Delft University of Technology

Study of Asphalt Binders Fatigue with a New Dynamic Shear Rheometer Geometry

Apostolidis, Panos; Kasbergen, Cor; Bhasin, A.; Scarpas, Athanasios; Erkens, S.

Publication date
2018
Document Version
Accepted author manuscript
Published in
Transportation Research Board 97th Annual Meeting

Citation (APA)
Apostolidis, P., Kasbergen, C., Bhasin, A., Scarpas, A., & Erkens, S. (2018). Study of Asphalt Binders
Fatigue with a New Dynamic Shear Rheometer Geometry. In Transportation Research Board 97th Annual
Meeting: 2018-1-7 to 2018-1-11, Washington DC, United States [TRB 18-02473] (Transportation Research
Record). Transporation Research Board (TRB).
Important note
To cite this publication, please use the final published version (if applicable).
Please check the document version above.

Copyright
Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent
of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.
Takedown policy
Please contact us and provide details if you believe this document breaches copyrights.
We will remove access to the work immediately and investigate your claim.

This work is downloaded from Delft University of Technology.

For technical reasons the number of authors shown on this cover page is limited to a maximum of 10.
 Study of Asphalt Binders Fatigue with a New Dynamic Shear Rheometer
Geometry
1 P. Apostolidis1 , C. Kasbergen1 , A. Bhasin2, A. Scarpas1,3 , S. Erkens1
2
1
3 Section of Pavement Engineering
4 Faculty of Civil Engineering and Geosciences,
5 Delft University of Technology
6 Stevinweg 1, 2628 CN Delft, the Netherlands
7
2
8 Department of Civil, Architectural and Environmental Engineering
9 University of Texas at Austin
10 301 E Dean Keeton Stop, C1761 Austin, Texas 78712, the USA
11
3
12 Department of Civil Infrastructure and Environmental Engineering
13 Khalifa University of Science and Technology
14 Abu Dhabi, United Arab Emirates
15
16 Corresponding author:
17 P. Apostolidis
18 E-mail: [email protected]
19
20
21
22
23
24
25
26
27
28
29
30 Total Number of Words
31
Words in abstract = 229 words
Words in text = 3949 words
Words in references = 730 words
Figures (10x250) = 2500 words equivalent
Total = 7408 words equivalent
32
33 Submitted for presentation for the 97nd meeting of the Transportation Research Board and
34 publication for the Transportation Research Record: Journal of the Transportation Research
35 Board.
36

1
37 Abstract: With the effort to predict precisely the lifetime of asphalt binders and subsequently
38 optimize their utilization in a more economical way, the objective of this study was to introduce
39 a new methodology to improve the fatigue characterization of asphalt binders through a new
40 dynamic shear rheometer (DSR) sample testing geometry. Initially, numerical analyses were
41 performed to study the geometry-related issues of standard DSR sample on time sweep tests and
42 assisted on the effort to increase the understanding of DSR damage phenomena of asphalt
43 samples. On the basis of these numerical analyses, a new testing geometry, the parallel hollow
44 plate, was developed and its test results compared with the standard sample testing geometry. A
45 single type of asphalt binder was assessed using amplitude sweep tests. The obtained results
46 demonstrated a significant difference between the fatigue of the two sets of DSR sample
47 geometries. On the basis of these, time sweep tests were conducted for the same sample
48 geometries and the results demonstrated that the new testing geometry yields material response
49 consistency under different loading conditions. The lifetime prediction of the standard parallel
50 plates showed a significant difference with the newly developed DSR sample testing geometry
51 by overestimating the total number of cycles until asphalt binder failure. The new testing
52 geometry allowed the isolation of the damage area of asphalt binder by localizing the shear
53 stresses in the samples’ periphery.
54
55

