As CT is a non-destructive testing method, the same specimens were scanned continuously at 0, 7, 14, 21, and 28 days of age. Before scanning, samples were removed from the water and salt water, wiped with a towel to remove their surface moisture, and then subjected to CT scanning. The CT-scanned specimens' test environment is shown in Table  6. The conditions for the soaking and drying-wetting cycle are the same as those in Table 4. Each specimen was scanned 42 times. Given that compaction affected the specimens' upper and lower ends greatly, 40 middle CT images were selected for experimental analysis, and the specimens' original CT images were obtained. Some of the scanned images are superimposed, as shown in Figure 3.  The initial asphalt mortar, aggregate, distilled water, and a 20% salt solution, respectively, were put in small glass bottles (22 × 60 mm), and their corresponding CT values were obtained by CT scanning. The scanning sample is shown in Figure 2. Four standard Marshall specimens were selected and referred to as AC1, AC2, AC3, and AC4. The test environment is shown in Table 6. As CT is a non-destructive testing method, the same specimens were scanned continuously at 0, 7, 14, 21, and 28 days of age. Before scanning, samples were removed from the water and salt water, wiped with a towel to remove their surface moisture, and then subjected to CT scanning. The CT-scanned specimens' test environment is shown in Table 6. The conditions for the soaking and drying-wetting cycle are the same as those in Table 4. Each specimen was scanned 42 times. Given that compaction affected the specimens' upper and lower ends greatly, 40 middle CT images were selected for experimental analysis, and the specimens' original CT images were obtained. Some of the scanned images are superimposed, as shown in Figure 3.

Image Processing
Because this paper focused on voids, image processing focused on the segmentation of the asphalt mortar/void interface. Because the specimens were immersed, there was water (brine) at the interface between the asphalt mortar and the specimens' open voids that affect the image processing results. The CT images of the asphalt mortar, aggregate, and salt solution at various concentrations are shown in Figure 4, and the CT values are presented in Table 7.

Asphalt mortar
Aggregate 0% 5% 10% 15% 20%   Table 7, it can be seen that the asphalt mortar and solutions' CT values were similar, such that the gray histogram of the asphalt mixture's CT image does not show obvious double peaks, and thus, the double peaks method is not suitable for segmentation. Therefore, in this paper, the gap and asphalt mortar were separated by calculating the T value using the Otsu method over the range of

Image Processing
Because this paper focused on voids, image processing focused on the segmentation of the asphalt mortar/void interface. Because the specimens were immersed, there was water (brine) at the interface between the asphalt mortar and the specimens' open voids that affect the image processing results. The CT images of the asphalt mortar, aggregate, and salt solution at various concentrations are shown in Figure 4, and the CT values are presented in Table 7.

Image Processing
Because this paper focused on voids, image processing focused on the segmentation of the asphalt mortar/void interface. Because the specimens were immersed, there was water (brine) at the interface between the asphalt mortar and the specimens' open voids that affect the image processing results. The CT images of the asphalt mortar, aggregate, and salt solution at various concentrations are shown in Figure 4, and the CT values are presented in Table 7.

