You are on page 1of 9

Construction and Building Materials 125 (2016) 72–80

Contents lists available at ScienceDirect

Construction and Building Materials
journal homepage: www.elsevier.com/locate/conbuildmat

Characterization of linear viscoelastic, nonlinear viscoelastic and damage
stages of asphalt mixtures
Rong Luo Ph.D., P.E., Professor a,⇑, Hanqi Liu Ph.D. Candidate, Graduate Research Assistant a,
Yuqing Zhang Ph.D., Lecturer b
a
School of Transportation, Wuhan University of Technology, 1178 Heping Avenue, Wuhan, Hubei Province 430063, China
b
School of Engineering and Applied Science, Aston University, MB153A, Aston Triangle, Birmingham B4 7ET, UK

h i g h l i g h t s 

Linear viscoelasticity was strictly differentiated from the nonlinearity. 
Material properties in linear viscoelastic stage were the reference properties. 
Viscoelastic stress, reference modulus & pseudostrain were rigorously established. 
The sole linear viscoelastic effect was eliminated to determine pseudostrains. 
Dissipated pseudostrain energies were determined for representative loading cycle.

a r t i c l e i n f o a b s t r a c t

Article history: It has been demonstrated that asphalt mixtures experienced linear viscoelastic stage, nonlinear viscoelas-
Received 10 February 2016 tic stage and damage stage when subjected to controlled-strain repeated direct-tension (RDT) tests with
Received in revised form 28 May 2016 increasing strain levels. However, the linear viscoelastic properties of asphalt mixtures are usually mud-
Accepted 10 August 2016
dled up with their nonlinear viscoelastic properties. These confusions directly lead to the incorrect deter-
mination of the pseudostrains and dissipated pseudostrain energies (DPSEs) in the nonlinear viscoelastic
stage and damage stage. This study investigated the material properties of fine aggregate mixture (FAM)
Keywords:
specimens in all three stages. These three stages were differentiated and characterized in terms of the
Asphalt mixture
Linear viscoelasticity
viscoelastic stress, pseudostrain and DPSE. The definitions of viscoelastic stress, reference modulus and
Nonlinear viscoelasticity pseudostrain were rigorously established to assure that the material properties in the linear viscoelastic
Viscoelastic stress stage were the reference properties and that the sole linear viscoelastic effect was eliminated when
Pseudostrain determining the pseudostrain and DPSE in the three stages. The characteristics of the DPSE in the three
Dissipated pseudostrain energy stages were found to be: (1) the DPSE of any loading cycle was zero in the linear viscoelastic stage; (2) in
the nonlinear viscoelastic stage, the DPSE of each loading cycle remained approximately the same with
the growth of the number of loading cycles, and the DPSE increased to a larger value when the strain level
of the RDT test increased to a higher level; (3) in the damage stage, the DPSE of the loading cycle
increased as the number of loading cycles increased. This study strictly distinguished the linear viscoelas-
ticity from the nonlinear viscoelasticity of the asphalt mixtures, which is critical for the accurate deter-
mination of the DPSE spent in overcoming the nonlinear viscoelasticity and in developing damages, such
as cracking and permanent deformation, in the asphalt mixtures.
Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction been demonstrated that, when subjected to typical controlled-
strain repeated direct-tension (RDT) tests, an asphalt mixture
Paving asphalt mixtures are complex composite materials that experiences multiple stages as the strain level increases, which
may exhibit different properties at different strain levels. It has include: (1) undamaged stage, consisting of the linear viscoelastic
stage and the nonlinear viscoelastic stage; and (2) damage stage
⇑ Corresponding author. [1,2]. These stages have the following characteristics:
E-mail addresses: rongluo@whut.edu.cn (R. Luo), hanqiliu@whut.edu.cn (H. Liu),
y.zhang10@aston.ac.uk (Y. Zhang).

http://dx.doi.org/10.1016/j.conbuildmat.2016.08.039
0950-0618/Ó 2016 Elsevier Ltd. All rights reserved.