2
 56 1. INTRODUCTION
57 Due to the extremely complex nature of asphalt binders, it is difficult for the infrastructure
58 designers to accurately predict the lifetime of the pavement structures. Taking into account the
59 higher and heavier traffic on the highways in the last decades, the implementation of new asphalt
60 binders for the highway network has been increased remarkably resulting in higher initial costs
61 for the pavement construction. Also, difficulties have appeared to estimate serviceability and to
62 plan maintenance operations during pavements service life. Various parameters affect the
63 performance prediction of asphalt binders and instead of the progress in the testing techniques,
64 the challenge of precisely characterizing the binders fatigue life still needs to be addressed.
65 Fatigue damage as one of the main asphalt binders distress modes can be described as the
66 material degradation process because of repeated loading by which the micro-cracks grow and
67 the coalescence to macro-cracks. Typically, fatigue in asphalt mixes is studied by subjecting the
68 test material to some form of cyclic stresses at a lower level than the ultimate strength and then
69 determining the relative change in their mechanical properties, such as stiffness and strength.
70 Therefore, having a test method that can predict the mechanical degradation of material, will
71 allow the understanding of the exact damage mechanisms in detail and subsequently to optimize
72 the utilization of asphalt binders in a more feasible way.
73 However, the asphalt binder fatigue characterization is not an uncharted territory for the
74 paving industry. Several laboratory studies have been conducted to provide understanding of the
75 degradation mechanism due to repeated stresses and ranking of the binders’ susceptibility to
76 resist these stresses. Unfortunately, the results from the tests are not predicting precisely the field
77 performance of asphalt mixes. As a result, the need for improvement of testing methods for
78 quality control of asphalt binders in terms of lifetime estimation has been increasing. Dynamic
79 shear rheometer (DSR) has been introduced to be used for fatigue characterization binders (1-3).
80 Nevertheless, a satisfactory link between the measured binder fatigue response with using DSR
81 and the potential field material performance over a range of various operational conditions is still
82 under investigation.
83 Within this framework, a study has been initiated to evaluate a potentially more appropriate
84 DSR fatigue testing method. The new Parallel Hollow Plate (PHP) system was designed and
85 developed with an outer diameter of 25 mm, as the standard geometry of Parallel Plates (PP) of
86 DSR, but with a concentric hollow area of 19 mm diameter and 0.1 mm depth. After filling the
87 inner hollow area with a silicon paper, the new testing system was used to explore the impact of
88 mechanical performance of asphalt binder. For the selected new geometry system, after carrying
89 out assessment of the repeatability of the test results, different dynamic shear measurements were
90 conducted to evaluate the material response. The experimental results demonstrated the
91 important variations on the binder performance at low and high cyclic torque level tests between
92 the new and the standard DSR apparatus. This comparison underlines the significance of the
93 geometry for DSR plates for a more accurate material characterization and the upcoming need to
94 minimize the geometry-related issues by localizing the shear damage in the tested material.
95
96 2. FATIGUE IN ASPHALT BINDER
97 Fatigue damage in asphalt is the material degradation due to repeated loading by which the
98 cracks grow and the material losses its capability to resist more loads. Significant effort has been
99 spent on evaluating the asphaltic materials fatigue life and thus several methods have been
100 developed through this process. These methods differ mainly in terms of the fatigue damage
101 approaches and testing configurations, such as the sample geometry, loading conditions, etc.

3
102 Herein, emphasis is given in assessing the fatigue performance of asphalt binders and for this
103 reason the state-of-the-art of DSR utilization as fatigue characterization tool is discussed.
104
105 2.1 Fatigue Damage Approaches
106 Fatigue life of asphalt binders has been thoroughly examined and several approaches, such as,
107 energy-related, mechanistic approaches and phenomenological, have been utilized to evaluate the
108 material response under cyclic load repetitions and to determine the remaining life of the
109 material.
110 Among the energy-related approaches, the energy ratio as function of the number of cycles
111 and the complex shear modulus for the different controlled modes has been applied as fatigue
112 life criterion (4). Especially, in the stress controlled mode, fatigue life of the material is defined
113 as the point when the energy ratio reaches the peak in the relationship of energy ratio versus the
114 number of cycles. On the other hand, in the strain controlled mode, the fatigue life is defined as
115 the number of load cycles at which the slope of energy ratio deviates from a straight line.
116 Another energy approach is the dissipated energy ratio which is defined as the ratio of the
117 difference between the dissipated energy for the successive load cycles to the dissipated energy
118 of the previous cycles (5, 6). The dissipated energy ratio is the area inside the hysteric loop (7, 8)
119 and the fatigue life of the material is considered as the transition point where the dissipated
120 energy ratio starts to increase rapidly from an approximately constant value (6). Similarly, the
121 dissipated strain energy approach has been used by converting the actual strain to an equivalent
122 pseudo-strain in order to remove the viscoelastic contribution (2, 9) and to quantify the damage
123 manifestation using mechanistic approaches, such as continuum damage and fracture mechanics
124 (10-12).
125 Finally, phenomenological approaches are the most used to define the fatigue life. One
126 example of such an approach is the determination of the fatigue as the number of cycles when
127 the complex modulus decreases to 10 % and 50% of the initial complex modulus for stress and
128 strain controlled testing modes, respectively (13, 14). However, the failure criterion of 50 %
129 complex modulus reduction is irrelevant to the damage accumulation since this value is arbitrary
130 and varies at different loading modes. Others considered fatigue life of asphalt as the point at
131 which the stress level changes rapidly (15) but this approach is sensitive to the test loading
132 conditions. In this study, the total number of fatigue cycles until complete failure of the sample
133 or end of test is used as fatigue life criterion (16).
134
135 2.2 DSR Fatigue Damage Characterization
136 The DSR is commonly used as a standard performance testing equipment to characterize the
137 viscoelastic properties of asphalt binders (17-19). Additionally, to evaluate the fatigue damage
138 mechanism and to predict the fatigue life in asphalt binders, the utilization of DSR has been
139 introduced using the oscillatory time sweep (TS) test (1-3). Previous researchers have
140 demonstrated that damage initiates at the outer periphery of the material and propagates through
141 the sample with increasing number of loading cycles. Thus, damage results in a reduction of the
142 radius of the test sample. Specialized imaging techniques have been used to demonstrate the
143 fatigue damage during DSR testing and the obtained images clearly demonstrate non- uniform
144 damage with fracture at the outer edge of the testing plates and an intact center (2, 3, 20).
145 Others who also studied the phenomena of fatigue with DSR have shown damage propagation
146 as hairline cracks propagating towards the center accompanied by modulus decrease (21). The
147 fatigue damage mechanism does not include the internal damage because the edge fracture is

