- Email: [email protected]
Micromechanical shear modulus modeling of activated crumb rubber modified asphalt cements
Construction and Building Materials 150 (2017) 56–65
Contents lists available at ScienceDirect
Construction and Building Materials journal homepage: www.elsevier.com/locate/conbuildmat
Micromechanical shear modulus modeling of activated crumb rubber modified asphalt cements Jose R. Medina a,⇑, B. Shane Underwood b a b
Arizona State University, 229BA ISTBII, Tempe, AZ 85287, USA Arizona State University, 229B ISTBII, Tempe, AZ, USA
h i g h l i g h t s Absorption of the lighter fractions by the crumb rubber particles. A method to quantify the swelling and splitting of the rubber particles. Use of micromechanical models to accurately predict the dynamic shear modulus of rubberized asphalt cement.
a r t i c l e
i n f o
Article history: Received 9 August 2016 Received in revised form 10 May 2017 Accepted 21 May 2017
Keywords: Crumb rubber Rheology Modified binder Activated rubber Activated mineral binder stabilizer (AMBS) Crumb rubber splitting Composite models Micromechanical models
a b s t r a c t The rheological properties of three asphalt cements containing reacted and activated rubber (RAR) are evaluated to quantify the relative effects of swelling, splitting, and absorption. Rheological testing and electron microscopy are used to measure the dynamic shear modulus, |G⁄|, and rubber particle changes respectively. It is found that |G⁄| increased for all RAR and that particles swell 15.8–49.3% with those in softer asphalt showing the greatest swelling. Micromechanical models are used to predict |G⁄| of the materials. The Hashin and Christensen models are found to accurately predict the measured moduli after accounting for the swelling, splitting, and absorption. Ó 2017 Elsevier Ltd. All rights reserved.
1. Introduction The use of crumb rubber with and without physical and chemical modification to improve the engineering performance of asphalt concrete has been well documented over the last 40 years [1–8]. Within these documented studies, the performance benefits from incorporating rubber are: greater rutting resistance, mitigation of thermal reflective cracking, increased resistance to fatigue cracking, and noise reduction. It is often cited that the reason for the increased performance of these rubber modified asphalts is that during the blending process, the crumb rubber particles absorb the aromatic oils and resins from the asphalt cement leaving the binder with a higher concentration of asphaltenes while at the same time storing the lighter resins for times when they are ⇑ Corresponding author. E-mail addresses: [email protected] (J.R. Medina), [email protected] (B.S. Underwood). http://dx.doi.org/10.1016/j.conbuildmat.2017.05.208 0950-0618/Ó 2017 Elsevier Ltd. All rights reserved.
needed for example to accelerate healing. This diffusion of lighter fractions is postulated to result in swelling of the rubber particles in a time and temperature dependent process that can have a substantial impact on the volume occupation and thus performance [9,10]. The concept of rubber swelling is somewhat controversial. It is documented that crumb rubber particles can swell by 3–5 times their original size [9,11,12], but very little peer reviewed literature has directly quantified the precise volume increase. Kutay et al. quantified the structural change in crumb rubber modified asphalt by using X-ray microtomography and image analysis and found that the volumetric percentage of rubber was less after the mixing process than was originally added. However, the researchers also suggested that this volume decrease was an artifact of melting or splitting of the crumb rubber during mixing and thus concluded that the crumb rubber particles will increase their size two to three times of their original size [13]. Airey et al. also studied this phenomenon using a basket drainage method to first separate the binder and rubber particles after mixing with rubber and
57
J.R. Medina, B.S. Underwood / Construction and Building Materials 150 (2017) 56–65
performed rheological testing as well as asphaltene content test (did not specified which method they used) on the pre-blending and post–blended asphalt to measure the loss of SARA fractions during mixing. Based on this testing, they concluded that the change in mass of crumb rubber particles was independent of the original source of the asphalt and only marginally dependent on the penetration grade of the asphalt cement used. Specifically, they found that the post-rubber mixing residue from the low penetration asphalt had a greater amount of lighter fractions remaining and thus the rubber in these asphalts must have taken in less of these fractions [9]. Both studies provide evidence of swelling and degradation of the rubber during mixing, but have some limitations. In the Kutay study, the authors must infer that swelling is negated by splitting without directly quantifying the number of particles that have split or disintegrated. The Airey et al. investigation did not directly measure the swelling and inferred its extent based on only mass changes. The importance of estimating the swelling of rubber particles in asphalt is that if known, then it may be possible to apply established computational and analytical composite models with some level of accuracy to estimate the modulus of the crumb rubber modified asphalt cement and by consequence design rubber blends for maximum benefit. Some models to describe the effect of rubber particles in asphalt (notably as a function of diameter and surface area) have been developed [14]. However, these approaches are based on empirical correlations between rubber related variables and resultant composite response, which limit their ability to provide direct and generalized insight into the impact of swelling and other physical phenomenon. Conversely, there are many micromechancial models, which do have the potential to provide such insights, that have been used to predict the viscosity and/or shear modulus of filled, viscous composites like rubber modified asphalt [15]. The models themselves attempt to reflect the impacts of the volume related impacts of particulates in a binding matrix using analogous physical systems. This, interpretation of the predictions in systems where particle actively modify the binding matrix can be challenging. The most well-known of these models is probably the Einstein function. This function models the system assuming non-interacting rigid spherical particles that are highly dispersed in a fluid medium. Physically this model is representative of systems that have very low particulate concentrations (<5%), and studies suggest that the intrinsic viscosity from the function is not constant and it is related to the physical characteristics of the mineral filler ranging from 2.5 to 4.9 [16]. Others have offered additional modifications to this function, for example Roscoe modified this model by considering the different sizes of mineral fillers [17]. Despite this and other modifications studies agree that the composite models derived from Einstein’s original function work well for some suspension or matrix but do not show an accurate representation of asphalt systems and in most cases under predict the modulus of the composite [16,18–21]. However, such under predictions can be the construed because of the model missing important physical parameters. Bahia et al. [12] used the Einstein and Mooney model to predict the viscosity of crumb rubber modified asphalt cement. As other studies have suggested, the models underestimated the viscosity values, and the authors used this finding as an indication that the interaction between the rubber and asphalt is not one involving simple solid inclusions. Consequently Bahia et al. propose that there must be a swelling mechanism and a change in the matrix (i.e. asphalt cement) [19]. Jamrah et al. used the Rule of Mixtures, Inverse Rule of Mixtures and the Differential model to backcalculate the modulus of the swollen rubber assuming that each model correctly described the composite system. Citing the previous study from Kutay et al. the volume concentration of rubber was adjusted to 300% of the original volume [22]. However, these models (as well as the Einstein derived
forms) have been shown to systematically underestimate the stiffening behavior of a related system consisting of asphalt cement and mineral particles smaller than 0.075 mm (known as asphalt mastic) [20,21,23]. In this system, the particulate phase does not swell or change its properties in the presence of the asphalt and yet, the models show a systematic under prediction of the composite modulus. Thus, while most studies agree about the primary mechanisms of interaction between the asphalt cement and the crumb rubber, that is the preferential absorption and the swelling. They differ in quantifying the volume expansion and linking these mechanisms to the modulus of the composite system. Consequently, it is difficult to know what course of action to make when choosing or designing a rubber blend in order to maximize the potential performance benefit. The objective of this paper strives to fill in this knowledge gap in understanding the multi-physical phenomena in asphalt mastics by reporting on a research study with the following objectives; 1) to evaluate the rheological properties of three asphalt cements modified with reacted and activated rubber (RAR), 2) estimate rubber swelling and potentially particle splitting, and 3) to quantify the relative effects of particle swelling, absorptions, and swelling/splitting in a composite system consisting of asphalt cement and RAR. 2. Micromechanical models Although many functions and techniques exist to calculate and predict the response of asphalt composites, analytical micromechanical models that are based on the material properties (Poisson’s ratio, modulus and volume fraction) and particle interaction via effective medium methods have proven to be most successful [20]. Many such models exist, but the most well-known and the ones used in this study are the Hashin Composite Sphere model, Eq. (1), and the Christensen model, Eq. (2). Both of these models are classified as effective medium models because they essentially solve the mathematical problem of a single particle within an effective medium, which itself has properties that are consistent (either in average energy or average displacement) with the matrix of the real composite. Note that the Christensen model has the form of a quadratic function, and that the parameters A, B, and C are calculated from known values [24]. Once these parameters are calculated, then the quadratic formula is used to calculate the composite shear modulus. It should be noted that other models were also evaluated during the study including the Einstein model [25], the Roscoe model [26], Mori-Tanaka scheme developed as applied by Benveniste [27], and the Differential method [28,29]. These models yielded values very similar to the Hashin and Christensen result and are not shown here in the interest of brevity, but their formulations are shown in the Appendix.
15ð1 mm Þ jGGrj 1 C v jG jc b i h ¼1þ jG jb 7 5mm þ 2ð4 5mm Þ jGGrj jGGrj 1 C v b
A
jG jc jG jb
2
jG jc þC ¼0 þB jG jb
ð1Þ
b
ð2Þ
where Gc is the modulus of the composite in kPa, Gr is the modulus of the rubber [30] in kPa, Gb is the modulus of the matrix (the asphalt) in kPa, mr is the Poisson’s ratio of the rubber (assumed equal to 0.5 for this study) [31,32], mm is the Poisson’s ratio of the matrix, Cv is the volumetric concentration of the filler, and A, B, C are functions of Gr, Gb, mr, mm, and Cv (shown in Appendix). The need and significance of using these models in this paper is that they permit a more direct interpretation of the multiphysical phenomenon that take place in the rubber based asphalt systems.
