Significant Findings from Full-Scale Accelerated Pavement Testing (2012)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

53 chapter three Pavement materials and structure evaluation introduction This chapter provides detailed findings on the specific pave- ment materials being investigated with f-sAPT, as well as the responses obtained from different pavement structures and the effect of loading and environmental conditions on the test results. The chapter covers the full range of applicable materi- als (HMA, concrete, stabilized and granular—including novel materials) and combinations of materials (i.e., traditional pavement structures, warm mix technologies, thin reinforced concrete, etc.). The chapter begins with a summary of the questionnaire results regarding materials and pavement structure, followed by an analysis of the available literature. The literature is divided into sections covering HMA, concrete, granular mate- rials, pavement systems, and miscellaneous topics. Under each of the major materials, the information is grouped into subsections of similar topics (i.e., overlays, cracking, etc.). It should be noted that all models and outputs referenced and discussed in this section are only applicable to the spe- cific materials and conditions under which the specific test- ing has been conducted. Space does not allow all these condi- tions to be repeated for each of the references, and the reader is urged to consult with the original reference to ensure that details on these aspects are well understood before the infor- mation in this section is applied or referenced. The most common surfacing material evaluated by respon- dents is HMA followed by traditional concrete, with most of the other surfacing types only being evaluated by less than 30% of the respondents (Figure 38). WMA was interestingly (as a relatively novel material) the third-most evaluated sur- facing. For pavement base layers, granular materials have been evaluated the most often, followed by a group of base materials including asphaltic, cemented, and recycled (Fig- ure 39), whereas the various unstabilized materials were most often used for subbase and subgrade materials (Figure 40). In response to the structural distress types evaluated for HMA surfacings, the traditional cracking (incorporating all types of structural fatigue) and rutting was the most often evaluated (Figure 41), as was also observed from the topics covered in the available literature. A similar picture evolved for concrete surfacings, with cracking being most often eval- uated (Figure 42). Permanent deformation was most often evaluated as the structural distress type for base and subbase layers (Figure 43). In terms of functional distress types, fewer respondents include it in their evaluations (as may be expected with f-sAPT devices typically focusing on structural evaluation aspects of pavements), with the second-most selected option indicated that functional distress is not deemed applicable for evaluation using f-sAPT. This is probably influenced by the large number of f-sAPT devices that use short pavement sec- tions for testing (as opposed to test tracks). Most respondents indicated rutting to be the functional safety aspect most often evaluated (similar to asphalt surfacing distress types). Most programs used wheel load as the load characteristic to which they related f-sAPT data, with the tire-related prop- erties (inflation pressure, type, contact stress) starting to play a more prominent role (Figure 44). The material properties used to explain performance (Figure 45) are mostly typical properties that are known to affect structural performance of materials. Respondents most commonly rated unconven- tional materials and compaction as the aspects of pavement engineering that may enhance construction and rehabilitation of pavements in their f-sAPT programs. Hot mix asPHalt and related materials rutting Mulvaney (2004) analyzed rutting on MnROAD’s original 14 HMA mainline test sections, showing that cells with the softer PG58-28 binder have rutted more than the stiffer PG64-22 binder cells, with rutting forming in the upper lifts of the HMA. Fifty percent of the rutting occurred in the first two years after construction, indicating that rutting is not lin- ear with time or traffic level. Although the driving lane has rutted 1.5 times more than the passing lane, the rutting is not linear with the number of ESALs (driving lane experienced four times the ESALs of the passing lane). Sivasubramaniam et al. (2004) evaluated the performance of a Superpave HMA mixture from f-sAPT compared with the PURWheel laboratory wheel tracker and the Indiana DOT/ Purdue University Accelerated Pavement Tester. It was observed that well-designed and constructed fine-graded mix- tures may be appropriate for use with heavy traffic applications and in hot climates. The rutting performance of coarse-graded

54 0 5 10 Ho t m ix a sp ha lt Co ncr ete Wa rm m ix a sp ha lt Oth er Un pa ve d Bit um en su rfa ce tre atm en t Co mp osi tes Ult rat hin wh ite top pin g Co ld m ix a sph alt Ult ra- thi n r ein for ce d c on cre te 15 20 25 30 N um be r o f r es po nd en ts FIGURE 38 Surfacing materials evaluated with f-sAPT. 0 5 10 Gr an ula r As ph alt Ce me nt sta bili zed Re cyc led Em uls ion st ab iliz ed Sa nd Cla y Co mp osi te Oth er 15 20 25 N um be r o f r es po nd en ts FIGURE 39 Base materials most often evaluated by f-sAPT respondents. mixtures appeared to be slightly better compared with fine-graded mixtures (although no statistical difference was observed). A simple power model was successfully fitted to HMA rutting data originating from the f-sAPT and the PURWheel. Muraya (2004) evaluated the resilient and rutting behav- ior of two asphalt motorway test pavements subjected to LINTRACK (Netherlands) f-sAPT loads to analyze the pos- sibility of back-calculating the measured rutting on the basis of cyclic triaxial tests. Different trends were observed in the measured and calculated rutting behavior, with the calculated rutting (based on material characterization by resilient defor- mation triaxial testing) being much lower than the measured rutting in the test pavements. The poor correlation between the triaxial test results and the measured pavement response may be improved by better simulation of the stresses occur- ring in the pavement. The simulation of pavement stresses can be enhanced by the application of different pulse dura- tions for both the vertical and horizontal stresses.

55 0 2 4 6 8 10 12 14 16 18 20 Granular Clay Sand Cement stabilized Emulsion stabilized Other stabilized N um be r o f r es po nd en ts Axis Title FIGURE 40 Subbase and subgrade materials evaluated by respondents. 0 5 10 15 20 Cra cki ng Ru ttin g Ag gre ga te los s Ble ed ing Oth er Mo istu re da ma ge / s trip pin g Ag ing Lo w t em pe rat ure cr ac kin g 25 30 35 40 N um be r o f r es po nd en ts FIGURE 41 Structural distress types evaluated for asphalt surfacings. 0 5 10 Cra cki ng Jo int fa ilur e Lo ad tra ns fer fa ilur e Fa ult ing Cu rlin g a nd wa rpi ng Sp alli ng Ero sio n o f s up po rt Pu nc ho uts De bo nd ing 15 20 25 30 N um be r o f r es po nd en ts FIGURE 42 Structural distress types evaluated for concrete surfacings.

56 0 5 10 Pe rm an en t d efo rm ati on Sh ea r Cra cki ng Cru shi ng Sw elli ng Fro st/t ha w d am ag e Co llap sin g Ca rbo na tion 15 20 25 N um be r o f r e sp o n de nt s FIGURE 43 Structural distress types evaluated for base and subbase layers. FIGURE 44 Load characteristics related to f-sAPT performance. 0 5 10 Ap plie d w he el loa d Tir e i nfl ati on pr es su re Tir e w an de r p att ern Lo ad co nfi gu rat ion Sp ee d Ov erl oa din g Tir e t yp e Tir e c on tac t s tre ss Ve hic le- pa ve me nt dy na mi cs Pa ve me nt rou gh ne ss Re st pe rio ds Su spe nsi on sy ste m 15 20 25 30 N um be r o f r es po nd en ts FIGURE 45 Material properties used to explain performance. 0 5 10 Bin de r ty pe Gr ad atio n Sti ffn ess De ns ity Co mp res siv e s tre ng th Vo lum etr ic p rop ert ies Bin de r c on ten t Vis co -el as tic pro pe rtie s Mo istu re co nte nt Te ns ile str en gth Fle xu ral st ren gth Bin de r a gin g in de x Att erb erg lim its Bin de r F ilm th ick ne ss Po iss on ’s r ati o Ero dib ility 15 20 25 N um be r o f r es po nd en ts

57 Park et al. (2004) proposed an elastic-visco-plastic con- tinuum model based on Perzyna’s visco-plastic theory and the Drucker–Prager yield function for simulation of HMA rutting at WesTrack using compressive strength test data at different strain rates and confinement pressures at a test temperature of 140°F (60°C). A Temperature Conversion Factor accounting for temperature variation in the field was developed based on repeated triaxial tests performed at 86°F, 104°F, and 122°F (30°C, 40°C, and 50°C). The Temperature Conversion Factor was used as a simple way to convert traffic loading at various temperatures to its equivalent at a standard temperature. Analy- sis indicated the WesTrack HMA to be a strain-dependent and pressure-dependent material at high temperature. Wu (2004) analyzed the effects of HMA material model parameters and load assembly (tire pressure, wheel spacing) on Superpave mixture rutting. A three-parameter creep model was successfully used to capture the behavior of HMA mix- tures while analyzing rutting development of the f-sAPT sec- tions, showing that the predicted rutting depth was sensitive to all parameters of the creep model. Predicted rutting increased significantly with increasing tire inflation pressure and also with an increase in HMA layer thickness. Because the rut- ting depth is the summation of all creep strains throughout the depth of the pavement layer, this is a logical outflow from using the specific creep model. A set of multiple-regression equations for predicting rutting depths on Superpave mixture pavements has been developed. Villiers et al. (2005) developed an approach to integrate FWD and core data along with transverse profile measurements to assess the contributions of different pavement layers to rut- ting and identify the instability within the HMA layer. Forensic data from f-sAPT trenches and laboratory test results verified the approach. The importance of evaluating rutting develop- ment from each individual layer was shown in the development. Agardh and Busch (2006) evaluated different types of rutting depth prediction models for an incremental design process. This included a model that calculates the plastic strain through the whole pavement, and a model that is based on the critical response at a specific location in the pavement. Both models were calibrated using f-sAPT data and validated with theoretical sections and in-service roads. Both mod- els were found to overestimate the rutting depth at the two real pavements used for the evaluation. The energy model overestimated rutting depth compared with real sections for high traffic volumes. As the comparison was done for roads located on relatively weak pavements, this may have affected the rutting model calibration. Monismith et al. (2006) presented relationships between shear test results on field cores obtained prior to trafficking for four different mixes comprising 34 of the WesTrack sections, showing an excellent correspondence between the shear test results and rutting depths measured at specific numbers of load repetitions. These results suggest that the methodology utiliz- ing the results of test data obtained with the Repeated Simple Shear Test at Constant Height (RSST-CH) can be used effec- tively for both HMA mixture design and rutting prediction in a M-E analysis/design procedure. The test appears to capture the impact of modified binders on rutting resistance. The behavior of a flexible pavement using an elastic- visco-plastic constitutive relation that takes into consider- ation the influence of the temperature for the HMA has been evaluated using numerical analysis, a laboratory evaluation, and f-sAPT (Ali et al. 2006a, b). Numerical results show that the elastic-visco-plastic model reproduces the behavior of flexible pavements well, with the increase in the tempera- ture significantly affecting the deformation of the pavement. Increasing the temperature from 68°F to 86°F (20°C to 30°C) caused an increase of 40% in the compressive strain at the top of the subgrade layer. Comparison of the numerical and f-sAPT results with laboratory data from LAVOC (France) indicated that the numerical model reproduced the pavement rutting behavior correctly. FDOT evaluated the rutting performance of coarse and fine-graded Superpave mixtures under f-sAPT (Choubane et al. 2006) (also refer to polymer modification section— Gokhale et al. 2005). The results (supported by experience from NCAT) have shown that fine-graded mixtures can per- form at least as well as coarse-graded mixtures with respect to rutting, and FDOT therefore changed the July 2005 edition of its Superpave specification, allowing fine-graded mixtures for traffic level D and E mixtures, and requiring PG76-22 modified binder in the top structural layer of traffic level D mixtures and in both structural layers for traffic level E mix- tures. The information on the performance of HMA mixes under realistic loading conditions was obtained in a short time using f-sAPT, demonstrating that f-sAPT can produce cost-effectively early, reliable, and practical results while improving pavement technology and the understanding and prediction of pavement systems performance. An effort was made to develop an HMA rutting prediction model from NCAT Test Track data (Immanuel and Timm 2007). One approach consisted of a vertical strain-based rut- ting model built by relating the measured rutting to the vertical strain on the top of granular layers and the traffic volume. The second approach linked rutting with maximum shear strain in the HMA layer and the traffic volume. Both approaches accurately predicted rutting on a section-by-section basis, and although the vertical strain approach could not accu- rately predict rutting when sections were grouped accord- ing to binder modification, the shear strain approach could do this accurately. The shear strain-based rutting prediction model is applicable only when the asphalt concrete layer rut- ting is the leading source of pavement rutting. Mateos et al. (2008) (Spain) conducted f-sAPT on six sec- tions consisting of 6 in. (150 mm) HMA on different alterna- tives of a high-quality subgrade to improve the understanding

