EVOLUTION OF ASPHALT MIXTURE STIFFNESS UNDER THE COMBINED EFFECTS OF DAMAGE, AGING AND DENSIFICATION UNDER TRAFFIC

Transcription

1 EVOLUTION OF ASPHALT MIXTURE STIFFNESS UNDER THE COMBINED EFFECTS OF DAMAGE, AGING AND DENSIFICATION UNDER TRAFFIC prepared for the Annual Meeting of the Transportation Research Board by Angel Mateos, Dr. Civil Engineer Javier P. Ayuso, Civil Engineer Belen Cadavid Jáuregui, Physicist Jose Orencio Marrón Fernández, Chemist CEDEX Transport Research Center Autovía de Colmenar Viejo km. El Goloso (MADRID), SPAIN Ph +-- Fax +-- s: Revised in the light of the comments from TRB Committee AFD Full-Scale Accelerated Pavement Testing Word Count: words + tables + figures = words November, TRB Annual Meeting

2 Mateos, Ayuso, Cadavid and Marrón EVOLUTION OF ASPHALT MIXTURE STIFFNESS UNDER THE COMBINED EFFECTS OF DAMAGE, AGING AND DENSIFICATION UNDER TRAFFIC prepared for the Annual Meeting of the Transportation Research Board by Angel Mateos, Dr. Civil Engineer Javier P. Ayuso, Civil Engineer Belen Cadavid Jáuregui, Physicist Jose Orencio Marrón Fernández, Chemist CEDEX Transport Research Center ABSTRACT The long-term evolution of asphalt mixture stiffness in the field represents a complex process where three main factors play an essential role: damage, aging and densification under traffic. In order to adequately take them into account, not only must the level of each of these three factors be known, but also how each particular level modifies the modulus of the asphalt mixture for the temperature and frequency range corresponding to field conditions. The research herein presented is based on experimental data from four flexible sections tested at the CEDEX test track. The stiffness of the asphalt layer was periodically evaluated by means of falling weight deflectometer back-calculation and by conducting dynamic modulus testing in laboratory.,, loads were applied on the pavements during a -month period. This caused a high level of deterioration in all sections as well as a significant asphalt densification, and also provided time for some aging effects to develop. Based on the results from this full-scale test, different conclusions could be deduced concerning the evolution of damage, aging and densification and how this evolution modifies the stiffness of the asphalt mixture. Special attention has been paid in order to quantify the effects in terms of changes in the parameters of the original master curve. The model and methodology incorporated in CalME design procedure were successfully used to reproduce the evolution of asphalt layer modulus during the test. In particular, the assumptions of this model regarding how the dynamic modulus master curve is modified by each of the three factors were found to be valid for this experiment. TRB Annual Meeting

3 Mateos, Ayuso, Cadavid and Marrón INTRODUCTION The evolution of asphalt mixture stiffness in the field is not only related to damage. It is a well-known fact that aging and post-compaction under traffic represent major factors that must be considered as well. This has important implications in terms of auscultation and design. Regarding auscultation, actual asphalt damage may be hidden by aging and postcompaction effects, which should be removed before the actual modulus is compared to the undamaged values of the original material. In relation to pavement design, especially when incremental-recursive procedures are used, the effects that the three factors have on the stiffness of the asphalt mixture must be accounted for before pavement structural response is calculated under traffic loads. The need to consider aging and post-compaction is fully recognized in the AASHTO Mechanistic-Empirical Pavement Design Guide (MEPDG), which explicitly considers both effects by using the Global Aging System developed by Mirza and Witczak () and Witczak dynamic modulus predictive equation (). Problem Statement In order to adequately consider the effects of asphalt damage, aging and densification, the level of each of these three factors must be known, but also how each particular level modifies the stiffness of the asphalt layer for the temperature and the frequency range corresponding to field conditions. Numerous studies exist where the effects of damage, aging or densification, as single factors, have been evaluated in laboratory, either from field specimens or from artificial specimens where such effects were simulated in laboratory. However there are very few studies in which the joint effect of the three factors have been evaluated in the field. The authors believe that an important effort is still to be conducted in order to validate a sound model and a thorough methodology that can be applied to the field and discriminate between the three main factors affecting asphalt layer modulus during the service life of a flexible pavement. Objective The objective of this research is the understanding of the evolution of asphalt mixture stiffness during the service life of a flexible pavement and the validation of a model that can reproduce the main aspects of this complex process. The results herein presented are part of a research effort whose outcomes have been included in two papers. The first one () was focused on damage prediction, and in particular on the transition from a fatigue model calibrated in the laboratory to the field. The present second paper is focused on the evolution of asphalt layer stiffness during a full-scale test, as a consequence of the combined effects of damage, aging and densification under traffic. TRB Annual Meeting

