Numerical-Experimental Approach to Characterize Fracture Properties of Asphalt Mixtures at Low Temperatures

Numerical–Experimental Approach to Characterize Fracture Properties of Asphalt Mixtures at Low Temperatures Francisco Thiago Sacramento Aragão, Diego Arthur Hartmann, Yong-Rak Kim, Laura Maria Goretti da Motta, and Mohammad Haft-Javaherian the significant heterogeneity of the mixture microstructures; and the complex geometric characteristics of aggregate particles and their random distribution within an asphalt mixture. In an attempt to model the fracture process in asphalt mixtures, many researchers have adopted the microstructural modeling approach as an efficient and promising tool (1–12). Unlike other methodologies, microstructural modeling allows the characterization of the mechanical behavior of the composite from individual characteristics of the constituents, such as constitutive behavior, geometry, and volume fraction in the mixtures. This characteristic of microstructural models can lead to significant savings in time and the cost spent on experimental efforts, because the laboratory characterization of constituents in general requires less material and has its basis in faster procedures than the characterization of the entire composite. The success of computational models on the basis of the microstructural approach is directly related to the level of understanding of the physics of the problem modeled and how the model simulates these physical characteristics. Moreover, the proper characterization of the constituents is essential to obtain representative input parameters for the models. Several protocols have been proposed to characterize the fracture process in asphalt mixtures from experimental procedures. Among these protocols, the most commonly used are probably (a) the singleedge notched beam [SE(B)] test, (b) the disk-shaped compact tension test [DC(T)], (c) the indirect tension test, and (d) the semicircular bend (SCB) test. To obtain fracture characteristics, Wagoner et al. proposed a testing protocol that used the SE(B) geometry (13). Song et al. (6) and Kim et al. (14) used the methodology proposed by Wagoner et al. (13) to calibrate fracture parameters used in their microstructural finite element method and discrete element method simulations, respectively. The main problems with the use of SE(B) geometry to routinely obtain fracture properties of asphalt mixtures are that the fabrication of specimens in the laboratory becomes impractical and that often it is not viable to extract beam specimens from mixtures in the field. With the limitations of the SE(B) geometry, Wagoner et al. proposed a testing protocol with a DC(T), which was similar to the DC(T) geometry proposed in ASTM E399 but with a longer precrack to avoid premature fracture at the loading holes (15, 16). The new testing configuration was later standardized in ASTM D7313. Further studies have combined the fracture energy from the DC(T) tests conducted as proposed by Wagoner et al. (15, 16) (especially at low temperatures) with the strength of the asphalt mixtures obtained from indirect tension test specimens to perform their fracture ­simulations (6).

Fracture damage mechanisms are some of the most significant causes of structural failure in asphalt mixtures. Yet much research is still needed to understand properly the fracture process in such complex materials. The study reported in this paper investigated several experimental testing protocols available in the literature to characterize fracture properties of asphalt mixtures. Two bending tests (i.e., semicircular bending, single-edge notched beam) and one tension test (disk-shaped compact tension) were performed. An integrated approach that combined experimental tests and numerical simulations was applied to characterize fracture properties of a fine aggregate mixture. The experimental tests were simulated with a computational model on the basis of the finite element method, which was incorporated with material viscoelasticity and cohesive zone fracture. Two cohesive zone fracture parameters (i.e., cohesive strength, fracture energy) were determined through a calibration process until a good match between experimental and numerical results was observed. To illustrate the efficiency of the integrated numerical–experimental approach, fracture properties also were determined through a traditional methodology that used globally averaged material displacements far from the actual fracture process zone. The results indicated that different fracture properties at low temperatures might be obtained from simulations of a single test, regardless of the sample geometry or loading configuration. Through further testing and analysis, it is expected that the modeling approach employed in this work can provide meaningful insights into the effects of constituents on an overall mixture’s ­performance, with significant savings in experimental cost and time.