Using these (2) Cutting: the upper and lower part of the raw specimen were cut off using an automatic saw into a shorter specimen σ 40 mm in height. as presented in Fig. However. As the strain level varies.23 23.46 18. 16 No. le. the material properties vary describes the configuration and procedure of the controlled-strain with the increase of the number of loading cycles. . and ER (1) Mixing and compaction: the aggregate batch was mixed has been chosen to be the magnitude of the complex modulus at with the asphalt binder at the temperature of 135 °C. and DPSE. Rt 2. pseudostrain If using the pseudostrain defined in Eq. At any specific strain level.2]). 2(a). as shown in Fig. No. The next section a. these advantages are diminished or even lost using the aggregate surface area method with the optimum asphalt if the analysis does not make clear distinctions and boundaries content of the corresponding full asphalt mixture [6–9]. The following section details the differentiation and character- c.60 0. Viscoelastic Point Sieve No. The deformation of the asphalt mixture cannot be com. linear viscoelastic stage and in the damage stage.18 0. and data. these stages can be illustrated via the study and briefs the authors’ ongoing research on this subject.4. the asphalt mixture properties in different stages based on the test pletely recovered after unloading. Undamaged Stage Damage Stage (3) Coring: the shorter specimen were cored following the pattern illustrated using red circles in Fig.Based binder (graded based on the penetration) and fine limestone on the identification of these distinct stages.85 13. as presented in Fig. The final section summarizes the major findings of this ear viscoelastic effect [3]. The material properties change as the strain level varies. to assure the Viscoelastic Point A Table 1 Critical Linear Gradation of the fine aggregates in FAM specimens. 1. rVE ðtÞ has been considered to be the same as the measured stress in the nonlinear viscoelastic stage. Pa. 2(c) to obtain Nonlinear Viscoelastic cylindrical specimens to be tested.75 <0. The subsequent section presents the determination of b.5]. 2(d) shows an example of the FAM specimen. the asphalt mixture has different controlled-strain RDT tests and to characterize the associated DPSEs properties in the linear viscoelastic stage from those in the in the nonlinear viscoelastic stage and damage stage. linear viscoelastic properties of the asphalt mixture in typical c. 1 being cured at 121 °C for 2 h.46 tests (after [1. B Linear (4) Gluing: each end of a FAM specimen was glued to an end Viscoelastic platen using a 2 ton epoxy with the aid of a specially Critical Nonlinear Stage designed gluing jig. height.075 Fig. ER = reference modulus.18 mm.2. 200 PAN O εR (-No.2]. ization of the linear viscoelastic stage. the nonlinear viscoelastic properties The procedure of fabricating and preparing the FAM specimens are usually muddled up with the linear viscoelastic properties for testing was composed of five major steps as follows: [1. nonlinear viscoelastic stage: To address this research need. ven by the corresponding dissipated pseudostrain energies (DPSEs). there is an urgent need to rigorously determine the non- recovered after unloading. testing to determine aggregates passing No. 50 No. t = loading time. s. le. R. 2(b). s. nonlinear viscoelastic stage and damage stage in terms of the viscoelastic stress. such constant despite of the increase of the number of loading as the fatigue cracking and permanent deformation. The asphalt and more precise by using pseudostrain concepts in analyzing the binder content was calculated to be 8.30 0. viscoelastic stage and damage stage.97% by weight of aggregates test data. stress-pseudostrain curve. The deformation of the asphalt mixture is completely As a result. the material properties stay tion of the development of the damages in the asphalt mixture. Nonlinear viscoelastic stage: the material properties their material properties in the linear viscoelastic stage. Stress-pseudostrain curve of an asphalt mixture in controlled-strain RDT Individual retaining (%) 0 44. Linear viscoelastic stage: the material properties remain Mechanical Analyzer (DMA) to perform controlled-strain RDT tests unchanged if the strain level varies within this stage. 16 sieve with the opening of 1. These confusions directly lead to the incorrect determi. which were 12 mm in Stage diameter and 40 mm in height. In fact. 30 No. as shown in Fig.1. For example. 1. Luo et al. EðtÞ = relaxation modulus in the linear vis. cated in the laboratory using an unmodified #70 petroleum asphalt coelastic stage. (1) to eliminate the lin. Specimen fabrication ing to the measured strain history. 2. and eðtÞ = measured strain history. on fine aggregate mixture (FAM) specimens in order to investigate ii. s = a dummy variable. this study employed a Dynamic i. Fig. the mechanical properties of asphalt mixtures is made simpler The gradation of the fine aggregates is listed in Table 1. The DPSEs in these stages were also characterized for future applications to the prediction of the (2) Damage stage: damage development in asphalt mixtures. which are dri- cycles. 200) Sieve size (mm) 1. indicating any arbitrary FAM specimens for the controlled-strain RDT tests were fabri- time between 0 and t. cylindrical raw specimen 150 mm in diameter and 70 mm in nation of the pseudostrains and pseudostrain energies in the non. / Construction and Building Materials 125 (2016) 72–80 73 (1) Undamaged stage: incorrectly determined results could hardly make accurate predic- a. b. RDT tests. nonlinear change as the strain level varies. 2(e). after the critical nonlinear viscoelastic point (Point B) shown in Fig. between these stages. rVE ðtÞ = viscoelastic stress correspond. At any specific strain level. MPa. Configuration and procedure of the controlled-strain RDT rVE ðtÞ Eðt  sÞ @ e@ðssÞ ds eR ¼ ¼ 0 ð1Þ tests ER ER where eR = pseudostrain. MPa. the asphalt mixture was com- when calculating the pseudostrain in the nonlinear viscoelastic pacted using the Superpave Gyratory Compactor (SGC) into a stage [1.

