A study on fatigue damage of asphalt mixture under different compaction using 3D-microstructural characteristics

Abstract

The aim of this paper is investigating the microstructural characteristics of asphalt mixture under different compaction powers. In order to achieve this aim, a test track was built to provide asphalt mixture specimens and X-ray computed tomography (XCT) device was used to scan the internal structure. The aggregate particles and air-voids were extracted using Digital Image Processing (DIP), so the relationship between compaction and air-voids was determined at first, and then, the effect of aggregate particles on the morphology of air-voids can be evaluated, finally, fatigue properties of asphalt mixture with different air-void ratio were measured by indirect tensile fatigue test as well. The research results release the distribution of microstructures in asphalt pavement. 3D fractal dimension is an effective indicator to quantize the complexity of aggregate particles and air-voids; suffering the same compaction power, aggregates cause different constitutions of air-voids in asphalt mixture; investigation in this paper can present the essential relationship between microstructures and fatigue properties.

Keywords

asphalt mixture / microstructure / morphology / digital image processing / fatigue damage

Introduction

Air-voids have significant influence on the properties of asphalt pavement, they causes the fatigue damage under repeated load and aggravates the strength of asphalt mixture, causing the macro crack appears in asphalt pavement. In fact, asphalt mixture is a composite material that contains aggregates, air-voids and asphalt mastic, the interactions between them are complicated. Air-void characteristics, such as distribution, shape and volume, are affected by aggregates and asphalt mastic obviously. Therefore, determining the effect of aggregates on air-void characteristics can improve the understanding about microstructural impact for asphalt mixture.

In recent years, the microstructure has become an important research field in pavement engineering [1], mostly because it can explain the essential reason of pavement distress. Using X-ray Computed Tomography (XCT) apparatus, internal structures of asphalt mixture can be scanned and presented by CT gray image, additionally, Digital Image Processing (DIP) is used to extract interesting objectives, so the influence of microstructure on asphalt mixture can be investigated [2–8]. Previous research has shown that air-voids have significant influence on the strength and durability of asphalt mixture [9]. XCT scanning technology provides a useful method to understand the change of internal structure before and after damage, such as permanent deformation and fatigue damage, and the mechanisms that air-void affects the mechanical behavior of asphalt pavement are presented as well [10–13]. Coleri used XCT apparatus to scan the internal structure of asphalt pavement before and after high temperature deformation, then the air-voids change were measured and the effect of air-voids on rutting distress was evaluated [14,15], Sefidmazgi also analyzed the influence of internal microstructure on anti-deformation performance of asphalt mixture using XCT images and DIP [16]. Khan developed a 3D microstructure numerical model according to XCT data, indicating that air-voids can cause stress concentration, and the moisture damage in asphalt mixture was investigated [17,18]. Due to the action of stress concentration in asphalt mixture, air-voids cause cracks easily, in another word, air-voids deteriorate the mechanical performance of asphalt mixture [19,20].

In addition to construction condition, such as compaction power and paving temperature, the morphology and distribution of air-voids are affected by asphalt mastic and aggregates, Tashman’s research has proved that fine aggregates play an important role in setting the air-voids distribution [21]. However, research about the influence of coarse aggregates on air-voids is rare because it is hard to find a quantitative parameter to represent the complexity of aggregates, and the investigations about the relationship between microstructures and fatigue damage are less.

In this paper, the relationships between aggregates and air-voids under the different compaction power are analyzed. The spatial characteristics of microstructure are investigated using DIP, and 3D Box-counting method was developed to calculate the 3D fractal dimension of aggregates and air-voids. Besides, the comprehensive effects of aggregates and air-voids on fatigue performance were evaluated.