4
148 dominant, especially in oscillatory TS tests (5). However, these are not the only issues that are
149 encountered with the standard DSR test methods using a parallel plate; also the accuracy of
150 complex modulus is limited since the generated radial stress field is non- linear. Many aspects of
151 DSR fatigue characterization are elaborated with approximations and extrapolations analogous to
152 how Ptolemy used epicycles to explain the planets movements around the earth. The need for
153 improving the fatigue testing methods and the asphalt binders quality is urgently required
154 nowadays to resolve the inaccurate use and interpretation of DSR and to link the DSR measured
155 response of binders with the field pavement performance. In the following section, numerical
156 analyses are performed to study the geometry related effects of DSR sample testing on fatigue
157 damage. Also, the numerical simulations of fatigue damage will assist in the effort to increase
158 the understanding of damage phenomena of asphalt samples during DSR TS tests and to further
159 optimize the testing configurations for obtaining more realistic material properties.
160
161 3. NUMERICAL SIMULATION OF DSR FATIGUE DAMAGE
162
163 3.1 Model Parameters Determination
164 A damage model was developed to illustrate the damage distribution of asphalt binder during a
165 DSR TS test. The material parameters that were required as an input were modelled based on a
166 linear viscoelastic response. The complex modulus values of asphalt binder were determined
167 from frequency sweep tests in the standard PP DSR system. These tests were carried out over a
168 temperature and frequency range from -10 o C to 60 o C and from 0.1 Hz to 100 Hz, respectively.
169 Instrument compliance was measured and accounted for in these measurements. The asphalt
170 binder used was a commonly applied binder for porous asphalt mixes in Dutch roads, the
171 penetration grade 70/100 unmodified bitumen. By employing the frequency-temperature
172 superposition principle, the master curve in the frequency domain was defined (reference
173 temperature of 20°C).
174
175 3.2 Continuum Damage Model
176 After determining the material parameters, with the Prony series coefficients (G∞, G i and ρi)
177 obtained by fitting the experimental data with the storage modulus, the relaxation modulus could
178 be expressed in the time domain as follows
179
𝑛

𝐺 (𝑡) = 𝐺∞ + ∑ 𝐺𝑖 𝑒 −𝑡/𝜌𝑖 (1)

𝑖=1
180
181 where G(t) is the shear relaxation modulus in time domain, t is the loading time, G∞ is the long-
182 time equilibrium modulus, Gi are the spring constants in the generalized Maxwell model, ρi are
183 the relaxation times and n is the number of Maxwell components in the generalized model.
184 If it is assumed that the Poisson’s ratio of binder is time independent and that the material is
185 isotropic, the following expression that relates the G(t) to E(t) can be written as
186
𝐸 (𝑡) = 2 ∙ 𝐺 (𝑡) ∙ (1 + 𝜈) (2)
187
188 where E(t) is the relaxation modulus and 𝜈 is the Poisson’s ratio.