58 of mechanistic behavior of these pavements. The data were modeled using FEM where soils were assumed to be linear elastic and the HMA to be linear visco-elastic, with a con- stant bulk modulus and a time-dependent shear modulus. The mechanical stiffness of the HMA strongly depends on the applied load frequency, causing increases in the pave- ment rutting response for decreased vehicle speeds. General agreement was found between model and f-sAPT data for low temperatures; however, as the temperatures increase, the measured response increases more quickly than the model, resulting in an underestimation of the rutting at the high- est temperature—probably related to the linear visco-elastic model used for the HMA mix. Xu and Mohammad (2008) developed an M-E model for simulating HMA rutting depth under f-sAPT based on the power law of vertical strain. Total rutting depth is an integra- tion of the individual layer rutting. Evaluating the model with Accelerated Load Facility (ALF) (Louisiana) data has shown reasonable agreement between the measured and modeled rut, with layer thickness and modulus showing significant influences on rutting development. Azari et al. (2008) reported that HMA lanes were con- structed with six different asphalt binder types and two dif- ferent thicknesses and tested for rutting and fatigue crack- ing at the FHWA’s ALF site. Plant manufactured HMA specimens were compacted in the laboratory and subjected to dynamic modulus and flow number tests using the Simple Performance Tester (SPT) together with laboratory manufactured specimens. Good agreement was observed between permanent strains measured by the flow number test and the ALF rutting measurements for initial rutting. HMA rutting was also predicted using the MEPDG Level 1 and 3 analyses. The objective was to estimate HMA rut- ting using the MEPDG software and NCHRP 1-37A perma- nent deformation prediction models and compare the per- manent strains measured during the flow number test with the measured f-sAPT rutting. The Level 3 MEPDG pre- dictions were generally higher than measured ALF rutting (but in a similar range) while the Level 1 MEPDG predictions were significantly higher than the ALF measured rutting, and also significantly higher than Level 3 predictions. The Level 3 stiffness (from the NCHRP 1-37A stiffness predic- tion equation) significantly overestimated the stiffness of the ALF lanes, and it is concluded that the main reason for the overprediction of rutting (using Level 1 models) was the calibration of the permanent deformation model that was apparently performed using predicted stiffnesses from the NCHRP 1-37A stiffness equation rather than tested E* using the SPT. This also showed that HMA stiffness alone cannot explain field rutting sufficiently. Gibson et al. (2009) (FHWA) further found that the SST- RSCH (at fixed air voids) and Flow Number (as-built air voids) appeared to provide the strongest indication of HMA rutting based on statistical analysis of laboratory data. When incorporating Flow Number test data into HMA rutting pre- diction models for the MEPDG analysis, realistic compari- sons were found between f-sAPT and predicted rutting. Coleri et al. (2008) (California) demonstrated the appli- cability of a two-stage Weibull approach to simulate field rutting performance of HMA mixes as a function of binder type, air-void content, shear stress level, and temperature and calibration of the model using f-sAPT results. The ini- tial laboratory models are based on RSST-CH data and the results indicated that the approach is a reliable and promising technique for rutting performance prediction. Korea is in the process of developing an M-E pavement design method (Cho et al. 2010). As part of this development two 0.75 in. (19 mm) dense-graded HMA sections with air- void ratios of 7.3% and 10.6% and a temperature of 122°F (50°C) at 2 in. (50 mm) depth were evaluated under f-sAPT for the development of a HMA rutting performance-based model. The rutting in the HMA was directly linked with the different air-void ratios. The surface rutting amounted to approximately 65% (10.6% air-void) and 57% (7.4% air-void). A stone mastic asphalt mixture and a dense-graded air- field HMA mixture (both known for high shear and rutting resistance) were trafficked with 1,500 passes of a F-15E aircraft load cart at Tyndall Air Force Base, Florida, to evalu- ate the rutting performance of pelletized asphalt HMA (Saeed et al. 2010). The pelletized asphalt HMA section showed no rutting compared with the up to 0.9 in. (22 mm) rutting in the conventional HMA section, supporting the conclusion that it is feasible to produce airfield-quality HMA with pelletized asphalt using conventional HMA plants. Labo- ratory workability tests indicated that the DGA mixture compacted better with less energy than the stone mastic asphalt mixture. Gibson et al. (2010) evaluated the similarities and differ- ences in rutting susceptibility of HMA (FHWA APT site) fabricated by a Superpave Gyratory Compactor and field compaction rollers. The field-compacted mixtures tended to be less resistant to rutting at fewer cycles and smaller permanent strains, but exhibited more strain hardening than Superpave Gyratory Compactor samples at greater cycles and larger permanent strains. F-sAPT induced densification in the wheel paths and resulted in the characteristic upheaval humps on the side of the wheel path. None of the applied stress states or compaction methods was able to produce cor- responding volumetric consolidation. Currently, reversals in stress resulting from a rolling tire are hypothesized as the reason why laboratory tests that do not have stress reversals cannot reproduce field observations accurately. Wang and Zhou (2010) evaluated the effect of HMA pave- ment structures to HMA pavement rutting through f-sAPT on a circular track in Chongqing. Three test pavements with similar HMA wearing course, binder course, and subgrade,

59 and different base and subbase layers were evaluated. Rut- ting was found to occur mostly in the HMA layers. Mbakisya and Romanoschi (2010) evaluated existing mechanistic models that predict HMA rutting by comparing the computed permanent deformation to f-sAPT data from Kansas for six different asphalt mixes. The Drucker–Prager model overpredicted permanent deformation even after creep hardening was added to the model, whereas elasto-visco- plastic and creep models predicted total rutting that ranged between 95% and 102% of the f-sAPT values on the surface, subbase, and subgrade. HMA rutting was predicted to range between 28% and 78% of measured values. Coleri and Harvey (2011) demonstrated a reliability analysis approach for prediction of HMA rutting perfor- mance by evaluating reliability through considering the variability in laboratory test results, layer thicknesses and stiffnesses, and measured in situ performance (California). Each input parameter’s contribution to variability was assessed, and these variability distributions were used for rutting perfor- mance prediction and reliability evaluation of highway sec- tions. The results indicated that distributions of calibration coefficients calculated by using measured rutting depths were very similar to calibration coefficient distributions calculated by using thickness and stiffness variability. This suggests that variability in performance can be effectively predicted by using the thickness and stiffness variability for f-sAPT sections as thickness and stiffness variability were found to be the major factors controlling the rutting vari- ability. The variability of stiffness for bound and unbound layers was represented through lognormal distributions, whereas the variability in layer thicknesses was described by normal distribution functions. Precision of laboratory test results did not have much effect on predicted rutting performance when the effect of laboratory variability is simulated with construction or performance variability. It thus appears that a high level of variability in measured performance, thickness, and stiffness masks the effect of laboratory variability on calculated calibration coefficient distributions. Li et al. (2011) (South Korea) developed and calibrated an HMA rutting model based on shear properties obtained from Triaxial Compressive Strength and Repeated Load Per- manent Deformation tests on three types of HMA at various loading and temperature conditions. The model was cali- brated using f-sAPT data conducted at various temperatures. Cohesion decreased with an increase in temperature for a given mixture, while the friction angle was not significantly affected by temperatures higher than 104°F (40°C). Devia- tory stress, confining pressure, test temperature, and load fre- quency significantly affected the rutting of HMA and were thus incorporated in the rutting model. The proposed model has been shown to successfully predict the rutting of vari- ous HMA mixes under a range of loading and temperature conditions. Fatigue Hartman et al. (2001) (Ireland) performed f-sAPT on Dense Base Course Macadam (DBCM) supported by a weak sub- grade. Transverse horizontal tensile strains caused longitu- dinal cracks in the wheel path with the single front steering tires inducing maximum strains. The measured data were modeled using a linear elastic model and lower strains were calculated, while the expected life to crack initiation was also underestimated using the linear elastic model. The framework used for analytically based HMA pave- ment design methods has remained virtually unchanged for the past 40 years, utilizing the tensile strain in the HMA to deal with fatigue cracking and compressive strain on the subgrade for rutting (Brown et al. 2004). A vast amount of research has identified difficulties and shortcomings with this approach. The formation and growth of fatigue cracks causes the effective stiffness of the HMA to decrease and this effect is viewed as more important in design than details of the cracks. As an alternative approach a prediction method for stiffness deterioration can be used where the HMA stiff- ness modulus decreases as fatigue damage in the layer is accumulated until it reaches a terminal level when the fatigue life has been used up. This can be modeled in an exponen- tial relationship between normalized HMA layer stiffness and accumulated fatigue damage by back-calculating HMA layer stiffness moduli from surface deflection measurements obtained at various loading stages during f-sAPT. Rich-bottom pavements are designed to reduce the sensitiv- ity to tensile strain in the lower layers of a pavement by the addition of an extra 0.5% binder content. The effectiveness of a rich-bottom HMA section was evaluated at NCAT (Willis and Timm 2007). Although it was expected to perform better than control sections, it failed earlier than the control sections in fatigue. Theoretical and dynamic strain data analyses showed the possibility of slippage occurring in the section. Further analysis showed that cracking initiated above the rich-bottom layer with lower strains calculated at the base of the rich- bottom. Forensic investigations indicated that debonding at the layer interfaces was most likely the cause of early cracking in the rich-bottom section, which led to early fatigue cracking. F-sAPT conducted at NCAT showed bottom-up fatigue cracking that developed as a series of short transverse cracks that extended to the edge of the wheel path before interconnect- ing into the familiar alligator pattern (Timm and Priest 2008). This differs from observations on LTM sections, but is often observed in pavement experiments. Measurements showed the longitudinal strain to be higher than the transverse strain, partially owing to strain reversal before and after a given axle and between axles in an axle group. Timm and Priest (2008) recommended that strain reversal in the longitudinal direction should be included in fatigue transfer function calibration and that tandem axles could be considered separately, rather than as an axle group.

