The base asphalt was 90# Kunlun asphalt, and the relevant performance tests were conducted strictly according to Chinese standard E20-2011, Standard Test Methods of Bitumen and Bituminous Mixtures for Highway Engineering (Code of China, 2011). The results are listed in Table 1.

In recent years, natural asphalt has been extensively studied and widely used in asphalt pavements. To explore more substitutes with high yield, stable properties, and low cost, Chinese Qingchuan rock asphalt (QRA) was employed for the preparation of gussasphalt in this study, according to T0614-2011 and T0618-1993 of Chinese standard E20-2011 (Code of China, 2011). The ash content of QRA was 4.5%, asphalt aromatic was 31.49%, saturate was 29.45%, resin was 34.44% and asphaltene was 4.62%.

Different modifying agents were adopted in this study (Table 2), it is widely acknowledged that Styrene-butadiene-styrene (SBS) can enhance the elasticity of binder at high temperatures and increase flexibility of binder at low temperatures, terpene resin (TR) is easy to mix with asphalt after heating and melting, and during this process, the initial adhesion and aging resistance of asphalt were greatly improved as well. Furfural extraction oil (FEO) can improve the elongation of asphalt at low temperature (Liu et al., 2018; Bai et al., 2019). It can be used as a blending component in asphalt to optimize the composition and make thermoplastic elastomers more compatible in asphalt. SAW modifier can reduce the high-temperature asphalt viscosity and the mixing temperature, and improve the high-temperature performance of asphalt mix.

The CNAMGA binder was heated to a flowing state, and a drop of the sample was taken with a needle and placed on a glass slide washed with absolute ethyl alcohol. Cover slips were gently placed on the slide, which was placed in an oven at 135°C for 10 min. After removing and cooling, a suitable multiple objective lens (×400) was selected for microscopic observation.

Based on AASHTO T 315 (American Association of State Highway and Transportation Officials, 2012), dynamic shear rheometer (DSR) tests were conducted with a Gemini II ADS dynamic shear rheometer, selecting the strain-controlled mode with a frequency of 10 rad/s (1.59 Hz). Five test temperatures from 64 to 88°C at increments of 6°C were applied in the Superpave grading system. The applied strain values, γ, for the original specimen, rolling thin film oven test (RTFO-aged) specimen, and pressure aging vessel test (PAV-aged) specimen were 12, 10, and 1%, respectively. The original specimen and RTFO-aged specimen had a diameter of 25 mm and thickness of 1 mm, while those of the PAV-aged specimen were 8 and 2 mm, respectively. The test process was depicted in Figure 1.

Bending beam rheometer (BBR) tests were conducted using a CANNON bending beam rheometer to evaluate the low-temperature rheological properties, in accordance with the procedures outlined in AASHTO T 313 (2016). When the loads were applied to asphalt beams (125 × 6.25 × 12.5 mm) during the detection process at test temperature of −12, −18 and −24°C, the deflection of the beams was recorded, and the deflection–time curve was obtained and analyzed using a software program to calculate creep stiffness S and creep rate m. The test process was depicted in Figure 2.

The specimens were prepared according to the preparation scheme for the CNAMGA binder. Factor A (the QRA content) had three levels (10, 12.5, and 15.0%) and factor B (the composite modifier #1 content) had three levels (0, 3, and 5%), and various combinations were adopted to analyze the binder properties. The interaction between the factors was not considered, as the test arrangement shown in Table 3 indicates.

Asphalt penetration test (25°C, 100 g, 5 s), ductility test (10°C, 5 cm/min), softening point test (R&B), and rolling thin film oven test (RTFOT) (163°C, evaporation for 5 h) were performed according to T0604-2011, T0605-2011, T0606-2011, and T0608-1993 of E20-2011 (Code of China, 2011), respectively. The results are shown in Figure 3.

As the results shown in Figure 3 indicate, when the QRA content is the same, the softening point increases and the mass loss decreases, the penetration increases and then decreases, and the ductility gradually decreases (except for specimen 6) with an increase in the modifier #1 content. Referring to the technical requirements for modified hard asphalt binder in the Technical Guide on Design and Construction of Bridge Deck Paving of Highway Steel Box Girder, the specimen with the 10% QRA content and less than 5% modifier #1 content does not meet the penetration requirements (1–4 mm). The specimen with 10% QRA and 5% modifier #1 or specimen with 15% QRA and any content of modifier #1, does not meet the ductility requirement (≥10 mm). When the QRA content is 7.5% and the modifier #1 content is less than 5%, or the QRA content is 15% and the modifier #1 content is less than 3%, the specimen doesn’t meet the softening point requirement (≥72°C). All the specimens meet the mass-loss requirement (<0.5%). According to the above analysis, specimens with a 12.5% QRA content and 5% modifier #1 content meet the all the index requirements.