4 Mateos, Ayuso, Cadavid and Marrón Research Approach Four full-depth flexible sections have been included in this research. Section consisted of a mm thick asphalt layer placed on a reference subgrade. The capping layer, mm thick, was constituted by a non-plastic granular soil with a CBR over. The embankment was a plastic soil with a CBR of.. Sections, and had a mm thick asphalt layer and the corresponding subgrades were achieved by introducing medium-high quality soils between the capping layer and the embankment (Figure ). The asphalt layer consisted of two different mixtures. The wearing course, mm thick in every section, was an open-graded mixture with a maximum nominal aggregate size of mm, design air voids of.% and bitumen content of.% by mass of aggregates. Bitumen was highly modified by SBS, with a design penetration interval of - mm/ and a ring & ball temperature over C. This mixture is designated as BBTM B PMB -/ according to European normative EN, and constitutes the typical wearing course used in highways in Spain. The base was a conventional asphalt concrete with a nominal maximum aggregate size of mm, design air voids of.% and a plain bitumen content of.% by mass of aggregates. Bitumen had a design penetration interval of - mm/ and a ring & ball temperature in the interval of - C. This mixture is designated as AC base / according to EN, and is typically used in Spain as a base, under the wearing and binder courses. mm." Section A.M. Section Section Section Section Section A.M. AC AC mm." mm." mm." mm." CBR LL/IP max. ω opt. Soil > N.P... S EST cement stabilized soil Soil N.P... note: Sections and, provided with Soil. /... a cement stabilized capping layer, were not included in this Soil. /.. particular research. FIGURE Tested Pavement Sections. Sections were tested in CEDEX Test Track (), a linear-circular combined facility where two automatic vehicles are continuously moving in order to reproduce real traffic (Figure ). Fullscale sections are built along the straight lines of the facility by using conventional road building procedures. Sections are built inside a reinforced concrete test pit ( meters depth, TRB Annual Meeting

5 Mateos, Ayuso, Cadavid and Marrón meters wide), allowing the generation of a water table. Additional information about the test conducted in this research is listed below: - Beginning and end: -August- to -December- - Total number of passes:,, - Vehicles loading: kn dual wheel - Transverse distribution: mm - Vehicle speed: - km/h - Environmental conditions: Open air. Water table generated by rainfall. FIGURE Traffic Simulation Vehicle of CEDEX Full-Scale Test Track. The two asphalt mixtures, wearing and base courses, were initially tested in laboratory by conducting compression tests on cylindrical specimens prepared in laboratory (FMLC). The master curves were determined by using AASHTO-MEPDG format. A weighted average master curve was obtained for the total asphalt thickness. By using this average master curve, the deflections under the load of the falling weight deflectometer (FWD), calculated with a multilayer linear elastic program, would be the same as those obtained when the two different materials are used. This simplification was necessary, since the evolution of the stiffness of the asphalt layer was to be evaluated by means of FWD testing, from which discrimination of the two asphalt courses would be impossible in practice. This average master curve would be used as a reference of the unaged and undamaged asphalt layer. In order to fulfill the objective of this research, the required practical goal is the determination of the evolution of the master curve during the -month duration of the test. With that intention in mind, periodic structural evaluations were carried out for each section by using a KUAB FWD, which applied a kn load. Each section was evaluated in - points ( point per meter). Moduli for the pavement layers were back-calculated from deflections measured at,,, and mm from the loading plate center. EVERCALC was used, which is a well-known back-calculation software developed by Washington State Department of Transportation. TRB Annual Meeting