Asphalt mortar
Aggregate 0% 5% 10% 15% 20%   Table 7, it can be seen that the asphalt mortar and solutions' CT values were similar, such that the gray histogram of the asphalt mixture's CT image does not show obvious double peaks, and thus, the double peaks method is not suitable for segmentation. Therefore, in this paper, the gap and asphalt mortar were separated by calculating the T value using the Otsu method over the range of   Table 7, it can be seen that the asphalt mortar and solutions' CT values were similar, such that the gray histogram of the asphalt mixture's CT image does not show obvious double peaks, and thus, the double peaks method is not suitable for segmentation. Therefore, in this paper, the gap and asphalt mortar were separated by calculating the T value using the Otsu method over the range of CT values. The specific steps were as follows: Over the range of CT values, the asphalt mortar's histogram void and CT value, the void's segmentation threshold (i.e., target), and the asphalt mortar were recorded as T, and the proportion of the void pixel points to the pixel points in the segmentation area of the CT value histogram was recorded as ω0, while the average CT value was µ0. The proportion of pitch mortar pixels in the CT value histogram segmentation area was ω1, and the mean CT value was µ1. The segmented area's total mean CT value was recorded as µ and the variance between classes was recorded as g.
The total number of pixels in the segmentation area was M, the number of pixels with a CT value less than the threshold T was N0, and the number of pixels with a CT value greater than the threshold T was N1.
The equivalent formula is obtained by substituting Equation (5) into Equation (6): The threshold T, which maximizes the variance g between classes, was obtained by the traversal method.
The number of CT values in the sample CT image was counted, and a histogram of CT values was drawn. Equation (7) was used to calculate the CT values in the range of 2HU-998HU, and T = 532 HU was obtained. The segmentation diagram is shown in Figure 5. To verify this method's applicability, the gray value of the CT image was calculated by the Otsu method at the same time, and the threshold T1 = 74 was obtained. The value of T1 was converted to 766 HU, which was recorded as T1. The schematic diagram of the gray value and CT value conversion is shown in Figure 6. When the CT value exceeds 3250 HU, the gray value is 255, and when the CT value is less than −250 HU, the gray value is 0. When the CT value is between −250 HU and 3250 HU, the conversion formula between the CT and gray values is as follows: CT values. The specific steps were as follows: Over the range of CT values, the asphalt mortar's histogram void and CT value, the void's segmentation threshold (i.e., target), and the asphalt mortar were recorded as T, and the proportion of the void pixel points to the pixel points in the segmentation area of the CT value histogram was recorded as ω0, while the average CT value was μ0. The proportion of pitch mortar pixels in the CT value histogram segmentation area was ω1, and the mean CT value was μ1. The segmented area's total mean CT value was recorded as μ and the variance between classes was recorded as g.
The total number of pixels in the segmentation area was M, the number of pixels with a CT value less than the threshold T was N0, and the number of pixels with a CT value greater than the threshold T was N1.
The equivalent formula is obtained by substituting Equation (5) into Equation (6): The threshold T, which maximizes the variance g between classes, was obtained by the traversal method.
The number of CT values in the sample CT image was counted, and a histogram of CT values was drawn. Equation (7) was used to calculate the CT values in the range of 2HU-998HU, and T = 532HU was obtained. The segmentation diagram is shown in Figure 5. To verify this method's applicability, the gray value of the CT image was calculated by the Otsu method at the same time, and the threshold T1 = 74 was obtained. The value of T1 was converted to 766 HU, which was recorded as T1. The schematic diagram of the gray value and CT value conversion is shown in Figure 6. When the CT value exceeds 3250 HU, the gray value is 255, and when the CT value is less than -250 HU, the gray value is 0. When the CT value is between −250HU and 3250HU, the conversion formula between the CT and gray values is as follows:  After the three-dimensional reconstruction with T, the specimens' T1 segmentation thresholds and void values were calculated and compared with those measured by the surface drying method. The results are shown in Table 8. The voidage value obtained from the test was taken as the true voidage value, and the error analysis of the voidage value calculated was carried out. The results of the relative error analysis are shown in Table 9.  After the three-dimensional reconstruction with T, the specimens' T1 segmentation thresholds and void values were calculated and compared with those measured by the surface drying method. The results are shown in Table 8. The voidage value obtained from the test was taken as the true voidage value, and the error analysis of the voidage value calculated was carried out. The results of the relative error analysis are shown in Table 9.  After the three-dimensional reconstruction with T, the specimens' T1 segmentation thresholds and void values were calculated and compared with those measured by the surface drying method. The results are shown in Table 8. The voidage value obtained from the test was taken as the true voidage value, and the error analysis of the voidage value calculated was carried out. The results of the relative error analysis are shown in Table 9. After the three-dimensional reconstruction with T, the specimens' T1 segmentation thresholds and void values were calculated and compared with those measured by the surface drying method. The results are shown in Table 8. The voidage value obtained from the test was taken as the true voidage value, and the error analysis of the voidage value calculated was carried out. The results of the relative error analysis are shown in Table 9.
X-true voidage value; x-calculated voidage value; D-absolute error; Er-relative error  Tables 8 and 9 show that the voidage calculated by the local CT value Otsu method is closer to the voidage value measured. This shows that this image processing method can obtain the mixture's ideal void fraction under the condition of solution immersion. Therefore, this method was used for image segmentation in all of the gap analyses in this paper.