60 Howard and James (2009) evaluated general design guid- ance for bottom-up fatigue cracking of thin flexible pavements (located in Arkansas) using data collected as part of a study to evaluate geosynthetics in thin flexible pavements. All test traffic was channelized at a speed of 35 mph (56 km/h) using single-axle dump trucks (FHWA category 5). HMA strains were observed to not be normally distributed, whereas the earth pressures were normally distributed. Through FEM analysis it was determined that temperature gradients caused skew HMA strain data. After-analysis relationships were developed of damage as a function of mean strain (considering the measured temperatures) and strain distributions. The design thickness resulting from this approach does not account for any distress other than bottom-up fatigue cracking. The guidance is only valid for the low-volume roads portion of MEPDG and compli- ments previous work. Kutay et al. (2009) evaluated the possibility of using small sample sizes for push-pull tests to evaluate the Visco-Elastic Continuum Damage (VECD) characteristics of HMA. The difference in characteristics for regular and small size samples was observed to be negligible. The use of the uni-axial push- pull test on cylindrical samples is viewed as a novel alterna- tive for the costly and time-consuming bending beam test. A practical fatigue life formulation based on the VECD theory was derived and used to investigate differences in fatigue lives of different HMA layers at the FHWA’s ALF site. Unlike the fatigue life formulation implemented in MEPDG, the proposed VECD equation was able to capture the f-sAPT behavior well. Since the push-pull tests are compatible with the Asphalt Mix- ture Performance Tester (AMPT), it is possible and practical to calibrate the VECD model by performing an additional test in the AMPT and then calculating the fatigue life of the pave- ments based on these equations. This fully mechanistic proce- dure may be used in place of the empirical fatigue life formula- tion implemented in the MEPDG. Christofa and Madanat (2010) developed crack initiation models for flexible pavements by using data from the Perfor- mance Analysis for Road Infrastructure (PARIS) project data- base (located in the European Union), which consists of both Real-Time Load Testing data and f-sAPT data. Crack initiation models allow for optimal maintenance and rehabilitation activ- ities and optimal pavement design. The development showed that many of the explanatory variables that are expected to play a significant role in determining crack initiation (i.e., granular layer thickness) appear to have insignificant parameter esti- mates, probably a result of the small number of observations used. Analysis of the model predictions indicates that they are reasonable, although they overestimate crack initiation prob- ability in the early stages of pavement life. Mateos et al. (2011) quantified the beneficial effect of traffic rest periods through use of a variable shift factor based on the methodology developed in NCHRP 9-38 (Endurance Limit of Hot Mix Asphalt Mixtures to Prevent Fatigue Crack- ing in Flexible Pavements) (Prowell et al. 2010) and NCHRP 9-44 (Developing a Plan for Validating an Endurance Limit for HMA Pavements) (Bonaquist 2008). A shift factor of 475 was calculated when laboratory fatigue life was compared with f-sAPT data from CEDEX (Spain) indicating the num- ber of load repetitions to reach 20% alligator cracking. This variable shift factor was successfully applied to laboratory data incorporating the reduced rest period effect, followed by application of the California Mechanistic Empirical (CalME) fatigue model to adequately predict HMA deterioration in an f-sAPT experiment. The HMA fatigue model in the MEPDG is based on the original Asphalt Institute model with modifications. Wen (2011) proposed two new fatigue models based on FHWA ALF data. Inclusion of the dynamic modulus and accounting for HMA layer thickness and construction variations in the traditional strain-based fatigue model improved the effective- ness of the applied model. A combination of critical strain energy density and dynamic modulus in the fatigue model correctly characterized the effect of HMA layer thickness on fatigue life in the analyses, and this model was proposed as being potentially effective for bottom-up cracking analysis. modified Binders Sirin et al. (2003) used data from the Florida HVS to evalu- ate the rutting performance of unmodified and styrene buta- diene styrene (SBS)-modified Superpave HMA. Analysis showed that the pavement sections with two lifts of SBS- modified mixture outperformed those with two lifts of unmodi- fied HMA. Sections with a lift of SBS-modified HMA over a lift of unmodified HMA only showed approximately 20% higher rutting compared with the two lifts of modified mixture tested at 122°F (50°C). Two lifts of SBS-modified HMA tested at 149°F (65°C) showed high rutting resistance compared with two lifts of unmodified HMA tested at 122°F (50°C). Rutting development in the Asphalt Pavement Analyzer correlated with the f-sAPT rutting data. It was concluded that for pave- ments with the unmodified HMA, rutting was caused by a com- bination of densification and shoving, while the SBS-modified HMA experienced rutting primarily the result of densifica- tion of the HMA. The resilient modulus of the SBS-modified HMA was not significantly different from that of the unmodi- fied HMA at 77°F (25°C). Evaluation of the viscosity at 140°F (60°C) of the recovered binders showed the binders recovered from the SBS-modified HMA to be two to three times that of the recovered binders from the unmodified HMA. FDOT evaluated the effects of polymer modifiers on the performance of two fine-graded Superpave HMA mixes using the HVS (Gokhale et al. 2005). The sections with the SBS-modified mixture significantly outperformed those with the unmodified mixture, and it was determined that rutting in the unmodified mixture was primarily a function of shear flow while rutting in the SBS-modified mixture was due pri- marily to densification.

61 D’Angelo and Dongre (2006) evaluated the development of a high temperature performance-based binder specifica- tion to characterize both modified and unmodified HMA binders (using an HMA mixture from the FHWA ALF site). Both creep and recovery testing of the binders showed strong correlation with the mixture performance and showed the stress dependency of several polymer-modified systems. With stress dependency viewed as a key factor in determin- ing the relationship between binder performance and mix- ture performance, a multistress creep and recovery test was developed that distinguishes the difference between polymer systems and simplifies the testing. Al-Khateeb et al. (2007) conducted mechanistic analyses on the modified and unmodified asphalt HMA pavements of the FHWA’s ALF using KENPAVE, WINLEA, EVERSTRS, EVERFLEX, and VESYS 5W. Load frequency impacted the tensile fatigue stress and strain at the bottom of the HMA layer and further affected the major principal stress at the surface and bottom of the HMA layer, and the minor prin- cipal stress at the bottom of the HMA layer. Insignificant effects were observed in the rutting. The mechanistic analy- ses indicated that Multi-Layered Elastic Theory provided reasonable predictions for the measured tensile strains. VESYS 5W provided good correlations with the measured vertical deformation within the HMA. When using the same loading frequency in the analyses, similar tensile strain outputs were obtained for the various analysis options. Von Quintus et al. (2007) evaluated the benefits of using polymer-modified asphalt (PMA) based on field and f-sAPT data from nearly three dozen pavement sections in the United States. It was evident that the use of PMA definitely extends the service life of flexible pavements and HMA overlays, with pavements incorporating PMA mixtures found to have lower amounts of fatigue cracking, transverse cracking and rutting, and extended service lives of 5 to 10 years. ALF testing conducted in Beijing focused on evaluation of the rutting resistance of four different modified HMA mixes (Zejiao et al. 2010). Analysis showed that the SBS- modified HMA had the best rutting resistance of the HMA mixes evaluated, and that the rutting resistance of the bottom HMA layer for semi-rigid base pavements (commonly used in China) is crucial in the performance. The effects of binder properties on HMA mixes at inter- mediate temperatures were evaluated using a new materials characterization method and five binders (four modified) and five HMA mixes containing these binders with similar aggre- gates (Wen et al. 2011). All the HMA mixes originated from the FHWA f-sAPT facility. All the modified binders reduced the strength of the binder (except for terpolymer) but increased the failure strain and ductility of the binders (except for air- blown material). All the modification methods reduced both the dynamic moduli and indirect tensile strengths of the HMA mixes. A clear correlation was observed between the failure strain of the binders and that of the HMA, and also between the complex shear moduli of the binders and dynamic mod- uli of the HMA. No clear relationship could be observed between the strength and fatigue resistance of the binders and those of the HMA mixes. Although PMA has traditionally been used in upper pave- ment lifts to enhance rutting performance where tempera- tures, vertical stresses, and shear action are extreme, the need for thinner high performing pavements provides motivation for using higher modulus, fatigue-resistant materials in the lower lifts of a flexible pavement. Timm et al. (2011a) compared the structural responses of highly polymer modified (HPM) asphalt on the NCAT test track with a control section. The structural behavior of both HPM and the control section in response to temperature changes was well-characterized by exponential functions. At both the lowest reference temperature [50°F (10°C)] and intermediate reference temperature [80°F (27°C)] the HPM section experienced higher average strain compared with the control section (partly owing its thickness). This implies that the HPM section would have lower strain at intermediate temperatures if the thickness of the two materials is simi- lar. At the highest reference temperature [109°F (43°C)] the control section experienced higher average strains compared with the HPM, despite a 1.4 in. (35 mm) thickness advantage over the HPM. The aggregate base experienced higher aver- age vertical stresses in the HPM section than the control sec- tion at all temperatures. The f-sAPT data indicated that the HPM is effective at reducing HMA layer thickness without significant change in performance. The transportation pooled fund study TPF-5(019) (Full- Scale Accelerated Performance Testing for Superpave and Structural Validation) was initiated with the construction of 12 f-sAPT lanes with various modified HMAs at the FHWA’s APT site with the objective of validating and refining pro- posed changes in Superpave specifications to properly grade- modified HMA binders (Mitchell et al. 2004; Qi et al. 2008, 2010). Results of the first year’s f-sAPT generally ranked the performance of the modified binders much differently than the results from laboratory mixture performance tests (French Pavement Rutting Tester, Hamburg Wheel Testing Device, and Superpave shear tester). None of the tests cor- relating to ALF-generated rutting, as the rutting depths of most of the lanes were statistically the same, as the binders were designed to have nearly identical high temperature per- formance grades (Gibson et al. 2011). Rutting performance throughout some mixes depended on HMA layer thickness, and correlations after sorting the data according to layer thickness were moderate between the French Pavement Rutting Tester results and the 4-in. (100-mm)-thick HMA rutting depth, and high between RSST-CH results and 6-in. (150-mm)-thick HMA rutting depth. No overall relationship was found between G*/sind of the HMA binders and rutting depth.