58
J.R. Medina, B.S. Underwood / Construction and Building Materials 150 (2017) 56–65
Other, empirical driven approaches rely on the implicit results of many of these phenomenon, and can be used to predict how specific types of asphalt cement and particles will react (based on for example the particle size and surface area). This information can be useful, but since the physical phenomenon are only implicit in the empirical functions, physical understanding of the phenomenon is limited when empirical approaches are utilized. 3. Experimental 3.1. Materials 3.1.1. Asphalt cement This study included three types of Superpave graded asphalt cements: a PG 7010, PG 64-22 and PG 58-22 from a single supplier in the state of Arizona. These three binders were modified with RAR at a rate of 20% by mass of the total blend (RAR + asphalt cement) which is equivalent to 25% by mass of the base asphalt cement. The volume fraction of RAR in the blend was approximately 17.6%, which was calculated based on typical density of crumb rubber and asphalt binder [33]. The naming convention followed in this paper is to refer to the unblended asphalt as ‘‘Virgin” asphalt and the rubber-asphalt blends as simply ‘‘RAR Modified” asphalt. 3.1.2. Reacted and activated crumb rubber The crumb rubber used for this study was a reacted and activated rubber (RAR) composed of soft asphalt cement, fine crumb rubber and activated mineral binder stabilizer (AMBS). The soft asphalt cement used in RAR aims to mitigate some of the increase stiffening caused by the rubber, which consequently enables production of asphalt with RAR at common mixing and compaction temperatures. The crumb rubber used comes from truck and automobile scrap tires, which are processed and ground by ambient or cryogenic process. The RAR has a particle size distribution ranging from 0.595 mm to sub 0.075 mm as shown in Fig. 1 [34]. 3.2. Sample preparation for rheological testing In the course of this study samples were prepared for both rheological testing and for evaluating the swelling and splitting potential of the RAR modified asphalt. The following sections detail these steps taken in preparing these samples. 3.2.1. Rheological testing Asphalt cements modified with RAR were manufactured in the laboratory. The process was completed in two steps. First, the cement was heated in an oven for 2 h at 170 °C. Then RAR was mixed into the asphalt at the aforementioned dosage rates and mixing was performed using a low shear mixed at 400 RPM for 30 min. To maintain the temperature at 170 °C during mixing the asphalt and rubber rested on a heating plate and the temperature inside the mixing vessel was monitored. During mixing extreme care was taken to ensure that a surface vortex did not exist, as this cause excessive oxidation of the asphalt. This goal was achieved by placing the head of the mixing armature at a depth of 25 mm above the bottom of the
100
3.2.2. Swelling and splitting testing To investigate the potential for swelling and splitting, particles sizes retained on 0.420 mm and 0.297 mm sieves were used. These particles were chosen because they were easy to handle and represented the two sizes with the greatest individual masses in the blend. As described later, the basic process followed for investigating swelling was to compare the size of rubber particles before and after blending with asphalt cement. Thus, testing was performed on both un-mixed RAR and RAR samples after mixing. To prepare the un-mixed samples, particles were first obtained for each size and then sprinkled onto Scanning Electron Microscope (SEM) holders for imaging. Care was taken during this process so that particles were not touching one another and imaging could adequately capture the characteristics of individual particles. Next, single size particles were blended with the asphalt cements using the same procedure described above and at a volume content of 8.7%. The volume content of 8.7%, which is about half of the amount used for the samples tested for G⁄, was chosen so that the SEM samples would not be too crowded with rubber particles, which would have greatly complicated the image analysis. To estimate swelling, a sample of each of the three rubberized asphalt cements were smeared onto the SEM holders in thin films. By smearing the samples onto the holders, the particles were exposed and easy to identify. To estimate splitting the remaining of the RAR blends were sampled and stored in tins until testing. 3.3. Procedures 3.3.1. Dynamic shear rheometer The rheological properties of the virgin asphalts, the RAR modified asphalts, and the D-RAR asphalts were measured using a Dynamic Shear Rheometer (DSR) conforming to the requirements of the ASTM D7175 protocol. The rheological parameter of interest was the Dynamic Modulus, |G⁄| and this quantity was evaluated at three temperatures that corresponded to the high temperature performance grade for each binder and ±6 °C. The test was performed using 25 mm parallel plates with a 1 mm gap at 12% strain control mode, and a frequency of 1.6 Hz. All test results shown in this paper represent the average of 3 replicates. It should be noted that some research suggests using a 2 mm gap for rubber modified asphalts [12,19,35]. In this study only a 1 mm gap was adopted for all blends, which could have contributed to a lower than desirable repeatability. However, the authors believe that the effect was minimal for the rubber used in this study as the coefficient of variability was only slightly higher for the rubber modified asphalt than the non-modified asphalt (5% versus 2%). 3.3.2. Scanning electron microscope The SEM was used to evaluate potential swelling of the RAR particles after mixing with the asphalt cement. Images were taken before and after mixing with asphalt using a SEM-XL30 Environmental microscope under the following conditions: environmental mode, secondary electron signal and accelerating voltage of 15 kV. For both the pre-blended and post-blending cases 35 SEM scans were taken in order to develop a probabilistic distribution of particle sizes. 3.3.3. Evaluation of particle splitting To estimate the potential for particle splitting the number of particles per gram of asphalt were determined before and after mixing and were compared. Particle counts were first performed on the pre-blender RAR. Using a high precision scale, 15 samples of crumb rubber particles between 0.001 and 0.006 g were first weighed out. Then, the particles were spread evenly on a white piece of paper and manually counted. The resulting number was divided by the actual mass of the sample and averaged across the 15 samples. This process yielded a pre-blended content of 20158 particles per gram of 0.420 mm RAR (20648–19744), and 31585 particles per gram of 0.297 mm RAR (39025–32043). To estimate the number of particles in the RAR Modified asphalt, 15 samples were first soaked in heavy duty citrus solvent for 30 min. Then each sample was washed with acetone over a 0.075 mm mesh. After drying completed, the particles were spread onto a white piece of paper and manually counted and averaged.
90 80 70 60 50 40 30 20
4. Results and discussion
10
Sieve Size (mm) Fig. 1. Size distribution of RAR.
0.595
0.420
0.297
0.250
4.1. DSR test results 0.149
0
0.074
Percent Passing
asphalt vessel. After the mixing was complete, the asphalt cement was either poured into small sample tins for later testing or filtered though a 0.075 mm mesh. The material that was passed through the mesh was then poured into small sample tins for later testing. Asphalt that was drained through the 0.075 mm mesh is referred to as ‘‘Drained RAR” or D-RAR. All the samples were kept at room temperature and were tested within 24 h after the blending process.
As mentioned in the introduction, literature clearly shows that at high temperatures, crumb rubber modified asphalt cements shows an increase in modulus with respect to the Virgin asphalt cement and a decrease in phase angle [9,12,22]. The test results
59
J.R. Medina, B.S. Underwood / Construction and Building Materials 150 (2017) 56–65
in this study also suggest this behavior as the data in Figs. 2 and 3 show for all the asphalt cements and temperatures the modulus increased by a factor of 2.6–4.9 with respect to the virgin asphalt cement. Note that the error bars in Fig. 2 represents the minimum and maximum values measured. This figure also shows that D-RAR asphalt has a lower modulus than the RAR Modified asphalt, but more importantly a substantially higher modulus than the virgin asphalt. A t-test was performed to compare the mean modulus between virgin to RAR and D-RAR. The null hypothesis is that the means are equal (H0: lControl = lRAR or D-RAR) and a two-tailed assessment is used. In this case, the null hypothesis is rejected in all cases (and for many at significance levels in excess of 99%), meaning that the mean modulus is statistically different for all cases, Table 1. Furthermore, all RAR modified asphalt cements showed a phase angle reduction of 10–19% with respect to virgin asphalt cement. The phase angle reduction between virgin and D-RAR was only 3–5%. This trend exists across all temperatures tested and supports the hypothesis that the rubber particles are indeed modifying the Virgin asphalt cement [9,12]. This experiment cannot confirm that this modification is due to the preferential absorption of the lighter fractions, but such is the most common explanation by other researchers. It should be noted that although the D-RAR asphalt was filtered through the 0.075 mm sieve, rubber particles smaller than this size could still be present in the asphalt. However, based on the gradation of the RAR and content of RAR in the blend it is estimated that at most this content would be less than 1.5%. The authors feel that it is highly unlikely that this 1.5% volume content would be solely responsible for the differences between Virgin and D-RAR since that assumption would seemingly require one to believe that the 18% volume content alone is responsible for the
differences between the Virgin and RAR Modified asphalt. The modeling effort followed in the remainder of this paper thus, assume that the increase in modulus of the RAR modified asphalts with respect to the virgin asphalt cement is due to volume filling aspects of the rubber particles and the modification of the binding matrix (the D-RAR). 4.2. Swelling and splitting of crumb rubber 4.2.1. Scanning electron microscope imaging Typical SEM images of the pre- and post- blended RAR samples are shown in Fig. 4. From these images, analysis to estimate their size was conducted using the principle of equivalent sphere diameter. This principle states that the volume occupied by a particle can be equated to the volume of a sphere that has a diameter determined by setting the 2D projected area of the particle to that of a circle and then solving for the necessary diameter of the circle [13,37]. The property adopted in this study to estimate the equivalent sphere diameter (ED) was the area of the particles highlighted in Fig. 4. The area of 35 un-swelled individual rubber particles were measured using the built in functionality of the ImageJ software along with Eqs. (3) and (4) to first estimate the ED and then calculate the volume for each individual particle. The particle sizes were found to be distributed according to a standard Gaussian distribution and thus, a statistical distribution was determined for the particle volumes. The results for both the pre- and post-blended 0.420 mm and 0.297 mm RAR are shown in Fig. 5 (a) and (b) respectively.
ED ¼ 2
ð3Þ
p
14
14
12
D-RAR
8 6
D-RAR
10
|G*| (kPa)
RAR
Virgin
12
Virgin
10
RAR
8 6
4
4
2
2
0
0
64
70
76
58
64
Temperature (°C)
Temperature (°C)
(a)
(b) 14 12
Virgin D-RAR
10
|G*| (kPa)
|G*| (kPa)
rffiffiffiffi A
RAR
8 6 4 2
0 52
58
64
Temperature (°C)
(c) Fig. 2. |G*| results from DSR experiments on study materials; (a) PG 70-10, (b) PG 64-22, and (c) PG 58-22.
70
60
J.R. Medina, B.S. Underwood / Construction and Building Materials 150 (2017) 56–65
100
100
Virgin
90
D-RAR
RAR
D-RAR
RAR
80
Phase Angle (° )
80
Phase Angle (° )
Virgin
90
70 60 50
40 30
70 60 50 40 30
20
20
10
10 0
0
64
70
58
76
64
70
Temperature (°C)
Temperature (°C)
(b)
(a) 100
Virgin
90
D-RAR
RAR
Phase Angle (° )
80 70 60 50
40 30 20 10
0
52
58
64
Temperature (°C)
(c) Fig. 3. Phase angle results from DSR experiments on study materials; (a) PG 70-10, (b) PG 64-22, and (c) PG 58-22.