Furthermore, given the test results and rules for different collocations of the CNAMGA binder in Figure 3, a variance analysis (Jamshidi et al., 2012) was adopted to investigate the influencing factors of different technical indexes. The results are presented in Table 4.

According to the one-tailed hypothesis test in mathematical statistics, with a confidence interval (α = 0.05), when the factor degree of freedom is 2 and the error degree of freedom is 4, it can be seen from the F distribution table that the critical value is F _0.05 (2, 4) = 6.9. That is, there is a 95% certainty that F ≤ 6.9 will appear, which can be judged to have no significant impact on the experimental index (acceptable region). If F > 6.9 happens with only a 5% certainty, this indicates that the null hypothesis is not valid, and that this factor has a significant impact on the experimental index (rejection region). The critical value is F _0.01 (2, 4) = 18.0, with a confidence interval (α = 0.01). That is, there is a 99% certainty that 6.9 < F < 18.0 will appear, which can be judged to have a significant impact on the experimental index (acceptable region). If F > 18.0 occurs with only a 1% certainty, this indicates that the null hypothesis “this factor has a significant effect on the experimental index” is invalid, and this factor has a strongly significant impact on the experimental index (rejection region).

Note that because 6.9 < F _B < F _A < 18.0 with regard to penetration, it can be considered that the QRA and modifier #1 contents have a significant impact on the penetration of the binder. From F _B < 6.9 < F _A < 18.0 for the ductility, it can be seen that the QRA content has a significant effect, while the modifier #1 content has no significant effect on the ductility. The 6.9 < F _A < 18.0 < F _B for softening point indicates that the QRA content has a significant effect on the softening point, and the modifier #1 content has a strongly significant effect on it. That 18.0 < F _B < F _A for the mass loss illustrates that the QRA and modifier #1 contents have a strongly significant impact on the mass loss.

The modifier #1 has a significant effect on the other indexes, but for the ductility, it is slightly less effective in improving the low-temperature performance of the asphalt. It is speculated that the modifier #1 component may affect the compatibility and performance of the modified asphalt. The collocation of the modifier #1 is adjusted to SBS: SAW: TR: FEO = 2: 2: 0.5: 0.5, the TR content is reduced, and FEO, which can enhance the ductility at low temperature, is added. The objective of this new modifier (#2) is to reduce the consistency, improve the compatibility between the modifier and asphalt, and improve the low-temperature performance of the binder.

The microstructure of the modified asphalt significantly affects the properties of the modified asphalt. To confirm whether modifier #2 could live up to the expectations mentioned above, a fluorescent microscope is employed to observe and analyze the dispersibility and compatibility of the QRA and modifier in the binder. Six specimens (specimens 2’, 3’, 5’, 6’, 8’, and 9’) are remade with modifier #2 according to the collocation in Table 3. They are then subjected to fluorescent microscope tests, and the results are shown in Figure 4.

Specimens presents different swollen polymer-rich phase from the vision of fluorescent microscope. In these pictures, the swollen polymer-rich phase has a bright yellow appearance and the dark region denotes the asphalt-rich phase. The QRA contents of specimens 2’ and 3’ are 10%, those of specimens 5’ and 6’ are 12.5%, and those of specimen 8’ and 9’ are 15%. The modifier #2 contents of specimens 3’, 6’, and 9’ are 5%, and those of specimens 2’, 5’, and 8’ are 3%. The irregular shape of polymer-rich phase in Figure 4A indicates the poor compatibility of the components in specimen 2’. Figure 4C shows that many polymer-rich phase in small size evenly distribute in specimen 5’, implying better swelling comparing to specimen 2’. The distribution of polymers tends to sparse and inconspicuous in Figure 4E, indicating poor compatibility with higher content of QRA, the increasing addition of heavy fractions obstructs polymers swelled by light fractions. Compared with Figures 4A,B depicts the dispersed polymer-rich phase is proned to be spherical but distributes unevenly, implying modifier #2 plays a certain role in compatibility. The compatibility shown in Figure 4D is slightly poorer than that seen in Figure 4C, which is not yet obvious. It is not difficult to see the shape of polymer-rich phase is large in size, indicating agglomeration with higher content of modifier #2 and QRA from Figure 4F, and a potential phase separation may occur. The above phenomena prove that the modifier and QRA contents should coordinate with each other; mismatching will affect the dispersion of the different components in the binder. It can be concluded that the compatibility of the binders with the 12.5% QRA and 3% or 5% content of modifier #2 is better.