62 F-sAPT fatigue cracking performance data were compared with four binder parameters [conventional G*sind obtained from time or frequency sweep at low strains, G*SsindS obtained from dynamic shear rheometer (DSR) strain sweep at high strains, Essential Work of Fracture, and number of cycles to fatigue failure (Nf) obtained from a stress sweep] (Qi et al. 2006). The Essential Work of Fracture did not show a good correlation with f-sAPT data, whereas the current Superpave specification and proposed refinements showed reasonably good correlations. Although Nf from a stress sweep showed a better correlation, the trend was reversed. A total of nine binder tests were ultimately considered for fatigue cracking along with a more refined statistical selection procedure for the final report. Calculated Critical Tip Opening Displacement is a variation of binder performance test to control fatigue crack- ing above the eight other candidates. F-sAPT rutting perfor- mance data has been compared with four binder parameters [conventional G*/sind, JNR (nonrecovered compliance) obtained from DSR multistress creep recovery tests], G*/ (1 - 1/tand sind) obtained from DSR time or frequency sweeps and MVR (volumetric flow rate) obtained from the flow mea- suring device] (Qi et al. 2006). It was found that the term G*/ (1 - 1/tandsind) obtained from DSR time or frequency sweep showed the best correlation among the binder parameters in all cases for the thin and thick HMA pavements evaluated together and separately when outliers were removed. In stress-controlled push-pull tests on mixtures, Nf versus field number of wheel passes to 20% crack length revealed an inverse relationship, attributed to the testing mode where it was observed that the soft mixtures performed worse in stress-controlled tests, while they performed well in the field (Qi et al. 2008). Nf based on stress-controlled push-pull tests are not recommended for fatigue characterization. Strain- controlled Nf in mixture testing generally exhibited similar trends when compared with the field; however, owing to the nature of the testing where the strain was controlled at the actuator, the field ranking was not strongly captured, and therefore Nf based on actuator strain-controlled tests may not be sufficient to understand the field fatigue behavior. Choice of stress level appears to influence the correlation between laboratory test and pavement rutting with different pavement thicknesses. High-stress ratio-unconfined friction number tests showed better relationship with rutting in thin pavements [4 in. (100 mm)] than rutting in thick pavements [6 in. (150 mm)]. Low-stress ratio-confined friction number tests provided good relationship with rutting in thick pave- ments [6 in. (150 mm)] and no relationship in thin pavements [4 in. (100 mm)]. Approaches to improve low temperature fracture resis- tance of HMA pavements were evaluated using data origi- nating from MnROAD and other sources (Marasteanu et al. 2007). It revealed that two simple mixture tests can be used to provide relevant fracture properties of the HMA mixtures. The fracture toughness and energy obtained from these tests correlated best with the field distresses measured in the selected pavement sections. Even at low temperatures HMA mixtures were shown to be complex visco-elastic composite materials that are significantly temperature and loading rate dependent. The study demonstrated that the effect of tem- perature is significant as the behavior changes from brittle- ductile to brittle. When conducting low temperature tests on HMA mixtures, testing temperatures should be established relative to the expected low pavement temperature and/or rela- tive to the low temperature Superpave performance grading (PG) grade for the location of interest. The mixture coefficient of thermal contraction was shown to be a critical parameter for estimation of field performance for low temperature cracking and it was shown that the coefficients are affected by binder grades and by mixture variables. It was recommended that a specification be developed for selecting HMA mixtures with increased fracture resistance similar to the PG system for binders, as low temperature cracking performance cannot rely entirely on the PG of the asphalt. Hot mix asphalt overlays F-sAPT on dense-graded asphalt concrete (DGAC) and gap- graded asphalt rubber hot mix (ARHM-GG) at elevated tem- peratures was used to compare the rutting behavior under different tire types (Harvey and Popescu 2000). The 2.5 in. (62 mm) ARHM-GG and 2 to 3 in. (50 to 75 mm) DGAC overlays had similar performance, while the 1.5 in. (38 mm) ARHM-GG overlay had superior performance under dual radial tire loads. However, the ARHM-GG overlays had bet- ter performance than the DGACs overlay under wide-base single tire loads. Low Hveem stabilometer values obtained for the ARHM-GG mixture did not indicate the good perfor- mance of the mixture under HVS loading and indicated the need for improved methods of characterizing rutting perfor- mance of ARHM-GG mixes. The 46°F (8°C) temperature difference at 2 in. (50 mm) depth between two similar tests resulted in 140 times more load repetitions being applied before rutting failure on the cooler section compared with the hotter section, emphasizing the need to account for local pavement temperatures to be incorporated in the mixture design process. Kumar et al. (2006) assessed whether any improvement occurs in pavement performance when incorporating a 2-in. (50-mm) HMA binder layer as part of a second generation overlay of HMA pavements, deviating from the typical 1.5-in. (38-mm) HMA surface layer overlay with leveling adopted in Pennsylvania. Dynamic modulus mastercurves for the surface and binder layer mixes revealed that for operating temperatures, both the surface and binder layer mixtures would probably lead to similar performance. F-sAPT indicated that the inclusion of the binder layer does not significantly alter the rutting resistance of the pavement structure. This was confirmed through a sensitivity study modeling both pavement structures and empirical distress prediction mod- els using FEM. Increasing binder layer thickness appeared

63 to offer slight improvement in rutting resistance. MEPDG analysis showed increases in top down cracking with binder layer thickness increases up to a certain thickness, after which additional thickness increase led to reduced cracking. Bejarano et al. (2007) used the Caltrans Highway Design Manual Chapter 600 method to design an f-sAPT section to validate Caltrans overlay strategies for the rehabilitation of cracked HMA. Proper compaction of the subgrade and aggregate base layers was primarily affected by the water content in these layers. Data indicated that low aggregate base moduli were obtained in locations where low subgrade moduli were observed. The HMA layer had a significant effect on the behavior of the aggregate base and subgrade, providing a confining pressure that increased the modulus of the aggregate base and also providing additional cover to the subgrade. The HMA modulus was significantly affected by temperature. The f-sAPT data showed that the overall pave- ment performance appeared to be significantly influenced by the behavior of the aggregate base, with sections tested dur- ing the dry months lasting longer in fatigue and surface rut- ting than the sections tested during the wet months. Jones et al. (2007a) evaluated the Caltrans overlay strate- gies for the rehabilitation of cracked HMA through f-sAPT to develop improved rehabilitation designs for reflective crack- ing. The underlying pavement incorporated a 3.5-in. (90-mm) DGAC surface on a 16-in. (410-mm) Class 2 aggregate base. All layers showed considerable variation in thicknesses. Rutting occurred primarily in the underlying DGAC and not in the overlay (rutting experiments), whereas it occurred in both HMA layers in the reflective cracking experiments. Little rutting occurred in the base, with no rutting in the sub- grade. Cracks were generally clearly visible in the underly- ing DGAC layer. Air-void contents of cores removed from the wheel paths in the reflective cracking sections after f-sAPT were lower compared with those from cores outside the wheel paths. Jones et al. (2007b, c) used the f-sAPT and laboratory data in CalME simulations and continuum damage mechan- ics analyses to develop two sets of M-E models for reflec- tive cracking. One set of models has been incorporated into CalME (based on layer elastic theory), while the second set of models (intended as a research tool) is based on FEM analysis and continuum damage mechanics. The second model provides insight into crack propagation damage. Both mod- els were calibrated and validated using the laboratory and f-sAPT data resulting in pavement prediction models in terms of calculated versus measured deflections, changes in stiffnesses, and ranking of reflective cracking performance. Bonding was found to be a significant variable in predicting actual performance of several HVS test sections. The models successfully predicted the performance of mixes with both modified and rubberized binders. Although the M-E rutting models predicted the overall rutting performance ranking of each section, it did not fully capture the distribution of rutting between the overlay and the underlying asphalt layers. The data indicated that the modified binder mixes have a greater risk of HMA rutting under slow, heavy loads and hot condi- tions compared with the full-thickness conventional DGAC overlay. Ullidtz et al. (2008a) used CalME and data from f-sAPT and RSST-CH to perform virtual analyses on the f-sAPT sec- tion as if the supporting conditions were similar to enable a more consistent comparison of the various HMA overlays. A comprehensive laboratory and f-sAPT experiment focuses on a comparison of gap-graded, terminal-blend- modified binder mixes with gap-graded Rubberized Asphalt Concrete (RAC-G) and conventional DGAC (Jones et al. 2008). Analysis indicated that gap-graded mixes with modi- fied binder, and a combination of modified binder and 15% recycled rubber, will provide superior performance in terms of reflection cracking compared with the same half thickness of RAC-G and full thickness of DGAC used in thin overlays on cracked HMA pavements. Conventional dense-graded HMA was superior to all other mixes in terms of rutting per- formance, followed by the RAC-G and then the modified binder mixes (Jones et al. 2008). Pérez et al. (2008) evaluated the maintenance and reinforce- ment of damaged pavements using HMA overlays through application of numerical analysis and f-sAPT. The reflective crack propagation was recorded and visualized on the f-sAPT section. The role of bonding interface properties in the f-sAPT was confirmed using FEM analysis with calculations matching particularly well to the measured strain data. Theoretical crack evolutions (affected by bonding interface conditions) were also confirmed in the comparison between f-sAPT data and strain field calculation data. F-sAPT evaluation of thin HMA overlays over cracked asphalt pavement (California) illustrated the benefits of using f-sAPT for assessing these structures (Jones and Harvey 2009). It was found that clear failure criteria (related to the pavement management objectives and typical field con- ditions under which the treatments will be used) must be selected for good models to be developed. For thin overlays, cracking and rutting criteria were used. Much of the rutting under thin overlays may be related to the overlay stiffness and shear resistance, although much of the rutting will occur in the underlying layers if the overlays are less than 2.4 in. (60 mm) thick. dynamic modulus The dynamic modulus test has been proposed as the SPT to verify the performance characteristics of Superpave mix- tures and as a potential field quality control/quality assurance parameter. Dynamic modulus (E*) is an input to MEPDG and supports the NCHRP Project 1-37A (Development of the