Table 1 Summary of statistical analysis. Binder (PG)
Temperature (°C)
Condition
p-value for t-test comparison (H0: lControl = lRAR or D-RAR)
58-22
52
D-RAR RAR D-RAR RAR D-RAR RAR
0.0004 0.0008 0.0005 0.0011 0.0005 0.0013
D-RAR RAR D-RAR RAR D-RAR RAR
0.0006 0.0002 0.0004 0.0001 0.0002 0.0001
D-RAR RAR D-RAR RAR D-RAR RAR
0.0045 0.0022 0.0024 0.0011 0.0005 0.0012
58 64 64-22
58 64 70
70-10
64 70 76
V¼
3 4 d p 3 2
ð4Þ
where ED is the equivalent diameter in mm, A is the area of the particle in, mm2, and V is the equivalent sphere volume in mm3. Based on this analysis, the RAR 0.420 mm particles showed an increase in the average diameter of 0.023 mm, 0.067 mm and 0.034 mm from the PG 70-10, PG 64-22, and PG 58-22. These increases equate to an increase in volume for the 0.420 mm particles of 15%, 40% and 24% respectively. Similarly the RAR 0.297 mm
particles increased in diameter 0.024 mm, 0.060 mm and 0.081 mm (PG 70-10, PG 64-22, and PG 58-22), which equates to an increase in volume of 17%, 47% and 66% respectively. For both particle sizes it is found that the PG 70-10 results in the least amount of swelling while the PG 64-22 and PG 58-22 show somewhat conflicting results (higher swelling for PG 64-22 with the larger particles and less swelling with the smaller particles). This swelling is somewhat surprising as it is expected that the lower grade asphalt would consistently show a greater amount of swelling since it likely contains a higher proportion of lighter fractions. Especially since these asphalts were taken from the same supplier at the same time. To better understand this inconsistency with the expectation one must also consider the possibility of particle breakdown/splitting during the processing. 4.2.2. Crumb rubber splitting In addition to swelling, it has been suggested that rubber particles may expand so much and/or are agitated sufficiently during blending that they physically split into two or more pieces. If true, then it would mean that the above swelling values underestimate the true increase in rubber particle volume. This comparison was carried out by dividing the number of particles per gram after blending by the number of particles per gram of the unblended rubber, and for each individual size, reported in the Ratio column of Table 2. A ratio value greater than one indicates that the number of particles increased whereas a value of one indicates that the particles did not increase in number Examining the 0.420 mm cases first, it is found that, when mixed with binders PG 70-10 and PG 64-22 they did not result in particle splitting, but, the same particles mixed with PG 58-22 binder showed a 10% increase in the number of particles per gram. Similarly, for the 0.297 mm particles mixed with PG 70-10 and PG
61
J.R. Medina, B.S. Underwood / Construction and Building Materials 150 (2017) 56–65
Volume Increased
(a)
(b)
Volume Increased and Particle Splitting
(c)
(d)
Fig. 4. SEM images showing rubber particles pre- and post-blended RAR samples; (a) pre- blending RAR 0.420 mm, (b) post-blending RAR 0.420 mm, (c) pre-blending RAR 0.297 mm, and (d) post-blending RAR 0.297 mm.
0.5 RAR 0.420 A RAR PG 70-10 A RAR PG 64-22 A RAR PG 58-22
40%
30%
20%
0.3
0.2
0.1
10%
0% 0.00
RAR 0.297 A RAR PG 70-10 A RAR PG 64-22 A RAR PG 58-22
0.4
Frequency (%)
Frequency (%)
50%
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0 0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
Size, Volume (mm3)
Size, Volume (mm3)
(a)
(b)
Fig. 5. Particle volume distribution of RAR prior to blending (A RAR denotes that the distribution is for the post-blending particles) for; (a) 0.420 mm particles and (b) 0.297 mm particles.
Table 2 Summary of particle counts from splitting test. Asphalt cement
Particles size (mm)
Particles/gram
Ratio
Unblended (Theoretical)
0.420 0.297 0.420 0.297 0.420 0.297 0.420 0.297
20158 31585 20520 42115 19565 30051 21297 55731
NA NA 1.0 1.3 1.0 1.0 1.1 1.3
PG 70-10 PG 64-22 PG 58-22
58-22 there was an observed 30% increase in the number of particles, while the PG 64-22 asphalt cement showed no increase in the number of particles. The results show that there is some splitting of the RAR particles (either by pure volume increase or more likely by a combination of particle swelling, resultant softening, and then agitation during the mixing process) thus increasing the total volume of rubber particles above that estimated from swelling alone.
4.2.3. Estimating RAR volume content As noted earlier, although the RAR was incorporated into the asphalt cement at a dosage rate equivalent to 17.6% by volume,
62
J.R. Medina, B.S. Underwood / Construction and Building Materials 150 (2017) 56–65
4.3. Stiffening mechanism of the asphalt cement
Table 3 Summary of volume fraction of rubber particles in RAR modified asphalts. Binder
PG 70-10 PG 64-22 PG 58-22
Volume fraction (%) Initial
After swelling (no splitting)
After splitting and swelling
17.6 17.6 17.6
19.8 23.2 24.2
20.3 23.2 25.8
the mechanism of particle swelling and splitting during the mixing process results in an increase in volume occupied by the particles. To estimate this increase in volume the results from the SEM and counting trials were combined and used to adjust the original 17.6% volume content. During this process it was assumed that the change in volume for particles greater than 0.297 mm was equal to that of the 0.420 mm particles, and that for particles less than 0.297 mm it was equal to the swelling from the experiments on the 0.297 mm particles. Similarly it was assumed that the volume of particles greater than 0.297 mm increased by a factor of 1.15, 1.40 and 1.24 for asphalt cements PG 70-10, PG 64-22 and PG 58-22, while particles smaller than or equal to 0.297 mm increased by a factor of 1.18, 1.47 and 1.66 respectively for the same asphalts. With this assumption, and considering the splitting of the rubber particles, and using the typical gradation of RAR, the overall increase in volume was calculated to be 15.8%, 41.5% and 49.3% for binders PG 70-10, PG 64-22 and PG 58-22 respectively. Accounting for this much swelling and splitting of rubber particles of the RAR in the rubberized asphalt blend, the actual volume fraction is summarized in Table 3.
The test results demonstrate that the addition of RAR increase the modulus of the asphalt cement by a factor of 2.6–4.9 depending on the temperature and asphalt cement. However, these findings provide no fundamental insight in the mechanisms that cause this increase. To improve the understanding of the mechanisms that contribute to the stiffening, the Hashin and Christensen models were applied using the measured data and known volume contents (inclusive of swelling and splitting). For these models the matrix phase properties were estimated using the modulus of either the virgin asphalt cement or the D-RAR asphalt. The volumetric content and modulus of the inclusions phase (the rubber particles) was taken from that given in Table 3 and from published literature [28] respectively. The results are shown in Fig. 6 when virgin asphalt is set as the matrix phase and Fig. 7 when D-RAR is the matrix. When treating the matrix phase as the virgin asphalt, both models underestimated the results from 28% to 64% for all three asphalt binders, Fig. 6. However, based on the experimental results shown in Fig. 2 it is known that in the presence of the rubber particles the binding matrix properties are changed from those of the virgin asphalt. This effect was postulated by Bahia et al. [12], and the D-RAR experimental data provide some proof that it is in fact true. With this hypothesis in mind, the more accurate reflection of the binding matrix as it exists in the composite would be that of the D-RAR. The results when using the D-RAR moduli in the two models are shown in Fig. 7 and it can be seen that for the PG 70-10 asphalt cement, the Christensen model overestimates the modulus by 10.5% at 64 °C and underestimates by 0.8% and
14
14 12
8 6
10
8 6
4
4
2
2
0
0 64
70
58
76
64
70
Temperature (°C)
Temperature (°C)
(b)
(a) 14
Virgin D-RAR RAR Modified Christensen Hashin
12 10
|G*| (kPa)
|G*| (kPa)
10
Virgin D-RAR RAR Christensen Hashin
12
|G*| (kPa)
Virgin D-RAR RAR Christensen Hashin
8 6 4
2 0
52
58
64
Temperature (°C)
(c) Fig. 6. Micromechanical simulation results when using Virgin asphalt as the matrix phase; (a) PG 70-10, (b) PG 64-22, and (c) PG 58-22.
63
J.R. Medina, B.S. Underwood / Construction and Building Materials 150 (2017) 56–65
18
18
Virgin D-RAR RAR Christensen Hashin
|G*| (kPa)
14 12 10 8
16
Virgin D-RAR RAR Christensen Hashin
14
|G*| (kPa)
16
12
10 8
6
6
4
4
2
2
0
0 64
70
58
76
64
Temperature (°C)
Temperature (°C)
(a)
(b)
70
18 16
Virgin D-RAR RAR Christensen Hashin
|G*| (kPa)
14 12 10 8 6 4 2 0 52
58
64
Temperature (°C)
(c) Fig. 7. Micromechanical simulation results when using the D-RAR asphalt as the matrix phase; (a) PG 70-10, (b) PG 64-22, and (c) PG 58-22.
18
Predicted Modulus (kPa)
16 14 12 10 8 6 Line of Equality 2 Hashin - R = 0.95 2 Christensen - R = 0.84
4 2 0
0
2
4
6
8
10
12
14
16
18
Measured Modulus (kPa) Fig. 8. Measured dynamic modulus and predicted shear modulus calculated using the Hashin and the Christensen models.