Strain Level 20 °C. as illustrated environmental chamber at the temperature of 20 °C for at in Fig. There was a 900 s (15 min) rest period between two adjacent RDT tests in order to recover possible deformation in the previous RDT test [1. Test configuration The test protocol was designed as follows: The controlled-strain RDT tests were performed on the FAM specimens using the DMA. As the specimen was in place. 2. the speci. 5. quency of 2p rad/s (1 Hz). in each RDT test. Before the test. and The entire test procedure consisted of a sequence of consecutive (5) Curing: all specimens with end platens were cured in an controlled-strain RDT tests at different strain levels. and (a) Raw specimen (b) Shorter specimen after cutting (c) Making cores (d) FAM specimen (e) Gluing jig (f) Glued specimen for testing Fig.3. was within the linear viscoelastic stage. 3. Fig. point. Luo et al. / Construction and Building Materials 125 (2016) 72–80 vertical pedestals of the two end platens were aligned. A haversine strain curve was imposed on the specimen least 1 h to achieve the temperature equilibrium. 2 2. . which was able to control the test temperature in a range strain level was increased with an increment of 10 le until from 60 °C to 600 °C.2. The test temperature in this study was the specimen developed into the damage stage. Test procedure (f) presents an example of the test specimen glued to end platens. Procedure of fabricating and preparing FAM specimens for testing.74 R. the environmental chamber was (2) Subsequent RDT tests from Strain Level 2 to ðn  1Þ: the axial closed. 2. The test protocol was programmed in the software TRIOS ðn  2Þ corresponded to the critical nonlinear viscoelastic designed specifically for the DMA. which had 600 loading cycles with a loading fre- mens were therefore ready for testing.2]. as shown in Fig. 4). the (1) First RDT test: the initial strain level (Level 1) of the first RDT test specimen with end platens was mounted on the upper and test was selected to be 20 le to assure that the specimen lower tension fixtures that were attached to the DMA (see Fig.

Configuration of DMA tests. 2p rad/s. ER to be the magnitude of the complex firstly processed in the software MATLAB using the Fourier series modulus in the linear viscoelastic stage and EðtÞ to be the to filter possible noise [10]. jE j. sented in Fig. 1) and 80 le corresponded to the critical nonlinear viscoelastic point (Point B in Fig. as illustrated in Fig. Stress and strain waves. It was found that 30 le corresponded Level i to the critical linear viscoelastic point (Point A in Fig.The above data analysis was applied to each RDT test starting from Strain Level 1 that was 20 le. The magnitude of the complex pseudostrain and DPSE were calculated in all three stages as modulus. Controlled-strain RDT test protocol. regard- 3. Properties of test specimens in three stages ing the pseudo strain definition shown in Eq. 70 le and 200 le are pre- Fig. As a result. this study consid- ered rVE ðtÞ to be the linear viscoelastic stress corresponding to the The strain and stress data measured from each RDT test were measured strain history. as detailed in Intro- duction. the reference properties. Characterization of three stages Fig. the Level 2 Level 1 viscoelastic stress. as shown in Fig. Overview of Dynamic Mechanical Analyzer (DMA). 5. u. nonlinear viscoelastic stage and damage stage were identified for the asphalt mixture speci- Strain (µεε) mens tested in this study. Examples of the determined jE j and u at strain levels of 30 le. 900 Time (s) 4. pseudostrain and DPSE will be determined in the next section. 6. 6. based on which the viscoelastic stress. 6. r0 jE j ¼ ð2Þ Upper tension fixture e0 To determine the phase angle of the complex modulus. the linear viscoelastic stage. Luo et al. 3. According to their characteristics. respectively. pseudostrain and DPSE. Fig. The strain amplitude (e0 ) and stress relaxation modulus in the linear viscoelastic stage. Based on the measured specimen properties and the identifica- (3) Final RDT test: the final strain level (Level n) was chosen to tion of the linear viscoelastic. the time lag between the peaks and the time lag between the troughs of the strain and stress waves in the same loading cycle were identified to be Dtp and Dt t . 1). First of all. (3): fixture Dt p þ Dt t u¼ x ð3Þ 2 where x = loading frequency. 7. (1). in which u is converted into degrees for the conve- nience of visual comparison. of every cycle was then calculated to be: follows. amplitude (r0 ) of every loading cycle were therefore determined the material properties in the linear viscoelastic stage were based on the peaks and troughs of the strain wave and stress wave. respectively. / Construction and Building Materials 125 (2016) 72–80 75 σ Δt p σ0 σsN t ε ε0 Δtt t Fig. The jE j Level n and u of every loading cycle in each RDT test were therefore determined. Therefore. It was found that Dtp FAM specimen Environmental and Dt t were not exactly the same in most loading cycles. chamber the average value of Dt p and Dt t was used to compute the phase angle of the complex modulus in the corresponding loading cycle Lower tension as shown in Eq. R. the pseudostrain was rigorously defined in this study to eliminate the linear viscoelastic effect only. Based on the identification of the three stages. . nonlinear viscoelastic and damage be 200 le for the purpose of introducing sufficient damages stages. 4. 50 le. tiated in terms of the viscoelastic stress. In other words. these three stages were further characterized and differen- to the specimen.