Experiment and XCT scanning

Gradation design and test pavement construction

Initially, a Stone Mastic Asphalt (SMA) mixture with a nominal maximum aggregate size (NMAS) of 12.5 mm and bitumen of 6.9% was prepared and named as SMA-11S. The NMAS is defined as one sieve size larger than the first sieve to retain more than 10% of the aggregates. The SMA-11S mixture was prepared using a PG 50/70 binder and diabase aggregates, the asphalt mixture was mixed and paved at temperature of 170°C, test specimens of SMA-11S were fabricated by Marshall method, the gradation of the diabase aggregates are shown in Fig. 1 and critical material parameters are listed in Table 1.

Field test track was built at Institute of Highway Engineering of RWTH Aachen University and named as ISAC test track, as shown in Fig. 2. Diabase aggregates were mixed at temperature of 170°C. A miniature paver and roller were used to construct the pavement, guaranteeing the test track can reflect the internal structures of real pavement. The length, width and thickness were 26 m, 1.2 m and 0.3 m respectively, and pavement was paved on the graded aggregate base. The pavement was composed of three layers, thickness of each layer is 0.12 m before compression and 0.1 m after compression, so the lower layer suffered the greatest compaction power. In order to acquire the test specimens suffered different compaction power, cylindrical cores, which size were 150 mm in diameter and 300 mm in height, were drilled at different sections of test track after maintenance. Each cylindrical core was sliced into five test specimens which size were 40 mm in height and 100 mm in diameter according to the demand of FGSV-Nr.430 of Germany, and the same compaction power applied on different test specimens if them located at the same depth of test track, this operation is beneficial to evaluate the influence of aggregates on air-voids.

Fatigue test and XCT scanning

The characteristics of air-voids, such as morphology and distribution, have an obvious impact on strength of asphalt mixture, and those characteristics affect the fatigue performance as well in addition to aggregates. In this paper, influence of microstructures on fatigue damage was investigated, namely relationship between fatigue damage and microstructures was researched.

The fatigue test was conducted by Universal Test Machine (UTM-25), as shown in Fig. 3. Test temperatures were −10°C, 0°C and 10°C, so the fatigue damage at low temperature range can be measured. Stress control mode was utilized in test, the frequency of load, which follows the sinusoidal wave, was 10 Hz, and the range of load was 0.035 MPa–0.5 MPa to ensure the peak value of horizontal strain of test sample from 0.05‰ to 0.3‰. The test device is shown in Fig. 3(b). In order to evaluate the influence of microstructures on fatigue damage, test time for each specimen was 2 h.

XCT device was used to scan the test specimens before and after fatigue damage, then, the change state of internal microstructures can be analyzed. The scanning apparatus is Y.CT Precision S mode XCT device, as shown in Fig. 4.

Gray image can be obtained from XCT device directly, the resolution was 1024×1024 Pixel, and the size of each pixel was 80mm. Scanning interval of 0.1 mm was set to assure the precision of gray image, so four hundred gray images could be obtained from each test specimen. The gray images of asphalt mixtures that located in different depth of pavement are shown in Fig. 5.

Digital image processing

The main function of DIP technology used in gray image of asphalt mixture are extraction and analysis for microstructures, and the critical point of whole research system is to identify the different microstructures accurately. The Artificial Neural Network (ANN) has been developed to extract microstructures [22,23]. Besides, using the Otsu method, the gray image of asphalt mixture can be processed and transfer to binary image, so the different microstructures, such as aggregate particles and air-voids are identified [24,25]. The basic principle of Otsu divides the image into target and background, and the least squares method was used to calculate the optimal segmentation threshold. The maximum variance of two groups means minimum probability.

Assuming X is an image with L gray level, the amount of pixels which has i gray level is N_i and the total number of pixels in the whole image is represented by N, therefore, the probability density function is shown in Eq. (1):

(1)P_i=N_i/N.