Analysis of Splitting Strength Results
The results of the splitting strength and 28 d splitting strength loss are shown in Figures 9-11.
X-true voidage value; x-calculated voidage value; D-absolute error; Er-relative error  9 show that the voidage calculated by the local CT value Otsu method is closer to the voidage value measured. This shows that this image processing method can obtain the mixture's ideal void fraction under the condition of solution immersion. Therefore, this method was used for image segmentation in all of the gap analyses in this paper.

Analysis of Splitting Strength Results
The results of the splitting strength and 28 d splitting strength loss are shown in Figures 9-11.   Figure 9 shows the change trend of splitting strength of an asphalt mixture with the increase of age under the condition of direct immersion in five salt concentration (including 0%). It can be seen that the damage trend of salt erosion is similar to that of water erosion (0% salt concentration) under this erosion method. As the age increased, the strength of the specimens soaked in clear water decreased most rapidly at 28 days, by 11.7%, while the strength of the specimens soaked in 20% brine decreased most slowly, by 8.5%. Given that there is salt crystallization in 20% brine-immersed specimens that fills small voids, the splitting strength increased, which counteracts some of the damaged by water [29]. Figure 10 shows that the splitting strength decreased with increased age under the condition of wet-dry cycling. The higher the salt concentration is, the faster the rate of decline becomes. After 28 days, the specimens' strength decreased most slowly under the dry-wet cycle with clean water, 8.1%, and 26.9% under the dry-wet cycle with 20% brine. This is because under dry-wet cycling conditions, salt crystallization produces expansion pressure, and the higher the salt concentration, the greater that pressure, and the greater the damage to the asphalt mixture.
The results suggest that the strength of the samples under the dry-wet cycling with the salt solution decreased significantly faster than that of the samples immersed directly, as shown in Figure 11. This may indicate that there are more changes that have taken place in the asphalt mixture during the salt water drying and wetting cycle. In order to analyze the deterioration  Figure 9 shows the change trend of splitting strength of an asphalt mixture with the increase of age under the condition of direct immersion in five salt concentration (including 0%). It can be seen that the damage trend of salt erosion is similar to that of water erosion (0% salt concentration) under this erosion method. As the age increased, the strength of the specimens soaked in clear water decreased most rapidly at 28 days, by 11.7%, while the strength of the specimens soaked in 20% brine decreased most slowly, by 8.5%. Given that there is salt crystallization in 20% brine-immersed specimens that fills small voids, the splitting strength increased, which counteracts some of the damaged by water [29]. Figure 10 shows that the splitting strength decreased with increased age under the condition of wet-dry cycling. The higher the salt concentration is, the faster the rate of decline becomes. After 28 days, the specimens' strength decreased most slowly under the dry-wet cycle with clean water, 8.1%, and 26.9% under the dry-wet cycle with 20% brine. This is because under dry-wet cycling conditions, salt crystallization produces expansion pressure, and the higher the salt concentration, the greater that pressure, and the greater the damage to the asphalt mixture.
The results suggest that the strength of the samples under the dry-wet cycling with the salt solution decreased significantly faster than that of the samples immersed directly, as shown in Figure 11. This may indicate that there are more changes that have taken place in the asphalt mixture during the salt water drying and wetting cycle. In order to analyze the deterioration  Figure 9 shows the change trend of splitting strength of an asphalt mixture with the increase of age under the condition of direct immersion in five salt concentration (including 0%). It can be seen that the damage trend of salt erosion is similar to that of water erosion (0% salt concentration) under this erosion method. As the age increased, the strength of the specimens soaked in clear water decreased most rapidly at 28 days, by 11.7%, while the strength of the specimens soaked in 20% brine decreased most slowly, by 8.5%. Given that there is salt crystallization in 20% brine-immersed specimens that fills small voids, the splitting strength increased, which counteracts some of the damaged by water [29]. Figure 10 shows that the splitting strength decreased with increased age under the condition of wet-dry cycling. The higher the salt concentration is, the faster the rate of decline becomes. After 28 days, the specimens' strength decreased most slowly under the dry-wet cycle with clean water, 8.1%, and 26.9% under the dry-wet cycle with 20% brine. This is because under dry-wet cycling conditions, salt crystallization produces expansion pressure, and the higher the salt concentration, the greater that pressure, and the greater the damage to the asphalt mixture.
The results suggest that the strength of the samples under the dry-wet cycling with the salt solution decreased significantly faster than that of the samples immersed directly, as shown in Figure 11. This may indicate that there are more changes that have taken place in the asphalt mixture during the salt water drying and wetting cycle. In order to analyze the deterioration mechanism of an asphalt mixture performance under the condition of salt erosion from the microscopic point of view, the mixture's void evolution characteristics under the maximum and minimum damage conditions were investigated by CT scanning (i.e., 0%, 20% salt water concentration immersion and dry-wet cycle).