6 Mateos, Ayuso, Cadavid and Marrón A model that can account for the effects of damage, aging and post-compaction on the modulus master curve was adopted for this research. To the best knowledge of the authors of this paper, only one of these models (or collection of models) has been incorporated into mechanistic-empirical pavement design procedures, which is the model incorporated in CalME. CalME is a design procedure developed by the California Department of Transportation for new flexible pavements and rehabilitations (). The program incorporates the existing Caltrans empirical design methods, a classical mechanistic-empirical method and an incremental-recursive damage model based on mechanistic-empirical principles (). Actual CalME software has not been used for this research, whereas its models for consideration of damage/aging/densification were indeed followed. STRUCTURAL EVALUATION FROM FWD BACK-CALCULATION Figure shows asphalt layer back-calculated moduli for each section during the test. Laboratory stiffness data (average master curve) for Hz are also represented in the graphs; this frequency was shown to be representative of the load pulse applied by the KUAB FWD device with the particular weight-buffer configuration adopted for this study (). Several observations can be made from Figure : Damage evolution during the test, in terms of stiffness reduction, is clear for all sections at the low temperature range (< C), but that does not happen so distinctly for high temperatures, especially over C. Modulus values tend to converge to the original undamaged curve for temperatures around - C. This indicates that asphalt damage, which causes stiffness reduction, is not the only factor affecting mixture stiffness. The overall tendency during the test indicates an increase in the minimum modulus (for high temperatures) as well as a reduction of the slope of the curve modulus versus temperature. This can be clearly inferred from points corresponding to cycle million. An unexpected increase in modulus takes place in sections, and for series,-, when compared to the modulus in cycle,. The overall tendency described above can be fully explained by CalME asphalt stiffness model. This model uses the MEPDG format, but introduces a damage parameter (ω) as shown in equation below: loge logfr e [] where, E is dynamic modulus ω is asphalt damage r is reduced frequency,,, are parameters of the model When damage is null (ω = ), the model is identical to the reference average master curve. As damage approaches to, the viscous term in equation [] decreases, and for the maximum level of deterioration (ω = ) the model will represent a linear elastic material with a modulus TRB Annual Meeting

7 Mateos, Ayuso, Cadavid and Marrón of. It should be noted that the reduction of the viscous part as damage increases is equivalent to the reduction of the slope of the curve modulus versus temperature that took place for all sections during the test. Section Section Asphalt Layer Modulus (MPa) Cycle Cycle Asphalt Layer Modulus (MPa) Cycle Cycle Temperature (ºC) Temperature (ºC) Section Section Asphalt Layer Modulus (MPa) Cycle Cycle Asphalt Layer Modulus (MPa) Cycle Cycle Temperature (ºC) Temperature (ºC) CalME Cycle Lab. Hz CalME Cycle million note: Hz is the frequency representative of the actual FWD load pulse FWD (Cycle ) FWD (Cycle ) FWD (-) FWD (-) FWD (Cycle million) FWD (Cycle ) FWD (Cycle ) FWD (untrafficked line) FWD ( mill.) Corrected vs Aging FWD ( mill.) Corrected vs Aging and vs Post-Comp. CalME ( mill.) without Aging FIGURE Asphalt Layer Moduli Back-Calculated from FWD. CalME ( mill.) without Aging and without Post-Comp. CalME methodology introduces aging and post-compaction by increasing the parameter in equation []. This is equivalent to introducing a constant factor Δ that multiplies moduli. It is possible to determine ω and Δ from FWD back-calculated moduli as soon as data is available for a relatively wide temperature interval at a particular time (constant damage/aging/post-compaction levels). This could be done for cycle million since five FWD tests were conducted for each section. The good agreement between CalME and backcalculated moduli can be observed in Figure. The same approach was attempted for cycle, by conducting six FWD tests, but the resulting temperature range was relatively narrow and the extrapolation to high temperatures would have not been reliable. Parameters obtained for cycle million are displayed below. TRB Annual Meeting

8 Mateos, Ayuso, Cadavid and Marrón Cycle million Sec. Sec. Sec. Sec. Δ.... ^Δ.... ω.... Std. Error (MPa).... It should be remarked that attributing exclusively the increase in to aging and postcompaction and considering that damage is equivalent to a decrease in are assumptions of CalME model. Both have been validated by the authors of the program on the basis of laboratory and field studies. Nonetheless special attention has been paid in this research study in order to identify the pattern of variation of the master curve during the test as a consequence of damage, aging and post-compaction. CONSIDERATION OF ASPHALT AGING AND POST-COMPACTION UNDER TRAFFIC Aging Effects Aging is known to produce an increase in asphalt binder viscosity. According to MEPDG methodology, this increase can be accounted for by using Witczak dynamic modulus predictive equation (). This equation assumes that aging effects can be reproduced by decreasing the β location parameter in equation [], but and would remain unchanged. This is equivalent to introducing a horizontal shift factor towards decreasing frequencies in the master curve. The NCHRP Project - () represents another important reference where aging effects have been evaluated in laboratory by conducting temperature and frequency sweep tests. Aging caused a significant increase in asphalt mixture modulus, but the stiffness ratio aged/unaged did not appear to follow a clear pattern versus temperature and frequency. In particular, the ratio did not seem to depend on the slope of the curve log(e) versus log(r) as would have been expected from a horizontal shift factor. This indicates that the parameter in the master curve might increase as a consequence of aging. Aging effects on master curve were evaluated in this study by conducting dynamic modulus testing in the laboratory. Cores along an untrafficked line of the sections were extracted days after construction, which approximately corresponds to cycle million. The untrafficked line is. meters far from the center of the wheel path, so aging was the only factor affecting mixture stiffness. The master curve was determined and compared to that of the original material, corresponding to cores extracted days after construction (cycle ). The resulting aging factor (E aged /E unaged ) was not constant, but increased as reduced frequency decreased. For the frequency representative of the FWD pulse, Hz, the aging factor was././. for temperatures // C respectively. This pattern cannot be reproduced by only increasing or by introducing a horizontal shift factor. However, it should be indicated that the modulus of the base course mixture was the only one that could be evaluated, due to the reduced thickness of the wearing course. Additionally, the height of the specimens was around mm, which is relatively low for a mm diameter. Due to these uncertainties, the effect of aging was also evaluated from FWD back-calculation along the untrafficked line. Values obtained from cycles, to,, (between and days after construction) were compared to those obtained between cycles and, ( to days after construction). A slight increase in modulus was observed, but the differences were not statistically significant. TRB Annual Meeting