5
189 In continuum mechanics, the damage is defined as a function of any micro- mechanical change
190 that develops in a homogeneous continuum media. To include damage in the above described
191 material model the following damage evolution equation was proposed based on total dissipated
192 energy as
193
𝜉(𝑡) = 1 − 𝑒𝑥𝑝(−𝑘 ∙ 𝑊 (𝑡)𝑟 ) (3)
194
195 where 𝜉 is damage degradation of asphalt binder, t is time, W is the total dissipated energy and
196 both k and r are damage rate parameters. In incremental form Eq. (4) can be written as
197
𝜉(𝑡 + ∆𝑡) = 1 − (1 − 𝜉 (𝑡)) ∙ 𝑒𝑥𝑝(−𝑘 ∙ (𝑊(𝑡 + ∆𝑡)𝑟 − 𝑊 (𝑡)𝑟 )) (4)
198
199 where ∆𝑡 is the time increment. If the value of 𝜉 is zero it indicates no damage and if the value of
200 𝜉 is one it resembles full damage.
201 The total energy dissipation W can be computed in incremental form as
202
𝑛 𝑡+∆𝑡
𝑒𝑓𝑓
𝑊 (𝑡 + ∆𝑡) = 𝑊 (𝑡) + ∑ ∫ 𝑆𝑖 (𝜏): 𝐸̇ (𝜏) 𝑑𝜏 (5)
𝑖=1 𝑡
203
𝑆𝑖𝑒𝑓𝑓 (𝜏) = (1 − 𝜉 (𝜏)) ∙ 𝑆𝑖 (𝜏) (6)
204
205 where i is the index of the Maxwell component, n is the number of Maxwell components, 𝜏 is the
206 time integration variable, 𝑆𝑖 is the second Piola-Kirchhoff stress in the i-th Maxwell component,
𝑒𝑓𝑓
207 𝑆𝑖 is the effective or remaining second Piola-Kirchhoff stress in the i-th Maxwell component
208 after damage has been taken into account and 𝐸̇ is the total Lagrange-Green strain rate.
209 Using the midpoint integration rule Eq. (5) can be simplified to
210
𝑚
𝑆𝑖𝑒𝑓𝑓 (𝑡 + ∆𝑡) + 𝑆𝑖𝑒𝑓𝑓 (𝑡)
𝑊(𝑡 + ∆𝑡) = 𝑊(𝑡) + ∑ ( ) (𝐸 (𝑡 + ∆𝑡) − 𝐸(𝑡)) (7)
2
𝑖 =1
211
212 3.3 Numerical Implementation
213 The CAPA 3D system was utilized. Three user-defined 3D finite-element (FE) meshes were
214 created to study the damage distribution and the localization of asphalt sample deterioration in a
215 sinusoidal (oscillating) loading mode during a TS DSR test, Fig. 1. The first FE mesh
216 representing the standard DSR geometry of 2400 cubic elements was developed. This DSR
217 geometry comprises the two parallel plates in which the asphalt binder is located in between with
218 the top plate being subjected to torsion and the bottom plate being fixed. Similarly, the second
219 FE mesh of DSR geometry with a ring as top plate with inner and outer diameter o f 19 mm and
220 25 mm, respectively, was created of 2200 elements. This configuration was named one ring-type
221 testing system. Also, a third mesh called two rings-type testing geometry comprising of two rings
222 instead solid plates of 2000 elements was generated.
223 To assess the fatigue damage behaviour under the same applied torque (0.245 Nm), the load
224 level was converted to shear stress (τ) based on the testing geometries. According to the elastic

6
225 torsional theory, the shear stress (𝜏) calculations for plate-type and ring-type testing geometries,
226 Fig. 2, are given in the following equations
227
2𝑇
𝜏𝑝𝑙𝑎𝑡𝑒 −𝑡𝑦𝑝𝑒 = (8)
𝜋𝑅0 3
228
𝑇𝑅0
𝜏𝑟𝑖𝑛𝑔−𝑡𝑦𝑝𝑒 =
𝜋(𝑅0 4 − 𝑅𝑖 4 ) (9)
( )
2
229
230 where T is torque, R0 is the outer radius and Ri is the inner radius of the plate.
231
232 Numerical Predictions
233 In Fig. 3, the damage distribution within the specimen was obtained after subjecting the standard
234 plate-type model to a torque of 0.245 Nm at 10 Hz frequency. The results from this analysis
235 demonstrate that the material degradation during a PP DSR TS test differs across the sample
236 radius. Specifically, the top part of Fig. 3 visualizes the damage progress in time for the first six
237 TS cycles. With increased loading, it is apparent that the damage, as reflected by the different
238 colors in the figure, is concentrated in the outer periphery. Plotting the damage values versus
239 time gives the bottom graph of Fig. 3, where the damage increases more rapidly in the points
240 closer to the sample’s periphery. As can be observed, the damage of the inner area of binder is
241 not the same with the edge or close to the edge locations. The damage rate shows the inner part
242 of the testing binder is not affected significantly by the torsional induced damage of the plates.
243 Therefore, these results corroborate the previously mentioned mechanism of damage initiation at
244 the outer periphery of sample and the almost intact centre during a DSR fatigue test (5).
245 Fig. 4 compares the performance of the standard DSR geometric configuration and of the two
246 ring-type geometries. The new geometries show a higher magnitude of damage localized on the
247 ring area than the plate-type sample geometry for a given number of loading cycles (bottom of
248 Fig. 4). This difference is explained by the fact that the area that resists the applied torque is
249 limited in the ring-type geometry compared to the standard system. Additionally, the impact of
250 top rotating part on the shear stress field and the subsequent damage propagation generated by
251 the applied torque across the sample radius is shown in Fig. 5. For the ring-type geometries, the
252 stress flow field appeared only on the outer sample periphery with very limited and no inward
253 stress propagation for the one ring-type and two rings-type sample geometries, respectively. The
254 edge damage phenomenon to the ring- type geometry is occurs earlier than the plate-type sample
255 geometry on account of the higher stress magnitude.
256 Additionally, the stress and damage difference across the sample thickness at three different
257 points at a certain time period is demonstrated in Fig. 6. It is obvious that the standard geometry
258 shows significant variation in damage across the sample thickness at all these points. The one
259 ring-type geometry has a bit less damage at the same location than the damage in the two rings-
260 type testing configuration. All these predicted results reinforce recent studies on the lack of
261 accuracy of standard DSR sample testing geometry and the limitations of this system on
262 providing true material properties (22-25).
263