Calculating the Evolution Value of Void Characteristics
Evolution values for the AC1, AC2, AC3, and AC4 specimens' void characteristics were calculated as shown in Tables 10-13.  The influence degree of each erosion environment on void ratio and opening void ratio is similar to that of splitting strength.

Relevance between Open Void and Splitting Strength
The experiments demonstrate that compared with the change of total voids, the loss of splitting strength is highly correlated with the change of open void fraction, as illustrated in Figure 12. This may indicate that the open voids are most likely to be damaged when the asphalt mixture is in an erosive environment, leading to the performance deterioration of an asphalt mixture.  It can be seen from Figure 13b that the number of closed voids of an asphalt mixture in four erosion environments also decreases with the increase of erosion age. Under the condition of direct soaking, the decrease of the number of closed voids is concentrated in the early stage of erosion, and tends to be gentle in the later stage. However, under the condition of dry-wet cycles, the number of closed voids in asphalt mixture tends to decrease rapidly in the later stage of erosion. These results may indicate that the erosion from salt water to the asphalt mixture's void structure is from the outside to the inside.  It can be seen from Figure 13b that the number of closed voids of an asphalt mixture in four erosion environments also decreases with the increase of erosion age. Under the condition of direct soaking, the decrease of the number of closed voids is concentrated in the early stage of erosion, and tends to be gentle in the later stage. However, under the condition of dry-wet cycles, the number of closed voids in asphalt mixture tends to decrease rapidly in the later stage of erosion. These results may indicate that the erosion from salt water to the asphalt mixture's void structure is from the outside to the inside.   On the whole, the results of void characteristics may indicate that salt crystallizes on the pore wall of the opening space in an asphalt mixture under the condition of dry and wet salt water (Figure 15), resulting in expansion stress, which leads to the generation of microcracks on the pore wall. Micro cracks are the beginning of failure [30,31]. With the increase of erosion age, the opening voids will run through other voids from the outside to the inside, resulting in the decrease of the number of voids and the increase of opening voids, as illustrated in Figure 16. As mentioned in section 4.3, the increase of opening voids directly leads to the deterioration of an asphalt mixture performance. Meanwhile, the early salt erosion deterioration of an asphalt mixture performance is due to the expansion stress produced by the crystallization of salt on the pore wall, which is likely to lead to the expansion of the opening voids to the interior of an asphalt mixture, and then connect with other voids, which aggravates the deterioration of an asphalt mixture performance. In high-humidity coastal areas, the early damage of an asphalt mixture exposed to salt may be mainly physical damage.     On the whole, the results of void characteristics may indicate that salt crystallizes on the pore wall of the opening space in an asphalt mixture under the condition of dry and wet salt water (Figure 15), resulting in expansion stress, which leads to the generation of microcracks on the pore wall. Micro cracks are the beginning of failure [30,31]. With the increase of erosion age, the opening voids will run through other voids from the outside to the inside, resulting in the decrease of the number of voids and the increase of opening voids, as illustrated in Figure 16. As mentioned in section 4.3, the increase of opening voids directly leads to the deterioration of an asphalt mixture performance. Meanwhile, the early salt erosion deterioration of an asphalt mixture performance is due to the expansion stress produced by the crystallization of salt on the pore wall, which is likely to lead to the expansion of the opening voids to the interior of an asphalt mixture, and then connect with other voids, which aggravates the deterioration of an asphalt mixture performance. In high-humidity coastal areas, the early damage of an asphalt mixture exposed to salt may be mainly physical damage.  