As a result. Eq. (4) and curve in an RDT test in the linear viscoelastic stage. Luo et al. and (10): rs. both EðtÞ and rs. (10) and ð8Þ (14): where rVE1 ðtÞ = viscoelastic stress corresponding to e1 ðtÞ. rVE ðtÞ was simplified as: 1200 1190 rVE ðtÞ ¼ e0 EðtÞ  e0 jE jLVE cosðxtÞ ð10Þ 1180 (2) Nonlinear viscoelastic stage: at a specific strain level where the phase angle was uNL .2. respectively: amplitude of the same RDT test. rVE ðtÞ was formulated as: 1170 100 200 300 400 500 600 rVE ðtÞ ¼ e0 EðtÞ  e0 jE jLVE cosðxt  uNL þ uLVE Þ ð11Þ Number of Loading Cycles (N) (3) Damage stage: the formulation of rVE ðtÞ in a specific loading 30 µε 50 µε 70 µε 200 µε cycle with a phase angle of uD was developed as: (a) Magnitude of complex modulus rVE ðtÞ ¼ e0 EðtÞ  e0 jE jLVE cosðxt  uD þ uLVE Þ ð12Þ 18 The above formulations of rVE ðtÞ will be used to determine the pseudostrain and DPSE in the following subsections.LVE ðtÞ nitude of the complex modulus in the linear viscoelastic stage.LVE ðtÞ were considered to be constants em ðtÞ ¼ e0 ½1  cosðxt  uÞ ¼ e0  e0 cosðxt  uÞ ð4Þ within the loading cycle. jE jLVE = mag. Viscoelastic stress EðtÞ ¼ jE jLVE  ð14Þ e0. (1). respectively: 1220 1210 (1) Linear viscoelastic stage: since u was equal to uLVE . eR ðtÞ ¼ rVEERðtÞ  EðtÞ = relaxation modulus in the linear viscoelastic stage. e0. 7.LVE ðtÞ 4. Pa. Therefore.LVE = absolute value of the downward shift of the stress curve in the Nth loading cycle of the same RDT test. (13) was firstly established for the linear 100 200 300 400 500 600 Number of Loading Cycles (N) viscoelastic stage: 30 µε 50 µε 70 µε 200 µε rVE ðtÞ ¼ rm ðtÞ ð13Þ (b) Phase angle of complex modulus According to Eq. Pa. MPa. e0. nonlinear viscoelastic and damage stages. le.LVE ðtÞ = absolute value of the downward shift of the stress measured in each RDT test were firstly simulated using Eqs. Pa. ¼ e0 EðtÞe0jEjE j jLVE cosðxtÞ rVE2 ðtÞ = viscoelastic stress corresponding to e2 ðtÞ. (1) The viscoelastic stresses corresponding to e1 ðtÞ and e2 ðtÞ were with the terms defined at the beginning of this section. EðtÞ was fur- 14 ther derived based on the formulation of rVE ðtÞ in the viscoelastic stage. (13). Pa. 1240 (9). the viscoelastic stress corresponding to em ðtÞ was the difference between rVE1 ðtÞ and rVE2 ðtÞ: 1260 rVE ðtÞ ¼ e0 EðtÞ  e0 jE jLVE cosðxt  u þ uLVE Þ ð9Þ Magnitude of Complex Modulus (MPa) 1250 According to the general formulation of rVE ðtÞ presented in Eq.LVE e0 jE jLVE cosðxtÞ ¼ uLVE = phase angle of the complex modulus in the linear viscoelastic jE jLVE . Pa. (5) Fig. 17 Phase Angle (°) 16 4. which was a sinusoidal strain history. Linear viscoelastic stage rVE2 ðtÞ ¼ e0 jE jLVE cosðxt  u þ uLVE Þ ðDerivation detailed in AppendixÞ The pseudostrain was calculated based on Eqs. Pseudostrain 15 To calculate the pseudostrains in all three stages. and Based on the determination of the relaxation modulus. (4) was then re-arranged as [2]: cycle of an RDT test in the linear viscoelastic stage. Since rVE ðtÞ and rm ðtÞ were exactly the same in the linear 13 viscoelastic stage. e0 jE jLVE  e0. Pa.76 R.  LVE  ð16Þ rs. where e1 ðtÞ ¼ e0 . the e2 ðtÞ ¼ e0 cosðxt  uÞ.1. / Construction and Building Materials 125 (2016) 72–80 1270 stage. MPa.1. rad. le. r0N = stress amplitude of the Nth loading cycle. which was a constant strain history. t = time. specific formulations of rVE ðtÞ were established for the linear 1230 viscoelastic. rm ðtÞ = measured rsN. pseudostrains in the three stages were determined using Eq.LVE rsN = absolute value of the downward shift of the stress curve in where EN = the value of the relaxation modulus in the Nth loading the Nth loading cycle. le. the value of EðtÞ in the Nth loading cycle of a specific RDT test in the linear viscoelastic stage rm ðtÞ ¼ r0N ½1  cosðxtÞ  rsN ð5Þ was determined to be: where em ðtÞ = measured strain.LVE To determine the viscoelastic stress. le. Pa. Examples of determined magnitudes and phase angles of complex moduli. which determined to be: are detailed as follows.2. For a specific loading cycle in any RDT test. Eq. rVE1 ðtÞ ¼ e1 ðtÞEðtÞ ¼ e0 EðtÞ ð7Þ 4.LVE EN ¼ jE jLVE  ð15Þ stress. EðtÞ was then derived based on Eqs. s.LVE = strain (5). the strain and stress waves where rs. MPa. and em ðtÞ ¼ e1 ðtÞ  e2 ðtÞ ð6Þ rsN.