7.32% at 70 °C and 76 °C. For the PG 64-22 asphalt this model overestimates the modulus by 11% at 58 °C and underestimates the modulus by 2% and 10% at 64 °C and 70 °C respectively. Results for the PG 58-22 asphalt cement Christensen model overestimates the results by 33, 18 and 12% for temperatures 52, 58, and 64 °C respectively. The results from Hashin model are also close to the measured modulus of the RAR modified asphalt cement. For the PG 70-10 asphalt cement, the model overestimates the modulus
by 4% and underestimates the modulus by 7% and 13% for temperatures of 64, 70 and 76 °C respectively. Moreover, the model overestimates the PG 64-22 asphalt binder stiffness at 58 °C by 2% and underestimates by 10% and 17% at temperatures of 64 and 70 °C respectively. The PG 58-22 asphalt cement overestimates the modulus at 52 and 58 °C by 19 and 5% and the percent difference at 64 °C is 0.4%. To compare the Hashin and the Christensen models, the results from both models are demonstrated in Fig. 8 by comparing the |G⁄| values calculated from the models to the measured |G⁄|. In general, both models show good agreement with the measured data. The Hashin model fit the results well having a coefficient of determination of 0.95 while the Christensen model is 0.84. Based on the temperature range used in this study, the models over-predict the shear modulus at lower temperatures and under-estimates the values at higher temperatures. It is seen that as the temperature decreases, the error increases. The authors believe that this increasing error is due, in part to the limitation of some micromechanical models which are based on physical properties of the constituents of the composite and disregard physicochemical effects, interaction between particles and gradation, and make quasistatic approximations of the viscoelastic properties of the binder. This latter issue can be readily addressed through derivation of a linear viscoelastic form of the classic micromechanical models. However, for the purpose of quantifying mechanisms and designing blends to control these mechanisms the simplification appears reasonable. Despite the slight underprediction of moduli at lower temperatures these models produce what are judged to be reasonable results (less than 19% error in the case of the Hashin model and approximately 30% error in the case of the Christensen model)
64
J.R. Medina, B.S. Underwood / Construction and Building Materials 150 (2017) 56–65
that permit quantification and with future study designing these composites [38]. 5. Conclusions In this study, rheology test was performed on three different asphalt cements modified with RAR to have a better understanding of the swelling and splitting mechanisms that take place during the blending process of asphalt cement and crumb rubber. SEM images and manual particle count were taken of the crumb rubber particles before and after blending to estimate the swelling and splitting potential, and evaluate the stiffening using micromechanical models. More specific conclusions that can be drawn from this research study are as follow: The RAR modified asphalt cement showed in all cases an increase in modulus with respect to the Virgin asphalt cements used. Drained RAR asphalt cement for all three cases also showed an increase in modulus with respect to Virgin asphalt confirming the hypothesis that the lighter fractions are absorbed by the crumb rubber and the residual binder is left with a higher concentration of asphaltenes. The rubber particles in the PG 58-22 modified asphalt cement showed the greatest increase in volume by a factor of 1.49 meaning that the actual volume in the blend increased from 17.6% to 25.8%, while the rubber particles showed an increase in volume by a factor of 1.43 and 1.15 which represent an increase in volume content in the blend from 17.6% to 23.2% and 20.3% for the PG 64-22 and PG 70-10 asphalt cement respectively. These results are expected since it is assumed that softer asphalt cement would have higher lighter fraction content and therefore more interaction between the rubber particles and the asphalt cement. The two micromechanical models included in this paper showed a good agreement with the measured modulus of the RAR modified asphalt cement. The Hashin model yielded a difference in the range of 0%–19% for all three asphalt cement and temperatures. The Christensen model yielded a difference of 1%–33%. The trend observed in both models was that at lower temperatures the models overestimated the measured modulus values but as the temperature increases the predictions from the models underestimated the results. In this study, the authors were able to quantify the swelling and splitting of the rubber particles, and integrate these finding in micromechanical models to accurately predict the dynamic shear modulus of the RAR modified asphalt cement. With a better insight of the interaction between the RAR particles and the asphalt cement it is possible to use these models to predict engineering properties with the objective to maximize the potential benefits of crumb rubber modified asphalt blends. Acknowledgements Authors will like to thank Dr. George Souza, Mr. Robert McGennis from Holly Frontier and Dr. Kamil Kaloush for providing the materials and technical support to develop this research study. Appendix A. Composite and micromechanical models used in the analysis: Einstein model,
jG jc ¼ 1 þ 2:5C v jG jb
ðB1Þ
Roscoe Model,
jG jc ¼ ð1 C v Þ2:5 jG jb
ðB2Þ
Christensen Model, h i G Gp 7=3 A ¼ 8 Gmp 1 ð4 5mm Þk1 C 10=3 v 2 63 Gm 1 k2 þ 2k1 k3 C v þ G Gp 2 252 Gmp 1 k2 C 5=3 v 50 Gm 1 7 12mm þ 8mm k2 C v þ 4ð7 10mm Þk2 k3 ðB3-1Þ
h i G Gp 7=3 B ¼ 4 Gmp 1 ð1 5mm Þk1 C 10=3 v þ 4 63 Gm 1 k2 þ 2k1 k3 C v G Gp 504 Gmp 1 k2 C 5=3 v þ 150 Gm 1 ð3 mm Þmm k2 C v þ 3ð15mm 7Þk2 k3 ðB3-2Þ
h i G G C ¼ 4 Gmp 1 ð5mm 7Þk1 C 10=3 2 63 Gmp 1 k2 þ 2k1 k3 C 7=3 v v þ G Gp 2 252 Gmp 1 k2 C 5=3 v þ 25 Gm 1 mm 7 k2 C v ð7 þ 5mm Þk2 k3 ðB3-3Þ Gp Gp 1 49 50mp mm þ 35 mp 2mm þ 35 2mp mm Gm Gm
k1 ¼
ðB3-4Þ k2 ¼ 5mp
Gp Gp 8 þ7 þ4 Gm Gm
ðB3-5Þ
Gp ð8 10mm Þ þ ð7 5mm Þ Gm
k3 ¼
ðB3-6Þ
Mori-Tanaka model,
jG jc 5C v ðGr jG jb Þ ¼1þ 5jG jb þ 2ð1 C v ÞðGr jG jb Þ jG jb
ðB4Þ
Differential Method,
jG jc ¼ Gr ð1 C v ÞðGr jG jb Þ Z
jG jc jG jb
2=5 ðB5Þ
a max
Gc ¼ a min
Gc ¼
jG jc ðaÞdPðaÞ
Nþ1 1X ½Gc ðai Þ þ Gc ðaiþ1 Þ½Pðai Þ P ðaiþ1 Þ 2 i¼1
ðB6Þ
ðB7Þ
References [1] M. Abdelrahman, Controlling low temperature properties of asphalt rubber binders, Proceedings of the 2nd Transportation Specialty Conference, Canadian Society for Civil Engineering, Halifax, 1998. [2] T.C. Billiter, R.R. Davison, C.J. Glover, J.A. Bullin, Physical properties of asphalt rubber binders, Transportation Research Record 1586, TRB, Washington, DC, 1997, pp 51–59. [3] J.S. Coplantz, M.T. Yapp, R.N. Finn, Review of relationships between modified asphalt properties and pavement performance, SHRP-A-631, 1993. [4] F.L. Roberts, P.S. Kandhal, E.R. Brown, D.Y. Lee, T.W. Kennedy, Hot mix asphalt materials, mixture design and construction, first ed., NAPA Eduacation Foundation, Lanham, Maryland, 1991. 68–382. [5] J.L. McQuillen, H.B. Takallou, R.G. Hicks, D.C. Esch, Economic analysis of rubber modified asphalt mixes, J. Transp. Eng. 114 (1988) 259–277. [6] S.A. Dantas Neto, M.M. Farias, J.C. Pais, P.A. Pereira, L. Picado Santos, Behavior of asphalt rubber hot mixes obtained with high crumb rubber content, in: Proceedings of the Asphalt Rubber 2003 Conference, Brasilia, Brazil, 2003. [7] G.B. Way, Asphalt-Rubber 45 Years of Progress, in: Proceedings Asphalt Rubber Conference, Germany, 2012. [8] S. Kocevski, S. Yagneswaran, F. Xiao, V.S. Punith, D.W. Smith, S. Amirkhanian, Surface modified ground rubber tire by grafting acrylic acid for paving applications, Constr. Build. Mater. 34 (2012) 83–90.
J.R. Medina, B.S. Underwood / Construction and Building Materials 150 (2017) 56–65 [9] Gordon D. Airey, Mujibur M. Rahman, Andrew C. Collop, Absorption of bitumen into crumb rubber using the basket drainage method, Int. J. Pavement Eng. 4 (2) (2003) 105–119. [10] D. Dong, X. Huang, X. Li, L. Zhang, Swelling process of rubber in asphalt and its effect on the structure and properties of rubber and asphalt, Constr. Build. Mater. 29 (2012) 316–322. [11] Feipeng Xiao, J. Putman Bradley, N. Amirkhanian Serji, Laboratory investigation of dimensional changes of crumb rubber reacting with asphalt binder, Road Mater. Pavement Des. (2006). [12] Hussain U. Bahia, Robert Davies, Effect of crumb rubber modifiers (CRM) on performance related properties of asphalt binders, Asphalt Pavement Technol. 63 (1994). 414-414. [13] M.E. Kutay, H. Ozturk, Internal structure characteristics of crumb rubber modified asphalt binders: an analysis using 3D X-ray Microtomography Imaging, in: Proceedings Asphalt Rubber Conference, Germany, 2012. [14] J. Shen, S. Amirkhanian, F. Xiao, B. Tang, Influence of surface area and size of crumb rubber on high temperature properties of crumb rubber modified binders, Constr. Build. Mater. 23 (2009) 304–310. [15] S. Ahmed, F.R. Jones, A review of particulate reinforcement theories for polymer composites, J Mater. Sci. 25 (12) (1990) 4933–4942. [16] E. Hesami, D. Jelagin, N. Kringos, B. Birgisson, An empirical framework for determining asphalt mastic viscosity as a function of mineral filler concentration, Constr. Build Mater. 35 (2012) 23–29. [17] R. Roscoe, The viscosity of suspensions of rigid spheres, Br. J. Appl. Phys. 23 (9) (1952) 1020–1024. [18] Akshay Gundla, Use of Micro-Mechanical Models to Study the Mastic Level Structure of Asphalt Concretes Containing Reclaimed Asphalt Pavement (MS Thesis), Arizona State University, Tempe, 2014. [19] Hussain U. Bahia, Robert Davies, Factors Controlling the Effect of Crumb Rubber on Critical Properties of Asphalt Binders vol. 64 (1995) 130–162. [20] B.S. Underwood, Y.R. Kim, A four phase micro-mechanical model for mastic modulus, Mech. Mater. 25 (2014) 13–33. [21] N. Shashidhar, A. Shenoy, On using micromechanical models to describe dynamic mechanical behavior of asphalt mastics, Mech. Mater. 34 (10) (2002) 657–669. [22] Jamrah Anas, M. Emin Kutay, Sudhir Varma. Backcalculation of Swollen Crumb Rubber Modulus in Asphalt Rubber Binder and its Relation to Performance, in: 94th Transportation Research Board Annual Meeting. United States, 2015.
65
[23] E. Aigner, R. Lackner, C. Pichler, Multiscale predictions of viscoelastic properties of asphalt concrete, J. Mater. Civil Eng. 21 (12) (2009) 771–780. [24] R.M. Christensen, K.H. Lo, Solutions for effective shear properties in three phase sphere and cylinders models, J. Mech. Phys. Solids 27 (4) (1979) 315– 330. [25] A. Einstein, New determination of molecular dimensions, Ann. Phys. 19 (2) (1906) 289–306. [26] R. Roscoe, The viscosity of suspensions of rigid spheres, Br. J. Appl. Phys. 3 (8) (1952) 267. [27] Y. Benceniste, A new approach to the application of mori-tanaka’s theory in composite materials, Mech. Mater. 6 (2) (1987) 147–153. [28] X. Shu, B. Huang, Michromechanics-based dynamic modulus prediction of polymeric asphalt concrete mixtures, Compos. Part B Eng. 39 (4) (2008) 704– 713. [29] X. Shu, B. Huang, Predicting dynamic modulus of asphalt mixtures with differential method, Road Mater. Pavement Des. 10 (2) (2009) 337–359. [30] A. Lion, Thixotropic behavior of rubber under dynamic loading histories: experimental and theory, J. Mech. Phys. Solids 46 (5) (1998) 895–930. [31] E. Guth, Theory of filler reinforcement, J. Appl. Phys. 16 (1) (1945) 20–25. [32] Y. Lin, H. Sheng-Jye, Temperature predictions of rolling tires by computer simulation, Math. Comp. Simul. 67 (3) (2004) 235–249. [33] Carl Thodesen, Khaldoun Shatanawi, Serji Amirkhanian, Effect of crumb rubber characteristics on crumb rubber modified (CRM) binder visocity, Constr. Build. Mater. 23 (1) (2009) 295–303. [34] J. Medina, B.S. Underwood, Use of particulate composite models and crumb rubber swelling to estimate stiffness of rubberized asphalt binders, in: Proceedings ISAP Symposium, 2016. [35] J Richards Willis, Clayton Plemons, Pamela Turner, Carolina Rodezno, Tyler Mitchell, Effect of Ground Tire Rubber Particle Size and Grinding Method on Asphalt Binder Properties, National Center for Asphalt Technology, Auburn, 2012. [37] W. Pabst, E. Gregorova, Characterization of Particles and Particle Systems, ICT Prague, 2007. [38] H.M. Yin, W.G. Buttlar, G.H. Paulino, H. Di Benedetto, Assesment of existing micro-mechanical models for asphalt mastics considering viscoelastic effects, Road Mater. Pavement Des. 9 (1) (2008) 37–57.