9 Mateos, Ayuso, Cadavid and Marrón The main conclusion that can be deduced from both laboratory and FWD testing along the untrafficked line is that the modulus of the asphalt mixture did not exceedingly change due to aging during this test. This result was somewhat expected due to the relatively short duration of the test in comparison to typical useful life of an in-service pavement. Although evidence of hardening was found, it is clear that it would only explain a minor part of the increase in that was deduced in the wheel path for cycle million. Such increase would be mainly related to densification under traffic. The reduced magnitude, together with testing limitations and variability, made the quantification of aging effects on asphalt mixture modulus very difficult. As a consequence, a constant aging factor of. was assumed for simplicity, which was the value obtained in laboratory for C and Hz. This can be reproduced by increasing in log(.) in equation []. It should be highlighted that this particular value of the factor applies to day in comparison to day. But it is necessary to determine the evolution of the factor during the test. In order to do this, CalME aging model takes into account temperature effects, but a simpler model was here adopted due to the relatively low impact of aging on asphalt mixture for this particular test. The model, that is included in a previous CalME version, is formulated below: t ln AGING E AgeA Et AgeA lnt [] where, t is age in days t is age of the mixture when the modulus was initially determined ( days after construction) AgeA is model parameter; calibrated by assuming a ratio of. for day Once the evolution of aging was known, it was possible to correct FWD back-calculated moduli from aging, as presented in Figure for cycle million. Evaluation of Post-Compaction under Traffic Densification under traffic is known to produce an increase in asphalt mixture modulus. According to CalME, this effect can be accounted for by increasing in equation. This is equivalent to assuming a post-compaction factor (E Post-Compacted /E Original-Voids ) that is constant versus reduced frequency. Two research studies have been found in which this assumption is validated in laboratory. Masad et al. () obtained the relaxation modulus curves for a particular asphalt mixture at three different air voids contents, between % and %. The three curves were parallel in log(e) versus log(t) space, which means that the corresponding dynamic modulus curves were also parallel in the log(e) versus log(r) space. Witczak dynamic modulus predictive equation () constitutes the second reference. According to this formulation, which explicitly considers air voids content, only the parameter of the master curve would change when densification takes place. However, in other laboratory studies the effect of air voids content did not result in a constant densification factor. Seo et al. () showed, for a particular asphalt concrete, that post-compaction can affect all parameters of the master curve. The combined effect resulted in a densification factor of. for the highest t TRB Annual Meeting