The pictures in aforementioned statement are only partial visions under fluorescent microscope, which are representative, while sometimes a partial vision is hard to depict all-inclusive information. Extensive binarization treatment and mathematical statistics in fluorescent photomicrograph can present a more comprehensive result. Image-Pro Plus software (version 6.0), a powerful 2D and 3D image processing software, was employed to implement quantitative analysis. The area of the fluorescent part (that is, the pixels with gray values ≥175) in the photomicrograph was analyzed. Taking specimens 5’ and 6’ as examples, Figure 5 presents the process of binarization treatment.

With binarization treatment, the sparse and uneven distribution of polymer-rich phase in specimen 5’ tends to be more conspicuous than that of specimen 6’. In this way, the proportions of the fluorescence areas in the total areas of observation of different specimens could be quantitatively compared. Samples are observed multiple times using different fields of view, and mathematical statistics are used to determine the results, which are shown in Figure 6.

As shown in Figure 6, with the increase addition of modifier #2 in the specimens with 10% content of QRA and specimens with 15% contents of QRA, the fluorescence spots area gradually decreases, while that of specimens with 12.5% content of QRA increases with the addition of modifier #2. This proves that higher modifier and QRA contents do not yield better results, the role of modifier #2 on the specimens with 12.5% content of QRA can be played to the better extent. Note that the fluorescence spots of the specimen with 12.5% QRA and 5% modifier #2 are the densest and the area proportion is the largest, as a result of the sufficient cross-linking and grafting reactions occurring in the binder. Therefore, the QRA and modifier contents should match each other, or the larger agglomeration or no fluorescent spots can be found under the sampling field. The statistical results show that the compatibility of the binder with the 12.5% QRA content and 5% modifier #2 is superior, which could be considered as the main research object in next characterization of macro-properties.

The shear strength of asphalt concrete mainly depends on the bonding force caused by the interaction between the binder and aggregate and the internal friction angle caused by the intercalation of the aggregates in the asphalt mix. When the temperature goes up and shear stress increases, the asphalt binder viscosity decreases and the flow mobility increases, and the migration of binder to the pavement surface leads to a reduction on the internal friction angle of asphalt mix. The integrity and shear strength of the mix deteriorate, leading to pavement rutting and other pavement diseases (Al-Khateeb and Al-Akhras, 2011; Ali et al., 2017), therefore high-temperature performance is important for the serviceability of asphalt concrete. The rutting factor G*/sinδ can be considered to be a characterization parameter for the high-temperature performance. A larger value indicates a worse high-temperature fluidity, and a higher anti-rutting performance. Based on fluorescent microscope test, the compatibility of the binder with the 12.5% QRA content is regarded as superior, then specimens 4, 5’, and 6’ are subjected to DSR test, and the results are depicted in Figure 7.

As anticipated, complex modulus G* before and after aging decreases with the rising temperature, phase angle δ before and after aging increases with the rising temperature. At the same temperature, the larger G* of the binder with modifier #2 appears than that of binder without modifier #2, indicating that binder with modifier #2 has better deformation resistance, but the G* of binder with 3% modifier #2 before aging has not much difference with that of binder without modifier #2 before aging. And the δ of modified binder is less than that of unmodified binder, the δ increases as the temperature increases for all samples. With the increasing dosage of modifier #2, the δ decreases at the same temperature, which indicates that the ratio of the elastic component in asphalt increases, the elastic recovery performance of asphalt has been improved. According to Superpave^TM, rutting factor G*/sinδ is limited to no less than 1.0 kPa for unaged binders or 2.2 kPa for RTFO-aged binders (Jamshidi et al., 2012; Fazaeli et al., 2016; Kim et al., 2017b). As shown in Figures 7C,D, specimen 4 at 88°C before and after aging doesn’t meet the requirements. Specimen 6’, which has the largest G*/sinδ value, has a superior high-temperature performance. The incorporation of the modifier changes the colloid structure of the asphalt, increasing the asphaltene content, which could harden the asphalt binder. Harder asphalt shows less conversion from elasticity to viscosity at high temperatures, enhancing the resistance to shear deformation at high temperatures. The composition of natural asphalt is similar to that of petroleum asphalt, but the content of hard components (resin and asphaltene) is higher in natural asphalt. Almost every large molecule of asphaltene contains the polar functional groups of carbon, hydrogen, oxygen, nitrogen, and sulfur, which can produce a strong adsorption force on the surface and improve the high-temperature performance of the asphalt. The performance grade (PG) at high or intermediate temperatures is increased by one level. From the above analysis, it can be seen that the high-temperature performance of the binder with 5% modifier #2 is better.