7
264
265 4. IMPROVING DSR FATIGUE DAMAGE CHARACTERIZATION
266 On the basis of the evidence from past research and the predicted results from implementing the
267 previously described continuum damage model, the main objective of this part of the study is to
268 introduce a new methodology to accurately characterize the fatigue performance of asphalt
269 binders through a new DSR testing system. Different dynamic shear measurements were
270 performed to assess the material response by using the standard PP and the newly developed
271 PHP configuration. The ability of the new geometry to characterize the asphalt binder fatigue has
272 been evaluated as well.
273
274 4.1 Test Methods
275 The standard DSR sample geometry is the PP with smooth polished surfaces with a typical
276 diameter of 25 mm. A new sample testing geometry was designed and manufactured on the basis
277 of the previous numerical analyses. Similar to the one ring-type geometry, the new sample
278 geometry named Parallel Hollow Plates (PHP) has an outer diameter of 25mm with a concentric
279 hollow space of 19 mm diameter and 0.1 mm depth. The testing procedure is shown in Fig. 7.
280 The DSR setup was utilized for testing with the conventional PP and the new PHP, both with 1
281 mm gap in accordance with the Superpave specifications, and obtaining the material response.
282 After filling the inner hollow space of PHP with a silicon paper, the new testing system was used
283 to explore the impact of mechanical performance of asphalt binder. A zero gap between the
284 upper and lower plates was established and after reaching it, a 1 mm gap was set by moving the
285 plates apart.
286
287 Amplitude Sweep Measurements
288 For obtaining the dynamic material response for very short loading time, a varying torque signal
289 is applied with a fixed sinusoidal oscillatory frequency. In this study, a cyclic strain-controlled
290 torque was applied throughout the test causing a constant rotational strain. These DSR
291 experiments resulted in amplitude sweep results for the two different sample geometries at 35 o C
292 for further comparison. Also, these results were used to determine the linear viscoelastic range
293 and the level of applied torque of 10 Hz frequency for conducting the TS studies in the latter
294 step.
295
296 Time Sweep Measurements
297 The material damage manifests as a decrease in complex modulus and an increase in phase angle
298 in asphalt binder. In this study, the damage was quantified as the reduction in complex modulus
299 measured during the cyclic loading test with DSR. The TS torque-controlled loading mode was
300 used to evaluate the binder fatigue life and the performance difference between the two sample
301 testing geometries. During these tests, the samples were subjected to a sinusoidal loading mode
302 with a fixed frequency of 10 Hz at 35 o C.
303
304 4.2 Test Results
305
306 Amplitude Sweep Results
307 For the selected geometries, after carrying out assessment of the repeatability of the test results,
308 different dynamic shear measurements were conducted to evaluate the material response using an
309 amplitude sweep test. Fig. 8 depicts the variation in viscoelastic properties versus applied torque