64 2002 Guide for the Design of New and Rehabilitated Pavement Structures) predictive performance models. The parameter is thus important, although the test is time-consuming. Reliable prediction of the compressive modulus of HMA mixes is essen- tial in material characterization and pavement performance prediction. Azari et al. (2007) compared the E* values of the ALF (FHWA sections) pavement mixtures with E* values predicted from the NCHRP 1-37A and the revised Witczak models and indicated that both models made reasonably good predictions of ALF moduli at high values. However, it overestimated the predictions for the lower modulus val- ues that correspond to low frequency and high temperatures. These are critical in rutting analysis of pavements, and fail- ing to capture these values could increase the risk of prema- ture rutting failure of HMA. The binder shear modulus (G*b) was found to be highly correlated with E* and justifies inclu- sion of G*b in the new predictive model. The phase angle of the binder was however found to be insensitive to the E* of the modified mixtures over the practical range of HMA mixture moduli. Based on a comprehensive laboratory evaluation of actual HMA materials obtained from MnROAD, FHWA-ALF, and WesTrack (all f-sAPT sites), Zhou and Scullion (2003) rec- ommended three SPTs for permanent deformation. These are the dynamic modulus term (E*/sind), flow time, and flow number. Data from prematurely rutted sections on US-281 in Texas (used to validate the recommendation) showed that dynamic modulus (E*), E*/sind, and flow number can effec- tively distinguish the good mixtures from the bad. Clyne et al. (2004) evaluated the use of the complex dynamic modulus as a design and performance parameter for four of the MnROAD test cells. Master curves were constructed for each mixture and compared with FWD field data. It was observed that the dynamic modulus decreases with an increase in test temperature for the same mixture (constant load frequency), while the dynamic modulus increases with the increase of test frequency (constant test temperature). The softest binder (PG 58-40) had the lowest dynamic modulus and the mixtures with stiffer binders had higher dynamic moduli. The modulus backcalculated from FWD data correlated well with laboratory dynamic modulus data. The accuracy and robustness of the Witczak empirical model for estimating the dynamic modulus inputs in the MEPDG methodology have been evaluated through a set of sensitivity and validation studies (Schwartz 2005). Vali- dation of the Witczak model against an independent set of laboratory dynamic modulus data for 26 mixtures (originat- ing from FHWA ALF, MnROAD, and WesTrack f-sAPT sections) found agreement between predicted and measured dynamic moduli nearly as good as for the calibration data set. Predicted rutting for the hypothetical pavement design using Level 1 versus Level 3 dynamic moduli inputs were in good agreement. The predicted rutting using the Level 3 inputs tended to be slightly lower than the corresponding pre- dictions using Level 1 inputs by an average of 12% for HMA rutting and 6% for total rut, consistent with the tendency of the predictive model to overestimate dynamic modulus rela- tive to laboratory measured values. NCHRP 9-19 (Superpave Support and Performance Mod- els Management) identified the confined dynamic modulus as one of three favorable indicators for evaluating the rutting potential of a mixture. Dynamic modulus testing at multiple confining pressures takes too long to be used routinely by state highway agencies. Experimental results presented by Lacroix et al. (2011) indicated that the linear visco-elastic properties of an HMA mixture are not affected by differ- ent confinements, and that all the confining stress effects manifest themselves in the elastic modulus at equilibrium. A method is proposed that uses a Prony series representation of the dynamic modulus curve and master curve shift fac- tors obtained from unconfined testing. This method employs the elastic modulus values predicted from a modified version of the universal material model to predict dynamic modulus values at different levels of confinement. The applicability of the method was verified for the AMPT using three dif- ferent mixtures that have been selected to highlight any dif- ferences in confined behavior resulting from various binder and aggregate structures. The mixture used in developing the procedure is the control mixture used at the FHWA ALF facility. This reduced testing protocol provides reasonable results, with most errors less than 20%. Hot mix asphalt modeling A 3-D dynamic FEM procedure incorporating rate-dependent visco-plastic models was developed to predict rutting of HMA pavements under ALF (Louisiana) loading (Huang et al. 2001). The extended Drucker–Prager elasto-plastic model was used to describe the aggregate base and subgrade. Although the tradi- tional 2-dimensional static FEM analysis was unable to simu- late the dynamic nature of the traffic load and the correspondent pavement responses, the results indicated that the 3-D model described reflected pavement responses well and predicted pavement rutting with reasonable accuracy. Seibi et al. (2001) evaluated HMA behavior under high rates of loading using uni-axial compression laboratory tests and f-sAPT (FHWA ALF). It was shown that HMA under high temperatures and high rates of loading exhibit an elastic- visco-plastic mechanical behavior reflecting the rate sensi- tivity of the material. Kim et al. (2006) applied the visco-elasto-plastic contin- uum damage model to uni-axial compression data (FHWA ALF control mixture). The model is found to significantly overpredict the strains in the monotonic tests and the loading portions of the repetitive creep and recovery tests. Investiga- tion suggests that the stiffening effect of aggregate interlock should be taken into account for a more accurate model of the

65 compression behavior of HMA. This was done successfully by using the ratio of the predicted strain rate to the measured strain rate. Kim et al. (2008) subsequently compared the strain responses of a 3-D visco-elastic model with f-sAPT data and showed that the predicted strains agreed well with the field measured strains. You and Buttlar (2006) developed a clustered distinct ele- ment modeling approach for modeling HMA microstructure in a research environment. It involves processing of high- resolution optical images to create a synthetic, reconstructed mechanical model that captures the complex morphology of the HMA. The model is applied to predict the dynamic modulus of specimens tested in cyclic, uni-axial compres- sion. A coarse HMA mixture from the FHWA f-sAPT site and a fine-grained HMA surfacing used on Illinois highways were modeled. Both the aggregate and mastic phases were modeled. The distinct element modeling approach provided a reasonable portrayal of the force chains developed in the aggregate skeleton and is shown to be a fundamental way of looking at the complex behavior of HMA under different loading and temperature conditions. Al-Qadi et al. (2008a, b) modeled f-sAPT responses of an HMA pavement to vehicular loading using a 3-D FEM model. The model incorporated measured 3-D tire–pavement con- tact stresses, a continuous moving load, and visco-elastic HMA properties. The tensile strain at the bottom of HMA decreased significantly as the HMA thickness increased, while the vertical shear strain variation was less significant. The longitudinal tensile strain at the bottom of HMA pro- vided the critical response for thin and medium-thickness HMA layers, although the critical response for a thick HMA layer is the vertical shear strain between 3 and 4 in. (76 and 100 mm) below the HMA surface. Underwood et al. (2010) proposed a more fundamentally appropriate method for modeling the performance of HMA based on coupling the response and material models, causing the effect of pavement degradation to be explicitly taken into account in the material response. The process requires sig- nificant computational time and effort. A numerical study of the behavior of HMA pavements subjected to f-sAPT wheel loads incorporating three pavement response tools [linear elastic analysis (LEA), linear visco-elastic analysis (LVEA), and FEM analysis] suggested that the use of the FEM for complete pavement response and performance prediction is still too computationally intensive for routine pavement design and analysis and that LEA or LVEA will be important to mechanistic pavement analysis for the foreseeable future. For pavement responses at the bottom of the HMA layers, LEA can be performed such that it yields reasonably accu- rate values as compared with LVEA. However, care must be taken in selecting the appropriate modulus value to use. NCHRP 9-19 Project, Superpave Support and Perfor- mance Models Management did not propose standard triaxial testing conditions for the magnitude of the deviator and con- fining stresses. Hajj et al. (2010) developed recommenda- tions for the selection of the deviator and confining stresses that best simulate the stress conditions encountered in the pavement under traffic loads through extensive mechanistic analyses (3D-Move model) of three different HMA struc- tures subjected to moving traffic loads at a range of speeds and braking and non-braking conditions. Prediction equa- tions for estimating the anticipated deviator and confining stresses have been developed and validated successfully on a HMA mixture from WesTrack as a function of pavement tem- perature, vehicle speed, and HMA stiffness. When braking conditions were incorporated into the analysis, an increase of 40% was observed for the deviator stress. Bonding and tack coats Inadequate HMA layer interface bonds typically lead to slip- page cracking. The interface condition significantly affects the distribution of stresses and strains in pavement structures. Romanoschi and Metcalf (2000) proposed a constitutive model for HMA interface characterization based on labora- tory-based shear-stress-displacement curves. The permanent shear displacement increases linearly with the increasing number of load repetitions under fatigue testing. The shear stress and displacement are proportional until the shear stress equals the shear strength and the interface fails under direct shear testing with normal load. Interface bonding between HMA overlays and PCC pave- ments affects the service significantly. Leng et al. (2008) developed a direct shear test device to investigate the char- acteristics of this interface and determine the interface shear strength, including parameters such as HMA type, tack coat type, tack coat application rate, PCC surface texture, and tem- perature. Finely graded HMA provided better interface shear strength than coarsely graded HMA, whereas asphalt emul- sion provided better interface shear strength than cutback bitumen. An optimum tack coat application rate of 0.04 gal/ yd2 (0.18 l/m2) was found independent of tested parameters. Tining direction in the PCC did not affect the interface shear strength at 68°F (20°C). Leng et al. (2009) added f-sAPT evaluation of HMA over PCC overlays to evaluate the data obtained from laboratory tests on tack coat performance. The measured strain response data validated the laboratory findings, with the optimum tack coat application rate sec- tion providing the smallest strain responses compared with sections with higher and lower tack coat application rates. The f-sAPT section without tack coat had the highest rutting compared with other sections. Milled PCC surfaces provided lower rutting compared with a smooth and transverse tined surface, while a deep cleaned surface showed lower rutting depth than a roughly cleaned surface. Mohammad et al. (2010) quantified the effects of tack coat type, tack coat application rate, surface type and condition