Selecting a threshold k, the pixels of image can be divided: C₀ is a group of pixels with a gray level of [0, 1,…,k−1], and C₁ is a group of pixels with a gray level of [k,k−1,…,L−1]. The definition of the maximum between-class varianced²(k) is represented by Eq. (2):

(2){\delta }^{\mathrm{2}}(k)={\omega }_{\mathrm{0}}{(\mu -{\mu }_{\mathrm{0}})}^{\mathrm{2}}+{\omega }_{\mathrm{1}}{(\mu -{\mu }_{\mathrm{1}})}^{\mathrm{2}},

where, m is average gray level of whole image, w₀, m₀, w₁, and m₁ represent pixel proportion and average gray level of C₀ and C₁ respectively:

(3)\mu ={\displaystyle \sum _{i=\mathrm{0}}^{L-\mathrm{1}}i{P}_i},{\omega }_{\mathrm{0}}={\displaystyle \sum _{i=\mathrm{0}}^{k-\mathrm{1}}{P}_i},{\mu }_{\mathrm{0}}={\displaystyle \sum _{i=\mathrm{0}}^{k-\mathrm{1}}i{P}_i}/{\omega }_{\mathrm{0}},{\mu }_{\mathrm{1}}={\displaystyle \sum _{i=k}^{L-\mathrm{1}}i{P}_i}/{\omega }_{\mathrm{1}},{\omega }_{\mathrm{1}}=\mathrm{1}-{\omega }_{\mathrm{0}}.

k varies from 0 to L−1, then, the corresponding d²(k) is calculated and the maximum one is the optimal segmentation threshold.

The scanning mechanism of XCT mainly depends on the material’s density. Materials can be identified because each them has an especially density, and then, densities determine the different gray values. Therefore, CT gray image shows the internal structures of asphalt mixture through different gray values that represent aggregates, air-voids and asphalt mastic, respectively. However, gray values distribution of original XCT images shows that gray value is not homogeneous but gradually increases from image center to boundary due to ray attenuation, scatter and noise, as shown in Fig. 6, therefore, the greater the radius, the greater the brightness from a visual point of view is.

Otsu cannot manage the problem that uneven gray distribution of XCT image scanned from asphalt mixture. In this research, an improved Otsu method was developed to extract the binary image of microstructures. The processing steps are listed below:

1) Due to the feature of gray distribution shown in Fig. 6, the original CT image was divided into annular regions with different radius byMatlab application, as illustrated in Fig. 7(a). A contrast enhancement program was compiled and used in each annular region to improve the range of gray value that belongs to different microstructures. Contrast of annular region after contrast enhancement is shown in Fig. 7(b), it shows that different microstructures can be identified effectively using an appropriate threshold.

2) According to Fig. 7(b), although the gray distribution of internal structures of asphalt mixture is confusion, for each annular region, the gray distribution is even, and two thresholds used to separate air-voids, asphalt mastic and aggregates are determined easily. In this paper, aggregates, air-voids and asphalt mastic of each annular region were extracted by Otsu respectively. In this step, each annular region was conducted individually because gray values in the same annular region are uniform, and then, the binary components obtained from each annular region were assembled into a whole binary structure. The binary images are presented in Fig. 8.

3) The neighboring aggregate particles are connected with each other in binary images due to the similar gray values. Watershed algorithm base on distance transform was used to separate the connection.

Discussion and analysis

The feature of air-void distribution

The binary image consists of pixels which size is 80mm, therefore, the air-voids area at different depths of pavement can be calculated by the quantity of pixel belonged to air-voids, and change of air-voids ratio are inferred, as shown in Fig. 9.

Figure 9 indicates the range of air-voids ratio along with the pavement. Due to different compaction powers, air-voids ratio decrease along with the depth of asphalt pavement, air-voids ratio are decreased significantly during the range 0 mm to 150 mm, especially in the range from 90 mm to 150 mm, all compaction degrees (Ratio of mixture’s density to mixture’s maximum density) of asphalt mixtures are increased obviously. The curves show that discrepancies of air-voids ratio nearby the pavement surface are great, such as the air-voids ratio at depths of 30 mm and 90 mm. However, air-voids ratio present similar value at depth of 150 mm.