LVE ðtÞ ð20Þ rm ðtÞ vs.LVE ðtÞ ¼ e0 ½1  cosðxt  uNL þ uLVE Þ  e0. (17) was then established for the Nth loading cycle of level of 80 le in the nonlinear viscoelastic stage. Damage stage eRN ðtÞ ¼ e0 ½1  cosðxtÞ  ð17Þ The pseudostrain in the damage stage was formulated based on jE jLVE Eqs. Nonlinear viscoelastic stage e0 rsN. pseudostrain in the nonlinear viscoelastic stage (Strain level = 30 le). Measured stress vs.LVE jE jLVE Fig. R. whose ¼ jE jLVE center was not located at the origin either.LVE e0 jE jLVE cosðxtuD þuLVE Þ a strain level of 30 le. The rm ðtÞ vs. Based on Eq. The area of this ellipse was the DPSE spent overcoming DPSE ¼ rm ðtÞ dt ð22Þ the sole effect of the nonlinear viscoelasticity since the entire linear t1 @t viscoelastic effect was eliminated already with the aid of the pseu. ðuNL  uLVE Þ stayed unchanged as the number of loading cycles increased As previously explained. Pseudostrain at 101st Loading Cycle Measured Stress vs. eR ðtÞ e0 jE jLVE  rs.LVE ðtÞ linear viscoelastic effect was successfully removed in the linear ¼ e0 ½1  cosðxt  uD þ uLVE Þ  e0. (5)–(19) indicated that eR ðtÞ was no longer in phase the RDT test with a strain level of 200 le in the damage stage.LVE eRN ðtÞ ¼ e0 ½1  cosðxt  uD þ uLVE Þ   ð21Þ The pseudostrain in the nonlinear viscoelastic stage was deter.12]: ellipse-shaped hysteresis loop.LVE  jE jLVE was the DPSE spent for the following purposes: The pseudostrain formulation for the Nth loading cycle of an  Overcoming the nonlinear viscoelastic effect.3. it eR ðtÞ ¼ rVEERðtÞ  became a straight line instead of a hysteresis loop. eR ðtÞ graph of the 101st loading cycle of Comparing Eqs.2. (11) and (14): When comparing Eqs. eR ðtÞ hysteresis loop. (22) [1. / Construction and Building Materials 125 (2016) 72–80 77 Since e0 ¼ e0. the DPSEs of representative loading cycles in dostrain formulation.LVE deformation.LVE jE jLVE mined based on Eqs. the rm ðtÞ vs. the pseudostrain was formulated as: 4. (5)–(17) showed that eR ðtÞ was in phase with rm ðtÞ. The rm ðtÞ vs. . respectively.LVE ðtÞ e0 jE jLVE cosðxtuNL þuLVE Þ ð18Þ e0. (12) and (14): Comparing Eqs. Measured stress vs. eR ðtÞ graph of the 101st loading cycle of the RDT test with e0 jE jLVE  e0. the linear viscoelastic. Luo et al.11.2. eR ðtÞ graph of any loading cycle exhibited an of the DPSE was shown in Eq. 9 presents an example of the rm ðtÞ vs. whose center was not located at Z t2 @ eR ðtÞ the origin. 30 80 25 60 20 Measured Stress (kPa) Measured Stress (kPa) 15 40 ER= E* LVE 10 1 20 5 0 0 -40 -20 0 20 40 60 -15 -10 -5 0 5 10 15 20 25 -5 -20 -10 -40 -15 -60 -20 Pseudostrain (με) Pseudostrain (με) Measured Stress vs. The area of this ellipse e0 rs. the value of the DPSE was the area of because of the characteristics of this stage as stated in previous sec.2. Pseudostrain at 101st Loading Cycle Fig. The phase ¼ e0 EðtÞe0 jE  jLVE cosðxtuNL þuLVE Þ angle was ðuD  uLVE Þ. The mathematical formulation tions.2. (5)–(21). Fig. the RDT test: rsN. Dissipated pseudostrain energy (DPSE) specific RDT test in the nonlinear viscoelastic stage. Eq. which validated the pseudostrain formulation established in this study.LVE  jE jLVE viscoelastic stage. 8. and RDT test in the nonlinear viscoelastic stage was then derived as:  Developing damages such as cracking and permanent e0 rsN.LVE for a specific RDT test in the linear viscoelastic eR ðtÞ graph at the 101st loading cycle of the RDT test with a strain stage. which was increasing as the number of load- jE jLVE   ing cycles increased in the destructive RDT test. it was obviously concluded that eR ðtÞ ¼ rVEERðtÞ eR ðtÞ was out of phase with rm ðtÞ in the damage stage. if plotting the measured stress versus pseudostrain. As a result. level = 80 le). and this straight ¼ e0 EðtÞe0 jE jLVE cosðxtuD þuLVE Þ jE jLVE line passed through the origin.LVE 4. This graph demonstrated that the entire ¼ jE jLVE e0 rs. (1). (22). The phase angle between eR ðtÞ and rm ðtÞ was ðuNL  uLVE Þ. e0. For the Nth loading cycle of an RDT test in the damage stage. with rm ðtÞ. (1). 8 presents an example of the   rs. eRN ðtÞ ¼ e0 ½1  cosðxt  uNL þ uLVE Þ   ð19Þ e0. pseudostrain in the linear viscoelastic stage (Strain Fig. Fig. 9.LVE hysteresis loop of any loading cycle also exhibited an ellipse. nonlinear viscoelastic and damage stages were determined. 10 shows the rm ðtÞ vs.3. In a 4. which was larger than zero since uNL > uLVE .