On the whole, the results of void characteristics may indicate that salt crystallizes on the pore wall of the opening space in an asphalt mixture under the condition of dry and wet salt water (Figure 15), resulting in expansion stress, which leads to the generation of microcracks on the pore wall. Micro cracks are the beginning of failure [30,31]. With the increase of erosion age, the opening voids will run through other voids from the outside to the inside, resulting in the decrease of the number of voids and the increase of opening voids, as illustrated in Figure 16. As mentioned in Section 4.3, the increase of opening voids directly leads to the deterioration of an asphalt mixture performance. Meanwhile, the early salt erosion deterioration of an asphalt mixture performance is due to the expansion stress produced by the crystallization of salt on the pore wall, which is likely to lead to the expansion of the opening voids to the interior of an asphalt mixture, and then connect with other voids, which aggravates the deterioration of an asphalt mixture performance. In high-humidity coastal areas, the early damage of an asphalt mixture exposed to salt may be mainly physical damage.

Findings and Conclusions
The findings may be summarized as follows: • In this paper, the local CT value Otsu method is proposed to conduct threshold segmentation of CT images of an asphalt mixture under the condition of solution immersion. By using this threshold to reconstruct three-dimensional specimens, the ideal void structure can be obtained.

Research Prospects
• At present, for the damage mechanism of an asphalt mixture under salt erosion environment, this paper studies it just in a short period, and will further simulate half a year or even longer salt erosion ages in the next work. • It is important to realize the virtual test of an asphalt mixture and the prediction of pavement service life under salt erosion environment, which is also the next step of the research.

Findings and Conclusions
The findings may be summarized as follows: • In this paper, the local CT value Otsu method is proposed to conduct threshold segmentation of CT images of an asphalt mixture under the condition of solution immersion. By using this threshold to reconstruct three-dimensional specimens, the ideal void structure can be obtained.

Research Prospects
• At present, for the damage mechanism of an asphalt mixture under salt erosion environment, this paper studies it just in a short period, and will further simulate half a year or even longer salt erosion ages in the next work. • It is important to realize the virtual test of an asphalt mixture and the prediction of pavement service life under salt erosion environment, which is also the next step of the research.

Findings and Conclusions
The findings may be summarized as follows: • In this paper, the local CT value Otsu method is proposed to conduct threshold segmentation of CT images of an asphalt mixture under the condition of solution immersion. By using this threshold to reconstruct three-dimensional specimens, the ideal void structure can be obtained.

Research Prospects
• At present, for the damage mechanism of an asphalt mixture under salt erosion environment, this paper studies it just in a short period, and will further simulate half a year or even longer salt erosion ages in the next work.

•
It is important to realize the virtual test of an asphalt mixture and the prediction of pavement service life under salt erosion environment, which is also the next step of the research.