8
310 at 10 Hz frequency and 35 o C. The effect of the new testing geometry is demonstrated as well.
311 The torque amplitude was increased in small amounts instead of large steps in each cycle. From
312 the data, it can be observed that the complex modulus drops and phase angle increases first when
313 the material was tested using the PHP configuration. The limited area in the outer periphery of
314 the PHP caused quicker degradation than the PP system when the applied torque was increased.
315 Thus, it is obvious that the material degradation rate is a function of the damaged area for an
316 amplitude sweep test and subsequently of the testing geometry.
317
318 Time Sweep Results
319 The fatigue life of asphalt binder is influenced by various factors, such as temperature, loading
320 level and frequency. In this study, the testing was done at 35 o C in which the initial complex
321 shear modulus was 0.5 MPa. Very different fatigue performances were observed between PP and
322 PHP geometries. By applying a torque level of 50 mNm, the complex shear modulus versus the
323 number of cycles is demonstrated Fig. 9. As expected, with increasing number of fatigue cycles,
324 the complex modulus of PHP dropped first since the tested area was limited indicating the faster
325 occurrence of damage. In addition, the shear modulus reported using the PP geometry is in fact
326 an average of the damaged periphery and the intact core.
327 Fig. 10 shows the fatigue life curves for PP and PHP DSR geometries. Here, the most
328 commonly applied fatigue life criterion is considered to be the number of loading cycles at which
329 the complex shear modulus reaches its lowest value Since failure happened only at the sample
330 periphery in the PHP system, PHP appeared to result in a shorter binder fatigue life for different
331 applied torque levels. The propagation of the micro-cracks from the edges to the internal area of
332 sample using the standard geometry produced more cycles in the TS tests. The TS results of the
333 newly developed sample testing geometry indicate the importance in characterizing the fatigue
334 performance accurately. According to these results, the fatigue resistance offered by the PP in a
335 TS test was influenced as an artifact of the geo metry. However, in addition to the various models
336 that are utilized to successfully predict fatigue life of material, the precise testing to obtain
337 accurate material properties should be a priority.
338
339 5. SUMMARY OF FINDINGS AND FUTURE WORK
340 From the perspective of pavement design, it is important to be able to predict the fatigue life of
341 an asphalt binder as a result of cyclic loading over time. This study proposed a new testing
342 geometry to more accurately predict the binders fatigue life. On the basis of analyses and test
343 results collected in this study, it could be stated that a less geometry-dependent measurement of
344 fatigue damage was achieved using the newly developed DSR configuration showing the
345 importance of using precise testing systems for the accurate material performance predictions.
346 The damage continuum model which was developed to demonstrate the non-uniform damage
347 distribution of asphalt binder subjected to sinusoidal loads with the standard sample geometry
348 showed that the damage was localized in the sample periphery, keeping the center intact. The
349 visualization of the concentration of damage during the fatigue testing with DSR was used as
350 evidence to manufacture a new testing configuration with an inner hollow space in the center of
351 the bottom plate. After conducting TS experiments using PP and PHP configurations, the fatigue
352 life predictions of the two geometries showed a significant difference with the edge damage
353 phenomenon happening earlier for the PHP than the damage with the PP. The very different
354 observed fatigue performances were derived by the fact that the new sample testing geometry

9
355 allowed the isolation of the material damage by localizing the shear stresses in the sample’s
356 periphery.
357 Further study is needed to maximize the damage by increasing the diameter of inner hollow
358 space and also the test loading and environmental conditions should be expanded to provide
359 more realistic fatigue predictions. Moreover, extensive experimental programs are required to be
360 performed in order to develop transferring functions to convert the results of the new geometry
361 to the results derived from the standard DSR geometry for modified and unmodified binders.

362 REFERENCES
363 1. Kim, Y.R., D.N. Little, R.L. Lytton. Use of Dynamic Mechanical Analysis to Evaluate the
364 Fatigue and Healing Potential of Asphalt Binders in Sand Asphalt Mixtures. Journal of the
365 Association of Asphalt Paving Technologists, 71, 2002, pp. 176–205.
366 2. Masad, E., V.T.F. Castelo Branco, D.N. Little, R. Lytton. A Unified Method for the
367 Analysis of Controlled-Strain and Controlled-Stress Fatigue Testing. International Journal
368 of Pavement Engineering, 9 (4), 2008, pp. 233–246.
369 3. Hintz, C., H. Bahia. Simplification of Linear Amplitude Sweep Test and Specification
370 Parameter. In Transportation Research Record, No. 2370, Transportation Research Board of
371 the National Academies, Washington, D.C., 2013, pp. 10–16.
372 4. Van Dijk, W. Visser. Energy Approach to Fatigue for Pavement Design. Journal of the
373 Association of Asphalt Paving Technologists, 46, 1977, pp. 1–40.
374 5. Keentok, M., S. Xue. Edge Fracture in Cone-Plate and Parallel Plate Flows. Rheological
375 Acta, 38, 1999, pp. 321-348.
376 6. Ghuzlan, K.A. S.H. Carpenter. Energy-derived, Damage-based Failure Criterion for Fatigue
377 Testing. In Transportation Research Record, No. 1723, Transportation Research Board of
378 the National Academies, Washington, D.C., 2000, pp. 141–149.
379 7. Ghuzlan, K.A., S.H. Carpenter. Fatigue Damage Analysis in Asphalt Concrete Mixtures
380 using the Dissipated Energy Approach. Canadian Journal of Civil Engineering, 33 (7),
381 2006, pp. 890–901.
382 8. Daniel, J.S., W. Bisirri, Y.R. Kim. Fatigue Evaluation of Asphalt Mixtures using Dissipated
383 Energy and Viscoelastic Continuum Damage Approaches. Journal of the Association of
384 Asphalt Paving Technologists, 73, 2004, pp. 557–583.
385 9. Bhasin, A., V.T. Castelo Branco, E. Masad, D.N. Little. Quantitative Comparison of Energy
386 Methods to Characterize Fatigue in Asphalt Materials. Journal of Materials in Civil
387 Engineering, 21 (2), 2008, pp. 83–92.
388 10. Lee, H.J., J.S. Daniel, Y.R. Kim, Continuum Damage Mechanics based Fatigue Model of
389 Asphalt Concrete. Journal of Materials in Civil Engineering, 12, 2000, pp. 105–112.
390 11. Park, S.W., Y.R. Kim, R.A. Schapery. A Viscoelastic Continuum Damage Model and its
391 Application to Uniaxial Behavior of Asphalt Concrete. Mechanics of Materials, 24, 1996,
392 pp. 241–255.
393 12. Si, Z., D. Little, R. Lytton. Characterization of Microdamage and Healing of Asphalt
394 Concrete Mixtures. Journal of Materials in Civil Engineering, 14 (6), 2002, pp. 461–470.
395 13. The American Association of State Highway and Transportation Officials (AASHTO) TP8-
396 94, 2002. Method for Determining the Fatigue Life of Compacted Hot Mix Asphalt (HMA)
397 Subjected to Repeated Flexural Bending. Washington DC: AASHTO Provisional Standards.