66 (milled versus un-milled) on the interface shear strength using f-sAPT, and an experimental matrix of five tack coat materials applied at three application rates on four surface types in Loui- siana. The highest shear strength was obtained at an application rate of 0.155 gal/yd2 (0.7 l/m2) for all tack coat types. Although higher application rates may increase interface shear strength, excessive tack coat may migrate into the HMA mat during compaction and service, causing a decrease in the air-void con- tent of the mix. A direct relationship was observed between the existing surface roughness and the interface shear strength. The milled HMA surface thus provided the greatest interface shear strength followed by the PCC surface, existing HMA, and new HMA surface (smooth and unweathered). High modulus asphalt High modulus asphalt (HiMA) allows rutting reduction in base courses as very stiff asphalt is used in the mixture [based on the French standards for Enrobe’ ‘a Module E’leve’ (EME) (or HiMA) regarding rutting resistance] (Perret et al. 2004). In preparation for the inclusion of HiMA into Swiss standards, three test sections (two HiMA and one reference HMA for base layer), were evaluated at Laboratoire Central des Ponts et Chaussées (LCPC) (France) using f-sAPT at 122°F (50°C). The behavior of the test sections corresponded to LCPC labo- ratory rutting results and the section designed to have a strong rutting resistance had no rutting in the base layer, confirming that high mastic content can strongly reduce the resistance to rutting. HiMA is a well-known rutting-resistant solution for high- volume roads in France. Rohde et al. (2008) evaluated HiMA using f-sAPT and verified that the performance of the HiMA was superior to that of a reference HMA mixture produced with conventional binder (pen grade 50/70). The binder type was shown to be vital in the improvement of rutting resistance. Wojciech et al. (2010) compared conventional HMA base performance with HiMA base performance, focusing on rut- ting development and using f-sAPT, field tests (FWD and GPR), and laboratory tests (binder content, gradation, air voids, resistance to rutting, stiffness, and fatigue). No fatigue cracking was visible and only slight rutting occurred as a result of subbase compaction. aging Field cores were extracted from eight HMA pavements (two with polymer-modified and six with unmodified binders) at the FHWA’s pavement testing facility (from a 1993 FHWA ALF experiment) to evaluate in situ pavement aging in the laboratory (Al-Khateeb et al. 2006). The polymer-modified HMA pavements show comparatively lower aging than the unmodified HMA pavements. For unmodified binders, the pavements with stiffer binder tended to have higher aging indices than those with lower stiffness binders. Pavements with larger nominal maximum aggregate size also experi- enced higher aging than those with smaller nominal aggre- gate size. The aging index near the surface [upper 0.3 in. (7 mm)] was four to five times higher than the aging index at a depth of 5.5 in. (140 mm). Qi et al. (2010) evaluated the fatigue performance of HMA pavements with unmodified and modified binders using f-sAPT in early-aged and accelerated-aged conditions [origi- nating from the TPF-5(019) (Full-Scale Accelerated Per- formance Testing for Superpave and Structural Validation) experiment]. The aging process decreased the fatigue perfor- mance significantly, and the overall ranking of the fatigue performance was mostly the same before and after aging. Where bottom-up fatigue cracking occurred on early-aged HMA, top-down cracking was mostly evident on the aged sections. Dynamic modulusE*data of the in situ HMA before and after f-sAPT indicated that the largest difference in modulus was seen between the upper layers and the bot- tom layers with the stiffness of the upper layers increasing. The largest increase in mixture stiffness was after more than 5 years of natural exposure, while the accelerated-aging process did not appear to cause a measurable increase in stiffness. MnROAD (Marasteanu et al. 2008) evaluated the opti- mum time for applying surface treatments to existing HMA pavement structures. A detailed study on the MnROAD test track pavements included both standard and nontraditional tests on the existing HMA to determine specifically the surface condition of the pavements at different ages. The nontraditional tests included X-Ray Photoelectron Spectroscopy, Scanning Electron Microscopy, Fourier Transform Infrared Spectros- copy, and spectral analysis of the asphalt pavements. Stan- dard Bending Beam Rheometer, Direct Tension Test, DSR, and Semi-Circular Bend Testing were also conducted on the materials. No clear single test was identified in the analysis as suitable for all purposes. The Fourier Transform Infra- red Spectroscopy appeared to be the most sensitive to the binder material age with a significant positive correlation with pavement age. The DSR, Bending Beam Rheometer, and Direct Tension Test data analyses were less significant and many times led to contradictory results that could have been affected by the HMA emulsion application rates that were adjusted with the age of the pavement. recycled asphalt Pavement Hachiya et al. (2006) examined the effect of rejuvenating agents on recycled asphalt pavement (RAP) performance on airport pavement surface courses (Tokyo International Air- port). Similar properties were obtained for the RAP made with different rejuvenating agents, although the recycled binder properties varied. The performance of RAP at a 70% recycling rate satisfied the specifications for use as a surface course in airport pavements.

67 Mohammad et al. (2006) evaluated the effectiveness of using untreated RAP as a base material in place of crushed stone in a soil cement asphalt pavement structure using f-sAPT (Louisiana ALF), FWD, and laboratory character- ization. The laboratory tests results indicated similar resil- ient moduli for RAP and crushed stone, although there was some temperature susceptibility in the RAP. Both the field and laboratory evaluations confirmed that untreated RAP can successfully be used as base course in lieu of the conven- tional crushed stone base. The performance of HMA mixes with different percent- ages of RAP (0, 25, and 40%) was determined using f-sAPT (LAVOC, France) and laboratory testing in relation to stiff- ness, fatigue cracking, water sensitivity, and permanent deformation (Bueche et al. 2008). No failure was observed in these experiments and the performance for these specific mixes, with and without RAP, was equivalent. West et al. (2009) evaluated the performance of f-sAPT sections (NCAT) with 20% and 45% RAP, and a control sec- tion. All mixes contained the same component aggregates and RAP, while one 20% RAP mixture contained PG 67-22 binder and the other PG 76-22 binder. The 45% RAP mixes contained PG 52-28, PG 67-22, PG 76-22, and PG 76-22, plus 1.5% Sasobit. Overall binder stiffness in the RAP mixes had an impact on field compactability, showing that the 20% RAP mixes compacted easier than the 45% RAP mixes. The RAP test sections performed well in the f-sAPT despite air- voids contents that were not optimal. The data indicate that there does not appear to be a strong case to support the use of softer grade virgin binders for high RAP mixes. The effectiveness of the use of foam asphalt-stabilized RAP from Full-Depth Reclamation (FAS-FDR) as base material for flexible pavements was evaluated through f-sAPT on four pavements with a conventional granular base and three thicknesses of FAS-FDR (Romanoschi et al. 2004, 2010). The FAS-FDR base performed well under moder- ate moisture conditions and supports the efficient use of the RAP material contaminated with soil and aggregates. The f-sAPT indicated that 25 mm of FAS-FDR is equivalent to 1 in. (25 mm) of conventional Kansas AB-3 granular base. Four test sections that included both WMA and high RAP contents (50% RAP) were constructed with a control sec- tion at NCAT for evaluating the full-scale structural and fatigue characteristics in the context of M-E analysis (Timm et al. 2011c). Similar strain versus temperature behavior was observed for all sections. None of the experimental sections were statistically different from the control with respect to tensile strain at the lowest reference temperature [50°F (10°C)]. At the highest reference temperature of 109°F (43°C), three distinct groups were formed based on the aver- age strain levels with the control section at the highest strain and the high RAP sections at the lowest strain. The WMA sections were in between these extremes. No adverse effects were visible through utilizing high RAP contents and WMA with respect to strain response relative to the control. These high RAP and WMA sections appeared to carry loads more efficiently and reduce the strain levels at higher temperatures in comparison with the control section. Combination of the temperature-corrected strain responses and laboratory-derived beam fatigue transfer functions indicated that the RAP-WMA section perform best at 68°F (20°C), with approximately 3.6 times more cycles to failure than the control. The fatigue performance estimates were made using laboratory-determined transfer functions combined with strains corrected to the cor- responding laboratory temperature. Warm mix asphalt WMA test sections were constructed at the NCAT Test Track to assess the rutting performance using a chemical emulsion (Prowell et al. 2007). The sections were compacted after being stored in a silo for 17 hours, and the in-place densities were equal to or better than the control HMA surface layers, even when compaction temperatures were reduced by 46°F to 108°F (8°C to 42°C). The two WMA sections and HMA section showed excellent field rutting performance, whereas laboratory rutting susceptibility tests indicated similar per- formance for the WMA and HMA mixes. Increased moisture damage potential was indicated for the WMA mixes through laboratory tests. Three WMA and one control HMA section were evalu- ated on the Ohio f-sAPT facility (Sargand et al. 2008). Early rutting of the WMA section was higher than the control mix, after which the rutting rate was slightly less for the WMA than for the control HMA. The performance of an HMA control mixture was compared with three WMA mixes in California (chemical foamant, chem- ical emulsion, and organic wax). The WMA was produced and compacted at approximately 95°F (35°C) lower than the con- trol (Jones et al. 2010). Acceptable compaction was achieved on all WMA mixes. The outcome of f-sAPT indicated that the use of any of the three WMA technologies will not significantly influence rutting performance. Laboratory moisture sensitivity testing indicated that all the WMA mixes were potentially sus- ceptible to moisture damage, although no significant difference in the level of moisture sensitivity between the control mixture and mixtures with additives could be observed. Laboratory fatigue testing indicated that the WMA technologies should not influence fatigue performance. MnDOT initiated research into WMA to specifically eval- uate the performance of WMA in cold climates (MnROAD WMA 2011). As thermal cracking is the predominant dis- tress mode of HMA pavements in Minnesota, studies at the MnROAD facility explore WMA’s potential for bet- ter low-temperature cracking performance. It is anticipated that reduced oxidation levels at the mix plant should lead

68 to enhanced long-term pavement performance. Six cells were paved with WMA on the MnROAD Mainline, car- rying just under one million ESALs per year. The mix is a level 4 Superpave with PG 58-34 binder and 20% RAP from MnROAD millings. A chemical-based additive was used in the WMA production. The HMA mix was produced approxi- mately 50°F (10°C) cooler than normal HMA production. Equal compaction was achieved than on HMA with less effort. The WMA mixes produced a good Tensile Strength Ratio, indicating this mixture is not prone to moisture dam- age. It is anticipated that the WMA test data will be dissemi- nated to city and county engineers, consultants, contractors, and researchers. Foam Bitumen Long et al. (2004) reported on laboratory testing in conjunc- tion with f-sAPT (South Africa) on pavements recycled with foamed bitumen-treated materials, demonstrating that f-sAPT can be enhanced by conducting a comprehensive lab- oratory testing project in conjunction with the f-sAPT. The testing indicated that the cement provides more strength and shear resistance than the foamed bitumen, while the foamed bitumen provides more flexibility to the mix, requiring care- ful selection of the foam bitumen and cement contents during design to enhance the strength and shear resistance as well as the flexibility. Theyse and Long (2004) used the South African HVS to evaluate the performance and identify distress mechanisms of pavements that were recycled in-place and treated with foamed bitumen and emulsified bitumen. Bitumen stabiliza- tion with active filler initially increases the resilient modu- lus of the treated material; however, this increased resilient modulus is reduced through repeated loading to the point where the resilient modulus stabilizes at values associated with unbound materials (referred to as the effective fatigue life). The reduction rate is mainly determined by the traffick- ing load. An f-sAPT test (California) on a road rehabilitated using a Full-Depth Reclamation (FDR) process using foamed asphalt (2.5%) and portland cement (1.0%) was used to develop improved mixture and structural design and construction guidelines for FDR of cracked HMA (Theyse et al. 2006). The observed mode of distress differed between favorable conditions in summer and fall and unfavorable conditions in winter and spring, with the distress mode before the onset of winter consisting of gradual rutting of the pavement resulting in a terminal surface rutting with limited fatigue cracking. After the winter, the mode changed to a more rapid rutting rate and base shear failure in certain locations. Extensive fatigue cracking was also evident, but this was probably caused by a localized weak soft base layer. The pavement structure of the HVS test track showed sensitivity to high moisture contents in terms of elastic and plastic response. An f-sAPT experiment on foam bitumen pavements was conducted in the Canterbury Accelerated Pavement Testing Indoor Facility (CAPTIF) to study the effects of foam bitu- men on unbound granular materials. Three sections were constructed using foam bitumen contents of 1.2%, 1.4%, and 2.8%, plus a common active filler content of 1.0% cement. Two more pavements were constructed with only cement (1.0%) and only foam bitumen (2.2%) as well as a control section with untreated unbound material. Rutting develop- ment under f-sAPT showed that the addition of foam bitu- men significantly improved the performance of the materi- als with 1% cement that were studied in this research, while little difference was observed within the sections stabilized with only foam bitumen and only cement. The existing foam bitumen design methods are overconservative. Vertical com- pressive strains measured at the top of the subgrade indicated that surface deflection was controlled by the subgrade elastic response (Gonzalez et al. 2009). MnROAD is developing full depth reclamation (FDR) and cold inplace recycle processes using foamed asphalt with the intention of developing specifications for successful imple- mentation of foamed asphalt recycling techniques in Minnesota (Eller and Olson 2009). Several foamed asphalt cold inplace recycle projects are performing well in Minnesota. MnDOT data showed that the recycled pavement layer develops a relatively uniform strength despite the high variability inherent in most low-volume roads, and that the foamed asphalt forms a cohesive matrix when mixed with the fines from the reclaimed material. It is expected that further study of these projects will add to the knowledge base regarding constructability and performance. crumb rubber Caltrans uses overlays with DGAC and overlays with ARHM- GG (Harvey et al. 2001; Harvey and Bejarano 2001). The ARHM-GG overlays are typically designed to be half the thickness of DGAC overlays. F-sAPT data indicated that the use of half thickness ARHM-GG overlays is reasonable for reflection cracking. The ARHM-GG overlays had less rutting than the DGAC overlays. The performance of HMA with crumb rubber-modified binders (AR-HMA) as compared with similar mixes with conventional binder was evaluated under f-sAPT (Louisiana ALF) (Mohammad et al. 2000, 2004). The AR-HMA wear- ing course mixture showed similar rutting resistance to the conventional wearing course mixture, whereas the AR-HMA base course mixture had better rutting resistance than the conventional base mixture. No visible fatigue cracks were present during the f-sAPT. The pavement layers constructed with crumb rubber-modified binder in the base course could last almost twice as long as the conventional one. The constructability, laboratory, and field performance of different rubber-modified HMA mixes were compared