10 Mateos, Ayuso, Cadavid and Marrón modulus (highest reduced frequencies) and. for the lowest modulus (lowest reduced frequencies), when air voids decreased from % to %. Similar results were obtained by Rowe et al. () by testing different air voids contents, from.% to.%. The resulting densification factor was. for the highest modulus and tended to decrease for low reduced frequencies. The pattern observed in Figure clearly reflects an increase in the minimum modulus during the test, which necessarily entails an increase in. Whether densification also affects, β or or not is difficult to conclude from this test. According to CalME model, the increase in for cycle million due to post-compaction can be obtained by subtracting aging effects from the total increase that was previously obtained by fitting FWD back-calculated moduli, as listed below. Cycle million Sec. Sec. Sec. Sec. Δ Total.... Δ AGING log(.) log(.) log(.) log(.) Δ POST-COMPACTION.... ^Δ POST-COMP..... The importance of post-compaction in this test can be deduced from the high factors displayed above, between. and.. This means that asphalt modulus increased between % and % due to the densification that took place under traffic. In order to verify such densification, air voids were measured from cores extracted along the center of the wheel path at the end of the test. A minor reduction,.%, was determined for the base course mixture. Nevertheless, a very important reduction was verified for the open-graded asphalt mixture of the wearing course; this mixture had a design air voids content of.%, but cores extracted after construction indicated an air void content of.%. Cores extracted in cycle, and at the end of the test indicated that the air voids content had decreased to.% and.% respectively. It is very likely that this important densification is the main reason behind the increase in during this test. A strong negative correlation exists, R² =., between the post-compaction factor and asphalt layer thickness; actual average thickness of section,, and was measured with a ground penetration radar as,, and mm respectively. Once the effects of post-compaction were quantified in terms of for cycle million, an assumption was necessary in order to determine evolution during the test. The assumption adopted for this research was that a linear relationship exists between air voids content and the post-compaction factor ( Δ ), as represented in equation [] below. This assumption is supported by different laboratory studies: Seo et al. () obtained a linear relationship between asphalt modulus and air voids content; the same result was obtained by Rowe et al. (). Masad et al. () demonstrated this linear relationship by using micromechanical modeling. POST COMPACTION k ( AV AV ) [] where, AV is air voids content AV is initial air voids content k is model parameter TRB Annual Meeting

11 Mateos, Ayuso, Cadavid and Marrón Air voids content in equation [] should reflect a weighted mean of both wearing and base courses mixtures, since an unique average master curve was assumed for the total asphalt thickness. Nevertheless, it was previously shown that the increase in during this test was mainly due to densification of the wearing course. As a consequence, the evolution of was assumed to be linearly related to air voids content of this layer. Actual air voids were not measured during the test, but it is possible to use permanent deformation instead of (AV AV) in equation []. This only means a change in k parameter. Permanent deformation of the wearing layer was actually measured along the curves of the facility. Curves of CEDEX test track are provided with a reinforced concrete slab, and are used for testing functional characteristics of wearing courses, typically skid resistance, macro-texture and drainage capacity. However, permanent deformation is measured as well, as shown in Figure. Only the shape of the curves in this figure is important here, which was very similar for the four wearing courses. As densification factor is known for cycle million, as well as wearing course rutting depth, it is possible to obtain parameter k and then apply equation [] for the rest of the test. RD POST COMPACTION k RD n / n [] where, RD(n) is asphalt rutting depth at cycle n n is number of cycles for RD normalization k is model parameter It can be observed from Figure that the most permanent deformation took place at the beginning of the test (end of summer ) and during summer. This supports other research studies that indicate that most asphalt rutting takes place during the first two years of service () and also supports Hanson et al. research () that concludes that mechanical properties of asphalt cores remain constant after two years of densification. It was previously noted from Figure that an unexpected increase in modulus had taken place in sections, and after cycle,. But at this point, it can be fully explained by the densification that happened in summer, between cycles,-,, as shown in Figure. When aging and densification effects are removed, damage is almost constant for this period, as is shown in Figure below. Permanent Deformation (mm) BBTM B PMB -/ ( mm) BBTM B PMB -/ ( mm) PA PMB -/ ( mm) PA PMB -/ ( mm) Average Curve Asphalt Temp. C Asphalt Temp. Number of Cycles FIGURE Permanent Deformation of Wearing Courses in Curves of CEDEX Test Track. TRB Annual Meeting

12 Mateos, Ayuso, Cadavid and Marrón ASPHALT DAMAGE EVOLUTION DURING THE FULL-SCALE TEST Once the evolution of aging and post-compaction effects is determined, it is possible to use equation [] to calculate asphalt damage for each particular FWD test. In practice, equation [] was used instead for simplicity, and results are presented in Figure. E E i E Ei min where, E is FWD backcalculated modulus, after removing aging and postcompaction effects E i is initial modulus of the material E min is initial Equation [] can be directly deduced from equation []. In order to do it, it has been assumed that β and parameters of the master curve do not change during the test. However, another assumption exits, that is implicit in the methodology employed in this research: timetemperature correspondence does not change as a consequence of aging, damage or postcompaction. Chehab et al. () and Zhao and Kim () have shown that the time-temperature relationship obtained for the original asphalt mixture still applies for high levels of damage. Seo et al. () verified the same thing in relation to air voids content. No research study has been found in which the assumption has been validated for asphalt mixtures in relation to aging, but the results obtained from cores tested in laboratory indicated it was valid for this particular test. A linear relationship was assumed: log(a T ) = λ (T-T ref ), where a T is the shift factor and T is temperature. Testing conducted on cores extracted right after construction resulted in λ =. C - whereas testing conducted on cores extracted days later resulted in λ =. C -, which is virtually the same value. ω / cracking % % % % % % Section C R² =. Number of Cycles ω / cracking % % % % % % Section R² =. C Number of Cycles [] ω / cracking % % % % % % Section C Number of Cycles ω / cracking % % % % % % Section C Number of Cycles R² =. R² =. Damage (%) Actual Cracking (%) Cracking predicted by CalME (%) Asphalt Temp. FIGURE Asphalt Damage and Cracking Evolution During the Test. TRB Annual Meeting