10
398 14. Kim, Y.R., H-J. Lee, D.N. Little. Fatigue Characterization of Asphalt Concrete Using
399 Viscoelasticity and Continuum Damage Theory. Journal of the Association of Asphalt
400 Paving Technologists, Vol. 66, 1997, pp. 521–569.
401 15. Himeno, K., T. Watanabe, and T. Maruyama. Estimation of Fatigue Life of Asphalt
402 Pavement. Proc., Sixth International Conference on the Structural Design of Asphalt
403 Pavements, University of Michigan, 1987, pp. 272–289.
404 16. Tayebali, A.A., G.M. Rowe, J.B. Sousa. Fatigue Response of Asphalt-Aggregate Mixtures.
405 Journal of the Association of Asphalt Paving Technologists, 61, 1992, pp. 333–360.
406 17. Bahia, H.U., D.I. Hanson, M. Zeng, H. Zhai, M.A. Khatri, R.M. Anderson. NCHRP Report
407 459: Characterization of Modified Asphalt Binder in Superpave Mix Design. Transportation
408 Research Board of the National Academies, Washington, D.C., 2001.
409 18. Performance Graded Asphalt Binder Specification and Testing. Superpave Series No. 1 (SP-
410 1). Asphalt Institute, Lexington, Ky., 1996.
411 19. Bahia, H.U., D. Perdomo, P. Turner. Applicability of Superpave Binder Testing Protocols to
412 Modified Binders. In Transportation Research Record, No. 1586, Transportation Research
413 Board of the National Academies, Washington, D.C., 1997, pp. 16-23.
414 20. Anderson, D.A., Y.M. Le Hir, M.O. Marasteanu, J.P. Planche, D. Martin, G. Gauthier.
415 Evaluation of Fatigue Criteria for Asphalt Binders. In Transportation Research Record, No.
416 1766, Transportation Research Board of the National Academies, Washington, D.C., 2001,
417 pp. 48-56.
418 21. Soenen, H., B Eckmann. Fatigue Testing of Bituminous Binders with a Dynamic Shear
419 Rheometer. Proc., 2nd Eurasphalt and Eurobitumen Congress, Barcelona, Spain, 2000.
420 22. Carreau, P.J., D.C.R. De Kee, R.P. Chhabra. Rheology of Polymeric Systems-Principles and
421 Applications. Hanser/Gardner Publications, Cincinnati, 1997.
422 23. Motamed, A., H.U. Bahia. Influence of Test Geometry, Temperature, Stress Level, and
423 Loading Duration on Binder Properties Measured using DSR. Journal of Materials in Civil
424 Engineering, 23 (10), 2011, pp. 1422-1432.
425 24. Shan, L., S. Tian, H. He, N. Ren. Internal Crack Growth of Asphalt Binders during Shear
426 Fatigue Process. Fuel, 189, 2017, pp. 293-300.
427 25. Underwood, B.S. A Continuum Damage Model for Asphalt Cement and Asphalt Mastic
428 Fatigue. International Journal of Fatigue, 82, 2016, pp. 387-401.