69 with a DGAC control mix (Cook et al. 2006). Construction using all three processes (wet, dry, and terminal blend) was successful, and f-sAPT testing (California) indicated that all asphalt-rubber-modified mixes performed at least as well as the conventional DGAC mixture. Since the 1990s, asphalt rubber has been used in Brazil to delay reflection cracking in overlays (Núñez et al. 2006). An f-sAPT study was carried out by the Federal University of Rio Grande do Sul to quantify reflection cracking evolution in asphalt rubber and HMA overlays. Data suggested that the asphalt rubber overlay was five times more efficient than the HMA overlay in delaying reflection cracking. Hot mix asphalt aggregate Gradation In a comparison of rutting and fatigue performance results from WesTrack with the SuperPave volumetric design method (Wen et al. 2003), it was shown that the fine-graded WesTrack HMA pavements performed better than coarse-graded HMA. The fracture energy of work potential theory was found to cor- relate well with the performance from WesTrack. Particle size distribution greatly influences the rutting resistance, workability, permeability, and durability of HMA mixes. FDOT (Greene et al. 2011) developed a theoretical approach for evaluation and specification of aggregate gra- dations with the intent to ensure that the resulting mixes will have sufficient aggregate interlock to resist rutting. A mixture’s Dominant Aggregate Size Range (DASR) and the DASR porosity is used to determine the interactive range of particle sizes and whether good contact exists between those particles, with a porosity of greater than 50% differentiating between good and poor performing gradations. The method was validated using f-sAPT. Porous Hot mix asphalt Open-graded porous asphalt (OGPA) was evaluated at CAP- TIF and the outcome compared with a field trial and laboratory evaluation (Alabaster et al. 2008). The trial demonstrated that full-scale manufacture and surfacing construction with epoxy OGPA could be undertaken without significant modification to the plant or operating procedures. Traffick- ing resulted in early signs of surface abrasion in the con- trol section but not in the epoxy OGPA section. Early life rutting of the epoxy OGPA is not likely to be greater than that of equivalent standard materials. The field trial results confirmed the findings of the f-sAPT test. Research on the MnROAD test cells evaluating the cor- relation between tire-pavement-interaction-noise (TPIN) measured with the on-board sound intensity protocol, ESAL, Surface Rating, and age of pavement showed an increase in on-board sound intensity in bituminous segments over time. Individual models were developed for different surface types with no tenable relationship between friction number and TPIN (Khazanovich and Izevbekhai 2008). Pavement texture is an important parameter in TPIN. As pavements carry traffic load over the years measurable deg- radation occurs in texture. As the pavement is exposed to environmental and traffic elements, changes occur in ride quality measured by the International Roughness Index (IRI) as well as the Surface Rating. thermal cracking Al-Qadi et al. (2005) evaluated the measured strain magnitude associated with thermal fatigue through field measurements (Virginia Smart Road) and 3-D FEM that simulates thermal fatigue in flexible pavement. The model simulates a typical interstate pavement design [less than 8 in. (200 mm) HMA]. It showed (as confirmed by the field measurements) that the pavement’s response to thermal loading was associated with a very high strain range (maximum value of 350 µm/m). This confirms the hypothesis that the criticality of thermal fatigue arises from the high stress/strain level exhibited in each cycle rather than its frequency. Hot mix asphalt Pavement maintenance Zerfas and Mulvaney (2004) evaluated the performance of pre- ventative maintenance conducted on the MnROAD sections three months after application of micro-surfacing. The micro- surfacing did not fill the ruts as well as expected, probably because a metal strike off bar was not used during the scratch coat of the micro-surfacing, in combination with only a single pass of thin maintenance surfacing used on some sections, even though a two pass application was warranted. The primary goal of restoring the ride quality for the traveling public appeared to be met through application of the thin maintenance surfacing, with a 48% improvement in riding quality attained. Huurman et al. (2008) initiated a project aimed at devel- opment of a Lifetime Optimization Tool (LOT) that allows the initiation and progression of HMA raveling to be pre- dicted and the effects of mixture modifications and increas- ing traffic loads to be analyzed. F-sAPT (the Netherlands) was used to validate the predictions made with the FEM- based meso-mechanics model in the LOT. Data showed that the LOT model predictions were in good agreement with the observed performance and it was possible through applica- tion of the LOT to rate the f-sAPT sections’ raveling perfor- mance. Analysis showed that the aggregate type appeared to be the most important factor controlling raveling. Fiber reinforcement of Hot mix asphalt F-sAPT for performance evaluation and validation of HMA mixes (in Romania) stabilized with various fibers is described by Vlad and Andrei (2004). The rutting performance of the

70 HMA stabilized with various types of fibers is significantly improved in comparison with that of the control mixture with- out fibers. Rutting development is still affected by temperature. concrete and related materials General Rao and Roesler (2004b) evaluated the applicability of Min- er’s Law for evaluating several concrete fatigue transfer func- tions using f-sAPT data (California HVS). Responses result- ing from trafficking were modeled with FEM over a range of temperature differences. The results indicated that although it is possible to predict the location of the crack, Miner’s approach cannot be used to predict the timing or number of load repetitions corresponding to slab cracking. Strauss et al. (2005) used a continuously reinforced con- crete pavement (CRCP) inlay on a steep uphill section of an interstate highway that experienced considerable rutting damage owing to slow-moving heavy vehicles. The HMA was milled and replaced with 7 in. (180 mm) of CRCP inlay in the climbing lane. Although the test section had narrowly spaced transverse cracking prior to testing, no punch-outs could be created using the South African HVS. However, in areas where concrete was substandard, significant punch- outs occurred. Introduction of water into the supporting lay- ers caused loss of bond between the slab and the subbase and a resulting void. Structural failure of the CRCP followed an S-curve, with a high incidence of failure occurring initially (mainly as a result of construction issues) followed by a sta- ble period before the rate of failure increased again. Load transfer efficiency (LTE) is commonly considered as a measure of joint quality and it plays an important role in rigid pavement evaluation and design. Wadkar et al. (2010, 2011) used data generated from the f-sAPT at the FAA National Airport Pavement Test Facility (NAPTF) to evalu- ate stress-based LTE [LTE(S)] and deflection-based LTE [LTE(d)] of rigid pavement airfield joints. LTE(S) values from static loading were 45% lower than moving loads. LTE(d) was found to be similar when measured under a single-wheel or 4-wheel gear configuration and was lower for stabilized bases compared with joints over a nonstabilized base. Field measured LTE(S) under real moving aircraft gear loading was signifi- cantly higher than the 25% value assumed in current thickness design. The f-sAPT analysis provided evidence to the theory that stiffer bases may lead to lower LTEs. LTE(d) appears not be an indicator of joint quality and structural integrity of the pavement. Yeh et al. (2011) examined how matched and mismatched joints affect the performance of unbounded concrete overlays under f-sAPT at the FAA NAPTF. The gauges in slabs with matched joints yielded the smallest average daily response. The peak response magnitude was inadequate to explain the timing of crack initiation, and examination of other strain response characteristics (i.e., duration and recovery) may assist in explaining the damage rate. Upon formation of the first crack, the effective modulus drops significantly. No evi- dence was found to state whether the slabs with matched or mismatched joint will have higher initial moduli, consistent with the behavior of new untrafficked doweled joints. The total joint stiffnesses at mismatched joints were higher than those at matched joints for most cases. This supports the hypothe- sis that mismatching joints take advantage of the underlying support and thus improve LTE. The benefit of mismatching joints is not as significant when the underlying slab is in good condition. MnDOT evaluated the use of pervious concrete on porous bases for minimizing storm water intrusion from developments through evaluation on parking lots and through f-sAPT by con- structing cells with porous concrete on the low-volume loop of the MnROAD facility (Izevbekhai 2008). Although the under- standing of the performance of pervious concrete in northern climates is still rudimentary, the intention is to, through ade- quately evaluating pervious concrete in this climate, provide long-term performance monitoring through monitoring the changes in porosity and infiltration over time under standard measurable traffic loads, environmental effects, and deicing operation. ultra-thin and thin Whitetopping Rajan et al. (2001) identified through f-sAPT (Indiana DOT) that the joint spacing of 47 in. (1.2 m) was sufficient for good performance. Measured longitudinal strains indicated that as the wheel approached a given point in the pavement, the top and bottom of the overlay experienced a tensile and compres- sive strain respectively that quickly reversed when the wheel was exactly over the point. Wu et al. (2002) evaluated ultra-thin and thin whitetopping (UTW) repair and rehabilitation techniques using the FHWA ALF. As some UTW panels exhibited corner cracking dis- tress, panel removal and replacement was selected to repair some of the distressed panels. The f-sAPT performance indi- cated that this panel removal and replacement is an effective repair method for UTW. Galal et al. (2004a, b) developed models for the mecha- nistic design of UTW of an HMA and a PCC pavement based on f-sAPT (Indiana DOT). Initial results indicated that the equivalent thickness approach may provide a good prediction for the pavement stresses and strains. Linear-elastic analysis of the pavement stresses and strains suggested that the bond between the UTW and HMA is not uniform. Data from three sequentially placed surface strain gauges confirmed the wave effect in front of the tire and the strain reversal that exists was shown to increase with the number of load applications. The application of UTW over composite pavements does not have an adequate mechanism to account for analysis