13 Mateos, Ayuso, Cadavid and Marrón Figure provides information regarding a topic of the utmost importance: how asphalt damage develops as a function of loads and temperature during the service life of a flexible pavement. Different conclusions related to his topic could be extracted from this experiment by using CalME models, and can be found in reference (). Probably the most important one was that the greatest amount of damage took place for medium to low temperatures, but not during the warmest periods, as can be observed in Figure for every section. This performance was attributed to the beneficial effect of rest periods between loads. Figure shows FWD back-calculated moduli after removing aging and post-compaction effects. Damage evolution during the test is obvious now for the complete temperature range, although it is even more evident at low temperatures. Experimental data in Figure show that stiffness reduction due to damage, in absolute as well as relative terms, becomes higher as temperature decreases. This tendency supports the approach used by CalME in order to account for asphalt damage by including the ω parameter in equation []. The results from this model are also represented in Figure for different cycles. Section Section Asphalt Layer Modulus (MPa) Cycle Cycle Asphalt Layer Modulus (MPa) Cycle Cycle Temperature (ºC) Temperature (ºC) Section Section Asphalt Layer Modulus (MPa) Cycle Cycle Asphalt Layer Modulus (MPa) Cycle Cycle Temperature (ºC) Temperature (ºC) CalME Cycle Lab. Hz FWD (Cycle ) FWD (Cycle ) CalME () without Aging and without Post-Comp. FWD (-) FWD (-) CalME () without Aging and without Post-Comp. FWD (Cycle million) FWD (Cycle ) CalME ( mill.) without Aging and without Post-Comp. FWD (Cycle ) FWD (untrafficked line) FIGURE Asphalt Layer Moduli Back-Calculated from FWD, after Removing Aging and Post-Compaction Effects. TRB Annual Meeting

14 Mateos, Ayuso, Cadavid and Marrón PREDICTION OF ASPHALT CRACKING Cracking was periodically measured during the test and was quantified in terms of percentage of the length of sections presenting any cracks. The results are shown in Figure, together with the damage (ω) evaluated from FWD testing. The correlation between both variables is clear, as expected, but the existence of lurking variables must be taken into consideration. The most important missing variable is asphalt aging, that has been widely shown to correlate to asphalt cracking. Some studies even indicate that aging is actually the main variable that defines cracking, e.g., Kandhal et al. (), who indicated that cracking occurs when the consistency of asphalt reaches a certain value regardless of the mixture properties and original asphalt grade. Another drawback when predicting cracking by using approaches based on continuum mechanics (which is the generalized approach of the current design methodologies) is that the hypothesis of continuity is no longer valid when cracking process begins. It must be highlighted that damage values presented in Figure have been obtained by assuming continuity in the asphalt layer. In spite of the limitations discussed above, an effort was conducted in this research in order to relate continuum damage to cracking by using CalME model. This model includes two steps: the first one to determine damage at crack initiation (equation []) and the second one to predict cracking evolution from damage (equation []). i i h AC h ref α C ωi % Cr α C α C ωi ω Cr i Cr i where, ω is damage ω i is damage at crack initiation h AC is combined thickness of the asphalt layers Cr i is cracking (%) at initiation h ref, i, c parameters of the model Model parameters are case sensitive, although recommended values are included in CalME program for new and rehabilitated pavements. A sensitivity analysis was conducted in order to determine the applicability of these parameters for the sections, materials and environmental conditions in this particular test, but no clear improvement was achieved. The results slightly improved when a value of - was adopted for C instead of the original recommendation that was -. Hence this parameter was the only changed, and default recommended values were used for the rest. Cracking at initiation was also fixed at % according to CalME recommendations. The determination of h AC, the combined thickness of asphalt layers, was not a trivial issue in this particular case, due to the large differences in fatigue resistance between both wearing and base course materials. The bitumen content of the mixture BBTM B employed in the wearing course,.%, was higher than the content of the asphalt concrete employed in the base,.%. Besides, the quality of the binder was considerably higher, since a highly SBS [] [] TRB Annual Meeting