Contents lists available at ScienceDirect
Construction and Building Materials journal homepage: www.elsevier.com/locate/conbuildmat
Micromechanical shear modulus modeling of activated crumb rubber modified asphalt cements Jose R. Medina a,⇑, B. Shane Underwood b a b
Arizona State University, 229BA ISTBII, Tempe, AZ 85287, USA Arizona State University, 229B ISTBII, Tempe, AZ, USA
h i g h l i g h t s Absorption of the lighter fractions by the crumb rubber particles. A method to quantify the swelling and splitting of the rubber particles. Use of micromechanical models to accurately predict the dynamic shear modulus of rubberized asphalt cement.
a r t i c l e
i n f o
Article history: Received 9 August 2016 Received in revised form 10 May 2017 Accepted 21 May 2017
Keywords: Crumb rubber Rheology Modified binder Activated rubber Activated mineral binder stabilizer (AMBS) Crumb rubber splitting Composite models Micromechanical models
a b s t r a c t The rheological properties of three asphalt cements containing reacted and activated rubber (RAR) are evaluated to quantify the relative effects of swelling, splitting, and absorption. Rheological testing and electron microscopy are used to measure the dynamic shear modulus, |G⁄|, and rubber particle changes respectively. It is found that |G⁄| increased for all RAR and that particles swell 15.8–49.3% with those in softer asphalt showing the greatest swelling. Micromechanical models are used to predict |G⁄| of the materials. The Hashin and Christensen models are found to accurately predict the measured moduli after accounting for the swelling, splitting, and absorption. Ó 2017 Elsevier Ltd. All rights reserved.
1. Introduction The use of crumb rubber with and without physical and chemical modification to improve the engineering performance of asphalt concrete has been well documented over the last 40 years [1–8]. Within these documented studies, the performance benefits from incorporating rubber are: greater rutting resistance, mitigation of thermal reflective cracking, increased resistance to fatigue cracking, and noise reduction. It is often cited that the reason for the increased performance of these rubber modified asphalts is that during the blending process, the crumb rubber particles absorb the aromatic oils and resins from the asphalt cement leaving the binder with a higher concentration of asphaltenes while at the same time storing the lighter resins for times when they are ⇑ Corresponding author. E-mail addresses: [email protected] (J.R. Medina), [email protected] (B.S. Underwood). http://dx.doi.org/10.1016/j.conbuildmat.2017.05.208 0950-0618/Ó 2017 Elsevier Ltd. All rights reserved.
needed for example to accelerate healing. This diffusion of lighter fractions is postulated to result in swelling of the rubber particles in a time and temperature dependent process that can have a substantial impact on the volume occupation and thus performance [9,10]. The concept of rubber swelling is somewhat controversial. It is documented that crumb rubber particles can swell by 3–5 times their original size [9,11,12], but very little peer reviewed literature has directly quantified the precise volume increase. Kutay et al. quantified the structural change in crumb rubber modified asphalt by using X-ray microtomography and image analysis and found that the volumetric percentage of rubber was less after the mixing process than was originally added. However, the researchers also suggested that this volume decrease was an artifact of melting or splitting of the crumb rubber during mixing and thus concluded that the crumb rubber particles will increase their size two to three times of their original size [13]. Airey et al. also studied this phenomenon using a basket drainage method to first separate the binder and rubber particles after mixing with rubber and
57
J.R. Medina, B.S. Underwood / Construction and Building Materials 150 (2017) 56–65
performed rheological testing as well as asphaltene content test (did not specified which method they used) on the pre-blending and post–blended asphalt to measure the loss of SARA fractions during mixing. Based on this testing, they concluded that the change in mass of crumb rubber particles was independent of the original source of the asphalt and only marginally dependent on the penetration grade of the asphalt cement used. Specifically, they found that the post-rubber mixing residue from the low penetration asphalt had a greater amount of lighter fractions remaining and thus the rubber in these asphalts must have taken in less of these fractions [9]. Both studies provide evidence of swelling and degradation of the rubber during mixing, but have some limitations. In the Kutay study, the authors must infer that swelling is negated by splitting without directly quantifying the number of particles that have split or disintegrated. The Airey et al. investigation did not directly measure the swelling and inferred its extent based on only mass changes. The importance of estimating the swelling of rubber particles in asphalt is that if known, then it may be possible to apply established computational and analytical composite models with some level of accuracy to estimate the modulus of the crumb rubber modified asphalt cement and by consequence design rubber blends for maximum benefit. Some models to describe the effect of rubber particles in asphalt (notably as a function of diameter and surface area) have been developed [14]. However, these approaches are based on empirical correlations between rubber related variables and resultant composite response, which limit their ability to provide direct and generalized insight into the impact of swelling and other physical phenomenon. Conversely, there are many micromechancial models, which do have the potential to provide such insights, that have been used to predict the viscosity and/or shear modulus of filled, viscous composites like rubber modified asphalt [15]. The models themselves attempt to reflect the impacts of the volume related impacts of particulates in a binding matrix using analogous physical systems. This, interpretation of the predictions in systems where particle actively modify the binding matrix can be challenging. The most well-known of these models is probably the Einstein function. This function models the system assuming non-interacting rigid spherical particles that are highly dispersed in a fluid medium. Physically this model is representative of systems that have very low particulate concentrations (<5%), and studies suggest that the intrinsic viscosity from the function is not constant and it is related to the physical characteristics of the mineral filler ranging from 2.5 to 4.9 [16]. Others have offered additional modifications to this function, for example Roscoe modified this model by considering the different sizes of mineral fillers [17]. Despite this and other modifications studies agree that the composite models derived from Einstein’s original function work well for some suspension or matrix but do not show an accurate representation of asphalt systems and in most cases under predict the modulus of the composite [16,18–21]. However, such under predictions can be the construed because of the model missing important physical parameters. Bahia et al. [12] used the Einstein and Mooney model to predict the viscosity of crumb rubber modified asphalt cement. As other studies have suggested, the models underestimated the viscosity values, and the authors used this finding as an indication that the interaction between the rubber and asphalt is not one involving simple solid inclusions. Consequently Bahia et al. propose that there must be a swelling mechanism and a change in the matrix (i.e. asphalt cement) [19]. Jamrah et al. used the Rule of Mixtures, Inverse Rule of Mixtures and the Differential model to backcalculate the modulus of the swollen rubber assuming that each model correctly described the composite system. Citing the previous study from Kutay et al. the volume concentration of rubber was adjusted to 300% of the original volume [22]. However, these models (as well as the Einstein derived
forms) have been shown to systematically underestimate the stiffening behavior of a related system consisting of asphalt cement and mineral particles smaller than 0.075 mm (known as asphalt mastic) [20,21,23]. In this system, the particulate phase does not swell or change its properties in the presence of the asphalt and yet, the models show a systematic under prediction of the composite modulus. Thus, while most studies agree about the primary mechanisms of interaction between the asphalt cement and the crumb rubber, that is the preferential absorption and the swelling. They differ in quantifying the volume expansion and linking these mechanisms to the modulus of the composite system. Consequently, it is difficult to know what course of action to make when choosing or designing a rubber blend in order to maximize the potential performance benefit. The objective of this paper strives to fill in this knowledge gap in understanding the multi-physical phenomena in asphalt mastics by reporting on a research study with the following objectives; 1) to evaluate the rheological properties of three asphalt cements modified with reacted and activated rubber (RAR), 2) estimate rubber swelling and potentially particle splitting, and 3) to quantify the relative effects of particle swelling, absorptions, and swelling/splitting in a composite system consisting of asphalt cement and RAR. 2. Micromechanical models Although many functions and techniques exist to calculate and predict the response of asphalt composites, analytical micromechanical models that are based on the material properties (Poisson’s ratio, modulus and volume fraction) and particle interaction via effective medium methods have proven to be most successful [20]. Many such models exist, but the most well-known and the ones used in this study are the Hashin Composite Sphere model, Eq. (1), and the Christensen model, Eq. (2). Both of these models are classified as effective medium models because they essentially solve the mathematical problem of a single particle within an effective medium, which itself has properties that are consistent (either in average energy or average displacement) with the matrix of the real composite. Note that the Christensen model has the form of a quadratic function, and that the parameters A, B, and C are calculated from known values [24]. Once these parameters are calculated, then the quadratic formula is used to calculate the composite shear modulus. It should be noted that other models were also evaluated during the study including the Einstein model [25], the Roscoe model [26], Mori-Tanaka scheme developed as applied by Benveniste [27], and the Differential method [28,29]. These models yielded values very similar to the Hashin and Christensen result and are not shown here in the interest of brevity, but their formulations are shown in the Appendix.
15ð1 mm Þ jGGrj 1 C v jG jc b i h ¼1þ jG jb 7 5mm þ 2ð4 5mm Þ jGGrj jGGrj 1 C v b
A
jG jc jG jb
2
jG jc þC ¼0 þB jG jb
ð1Þ
b
ð2Þ
where Gc is the modulus of the composite in kPa, Gr is the modulus of the rubber [30] in kPa, Gb is the modulus of the matrix (the asphalt) in kPa, mr is the Poisson’s ratio of the rubber (assumed equal to 0.5 for this study) [31,32], mm is the Poisson’s ratio of the matrix, Cv is the volumetric concentration of the filler, and A, B, C are functions of Gr, Gb, mr, mm, and Cv (shown in Appendix). The need and significance of using these models in this paper is that they permit a more direct interpretation of the multiphysical phenomenon that take place in the rubber based asphalt systems.