Cracking is a major source of deterioration of asphaltic mixtures and pavements. To model the initiation and coalescence of microcracks and the propagation of the resulting macrocracks is a challenging task, given several complex characteristics inherent in these ­materials, such as the inelastic constitutive behavior of the asphalt matrix, which is composed of fine aggregates, asphalt binder, and entrained air voids; F. T. S. Aragão, D. A. Hartmann, and L. M. G. da Motta, Laboratório de Geotecnia/ Pavimentos, Programa de Engenharia, Civil, Universidade Federal do Rio de Janeiro/COPPE, Avenida Pedro Calmon, s/n, Cidade Universitária, Ilha do Fundão, Rio de Janeiro 21941-596, Brazil. Y.-R. Kim, 362N, and M. Haft-Javaherian, 362D, Whittier Research Center, Department of Civil Engineering, University of Nebraska, Lincoln, NE 68583. Corresponding author: F. T. S. Aragão, fthiago@ coc.ufrj.br. Transportation Research Record: Journal of the Transportation Research Board, No. 2447, Transportation Research Board of the National Academies, Washington, D.C., 2014, pp. 42–50. DOI: 10.3141/2447-05 42

Aragão, Hartmann, Kim, da Motta, and Haft-Javaherian

Another geometry that has been used in fracture tests of asphalt mixtures is the SCB (12, 17–21). SCB testing is more advantageous than other types of fracture tests because of its relatively simple testing configuration, more economical aspects in specimen fabrication (two testing specimens are produced from one cylinder sample, which can be obtained easily from a Superpave® gyratory c­ ompactor and field cores), and its repeatable testing results. Despite the limitations reported in the literature (e.g., the relatively small potential fracture area and the arching effects that result from the compressive stress state on the upper part of the samples), it has been indicated that as more knowledge is accumulated of the SCB testing characteristics, the simplified testing protocol can in fact become a promising and efficient methodology to routinely obtain fracture parameters of asphalt mixtures (16). Recently, Aragão and Kim proposed a numerical–experimental methodology on the basis of the finite element method and on the cohesive zone fracture concept to determine damage characteristics from the simulations of SCB tests (17). According to their study, that approach appears to be more theoretically sound than traditional methodologies that determine fracture properties, such as fracture energy, from globally averaged material displacements. This procedure may not be appropriate, because it erroneously includes other sources of energy dissipation in the calculation of the energy that should solely characterize the fracture process. As mentioned, several tests have been proposed to characterize the fracture damage behavior of asphalt mixtures. However, further investigation of appropriate procedures is required to analyze the information obtained from the experimental fracture tests and to accurately determine fracture properties. Ideally, such procedures should characterize the material damage characteristics locally at the fracture process zone. In addition, it is desirable that the testing and analysis protocols be designed in such a way that different fracture properties can be identified from a single fracture test that is simple and efficient. The study reported here explored the numerical–experimental approach proposed by Aragão and Kim to analyze and compare the fracture damage characteristics of asphalt mixtures tested with

43

several testing configurations [e.g., SCB, SE(B), DC(T)] (17). The ­specific objectives of this study were as follows: • To explore different Mode 1 fracture tests typically used in the literature to characterize the fracture damage behavior of asphalt mixtures in low-temperature conditions, • To compare the recently proposed numerical–experimental procedure with traditional methodologies that use globally averaged testing measurements in the calculation of fracture properties, and • To investigate the potential of the numerical–experimental approach to provide multiple fracture properties from simulations of a single laboratory test. Materials and Mixture As mentioned, computational microstructure models have been regarded as promising and efficient tools to predict the mechanical behavior of asphalt mixtures. Such models in general rely on mixture constituent characteristics (e.g., their constitutive behavior, geometry, volume fraction) to predict the overall behavior of the mixtures. Compared with other modeling methodologies, the computational microstructure approach presents several advantages: it provides detailed information on the distribution of stresses and strains within the mixture microstructure; it appears to be convenient to evaluate the changes in the overall mixture behavior, given any changes in the mixture composition and mechanical and geometric characteristics of mixture components; and it can be incorporated into a modeling framework for the prediction of damage-induced performance behavior of the mixtures, among others. To simulate the mechanical behavior of hot-mix asphalt (HMA) mixtures with the computational microstructure modeling approach, researchers have regarded such mixtures as composite m ­ aterials that consist of coarse aggregates, fine aggregate matrix (FAM) (i.e., asphalt cement, fine aggregate particles, mineral fillers, entrained air voids, optional additives). Figure 1 shows an example of a typical micro­ structure of an HMA paving mixture and identifies the most relevant phases that often are considered in microstructure models. The