15 Mateos, Ayuso, Cadavid and Marrón modified bitumen was used (PMB -/, according to EN denomination). No fatigue test was conducted for the wearing course material, but results from four-point bending tests indicated that the fatigue life of an asphalt concrete mixture (similar to the base used in this particular test) was multiplied by more than when the plain bitumen was substituted by PMB -/. In fact, this type of bitumen is frequently used in Spain in anticracking layers. Consequently, the use of this bitumen in the wearing course was expected to result in a significant cracking delay. In order to evaluate and quantify such delay, the actual asphalt thickness of each section was used in equation [] to predict damage at crack initiation. It was observed that predicted values were considerably smaller than actual damage values when cracking first appeared. Based on this result, the better quality of the mm wearing course was accounted for by increasing the combined asphalt thickness of the four sections in equation []. The increment, h, was determined by minimizing the sum of squared differences between actual and predicted damage values at crack initiation. The result h = mm was obtained, which is a very high value considering the reduced thickness of the wearing layer; this actually shows the excellent anti-cracking performance of this binder. CalME cracking predictions were calculated and compared to actual values measured during the test, as presented in Figure. With the only exception of section, the agreement between model and measured cracking was excellent, especially considering that the agreement was achieved by using default parameters. Concerning the apparent lack of agreement for section, authors believe it was related to the difficulty inherent in detecting cracks on open graded mixtures. The test was considered finished on December st, when,, loads had been applied. Along, vehicles were periodically in circulation, even though the test was finished, just for maintenance and update operations., additional cycles had been applied by the end of, which was not enough to produce any significant damage. Nevertheless, it provided enough time for cracks to open due to thermal effects. This meant an increase in cracking for all sections, but especially for section. The amount of additional cracking of this section cannot be explained by damage, which means that asphalt was highly deteriorated by the end of the test, but generalized cracking was simply not visible on the surface. It must be borne in mind that aging effects were not significant for this test, due to its relatively short duration. As a consequence, the extrapolation of these results to in-service pavements, where aging plays a major role, is not straight forward. The relatively short duration of the test, directly related to its accelerated nature, also represents one of its main drawbacks. Different procedures exist that can accelerate aging process in accelerated pavement testing (), but they were not employed for this particular test. The type of cracking that was developed on the surface of the pavements resembled the pattern of alligator skin, as expected from bottom-up fatigue. But slabs extracted after the test indicated that both bottom-up and top-down cracks were present. The presence of top-down cracks was not expected, due to the high quality of the wearing course mixture and to the clear relation existing between asphalt layer modulus decrease (i.e. ω) and cracking. No reliable explanation can be given at present, but it might indicate that top-down cracking is also related to the loss of support under the wearing course, which was caused from bottomup fatigue in this particular test. This hypothesis is also supported by the fact that no crack appeared along the curves of the facility, where the same wearing course was tested on a high modulus asphalt concrete binder layer placed on top of a reinforced concrete slab, especially considering the high shear stresses that develop under the tires of the vehicles due to the reduced radio ( m) of the curve. TRB Annual Meeting

16 Mateos, Ayuso, Cadavid and Marrón CONCLUSIONS The systematic evaluation of the bearing capacity of four flexible pavements during a fullscale test has provided valuable information concerning the evolution of asphalt damage, aging and densification under traffic. It has also provided information in order to determine how each one of these three factors modifies the stiffness of the asphalt mixture. In particular, the effects could be quantified in terms of changes in the parameters of the original master curve, for which the MEPDG format was used. The overall tendency observed for FWD back-calculated moduli during the test, consists on an increase in the minimum modulus (for high temperatures) as well as a reduction of the slope of the curve modulus versus temperature. Damage effects were reflected as a reduction of mixture modulus that was higher, in absolute and relative terms, for decreasing asphalt temperatures. This pattern is fully compatible with reducing in the master curve. Aging did not play a major role in this particular test, due to its relatively short duration in comparison to typical useful life of an in-service pavement. This was verified from FWD back-calculation and laboratory testing conducted on cores extracted along an untrafficked line of the sections. Densification under traffic was reflected as an increase in the minimum modulus. This pattern is fully compatible with increasing in the master curve. This is equivalent to introducing a constant densification factor that multiplies the modulus of the uncompacted asphalt mixture for any reduced frequency. Considerably high values, between. and., were obtained for the densification factors of the different sections. A strong negative correlation existed between these values and actual layer thickness, as was expected from a post-compaction process. The evolution of these factors during the test was related to permanent deformation of the wearing course. All previous observations are fully supported by CalME asphalt stiffness model. This model considers aging and densification under traffic by increasing, and takes damage into account by introducing the factor (-ω) that multiplies the parameter in the master curve. By using CalME model, the evolution of the stiffness of the asphalt layer could be successfully reproduced during this full-scale test as a combination of damage, densification under traffic and, to a minor extent, aging. CalME cracking model was successfully used in order to predict measured cracking from actual damage (ω). An excellent agreement was achieved by using default parameters recommended for new flexible pavements by this program. Cracking pattern on the surface resembled alligator skin, although both bottom-up and top-down cracks were present. Experimental evidence from this test indicates that topdown cracking is not only related to surface stresses but also to the loss of support under the wearing layer. TRB Annual Meeting