58
J.R. Medina, B.S. Underwood / Construction and Building Materials 150 (2017) 56–65
Other, empirical driven approaches rely on the implicit results of many of these phenomenon, and can be used to predict how specific types of asphalt cement and particles will react (based on for example the particle size and surface area). This information can be useful, but since the physical phenomenon are only implicit in the empirical functions, physical understanding of the phenomenon is limited when empirical approaches are utilized. 3. Experimental 3.1. Materials 3.1.1. Asphalt cement This study included three types of Superpave graded asphalt cements: a PG 7010, PG 64-22 and PG 58-22 from a single supplier in the state of Arizona. These three binders were modified with RAR at a rate of 20% by mass of the total blend (RAR + asphalt cement) which is equivalent to 25% by mass of the base asphalt cement. The volume fraction of RAR in the blend was approximately 17.6%, which was calculated based on typical density of crumb rubber and asphalt binder [33]. The naming convention followed in this paper is to refer to the unblended asphalt as ‘‘Virgin” asphalt and the rubber-asphalt blends as simply ‘‘RAR Modified” asphalt. 3.1.2. Reacted and activated crumb rubber The crumb rubber used for this study was a reacted and activated rubber (RAR) composed of soft asphalt cement, fine crumb rubber and activated mineral binder stabilizer (AMBS). The soft asphalt cement used in RAR aims to mitigate some of the increase stiffening caused by the rubber, which consequently enables production of asphalt with RAR at common mixing and compaction temperatures. The crumb rubber used comes from truck and automobile scrap tires, which are processed and ground by ambient or cryogenic process. The RAR has a particle size distribution ranging from 0.595 mm to sub 0.075 mm as shown in Fig. 1 [34]. 3.2. Sample preparation for rheological testing In the course of this study samples were prepared for both rheological testing and for evaluating the swelling and splitting potential of the RAR modified asphalt. The following sections detail these steps taken in preparing these samples. 3.2.1. Rheological testing Asphalt cements modified with RAR were manufactured in the laboratory. The process was completed in two steps. First, the cement was heated in an oven for 2 h at 170 °C. Then RAR was mixed into the asphalt at the aforementioned dosage rates and mixing was performed using a low shear mixed at 400 RPM for 30 min. To maintain the temperature at 170 °C during mixing the asphalt and rubber rested on a heating plate and the temperature inside the mixing vessel was monitored. During mixing extreme care was taken to ensure that a surface vortex did not exist, as this cause excessive oxidation of the asphalt. This goal was achieved by placing the head of the mixing armature at a depth of 25 mm above the bottom of the
100
3.2.2. Swelling and splitting testing To investigate the potential for swelling and splitting, particles sizes retained on 0.420 mm and 0.297 mm sieves were used. These particles were chosen because they were easy to handle and represented the two sizes with the greatest individual masses in the blend. As described later, the basic process followed for investigating swelling was to compare the size of rubber particles before and after blending with asphalt cement. Thus, testing was performed on both un-mixed RAR and RAR samples after mixing. To prepare the un-mixed samples, particles were first obtained for each size and then sprinkled onto Scanning Electron Microscope (SEM) holders for imaging. Care was taken during this process so that particles were not touching one another and imaging could adequately capture the characteristics of individual particles. Next, single size particles were blended with the asphalt cements using the same procedure described above and at a volume content of 8.7%. The volume content of 8.7%, which is about half of the amount used for the samples tested for G⁄, was chosen so that the SEM samples would not be too crowded with rubber particles, which would have greatly complicated the image analysis. To estimate swelling, a sample of each of the three rubberized asphalt cements were smeared onto the SEM holders in thin films. By smearing the samples onto the holders, the particles were exposed and easy to identify. To estimate splitting the remaining of the RAR blends were sampled and stored in tins until testing. 3.3. Procedures 3.3.1. Dynamic shear rheometer The rheological properties of the virgin asphalts, the RAR modified asphalts, and the D-RAR asphalts were measured using a Dynamic Shear Rheometer (DSR) conforming to the requirements of the ASTM D7175 protocol. The rheological parameter of interest was the Dynamic Modulus, |G⁄| and this quantity was evaluated at three temperatures that corresponded to the high temperature performance grade for each binder and ±6 °C. The test was performed using 25 mm parallel plates with a 1 mm gap at 12% strain control mode, and a frequency of 1.6 Hz. All test results shown in this paper represent the average of 3 replicates. It should be noted that some research suggests using a 2 mm gap for rubber modified asphalts [12,19,35]. In this study only a 1 mm gap was adopted for all blends, which could have contributed to a lower than desirable repeatability. However, the authors believe that the effect was minimal for the rubber used in this study as the coefficient of variability was only slightly higher for the rubber modified asphalt than the non-modified asphalt (5% versus 2%). 3.3.2. Scanning electron microscope The SEM was used to evaluate potential swelling of the RAR particles after mixing with the asphalt cement. Images were taken before and after mixing with asphalt using a SEM-XL30 Environmental microscope under the following conditions: environmental mode, secondary electron signal and accelerating voltage of 15 kV. For both the pre-blended and post-blending cases 35 SEM scans were taken in order to develop a probabilistic distribution of particle sizes. 3.3.3. Evaluation of particle splitting To estimate the potential for particle splitting the number of particles per gram of asphalt were determined before and after mixing and were compared. Particle counts were first performed on the pre-blender RAR. Using a high precision scale, 15 samples of crumb rubber particles between 0.001 and 0.006 g were first weighed out. Then, the particles were spread evenly on a white piece of paper and manually counted. The resulting number was divided by the actual mass of the sample and averaged across the 15 samples. This process yielded a pre-blended content of 20158 particles per gram of 0.420 mm RAR (20648–19744), and 31585 particles per gram of 0.297 mm RAR (39025–32043). To estimate the number of particles in the RAR Modified asphalt, 15 samples were first soaked in heavy duty citrus solvent for 30 min. Then each sample was washed with acetone over a 0.075 mm mesh. After drying completed, the particles were spread onto a white piece of paper and manually counted and averaged.
90 80 70 60 50 40 30 20
4. Results and discussion
10
Sieve Size (mm) Fig. 1. Size distribution of RAR.
0.595
0.420
0.297
0.250
4.1. DSR test results 0.149
0
0.074
Percent Passing
asphalt vessel. After the mixing was complete, the asphalt cement was either poured into small sample tins for later testing or filtered though a 0.075 mm mesh. The material that was passed through the mesh was then poured into small sample tins for later testing. Asphalt that was drained through the 0.075 mm mesh is referred to as ‘‘Drained RAR” or D-RAR. All the samples were kept at room temperature and were tested within 24 h after the blending process.
As mentioned in the introduction, literature clearly shows that at high temperatures, crumb rubber modified asphalt cements shows an increase in modulus with respect to the Virgin asphalt cement and a decrease in phase angle [9,12,22]. The test results
59
J.R. Medina, B.S. Underwood / Construction and Building Materials 150 (2017) 56–65
in this study also suggest this behavior as the data in Figs. 2 and 3 show for all the asphalt cements and temperatures the modulus increased by a factor of 2.6–4.9 with respect to the virgin asphalt cement. Note that the error bars in Fig. 2 represents the minimum and maximum values measured. This figure also shows that D-RAR asphalt has a lower modulus than the RAR Modified asphalt, but more importantly a substantially higher modulus than the virgin asphalt. A t-test was performed to compare the mean modulus between virgin to RAR and D-RAR. The null hypothesis is that the means are equal (H0: lControl = lRAR or D-RAR) and a two-tailed assessment is used. In this case, the null hypothesis is rejected in all cases (and for many at significance levels in excess of 99%), meaning that the mean modulus is statistically different for all cases, Table 1. Furthermore, all RAR modified asphalt cements showed a phase angle reduction of 10–19% with respect to virgin asphalt cement. The phase angle reduction between virgin and D-RAR was only 3–5%. This trend exists across all temperatures tested and supports the hypothesis that the rubber particles are indeed modifying the Virgin asphalt cement [9,12]. This experiment cannot confirm that this modification is due to the preferential absorption of the lighter fractions, but such is the most common explanation by other researchers. It should be noted that although the D-RAR asphalt was filtered through the 0.075 mm sieve, rubber particles smaller than this size could still be present in the asphalt. However, based on the gradation of the RAR and content of RAR in the blend it is estimated that at most this content would be less than 1.5%. The authors feel that it is highly unlikely that this 1.5% volume content would be solely responsible for the differences between Virgin and D-RAR since that assumption would seemingly require one to believe that the 18% volume content alone is responsible for the
differences between the Virgin and RAR Modified asphalt. The modeling effort followed in the remainder of this paper thus, assume that the increase in modulus of the RAR modified asphalts with respect to the virgin asphalt cement is due to volume filling aspects of the rubber particles and the modification of the binding matrix (the D-RAR). 4.2. Swelling and splitting of crumb rubber 4.2.1. Scanning electron microscope imaging Typical SEM images of the pre- and post- blended RAR samples are shown in Fig. 4. From these images, analysis to estimate their size was conducted using the principle of equivalent sphere diameter. This principle states that the volume occupied by a particle can be equated to the volume of a sphere that has a diameter determined by setting the 2D projected area of the particle to that of a circle and then solving for the necessary diameter of the circle [13,37]. The property adopted in this study to estimate the equivalent sphere diameter (ED) was the area of the particles highlighted in Fig. 4. The area of 35 un-swelled individual rubber particles were measured using the built in functionality of the ImageJ software along with Eqs. (3) and (4) to first estimate the ED and then calculate the volume for each individual particle. The particle sizes were found to be distributed according to a standard Gaussian distribution and thus, a statistical distribution was determined for the particle volumes. The results for both the pre- and post-blended 0.420 mm and 0.297 mm RAR are shown in Fig. 5 (a) and (b) respectively.
ED ¼ 2
ð3Þ
p
14
14
12
D-RAR
8 6
D-RAR
10
|G*| (kPa)
RAR
Virgin
12
Virgin
10
RAR
8 6
4
4
2
2
0
0
64
70
76
58
64
Temperature (°C)
Temperature (°C)
(a)
(b) 14 12
Virgin D-RAR
10
|G*| (kPa)
|G*| (kPa)
rffiffiffiffi A
RAR
8 6 4 2
0 52
58
64
Temperature (°C)
(c) Fig. 2. |G*| results from DSR experiments on study materials; (a) PG 70-10, (b) PG 64-22, and (c) PG 58-22.
70
60
J.R. Medina, B.S. Underwood / Construction and Building Materials 150 (2017) 56–65
100
100
Virgin
90
D-RAR
RAR
D-RAR
RAR
80
Phase Angle (° )
80
Phase Angle (° )
Virgin
90
70 60 50
40 30
70 60 50 40 30
20
20
10
10 0
0
64
70
58
76
64
70
Temperature (°C)
Temperature (°C)
(b)
(a) 100
Virgin
90
D-RAR
RAR
Phase Angle (° )
80 70 60 50
40 30 20 10
0
52
58
64
Temperature (°C)
(c) Fig. 3. Phase angle results from DSR experiments on study materials; (a) PG 70-10, (b) PG 64-22, and (c) PG 58-22.