Air Voids

Surface Course Base Course

FAM

Subbase Course (Optional) Subgrade (Existing Soil)

Coarse Aggregate Particles FIGURE 1   Three-phase microstructure of typical HMA mixture.

44

Transportation Research Record 2447

Sample Fabrication Figure 3 illustrates the process of sample fabrication and the tests performed in the laboratory for this study. As shown in the figure, cylindrical samples were fabricated with a Superpave gyratory compactor, from which different specimen geometries were extracted:

SGC Sample

FIGURE 3   Sample fabrication and laboratory tests.

Testing Samples

100 Percentage Passing

p­ resence of the FAM introduces a rate and temperature dependence to the mechanical properties of the whole composite because of the visco­elastic behavior of the material (12). The complexity is intensified if highly nonlinear responses, such as cracking, are involved. Because of these complex characteristics, computational microstructural modeling approaches, such as the one currently pursued in this work, in general have considered the FAM to be the most significant constituent of HMA mixtures. Thus the proper understanding and characterization of the constitutive and damage properties of FAM is crucial for its use as an input parameter in microstructural models. In addition, the testing repeatability of FAM is likely to be higher than that of highly heterogeneous HMA mixtures. This study used a FAM composed of a PG 70-28 binder and fine aggregates smaller than 2 mm. The FAM was designed on the basis of the volumetric characteristics of an HMA mixture that contained 4% of air voids (22). Figure 2 shows the gradations of the HMA and the corresponding FAM used in this study.

80 60

HMA FAM

40 20 0 0.01

0.10 1.00 10.00 Sieve Opening (mm) (log)

100.00

FIGURE 2   Gradation curves for HMA and FAM used in study.

small cylindrical cores that were used in the characterization of linear viscoelastic properties of the FAM and SCB, SE(B), and DC(T) specimens for the fracture tests. Superpave gyratory compactor samples of 150 mm in diameter and 50 mm in height were cored to produce the small samples (12 mm in diameter and 45 mm high) for the linear viscoelastic characterization. Other Superpave gyratory compactor samples of 150 mm in diameter and 170 mm in height were sliced to produce DC(T) and SCB samples or cut in the thickness direction to obtain SE(B) specimens.

Laboratory Testing

Aragão, Hartmann, Kim, da Motta, and Haft-Javaherian

45

Slices 25 mm thick were sawed in half to produce the SCB samples. Other slices were used to generate the DC(T) samples, with loading holes positioned as specified by ASTM D731. The SE(B) sample ­fabrication process was the most time-consuming, because the sides of the samples had to be cut carefully to guarantee the parallelism of the three pairs of opposite faces. Those samples measured 150 mm × 60 mm × 25 mm. Fracture tests in conventional asphalt mixtures with the SE(B) geometry in general use beams much larger than the ones fabricated in this study. According to Wagoner et al., knowledge about the size of the fracture process zone in asphalt mixtures has not been consolidated (13). It is common sense, however, that the size of the damage zone is a function of several variables and characteristics of the testing materials (e.g., temperature, crack propagation rate, air voids, binder ductility and content, maximum aggregate size). One hypothesis typically adopted is that the asphalt concrete sample must have a minimum dimension at least three or four times greater than the maximum aggregate size to ensure statistical validity of the results. Because most of the mixtures have a maximum aggregate size of 19 mm or less, a ligament of at least 76 mm may be adequate. In this study, however, only fine aggregates (