17 Mateos, Ayuso, Cadavid and Marrón ACKNOWLEDGEMENTS To Prof. Per Ullidtz, for his advice and for providing a sound basis for this study. To California Department of Transportation for the development of California Mechanistic- Empirical Pavement Design Program and the University of California for the technical support. REFERENCES () Mirza, M. W., and M. W. Witczak. Development of a Global Aging System for Short and Long Term Aging of Asphalt Cements. Journal of the Association of the Asphalt Paving Technologists, Vol.,. () Andrei, D., M. W. Witczak, and M. W. Mirza. Development of a revised predictive model for the dynamic (complex) modulus of asphalt mixtures. Inter Team Rep. NCHRP -A, Univ. of Maryland, College Park,. () Mateos, A., J. Ayuso, and B. Cadavid. Shift Factors for Asphalt Fatigue from Full- Scale Testing. th Annual Meeting of The Transportation Research Board, National Research Council, Washington, DC,. (pending publication in Journal of the Transportation Research Board) () APT Update. Third International Conference on Accelerated Pavement Testing, Centro de Estudios y Experimentación de Obras Públicas, Madrid, Spain. () CalME v.. Help file, California Department of Transportation, July. () Ullidtz, P., J. Harvey, B. W. Tsai, and C. L. Monismith. Calibration of Incremental- Recursive Flexible Damage Models in CalME Using HVS Experiments. UCPRC-RR- -, California Department of Transportation,. () Mateos, A. Modeling the Structural Response of Flexible Pavements from Full Scale Test Track Experimental Data. Ph.D. Thesis, Technical University of Madrid, Department of Mechanics of Continuum Media and Theory of Structures, Madrid,. () Houston, W. N., M. W. Mirza, C. E. Zapata, and S. Raghavendra. Environmental Effects in Pavement Mix and Structural Design Systems. NCHRP Report w, Transportation Research Board, Washington, DC,. () Masad, E., E. Kassem, A. Chowdhury, and Z. You. A Method for Predicting Asphalt Mixture Compactability and its Influence on Mechanical Properties. Report --, Texas Transportation Institute, College Station, Texas,. TRB Annual Meeting

18 Mateos, Ayuso, Cadavid and Marrón () Seo, Y., O. El-Haggan, M. King, S. Lee, and Y. R. Kim. Air Void Models for the Dynamic Modulus, Fatigue Cracking, and Rutting of Asphalt Concrete. Journal of Materials in Civil Engineering, Vol., No.,. () Rowe, G., S. Khoee, P. Blankenship, and K. Mahboub. Evaluation of Aspects of E* Test by Using Hot-Mix Asphalt Specimens with Varying Void Contents. Journal of the Transportation Research Board No., National Research Council, Washington, D.C.,. () Hanson, D. I., R. B. Mallick, and E. R. Brown. Five-year evaluation of HMA properties at the AAMAS test projects. Journal of the Transportation Research Board No., National Research Council, Washington, D.C.,. () Chehab G. R., Y. R. Kim, R. A. Schapery, M. W. Witczak, and R. Bonaquist. Time- Temperature Superposition Principle For Asphalt Concrete With Growing Damage in Tension State. Proceedings of the Association of Asphalt Paving Technologists, Vol., Association of Asphalt Paving Technologists, Colorado Springs, Colorado,. () Zhao, Y., and Y. R. Kim. Time-Temperature Superposition for Asphalt Mixtures with Growing Damage and Permanent Deformation in Compression. Journal of the Transportation Research Board, No., National Research Council, Washington, D.C.,. () Kandhal, P.S., and W.C. Koehler. "Significant Studies on Asphalt Durability: Pennsylvania Experience." Journal of the Transportation Research Board, No., National Research Council, Washington D.C.,. () Roque, R., A. Guarin, G. Wang, J. Zou and H. Mork. "Develop Methodologies / Protocols to Asses Cracking Potential of Asphalt Mixtures using Accelerated Pavement Testing." Final Report Project No.:, University of Florida, Gainesville, Florida. TRB Annual Meeting