Table 1 Summary of statistical analysis. Binder (PG)
Temperature (°C)
Condition
p-value for t-test comparison (H0: lControl = lRAR or D-RAR)
58-22
52
D-RAR RAR D-RAR RAR D-RAR RAR
0.0004 0.0008 0.0005 0.0011 0.0005 0.0013
D-RAR RAR D-RAR RAR D-RAR RAR
0.0006 0.0002 0.0004 0.0001 0.0002 0.0001
D-RAR RAR D-RAR RAR D-RAR RAR
0.0045 0.0022 0.0024 0.0011 0.0005 0.0012
58 64 64-22
58 64 70
70-10
64 70 76
V¼
3 4 d p 3 2
ð4Þ
where ED is the equivalent diameter in mm, A is the area of the particle in, mm2, and V is the equivalent sphere volume in mm3. Based on this analysis, the RAR 0.420 mm particles showed an increase in the average diameter of 0.023 mm, 0.067 mm and 0.034 mm from the PG 70-10, PG 64-22, and PG 58-22. These increases equate to an increase in volume for the 0.420 mm particles of 15%, 40% and 24% respectively. Similarly the RAR 0.297 mm
particles increased in diameter 0.024 mm, 0.060 mm and 0.081 mm (PG 70-10, PG 64-22, and PG 58-22), which equates to an increase in volume of 17%, 47% and 66% respectively. For both particle sizes it is found that the PG 70-10 results in the least amount of swelling while the PG 64-22 and PG 58-22 show somewhat conflicting results (higher swelling for PG 64-22 with the larger particles and less swelling with the smaller particles). This swelling is somewhat surprising as it is expected that the lower grade asphalt would consistently show a greater amount of swelling since it likely contains a higher proportion of lighter fractions. Especially since these asphalts were taken from the same supplier at the same time. To better understand this inconsistency with the expectation one must also consider the possibility of particle breakdown/splitting during the processing. 4.2.2. Crumb rubber splitting In addition to swelling, it has been suggested that rubber particles may expand so much and/or are agitated sufficiently during blending that they physically split into two or more pieces. If true, then it would mean that the above swelling values underestimate the true increase in rubber particle volume. This comparison was carried out by dividing the number of particles per gram after blending by the number of particles per gram of the unblended rubber, and for each individual size, reported in the Ratio column of Table 2. A ratio value greater than one indicates that the number of particles increased whereas a value of one indicates that the particles did not increase in number Examining the 0.420 mm cases first, it is found that, when mixed with binders PG 70-10 and PG 64-22 they did not result in particle splitting, but, the same particles mixed with PG 58-22 binder showed a 10% increase in the number of particles per gram. Similarly, for the 0.297 mm particles mixed with PG 70-10 and PG
61
J.R. Medina, B.S. Underwood / Construction and Building Materials 150 (2017) 56–65
Volume Increased
(a)
(b)
Volume Increased and Particle Splitting
(c)
(d)
Fig. 4. SEM images showing rubber particles pre- and post-blended RAR samples; (a) pre- blending RAR 0.420 mm, (b) post-blending RAR 0.420 mm, (c) pre-blending RAR 0.297 mm, and (d) post-blending RAR 0.297 mm.
0.5 RAR 0.420 A RAR PG 70-10 A RAR PG 64-22 A RAR PG 58-22
40%
30%
20%
0.3
0.2
0.1
10%
0% 0.00
RAR 0.297 A RAR PG 70-10 A RAR PG 64-22 A RAR PG 58-22
0.4
Frequency (%)
Frequency (%)
50%
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0 0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
Size, Volume (mm3)
Size, Volume (mm3)
(a)
(b)
Fig. 5. Particle volume distribution of RAR prior to blending (A RAR denotes that the distribution is for the post-blending particles) for; (a) 0.420 mm particles and (b) 0.297 mm particles.
Table 2 Summary of particle counts from splitting test. Asphalt cement
Particles size (mm)
Particles/gram
Ratio
Unblended (Theoretical)
0.420 0.297 0.420 0.297 0.420 0.297 0.420 0.297
20158 31585 20520 42115 19565 30051 21297 55731
NA NA 1.0 1.3 1.0 1.0 1.1 1.3
PG 70-10 PG 64-22 PG 58-22
58-22 there was an observed 30% increase in the number of particles, while the PG 64-22 asphalt cement showed no increase in the number of particles. The results show that there is some splitting of the RAR particles (either by pure volume increase or more likely by a combination of particle swelling, resultant softening, and then agitation during the mixing process) thus increasing the total volume of rubber particles above that estimated from swelling alone.
4.2.3. Estimating RAR volume content As noted earlier, although the RAR was incorporated into the asphalt cement at a dosage rate equivalent to 17.6% by volume,
62
J.R. Medina, B.S. Underwood / Construction and Building Materials 150 (2017) 56–65
4.3. Stiffening mechanism of the asphalt cement
Table 3 Summary of volume fraction of rubber particles in RAR modified asphalts. Binder
PG 70-10 PG 64-22 PG 58-22
Volume fraction (%) Initial
After swelling (no splitting)
After splitting and swelling
17.6 17.6 17.6
19.8 23.2 24.2
20.3 23.2 25.8
the mechanism of particle swelling and splitting during the mixing process results in an increase in volume occupied by the particles. To estimate this increase in volume the results from the SEM and counting trials were combined and used to adjust the original 17.6% volume content. During this process it was assumed that the change in volume for particles greater than 0.297 mm was equal to that of the 0.420 mm particles, and that for particles less than 0.297 mm it was equal to the swelling from the experiments on the 0.297 mm particles. Similarly it was assumed that the volume of particles greater than 0.297 mm increased by a factor of 1.15, 1.40 and 1.24 for asphalt cements PG 70-10, PG 64-22 and PG 58-22, while particles smaller than or equal to 0.297 mm increased by a factor of 1.18, 1.47 and 1.66 respectively for the same asphalts. With this assumption, and considering the splitting of the rubber particles, and using the typical gradation of RAR, the overall increase in volume was calculated to be 15.8%, 41.5% and 49.3% for binders PG 70-10, PG 64-22 and PG 58-22 respectively. Accounting for this much swelling and splitting of rubber particles of the RAR in the rubberized asphalt blend, the actual volume fraction is summarized in Table 3.
The test results demonstrate that the addition of RAR increase the modulus of the asphalt cement by a factor of 2.6–4.9 depending on the temperature and asphalt cement. However, these findings provide no fundamental insight in the mechanisms that cause this increase. To improve the understanding of the mechanisms that contribute to the stiffening, the Hashin and Christensen models were applied using the measured data and known volume contents (inclusive of swelling and splitting). For these models the matrix phase properties were estimated using the modulus of either the virgin asphalt cement or the D-RAR asphalt. The volumetric content and modulus of the inclusions phase (the rubber particles) was taken from that given in Table 3 and from published literature [28] respectively. The results are shown in Fig. 6 when virgin asphalt is set as the matrix phase and Fig. 7 when D-RAR is the matrix. When treating the matrix phase as the virgin asphalt, both models underestimated the results from 28% to 64% for all three asphalt binders, Fig. 6. However, based on the experimental results shown in Fig. 2 it is known that in the presence of the rubber particles the binding matrix properties are changed from those of the virgin asphalt. This effect was postulated by Bahia et al. [12], and the D-RAR experimental data provide some proof that it is in fact true. With this hypothesis in mind, the more accurate reflection of the binding matrix as it exists in the composite would be that of the D-RAR. The results when using the D-RAR moduli in the two models are shown in Fig. 7 and it can be seen that for the PG 70-10 asphalt cement, the Christensen model overestimates the modulus by 10.5% at 64 °C and underestimates by 0.8% and
14
14 12
8 6
10
8 6
4
4
2
2
0
0 64
70
58
76
64
70
Temperature (°C)
Temperature (°C)
(b)
(a) 14
Virgin D-RAR RAR Modified Christensen Hashin
12 10
|G*| (kPa)
|G*| (kPa)
10
Virgin D-RAR RAR Christensen Hashin
12
|G*| (kPa)
Virgin D-RAR RAR Christensen Hashin
8 6 4
2 0
52
58
64
Temperature (°C)
(c) Fig. 6. Micromechanical simulation results when using Virgin asphalt as the matrix phase; (a) PG 70-10, (b) PG 64-22, and (c) PG 58-22.
63
J.R. Medina, B.S. Underwood / Construction and Building Materials 150 (2017) 56–65
18
18
Virgin D-RAR RAR Christensen Hashin
|G*| (kPa)
14 12 10 8
16
Virgin D-RAR RAR Christensen Hashin
14
|G*| (kPa)
16
12
10 8
6
6
4
4
2
2
0
0 64
70
58
76
64
Temperature (°C)
Temperature (°C)
(a)
(b)
70
18 16
Virgin D-RAR RAR Christensen Hashin
|G*| (kPa)
14 12 10 8 6 4 2 0 52
58
64
Temperature (°C)
(c) Fig. 7. Micromechanical simulation results when using the D-RAR asphalt as the matrix phase; (a) PG 70-10, (b) PG 64-22, and (c) PG 58-22.
18
Predicted Modulus (kPa)
16 14 12 10 8 6 Line of Equality 2 Hashin - R = 0.95 2 Christensen - R = 0.84
4 2 0
0
2
4
6
8
10
12
14
16
18
Measured Modulus (kPa) Fig. 8. Measured dynamic modulus and predicted shear modulus calculated using the Hashin and the Christensen models.