To better understand the influence of compaction power, compaction degree is used to quantify the compaction power at different depths of pavement. Selecting three specimens drilled from different positions of test track to determine the relationship between depth of pavement and compaction degree, the results are shown in Fig. 10.

It can be concluded that, due to the construction method, the lower layer of test track suffers more compaction power, so compaction degree is greater than that of other layers. Therefore, more compaction power cause greater compaction degree, the air-voids are less because asphalt mixture is compressed.

Spatial morphology of aggregates and air-voids

The morphology and distribution of microstructures are very complicated, the geometrical method cannot describe the randomness and complexity of microstructures. Fractal dimension is regarded as an effective index to represent the morphology characteristics of irregular particles or structures, such as the micropattern of aggregate particles of asphalt mixture [26] and crack distress of cement concrete [27,28]. Box-counting is an important method of calculating the fractal dimension of plane image, however, most fractal dimension is mainly used to evaluate the morphology of 2D particles or structures, and it cannot reflect the spatial morphology. In order to provide a quantitative value to represent the complexity and distribution characteristics of microstructures, 3D Box-counting method was developed byMATLAB, as shown in Fig. 11.

Traditionally, fractal dimension is a non-integer number that measures the degree of fractal boundary fragmentation or irregularity over multiple scale. In this paper, cubes with side lengthr are used to divide the spatial model of microstructures into different portions, as shown in Fig. 11(a); the number of cubes which contain the microstructures isN(r), and N(r) is changed with the change of side r, as shown in Fig. 11(b) and Fig. 11(c). Least square method is utilized to fit the data of r and N(r) at double logarithmic coordinate system, and then, the slope of fitting curve D determined by Eq. (4) can be used to represent the complexity of microstructures:

(4)D=\underset{r\to \mathrm{0}}{\lim }{\displaystyle \frac{\log (N(r))}{-\log (r)}}.

Selecting fifteen specimens as research objectives, the 2D binary images of aggregates in each specimen were extracted at first and a spatial model was created using the squeeze of binary images byMATLAB application, and then, the 3D fractal dimension of aggregates can be calculated. The results are listed in Table 2.

Due to the same raw materials and gradation state, it can imagine that the constitution, morphology and distribution of aggregates in different asphalt mixtures are similar, meanwhile, it can be founded from Table 2 that fractal dimension of aggregates in different specimens and depths are similar as well. This phenomenon shows fractal dimension is an effective indicator to evaluate the internal structures of asphalt mixture, it provides a quantitative value to explain the complexity of internal structures.

Figure 12 show the relationship between 3D fractal dimension of air-voids and logarithm of corresponding volume and surface area respectively. Due to the morphology and distribution of air-voids, the 3D fractal dimensions of air-voids in different specimens show great discrepancy, nearly from 1.6 to 2.3. 3D fractal dimension increases with the increase of volume and surface area, the fitting curves show linear variation tendency. Besides, it can be concluded that the dispersion between 3D fractal dimension and volume illustrated in Fig. 12(a) is less compared with that shown in Fig. 12(b), indicating that 3D fractal dimension can describe the volume characteristics obviously.

In order to determine the influence of aggregates on air-voids, the 3D fractal dimension of aggregates and air-voids are compared and shown in Fig. 13.

Figure 13 presents the relationship of 3D fractal dimension between aggregates and air-voids. The same depth of test pavement means an identical compaction power, so the aggregates is the most important factor for air-voids shaping. 3D fractal dimension of air-voids increases with the increase of 3D fractal dimension of aggregates, and the relationship approximatively follows linear tendancy. The result shows that morphology and distribution of aggregates have obvious influence on airvoids.

The influence of microstructures on fatigue damage

Aggregate particles have a significant impact on performance of asphalt mixture, the stress state is mainly affected by aggregates. Furthermore, air-voids can be regarded as initial faults caused by the interaction between aggregate particles and asphalt mastic during molding. Therefore, Coupling effect of aggregates and air-voids may lead to the fatigue damage of asphalt mixture. In this paper, total fractal dimension which equal to the sum of fractal dimension of aggregates and air-voids was assumed to represents the coupling effect, and the changes of air-voids ratio (mainly caused by the cracks distress) before and after fatigue test were measured as well, as shown in Fig. 14.