70 le). respectively. rm ðtÞ and eR ðtÞ .2 terms of the integrand in Eq. the DPSE in a loading cycle in the damage stage was for- such as cracking and permanent deformation in the asphalt mix- mulated as: ture specimen. Pseudostrain at 101st Loading Cycle Fig. / Construction and Building Materials 125 (2016) 72–80 150 0. respectively. pseudostrain DPSE ¼ pr0N e0 sinðuD  uLVE Þ ð26Þ and DPSE based on the measurements of controlled-strain RDT Using Eqs. 1.78 R. ful elimination of the sole linear viscoelastic effect. (23) was simplified to be: DPSE ¼ pr0N e0 sinðuNL  uLVE Þ ð24Þ Fig. the 1.14 0. which demonstrates the coelastic stage were the reference properties.LVE  jLVE  dt ð25Þ This study investigated the material properties of FAM speci- @t mens in the linear viscoelastic stage. which were performed using determined for each RDT test in the nonlinear viscoelastic stage the DMA. (3) Damage stage It is clearly illustrated that the DPSE had sustained growth while When calculating the DPSE in a complete loading cycle in the the number of loading cycles was increasing. with the increasing number of loading cycles. pseudostrain in the damage stage (Strain level = 50. As a result. Z t 0 þ2xp DPSE ¼ fr0N ½1  cosðxtÞ  rsN g t0 5. Eq. eR ðtÞ hysteresis loop was in fact a straight DPSE (J/m3) line. the measured stant as the number of loading cycles increased. DPSE of loading cycles in an RDT test in damage stage (Strain level = 200 le).1 DPSE (J/m3) 50 0. With the success- (1) The DPSEs of all loading cycles stayed approximately con.LVE  jLVE (2) When the strain level increased to a higher level within this  dt ð23Þ stage. (22). 12 exhibits the determined DPSEs of loading cycles in the RDT test with a strain level of 200 le. This fact indicated that there 1.LVE  jEsN.LVE  jEsN. 11.06 -100 -50 0 50 100 150 0.02 -100 0 100 200 300 400 500 600 -150 Number of Loading Cycles (N) Pseudostrain (με) 50 με 60 με 70 με Measured Stress vs. nonlinear viscoelastic stage The definite integral was calculated to be: and damage stage. Therefore. 11. 2. only following characteristics: the linear viscoelastic effect was eliminated when determining the pseudostrain and DPSE in the three stages.12 100 Measured Stress (kPa) 0. the DPSE in a loading cycle of any RDT test in the linear vis. (24) and (26).04 -50 0. DPSE of loading cycles in RDT tests in nonlinear viscoelastic stage (Strain Fig. and stress versus pseudostrain in any loading cycle exhibited a . 10. an increasing in Eq. 12. reference modu- and the damage stage. (5) and (19). were presented 0 100 200 300 400 500 600 in Eqs. the DPSE of the loading cycle increased to a larger @t value.2 Since the rm ðtÞ vs.08 0 0.8 was no DPSE spent overcoming the linear viscoelastic effect. Conclusions n o e0 r @ e0 ½1  cosðxt  uD þ uLVE Þ  e0. the formulations of rm ðtÞ and eR ðtÞ in the integrand that. which is in the damage stage. the DPSE of a loading Number of Loading Cycles (N) cycle in this stage was calculated to be: Z t 0 þ2xp Fig. The definitions of viscoelastic stress. (22) were presented in Eqs. 2 coelastic stage was equal to zero. Conse- amount of DPSE accumulated to drive the development of damages quently. DPSE ¼ fr0N ½1  cosðxtÞ  rsN g t0 n o e0 r @ e0 ½1  cosðxt  uNL þ uLVE Þ  e0. This fact indicated damage stage. Since x = 2p rad/s. level = 200 le). Luo et al.4 For any loading cycle in the nonlinear viscoelastic stage. (5) and (21). Measured stress vs.6 (2) Nonlinear viscoelastic stage 1.4 (1) Linear viscoelastic stage 2. the DPSE of every loading cycle was tests at a variety of strain levels. These three stages were differentiated and characterized in terms of the viscoelastic stress. 60. The determined DPSEs of selected loading lus and pseudostrain were rigorously established in the analysis cycles in the RDT tests with different strain levels in the nonlinear in order to assure that the material properties in the linear vis- viscoelastic stage are presented in Fig.