Cracking not only destroys the continuity of pavement and reduces the driving quality, but also poses a water-induced distress concern, leading to the deterioration of the pavement bearing capacity and pavement damage. The low-temperature cracking resistance of asphalt pavement mainly depends on the tensile properties of the binder at low temperatures. In contrast to the high-temperature anti-rutting ability, the shrinkage cracking of asphalt pavement is usually caused by continuous aging, an increase in the stiffness modulus, and the gradual transformation of the low-temperature compliance into brittleness of the binder during its service life (You et al., 2011; Yao et al., 2012, Hui Yao et al., 2013). For comparison with the unadjusted modified binder, modifier #2 is used to make specimens 5’ and 6’ according to the collocation of specimens 5 and 6 with modifier #1, and the test results are shown in Figure 8.

Reflecting the release rate of the shrinkage stress caused by the viscoelastic flow of the asphalt, a larger creep rate m-value indicates a better stress dissipation capacity for the asphalt and its crack resistance. Asphalt with a small creep stiffness (S) has less internal stress, better low-temperature compliance, and less chance of cracking at low temperatures. As shown in Figure 8, with the same QRA content, the S value of the binder-added modifier is larger than that of the binder with no modifier, whereas the m-value of the binder with the modifier is smaller than that of the binder without the modifier. With an increase in the modifier content, S increases, and the m-value decreases. It can be seen from this phenomenon that the binder will be hardened and the stress relief capacity will be reduced after the modifier is added, which will reduce the low-temperature performance of the binder to a certain extent. At the same time, note that the S value of the binder with modifier #2 is smaller and the m-value is larger than that of the binder with the same content of modifier #1. The PG at low temperature is also improved by one level. Modifier #2 plays a role in enhancing the low-temperature performance of the binder, strongly proving that the adjusted modifier composition works effectively.

Based on American Association of State Highway and Transportation Officials (2010) and American Association of State Highway and Transportation Officials (2007), the results of the PG determination of asphalt binder are summarized in Table 5.

Table 5 shows that the addition of modifier #2 improved the PG at high or intermediate temperatures, and the influence of modifier #2 on the PG at low-temperature is stable. Reflecting the sensitivity of the asphalt binder to temperature changes (Rasmussen et al., 2002; Al-Hadidy and Tan, 2009; Walters et al., 2014), the temperature susceptibility (TS) has an important influence on the pavement performance. One of technical indicators of the TS is the penetration.

The penetration index (PI) proposed by Pfeiffer (Pfeiffer and Doormaal, 1936) is calculated by the linear regression of the penetration test results in a temperature range of 15–30°C. Obviously, the PI only reflects the TS at a normal temperature. It is meaningless for high or intermediate temperatures and low temperatures, which significantly affect the rheological properties of the binder. The G*/sinδ value with a change in temperature exactly indicates the TS of the asphalt at high or intermediate temperature ranges, and the S value with a change in temperature exactly reflects the TS value of the asphalt in the low-temperature range. With temperature T as the independent variable, lg (G*/sin) and lgS are considered as dependent variables, and can be defined as follows: l g ( G ∗ / sin ⁡ δ ) = − A 1 T + K 1 , l g S = − A 2 T + K 2 , where A ₁ and A ₂ are susceptibility factors, and K ₁ and K ₂ are regression constants.

Regression fitting was performed on the data of the three specimens, and the results are listed in Table 6.

As shown in Table 6, the fitting degree, R ² , values of all the equations were greater than 0.98, indicating that both G*/sinδ and S were highly correlated with temperature T. By covering the evaluated temperature, it is feasible for the binder to form an evaluation system for the TS. Generally speaking, a smaller absolute value for the susceptibility factor indicates a better resistance to TS. Among the three specimens, the TS at high or intermediate temperatures for specimen 4 is the best, while that at a low temperature is the worst. The TS at high or intermediate temperatures for specimen 5’ is the worst, while that at a low temperature is the best. Combining the above factors, the collocation of specimen 6’ could be identified as a better collocation with a superior comprehensive TS.