Figure 14 shows that total fractal dimension is an effective indicator and can evaluate the fatigue damage of asphalt mixture, namely the asphalt mixture which has a great total fractal dimension may appear more cracks after fatigue test. Moreover, the discreteness decreases with the increase of test temperature, it is probably that asphalt mastic presents great strength at low temperature, and the fatigue damage should be determined by the comprehensive effect of aggregates, air-voids and asphalt mastic.

Summary and conclusions

This paper study the influence of aggregates on air-voids under different compaction powers, XCT device was used to scanned the gray images of internal structures, and the relationship between microstructural characteristics and fatigue damage was evaluated. Results of this study can be summarized as follows.

1) The air-voids ratio decrease along with the depth of asphalt pavement, especially at the range from 90mm to 150mm, the asphalt mixture nearby the pavement surface presents a adverse fatigue performance because of high porosity.

2) 3D Box-counting method developed in this paper can provide an effective index to describe the morphology and distribution of microstructures; 3D fractal dimensions of different aggregates show similar value but that of air-voids presents obvious discrepancy due to the different air-voids ratios.

3) The aggregates have significant impact on the air-voids shape, aggregates with great 3D fractal dimension causes complex air-voids easily under the same compaction power.

4) Comprehensive effect of aggregates and air-voids on fatigue performance is important, the total 3D fractal dimension can be used to evaluate the fatigue damage.

References

[1] Hou Y, Wang  L, Pauli T ,  Sun W. Investigation of the asphalt self-healing mechanism using a phase-field model. Journal of Materials in Civil Engineering, 2015, 27(3): 04014118

[2] Apeagyei A K, Grenfell  J R A, Airey  G D. Observation of reversible moisture damage in asphalt mixtures. Construction & Building Materials, 2014, 60: 73–80

[3] Presti D. L, Hassan  N. A, Khan  R, Airey G. Reclaimed Asphalt aest specimen preparation assisted by image analysis. Journal of Materials in Civil Engineering, 2014, 27(8): C4014005

[4] Mohajeri M, Molenaar  A A A, Van de Ven  M F C. Experimental study into the fundamental understanding of blending between reclaimed asphalt binder and virgin bitumen using nanoindentation and nano-computed tomography. Road Materials and Pavement Design, 2014, 15(2): 372–384

[5] Arega Z A, Bhasin  A, Kesel T D . Influence of extended aging on the properties of asphalt composites produced using hot and warm mix methods. Construction & Building Materials, 2013, 44: 168–174

[6] Braz D, Barroso  R C, Lopes  R T, Motta  L M G. Crack detection in asphaltic mixtures by computed tomography. NDT & E International, 2011, 44(2): 195–201

[7] Moriyoshi A, Takahashi  N, Ikeda O ,  Kawashima M ,  Akabane T . Strain distribution in asphalt mixtures during the wheel tracking test at high temperatures. Construction & Building Materials, 2013, 40: 1128–1135

[8] Liu P, Hu  J, Wang D ,  Oeser M ,  Alber S ,  Ressel W ,  Canon Falla G . Modelling and evaluation of aggregate morphology on asphalt compression behavior. Construction & Building Materials, 2017, 133: 196–208

[9] Castillo D, Caro  S, Darabi M ,  Masad E . Studying the effect of microstructural properties on the mechanical degradation of asphalt mixtures. Construction & Building Materials, 2015, 93: 70–83

[10] Aliha M R M ,  Fazaeli H ,  Aghajani S ,  Moghadas Nejad F . Effect of temperature and air void on mixed mode fracture toughness of modified asphalt mixtures. Construction & Building Materials, 2015, 95: 545–555