Civ. R. When the strain level of the RDT test increased to a higher Z t level within the nonlinear viscoelastic stage.A. Little. the DPSE of ðA:2Þ each loading cycle remained approximately the same. (A. sim. . J.N. Characterization of fatigue damage in asphalt mixtures using pseudostrain energy. Luo. References plicity and accuracy to the characterization of material properties that are used in the design and construction of asphalt pavements. Pavement Eng. the pseudostrains and the DPSEs of To summarize. D. 9 (4) (2008) 233–246. Eng. Luo et al. rVE2 ðtÞ ¼ e0 jE jLVE cosðxt  u þ uLVE Þ ðA:6Þ mination of the DPSE spent in overcoming the nonlinear viscoelas- ticity and in developing damages in the asphalt mixtures. Lytton. Eng. [4] Z. 2015CB060100) and the Research Fund for the [7] E. R. ¼ e0 x 0 Eðt  sÞ sinðxs  uÞds Based on the formulations of the pseudostrain and DPSE.L. The test and analysis methods developed in this study provide clarity. n 2 ðt. (1984) 195–223. (A. R. R. Z t this fact indicated that more energy was spent overcom- E00 ðxÞ ¼ x EðnÞ cosðxnÞdn ¼ jE jLVE sin uLVE ðA:4Þ ing the larger nonlinear viscoelasticity at a higher strain 0 level. Fract. Luo. thesis). the storage spent overcoming the nonlinear viscoelasticity in every modulus E0 ðxÞ and the loss modulus E0 ðxÞ of the complex modulus cycle. the viscoelastic stress rVE2 ðtÞ corresponding to the nonlinear viscoelastic stage and damage stage can be rigor- the strain history e2 ðtÞ ¼ e0 cosðxt  um Þ is determined to be: ously determined while eliminating the sole linear viscoelastic effect of the asphalt mixtures. and Rt (2) Damage stage: the DPSE was spent in overcoming the non. Int. Luo. Application of surface energy measurements to evaluate The authors acknowledge the financial support of the ‘‘973 moisture susceptibility of asphalt and aggregates (Master of Science. the cor- responding viscoelastic stress rVE2 ðtÞ is derived as follows.2) is further formulated as: (3) Damage stage: with the growth of the number of loading cycles. Little. 2006. which indicated that the same amount of energy was According to the linear viscoelastic theory [13–15]. the DPSE of the loading cycle was increasing. Zollinger.F. Therefore. / Construction and Building Materials 125 (2016) 72–80 79 straight line passing through the origin in the linear viscoelastic Program of China for the start-up funds for purchasing the stage but an ellipse-shaped hysteresis loop. Little. whose center was laboratory equipment that is crucial to this research.L. Texas A&M 20120143110004). Lytton. J. 25 (2) (2013) Based on the definitions and formulations established in this study.1) is DPSEs of all loading cycles were determined for every RDT test at then re-arranged as: each strain level. Characterization of microdamage and healing of asphalt concrete mixtures. Lytton. Energy-based mechanistic approach to characterize distinguished from the DPSE for driving the damage development crack growth of asphalt mixtures. the Let n ¼ t  s. 25 (3) mixtures.N. Civ. Mater.F. Lytton. the DPSE for overcoming the nonlinear viscoelasticity is further [2] X. not located at the origin. Civ. V. which was used for different purposes in the nonlinear viscoelastic stage and damage stage: For a sinusoidal strain history e2 ðtÞ ¼ e0 cosðxt  um Þ. R. Eq. Castelo Branco. This is critical for the accurate deter. The area within the hysteresis loop was the Appendix A. [5] E. Masad. An Improved Method for the Dynamic Mechanical Analysis of Fatigue Failure of Sand Asphalt Mixtures. ASCE J. R. D. Masad. tÞ. 0Þ. Int.L. D. in an ongoing investigation for the purpose of establishing [3] R. FHWA/473630. Correspondence principles and a generalized J integral for large energy-based models for predicting damage evolution in asphalt deformation and fracture analysis of viscoelastic media. the DPSE of E0 ðxÞ ¼ x EðnÞ sinðxnÞdn ¼ jE jLVE cos uLVE ðA:3Þ every loading cycle increased to a larger value and stayed 0 unchanged as the number of loading cycles increased. in both nonlinear viscoelastic stage and damage stage.L. rVE2 ðtÞ ¼ 0 Eðt  sÞ @ e@2sðsÞ ds Rt linear viscoelastic effect and in developing damages such ¼ Eðt  sÞ @ e0 cosð@ sxsuÞ ds 0 ðA:1Þ as cracking and permanent deformation in the asphalt Rt mixture. E ðxÞ can be expressed as follows: b. As the number of loading cycles increased. then s ¼ t  n. Program” of the Ministry of Science and Technology of China Texas A&M University. and when s 2 ð0.L. 2005. The following characteristics of the DPSEs were observed in the three stages: R rVE2 ðtÞ ¼ e0 x 0t EðnÞ sin½xðt  nÞ  udðt  nÞ R0 ¼ e0 x t EðnÞ sin½ðxt  uÞ  xndn (1) Linear viscoelastic stage: the DPSE of any loading cycle was Rt ¼ e0 x 0 EðnÞ½sinðxnÞ cosðxt  uÞ  cosðxnÞ sinðxt  uÞdn zero. [6] C. 25 (9) (2013) 1198–1208. Mater. Texas Transportation Institute.T. R. ¼ e0 jE jLVE ½cos uLVE cosðxt  uÞ  sin uLVE sinðxt  uÞ ¼ e0 jE jLVE cosðxt  u þ uLVE Þ The findings of this study are capable of strictly differentiating ðA:5Þ the linear viscoelasticity from the nonlinear viscoelasticity of asphalt mixtures. V. 208–218. Special thanks are to the 1000-Youth Elite University. Eng. Mater. 14 (6) (2002) 461–470. Castelo Branco. [1] X. A unified method for the Acknowledgements analysis of controlled-strain and controlled-stress fatigue testing. Doctoral Program of Higher Education of China (Project No. (Project No. ASCE J. Lytton. Federal Highway Administration. ASCE J. R.J. Luo. hR i hR i ¼ e0 x EðnÞ sinðxnÞdn cosðxt  uÞ  e0 x 0 EðnÞ cosðxnÞdn sinðxt  uÞ t t (2) Nonlinear viscoelastic stage: 0 a. Derivation of Equation (8) DPSE. (1) Nonlinear viscoelastic stage: the DPSE was spent in over- coming the nonlinear viscoelastic effect only.N. which indicated that a larger amount of DPSE was spent developing rVE2 ðtÞ ¼ e0 E0 ðxÞ cosðxt  uÞ  e0 E00 ðxÞ sinðxt  uÞ damages including cracking and permanent deformation in ¼ e0 jE jLVE cos uLVE cosðxt  uÞ  e0 jE jLVE sin uLVE sinðxt  uÞ the asphalt mixtures. Eq. Si. Schapery.T. Therefore.