11
429 LIST OF TABLES AND FIGURES
430
431 FIGURE 1 Three-dimensional meshes of (a) standard plate-type, (b) one ring-type, (c) two
432 rings-type DSR sample testing geometries, and (d) the testing sample
433
434 FIGURE 2 Shear stress distribution on (a) standard plate -type and (b) one ring-type
435 sample geometry
436
437 FIGURE 3 Predicted development of damage along the radius of the standard DSR sample
438 testing geometry
439
440 FIGURE 4 Simulation of damage distribution of : (a) standard plate -type, (b) one ring-
441 type and (c) two rings-type DSR sample testing geometries at the end of the analyses
442
443 FIGURE 5 Pre dicted (a) stress and (b) damage distribution over the sample radius of
444 different DSR sample testing geometries at the end of the analyses
445
446 FIGURE 6 Predicted stress and damage distribution over the sample height at different
447 points over sample radius of the DSR sample testing geometries at the end of the analyses
448
449 FIGURE 7 PHP DSR sample testing system: (a) laser cutting of silicon paper, (b) sample
450 placed on the PHP DSR plates, and (c) view of top plate after test completion
451
452 FIGURE 8 Amplitude sweep results rheological properties versus torque for the different
453 sample testing geometries
454
455 FIGURE 9 Complex shear modulus versus number of cycles of different sample testing
456 geometries
457
458 FIGURE 10 Fatigue life curves of different sample testing geometries
459
460

12
 (a)

(b)

(c)

(d)
461
462 FIGURE 1 Three-dimensional meshes of (a) standard plate-type, (b) one ring-type, (c) two
463 rings-type DSR sample testing geometries, and (d) the testing sample
464

13
 (a) (b)
465
466 FIGURE 2 Shear stress distribution on (a) standard plate-type and (b) one ring-type
467 sample geometry
468

14
 (i) (ii) (iii)

(iv) (v) (vi)

469
2,00E-04
A (12.30 mm)
B (11.50 mm) (vi)
C (11,00 mm)
D (10.50 mm)
E (9.50 mm)
F (6.50 mm) (v)
G (3.00 mm)
1,50E-04

(iv)
Damage ξ

1,00E-04

(iii)

5,00E-05
(ii)

(i)
0,00E+00
0,00E+00 5,00E-02 1,00E-01 1,50E-01 2,00E-01 2,50E-01
Time [s]
470
471 FIGURE 3 Predicted development of damage along the radius of the standard DSR sample
472 testing geometry

15
 (a)

(b) (c)
473
474 FIGURE 4 Simulation of damage distribution of : (a) standard plate-type, (b) one ring-
475 type and (c) two rings-type DSR sample testing geometries at the end of the analyses
476

16
 (a)

(b)
477 FIGURE 5 Pre dicted (a) stress and (b) damage distribution over the sample radius of
478 different DSR sample testing geometries at the end of the analyses
479

17
 1,00

0,90

0,80

Thickness of sample [mm]

0,70

0,60

0,50

0,40

0,30

0,20

0,10

0,00
0,00E+00 5,00E-04 1,00E-03 1,50E-03 2,00E-03 2,50E-03 3,00E-03

480
Stress [MPa]
481 (a)
1,00

0,90

0,80
Thickness of sample [mm]

0,70

0,60

0,50

0,40

0,30

0,20

0,10

0,00
0,00E+00 2,00E-02 4,00E-02 6,00E-02 8,00E-02

482
Damage [1]
483 (b)
484 FIGURE 6 Predicted stress and damage distribution over the sample height at different
485 points over sample radius of the DSR sample testing geometries at the end of the analyses

18
 (a) (b) (c)
486 FIGURE 7 PHP DSR sample testing system: (a) laser cutting of silicon paper, (b) sample
487 placed on the PHP DSR plates, and (c) view of top plate after test completion
488

19
 1,60 73

Complex shear modulus [MPa] 1,40 72

1,20 71

Phase angle [deg]

PP: Complex shear modulus
PHP: Complex shear modulus
1,00 PP: Phase angle
70
PHP: Phase angle

0,80 69

0,60 68

0,40 67

0,20 66
0 50 100 150
489
Torque [mNm]
490
491 FIGURE 8 Amplitude sweep results rheological properties versus torque for the different
492 sample testing geometries

20
 N cycles
0,00E+00 2,00E+04 4,00E+04 6,00E+04 8,00E+04 1,00E+05
1

Complex shear modulus [MPa]

0,1

PP | 50mNm
PHP | 50mNm

0,01
493
494
495 FIGURE 9 Complex shear modulus versus number of cycles of different sample testing
496 geometries
497
498

21
 75
PP testing geometry
PHP testing geometry
70

65
Torque [mNm]

60

55

50

45
1,50E+03 1,50E+04 1,50E+05
N cycles
499
500
501 FIGURE 10 Fatigue life curves of different sample testing geometries
502

22