[11] Yan K, Ge  D D, You  L Y, Wang  X L. Laboratory investigation of the characteristics of SMA mixtures under freeze–thaw cycles. Cold Regions Science and Technology, 2015, 119: 68–74

[12] Khosravi H, Abtahi  S M, Koosha  B, Manian M . An analytical–empirical investigation of the bleeding mechanism of asphalt mixes. Construction & Building Materials, 2013, 45: 138–144

[13] Kodippily S, Henning  T F P, Ingham  J M, Holleran  G. Computed tomography scanning for quantifying chipseal material volumetrics. Journal of Computing in Civil Engineering, 2014, 28(3): 04014002

[14] Coleri E, Harvey  J T, Yang  K, Boone J M . A micromechanical approach to investigate asphalt concrete rutting mechanisms. Construction & Building Materials, 2012, 30: 36–49

[15] Coleri E, Harvey  J T, Yang  K, Boone J M . Micromechanical investigation of open-graded asphalt friction courses’rutting mechanisms. Construction & Building Materials, 2013, 44: 25–34

[16] Sefidmazgi N. R ,  Tashman L ,  Bahia H. Internal structure characterization of asphalt mixtures for rutting performance using imaging analysis. Road Materials and Pavement Design, 2012, 13(sup1 S1): 21–37

[17] Khan R, Grenfell  J, Collop A ,  Airey G ,  Gregory H . Moisture damage in asphalt mixtures using the modified SATS test and image analysis. Construction & Building Materials, 2013, 43: 165–173

[18] Khan R, Grenfell  J, Collop A . Micromechanical modelling of typical Dense Bitumen Macadam (DBM) from laboratory testing and X-ray Computed Tomography (CT). Construction & Building Materials, 2015, 77: 1–6

[19] Zhang Y Q, Luo  X, Luo R ,  Lytton R L . Crack initiation in asphalt mixtures under external compressive loads. Construction & Building Materials, 2014, 72: 94–103

[20] Liu P, Wang  D, Otto F ,  Hu J, Oeser  M. Application of semi-analytical finite element method to evaluate asphalt pavement bearing capacity. International Journal of Pavement Engineering, 2017 

[21] Tashman L, Pearson  B. Characterisation of stone matrix asphalt mixtures. International Journal of Pavement Engineering, 2012, 13(4): 297–309

[22] Zelelew H M, Papagiannakis  A T. A volumetrics thresholding algorithm for processing asphalt concrete X-ray CT images. International Journal of Pavement Engineering, 2011, 12(6): 543–551

[23] Kutay M E, Arambula  E, Gibson N ,  Youtcheff J . Three-dimensional image processing methods to identify and characterise aggregates in compacted asphalt mixtures. International Journal of Pavement Engineering, 2010, 11(6): 511–528

[24] Xu X Y, Xu  S Z, Jin  L H, Song  E. Characteristic analysis of Otsu threshold and its applications. Pattern Recognition Letters, 2011, 32(7): 956–961

[25] Mizushima A, Lu  R. An image segmentation method for apple sorting and grading using support vector machine and Otsu’s method. Computers and Electronics in Agriculture, 2013, 94: 29–37

[26] Arasan S, Yener  E, Hattatoglu F ,  Akbulut S ,  Hinislioglu S. The relationship between the fractal dimension and mechanical properties of asphalt concrete. International journal of civil and structural engineering, 2010, 1(2):165–70

[27] Baer J U, Kent  T F, Anderson  S H. Image analysis and fractal geometry to characterize soil desiccation cracks. Geoderma, 2009, 154(1–2): 153–163

[28] Farhidzadeh A, Dehghan-Niri  E, Moustafa A ,  Salamone S ,  Whittaker A . Damage assessment of reinforced concrete structures using fractal analysis of residual crack patterns. Experimental Mechanics, 2013, 53(9): 1607–1619

RIGHTS & PERMISSIONS

Higher Education Press and Springer-Verlag Berlin Heidelberg