140 (4) (2010) 445–452. Eng. Lytton. [9] Y. Moisture and aging damage evaluation of asphalt [13] J. Creep and Relaxation of Nonlinear Viscoelastic [10] MATLAB (computer software). USA. Viscoelastic Properties of Polymers.R. Mater. Materials: With an Introduction to Linear Viscoelasticity. Lai. R. ASCE J. ASCE J. [11] Y. Lytton. ASCE J. Mechanistic modeling of fracture in asphalt Mineola.L. Anisotropic characterization of crack growth in of water diffusion through fine aggregate mixtures.S. 2000. K. Zhang. Tong. Ferry. 25 (9) (2013) [15] A.S. Onaran. Dove Publications. mixtures under compressive loading.L. Wineman. R. Mechanical Response of Polymers: Introduction.L. / Construction and Building Materials 125 (2016) 72–80 [8] K. Lytton. 16 (5) (2015) 397–410. R. Civ. A. 1989. . 23 the tertiary flow of asphalt mixtures in compression. R. [14] W. R. 1189–1197. Experimental measurement [12] Y. Cambridge. Civ. Wiley. Findley. Natick.L. Luo. Pavement Eng. D. Little. Eng. Vasconcelos. Luo. Int.D. New York. Cambridge University Press. third ed.. (6) (2014) 682–694.N. J. MA. Lytton. R. UK. Rajagopal.N. Luo et al. Mech. USA. Mater.L. Luo.80 R. mixtures using the repeated direct tensional test method. Math Works. R. 1980. Bhasin. J. Zhang. New York. Eng. K.