7.32% at 70 °C and 76 °C. For the PG 64-22 asphalt this model overestimates the modulus by 11% at 58 °C and underestimates the modulus by 2% and 10% at 64 °C and 70 °C respectively. Results for the PG 58-22 asphalt cement Christensen model overestimates the results by 33, 18 and 12% for temperatures 52, 58, and 64 °C respectively. The results from Hashin model are also close to the measured modulus of the RAR modified asphalt cement. For the PG 70-10 asphalt cement, the model overestimates the modulus
by 4% and underestimates the modulus by 7% and 13% for temperatures of 64, 70 and 76 °C respectively. Moreover, the model overestimates the PG 64-22 asphalt binder stiffness at 58 °C by 2% and underestimates by 10% and 17% at temperatures of 64 and 70 °C respectively. The PG 58-22 asphalt cement overestimates the modulus at 52 and 58 °C by 19 and 5% and the percent difference at 64 °C is 0.4%. To compare the Hashin and the Christensen models, the results from both models are demonstrated in Fig. 8 by comparing the |G⁄| values calculated from the models to the measured |G⁄|. In general, both models show good agreement with the measured data. The Hashin model fit the results well having a coefficient of determination of 0.95 while the Christensen model is 0.84. Based on the temperature range used in this study, the models over-predict the shear modulus at lower temperatures and under-estimates the values at higher temperatures. It is seen that as the temperature decreases, the error increases. The authors believe that this increasing error is due, in part to the limitation of some micromechanical models which are based on physical properties of the constituents of the composite and disregard physicochemical effects, interaction between particles and gradation, and make quasistatic approximations of the viscoelastic properties of the binder. This latter issue can be readily addressed through derivation of a linear viscoelastic form of the classic micromechanical models. However, for the purpose of quantifying mechanisms and designing blends to control these mechanisms the simplification appears reasonable. Despite the slight underprediction of moduli at lower temperatures these models produce what are judged to be reasonable results (less than 19% error in the case of the Hashin model and approximately 30% error in the case of the Christensen model)
64
J.R. Medina, B.S. Underwood / Construction and Building Materials 150 (2017) 56–65
that permit quantification and with future study designing these composites [38]. 5. Conclusions In this study, rheology test was performed on three different asphalt cements modified with RAR to have a better understanding of the swelling and splitting mechanisms that take place during the blending process of asphalt cement and crumb rubber. SEM images and manual particle count were taken of the crumb rubber particles before and after blending to estimate the swelling and splitting potential, and evaluate the stiffening using micromechanical models. More specific conclusions that can be drawn from this research study are as follow: The RAR modified asphalt cement showed in all cases an increase in modulus with respect to the Virgin asphalt cements used. Drained RAR asphalt cement for all three cases also showed an increase in modulus with respect to Virgin asphalt confirming the hypothesis that the lighter fractions are absorbed by the crumb rubber and the residual binder is left with a higher concentration of asphaltenes. The rubber particles in the PG 58-22 modified asphalt cement showed the greatest increase in volume by a factor of 1.49 meaning that the actual volume in the blend increased from 17.6% to 25.8%, while the rubber particles showed an increase in volume by a factor of 1.43 and 1.15 which represent an increase in volume content in the blend from 17.6% to 23.2% and 20.3% for the PG 64-22 and PG 70-10 asphalt cement respectively. These results are expected since it is assumed that softer asphalt cement would have higher lighter fraction content and therefore more interaction between the rubber particles and the asphalt cement. The two micromechanical models included in this paper showed a good agreement with the measured modulus of the RAR modified asphalt cement. The Hashin model yielded a difference in the range of 0%–19% for all three asphalt cement and temperatures. The Christensen model yielded a difference of 1%–33%. The trend observed in both models was that at lower temperatures the models overestimated the measured modulus values but as the temperature increases the predictions from the models underestimated the results. In this study, the authors were able to quantify the swelling and splitting of the rubber particles, and integrate these finding in micromechanical models to accurately predict the dynamic shear modulus of the RAR modified asphalt cement. With a better insight of the interaction between the RAR particles and the asphalt cement it is possible to use these models to predict engineering properties with the objective to maximize the potential benefits of crumb rubber modified asphalt blends. Acknowledgements Authors will like to thank Dr. George Souza, Mr. Robert McGennis from Holly Frontier and Dr. Kamil Kaloush for providing the materials and technical support to develop this research study. Appendix A. Composite and micromechanical models used in the analysis: Einstein model,
jG jc ¼ 1 þ 2:5C v jG jb
ðB1Þ
Roscoe Model,
jG jc ¼ ð1 C v Þ2:5 jG jb
ðB2Þ
Christensen Model, h i G Gp 7=3 A ¼ 8 Gmp 1 ð4 5mm Þk1 C 10=3 v 2 63 Gm 1 k2 þ 2k1 k3 C v þ G Gp 2 252 Gmp 1 k2 C 5=3 v 50 Gm 1 7 12mm þ 8mm k2 C v þ 4ð7 10mm Þk2 k3 ðB3-1Þ
h i G Gp 7=3 B ¼ 4 Gmp 1 ð1 5mm Þk1 C 10=3 v þ 4 63 Gm 1 k2 þ 2k1 k3 C v G Gp 504 Gmp 1 k2 C 5=3 v þ 150 Gm 1 ð3 mm Þmm k2 C v þ 3ð15mm 7Þk2 k3 ðB3-2Þ
h i G G C ¼ 4 Gmp 1 ð5mm 7Þk1 C 10=3 2 63 Gmp 1 k2 þ 2k1 k3 C 7=3 v v þ G Gp 2 252 Gmp 1 k2 C 5=3 v þ 25 Gm 1 mm 7 k2 C v ð7 þ 5mm Þk2 k3 ðB3-3Þ Gp Gp 1 49 50mp mm þ 35 mp 2mm þ 35 2mp mm Gm Gm
k1 ¼
ðB3-4Þ k2 ¼ 5mp
Gp Gp 8 þ7 þ4 Gm Gm
ðB3-5Þ
Gp ð8 10mm Þ þ ð7 5mm Þ Gm
k3 ¼
ðB3-6Þ
Mori-Tanaka model,
jG jc 5C v ðGr jG jb Þ ¼1þ 5jG jb þ 2ð1 C v ÞðGr jG jb Þ jG jb
ðB4Þ
Differential Method,
jG jc ¼ Gr ð1 C v ÞðGr jG jb Þ Z
jG jc jG jb
2=5 ðB5Þ
a max
Gc ¼ a min
Gc ¼
jG jc ðaÞdPðaÞ
Nþ1 1X ½Gc ðai Þ þ Gc ðaiþ1 Þ½Pðai Þ P ðaiþ1 Þ 2 i¼1
ðB6Þ
ðB7Þ
References [1] M. Abdelrahman, Controlling low temperature properties of asphalt rubber binders, Proceedings of the 2nd Transportation Specialty Conference, Canadian Society for Civil Engineering, Halifax, 1998. [2] T.C. Billiter, R.R. Davison, C.J. Glover, J.A. Bullin, Physical properties of asphalt rubber binders, Transportation Research Record 1586, TRB, Washington, DC, 1997, pp 51–59. [3] J.S. Coplantz, M.T. Yapp, R.N. Finn, Review of relationships between modified asphalt properties and pavement performance, SHRP-A-631, 1993. [4] F.L. Roberts, P.S. Kandhal, E.R. Brown, D.Y. Lee, T.W. Kennedy, Hot mix asphalt materials, mixture design and construction, first ed., NAPA Eduacation Foundation, Lanham, Maryland, 1991. 68–382. [5] J.L. McQuillen, H.B. Takallou, R.G. Hicks, D.C. Esch, Economic analysis of rubber modified asphalt mixes, J. Transp. Eng. 114 (1988) 259–277. [6] S.A. Dantas Neto, M.M. Farias, J.C. Pais, P.A. Pereira, L. Picado Santos, Behavior of asphalt rubber hot mixes obtained with high crumb rubber content, in: Proceedings of the Asphalt Rubber 2003 Conference, Brasilia, Brazil, 2003. [7] G.B. Way, Asphalt-Rubber 45 Years of Progress, in: Proceedings Asphalt Rubber Conference, Germany, 2012. [8] S. Kocevski, S. Yagneswaran, F. Xiao, V.S. Punith, D.W. Smith, S. Amirkhanian, Surface modified ground rubber tire by grafting acrylic acid for paving applications, Constr. Build. Mater. 34 (2012) 83–90.
J.R. Medina, B.S. Underwood / Construction and Building Materials 150 (2017) 56–65 [9] Gordon D. Airey, Mujibur M. Rahman, Andrew C. Collop, Absorption of bitumen into crumb rubber using the basket drainage method, Int. J. Pavement Eng. 4 (2) (2003) 105–119. [10] D. Dong, X. Huang, X. Li, L. Zhang, Swelling process of rubber in asphalt and its effect on the structure and properties of rubber and asphalt, Constr. Build. Mater. 29 (2012) 316–322. [11] Feipeng Xiao, J. Putman Bradley, N. Amirkhanian Serji, Laboratory investigation of dimensional changes of crumb rubber reacting with asphalt binder, Road Mater. Pavement Des. (2006). [12] Hussain U. Bahia, Robert Davies, Effect of crumb rubber modifiers (CRM) on performance related properties of asphalt binders, Asphalt Pavement Technol. 63 (1994). 414-414. [13] M.E. Kutay, H. Ozturk, Internal structure characteristics of crumb rubber modified asphalt binders: an analysis using 3D X-ray Microtomography Imaging, in: Proceedings Asphalt Rubber Conference, Germany, 2012. [14] J. Shen, S. Amirkhanian, F. Xiao, B. Tang, Influence of surface area and size of crumb rubber on high temperature properties of crumb rubber modified binders, Constr. Build. Mater. 23 (2009) 304–310. [15] S. Ahmed, F.R. Jones, A review of particulate reinforcement theories for polymer composites, J Mater. Sci. 25 (12) (1990) 4933–4942. [16] E. Hesami, D. Jelagin, N. Kringos, B. Birgisson, An empirical framework for determining asphalt mastic viscosity as a function of mineral filler concentration, Constr. Build Mater. 35 (2012) 23–29. [17] R. Roscoe, The viscosity of suspensions of rigid spheres, Br. J. Appl. Phys. 23 (9) (1952) 1020–1024. [18] Akshay Gundla, Use of Micro-Mechanical Models to Study the Mastic Level Structure of Asphalt Concretes Containing Reclaimed Asphalt Pavement (MS Thesis), Arizona State University, Tempe, 2014. [19] Hussain U. Bahia, Robert Davies, Factors Controlling the Effect of Crumb Rubber on Critical Properties of Asphalt Binders vol. 64 (1995) 130–162. [20] B.S. Underwood, Y.R. Kim, A four phase micro-mechanical model for mastic modulus, Mech. Mater. 25 (2014) 13–33. [21] N. Shashidhar, A. Shenoy, On using micromechanical models to describe dynamic mechanical behavior of asphalt mastics, Mech. Mater. 34 (10) (2002) 657–669. [22] Jamrah Anas, M. Emin Kutay, Sudhir Varma. Backcalculation of Swollen Crumb Rubber Modulus in Asphalt Rubber Binder and its Relation to Performance, in: 94th Transportation Research Board Annual Meeting. United States, 2015.
65
[23] E. Aigner, R. Lackner, C. Pichler, Multiscale predictions of viscoelastic properties of asphalt concrete, J. Mater. Civil Eng. 21 (12) (2009) 771–780. [24] R.M. Christensen, K.H. Lo, Solutions for effective shear properties in three phase sphere and cylinders models, J. Mech. Phys. Solids 27 (4) (1979) 315– 330. [25] A. Einstein, New determination of molecular dimensions, Ann. Phys. 19 (2) (1906) 289–306. [26] R. Roscoe, The viscosity of suspensions of rigid spheres, Br. J. Appl. Phys. 3 (8) (1952) 267. [27] Y. Benceniste, A new approach to the application of mori-tanaka’s theory in composite materials, Mech. Mater. 6 (2) (1987) 147–153. [28] X. Shu, B. Huang, Michromechanics-based dynamic modulus prediction of polymeric asphalt concrete mixtures, Compos. Part B Eng. 39 (4) (2008) 704– 713. [29] X. Shu, B. Huang, Predicting dynamic modulus of asphalt mixtures with differential method, Road Mater. Pavement Des. 10 (2) (2009) 337–359. [30] A. Lion, Thixotropic behavior of rubber under dynamic loading histories: experimental and theory, J. Mech. Phys. Solids 46 (5) (1998) 895–930. [31] E. Guth, Theory of filler reinforcement, J. Appl. Phys. 16 (1) (1945) 20–25. [32] Y. Lin, H. Sheng-Jye, Temperature predictions of rolling tires by computer simulation, Math. Comp. Simul. 67 (3) (2004) 235–249. [33] Carl Thodesen, Khaldoun Shatanawi, Serji Amirkhanian, Effect of crumb rubber characteristics on crumb rubber modified (CRM) binder visocity, Constr. Build. Mater. 23 (1) (2009) 295–303. [34] J. Medina, B.S. Underwood, Use of particulate composite models and crumb rubber swelling to estimate stiffness of rubberized asphalt binders, in: Proceedings ISAP Symposium, 2016. [35] J Richards Willis, Clayton Plemons, Pamela Turner, Carolina Rodezno, Tyler Mitchell, Effect of Ground Tire Rubber Particle Size and Grinding Method on Asphalt Binder Properties, National Center for Asphalt Technology, Auburn, 2012. [37] W. Pabst, E. Gregorova, Characterization of Particles and Particle Systems, ICT Prague, 2007. [38] H.M. Yin, W.G. Buttlar, G.H. Paulino, H. Di Benedetto, Assesment of existing micro-mechanical models for asphalt mastics considering viscoelastic effects, Road Mater. Pavement Des. 9 (1) (2008) 37–57.