71 of the in situ composite pavement (Newbolds et al. 2005) applied f-sAPT (Indiana DOT) to determine that the equiva- lent thickness approach to model the existing composite pavement produces reasonable results for this analysis, indi- cating good agreement between the calculated and measured strains. De Larrard et al. (2005) showed through f-sAPT (France) that a thin high-performance concrete layer constructible with conventional tools and that the cracking behavior under accelerated loading is acceptable and that no damage was caused to the layer through traffic. Research on the applicability of UTW using f-sAPT (Changsha, China) focused on the dynamic and static strains to evaluate the characteristics of different pavement struc- tures (Cui et al. 2007). Data showed that the tensile strains are small at both the surface and the bottom of the panel when good bonding exists between the UTW and the under- lying HMA layer. 3-D FEM analysis of the sections con- firmed the location of maximum tensile stress as the location where cracking initiated in the pavement. F-sAPT at Kansas State University (Romanoschi et al. 2008) focused on determining the response and the failure mode of four pavement structures: two thin concrete overlay pavements [4- and 6-in. (100- and 150-mm)-thick on top of a 5-in. (125-mm)-thick PCC pavement (PCCP)] and two thin whitetopping pavements [4- and 6-in. (100- and 150-mm)- thick PCC overlays constructed on top of 5 in. (125 mm) HMA]. Three of the four pavements performed well during f-sAPT. The 4 in. (100 mm) TWT pavement developed a crack much earlier than the other three sections. No signifi- cant joint faulting was recorded. FEM analysis conducted with all materials modeled as linear-elastic and a perfect bond between the overlays and the supporting layers, showed that the magnitude and shape of computed and measured strains matched well before any f-sAPT loading was applied. All sections experienced loss of support under the joints during trafficking, causing an increase in the maximum longitudinal strains at mid-span, and suggesting that loss of bonding may be attributed to environmental factors. The length of the loss of support can be backcalculated from the strains computed with the FEM. The structural properties of pavement sections have a significant influence on the strain distribution and maximum strains for UTW constructed over composite pavements (Newbolds and Olek 2009). F-sAPT at the Indiana DOT indi- cated that the critical HMA strain location is at the top of the HMA layer. Measured strains in the bonded lanes well matched the theoretical values calculated using linear-elastic analysis. Snyder (2009) evaluated the performance of UTW [3 to 6 in. (75 to 150 mm)] under years of heavy traffic at MnROAD. The observed performance was partly the result of the pres- ence of good support by (and bond with) the underlying HMA layer. Joint layout should avoid placement of longitudinal joints in or near the wheel paths. The incidence of reflective cracking of well-bonded concrete placed over thermal cracks in HMA pavements appears to depend on the relative stiff- nesses of the concrete and HMA layers. Reflection cracking is less likely when the HMA thickness is less than double the concrete thickness. Seasonal resilient modulus variations strongly affected load-related strains in the UTW. Reflective cracks generally occurred more quickly in the driving lane than in the passing lane. Observed distresses in the UTW confirmed that bonding of whitetopping to the underlying asphalt is essential for good long-term performance. Vandenbossche and Barman (2010) observed in f-sAPT (MnROAD) that reflection cracking of UTW is a function of both temperature- and load-related stress, with reflection cracks developing earlier in the driving lane than the pass- ing lane. Increasing the panel size to move the wheel path away from the longitudinal joint had the same effectiveness in decreasing reflection cracking as increasing the thickness of the overlay by 1 in. (25 mm). It was also verified that the occurrence of reflection cracking is a function of the stiffness of the concrete relative to that of the HMA layer with data indicating that reflection cracking will develop in bonded UTW if the relative stiffness of the layers falls below 1. ultra-thin continuously reinforced concrete Pavements Ultra-thin CRCP (UTCRCP) [also referred to as ultra-thin heavy reinforced high performance concrete (UTHRHPC)] has been used successfully in Europe as a rehabilitation mea- sure on steel bridge decks and reported on at the 5th Interna- tional CROW [Centrum voor Regelgeving en Onderzoek in de Grond-, Water- en Wegenbouw en de Verkeerstechniek (Center for Regulatory and Research in Soil, Water and the Road and Traffic Engineering)] workshop in Istanbul (2004). The concept has been explored further in South Africa for use as a road strengthening measure by constructing experi- mental sections of 2 in. (50 mm) UTCRCP directly on top of both natural gravel and cement-treated natural materials. Kannemeyer et al. (2008) described f-sAPT to test the suitability of a UTCRCP overlay on a flexible pavement near the end of its life. Test sections were constructed using a different combination of support stiffnesses, construction joints in the UTCRCP layer and environment, by adding water. FEM was applied to evaluate the observed behavior numerically. Although high tensile stresses were indicated by the FEM at the bottom of the UTCRCP layer, cracking at this location of high tensile stress was not as prominent as circular surface cracks and thus tensile stress at the sur- face was used to simulate observed distress. The relative position of the longitudinal steel reinforcement was found not to be important, although placing it closer to the top of the UTCRCP layer reduces compressive stresses, thereby

72 reducing the risk of spalling and the access of surface water. Debonding that occurred between the UTCRCP layer and the support resulting from water entering the pavement led to an increased stress in the UTCRCP layer with the resulting void being detrimental to the performance of the UTCRCP. CSIRO, University of Pretoria, Gautrans, and the Cement and Concrete Institute in South Africa developed a low-cost version of the UTCRCP technology to be used in upgrading former township roads. It consists of placing a 2 in. (50 mm) layer of concrete reinforced with a 0.2 in. (5.6 mm) diameter steel mesh [grid size 8 in. (200 mm)] on top of the exist- ing road surfacing using labor-intensive methods. F-sAPT on several of these sections showed the technology to be adequate for low-volume traffic applications (e.g., residen- tial streets) with bus traffic. Test sections showed no dete- rioration after up to 700,000 equivalent 18 kip (80 kN) axles (E80s) of APT traffic, although some curling and warping was evident during the tests (Du Plessis et al. 2009a, b). Steyn et al. (2011) demonstrated the use of a time-series analysis technique for evaluation of the influence of tempera- ture and climate on the LTE across cracks in UTCRCP under f-sAPT. Analysis indicated that the loaded and unloaded deflections across a crack were better pavement performance predictors than LTE for the UTCRCP. It also indicated that temperature difference over the pavement influences the per- formance of UTCRCP. Precast concrete Precast concrete slabs offer an important advantage as a pavement rehabilitation option because they can be opened to traffic immediately upon installation, making them attrac- tive for use on heavily traveled highways and airfields where work windows for repairs or reconstruction are very short. Tayabji et al. (2008) synthesized the results of f-sAPT evaluation of the durability of precast concrete pavements conducted in California, the Netherlands, and France. Analy- sis indicated that all the precast concrete slabs systems showed potential to adequately resist high traffic volumes. Kohler et al. (2007, 2008, 2009) described an f-sAPT pro- gram in California during which the performance of a number of precast concrete slabs were evaluated. Slabs were instru- mented with displacement sensors and with thermocouples and deformations caused by temperature changes measured in the ungrouted and then in the grouted condition before any load was applied to the pavement. Slab corner curl was reduced from a range of ± 0.06 in. (1.5 mm) to ± 0.02 in. (0.5 mm) by grouting the joints and the voids under the slab. LTE increased from 5% to 40% (ungrouted transverse joints) to close to 100% after grouting. F-sAPT indicated that the precast slabs can be safely opened to traffic in the ungrouted condition, so that panels can be installed in consecutive nights rather than completing the entire installation at one time. The expected life of the slab pavements was estimated to be between 140 and 240 million ESALs, equivalent to between 25 and 37 years of service. The failure mechanism was similar to that observed in cast-in-place jointed concrete pavements. Data indicate that doweled jointed plain concrete pavement and precast concrete pavement system pavement of similar thickness may be expected to have similar perfor- mance and failure mechanisms. Nishizawa et al. (2007) evaluated long-term performance of a precast high strength fiber-reinforced concrete panel under f-sAPT (Japan) and compared the data with dynamic 3-D FEM data. Prefabricated panels 39 in. (1.0 m) by 67 in. (1.7 m) by 1.2 in. (30 mm) thick were placed over an exist- ing HMA pavement. It was observed that an unbonded area developed at joints between panels causing weak panel sup- port. Texturing on the bottom face of the panel effectively enhanced the bonding. No serious problems were identified under traffic loads, as long as a sound bonding was provided at the interface between the panels and HMA layer. Ashtiani et al. (2011) investigated the feasibility and effi- ciency of using precast concrete panel installation techniques using high density polymer (HDP) foam and flowable fill as leveling materials for runways (Air Force Research Labora- tory, Florida). The installation system impact on the perfor- mance of the repaired sections was characterized by LTE, joint stiffness, and deformation energy dissipated through the pavement foundation. Three precast concrete repair panels (conventional HDP, deep HDP injection, and flowable fill) were subjected to load applications using an F-15 loadcart. The results indicated significant increases in the deformation energy and considerable loss of joint stiffness with increas- ing number of load applications when flowable fill was used as leveling material, while precast panels installed with HDP foam performed better. The best performance was shown by the deep injection method. cracking of concrete slabs JPCP test sections (California) constructed using fast setting hydraulic cement concrete developed top-down transverse cracks under environmental influences before any f-sAPT traffic was applied to them (Heath and Roesler 2000; Heath et al. 2001; Du Plessis et al. 2002, 2005; Kohler and Roesler 2004). Initial strains were most likely derived from thermal contraction after construction. Strains at the top of the slabs increased significantly after 2 months, whereas the strains in the bottom of the slab remained constant. This increase in the top of the slab was a result of drying shrinkage. The differ- ential strains between the top and bottom of the slab resulted in slab bending stresses that exceeded the concrete strength and caused transverse cracking. The measured and FEM data showed significant curling of the slab corners owing to dif- ferential drying shrinkage. Load plus environmental stress

73 analysis showed excessive differential drying shrinkage result- ing in the critical failure location being at the corner of the slab, indicating either corner or longitudinal cracking as fail- ure modes. Evangelista and Roesler (2009) evaluated critical gear posi- tions that produce maximum tensile stresses on the surface of rigid airfield pavements given no initial curling using four individual aircraft gear geometries (dual, dual tandems, and two types of triple dual tandems) and four aircraft types (B-777, A-380, MD-11, and B-747). F-sAPT on rigid pave- ments at the FAA NAPTF and Airbus PEP showed top-down cracking occurring under certain combined loading and pave- ment geometry configurations. A 2-dimensional finite-element analysis has shown consideration of the full aircraft gear is necessary if the position and magnitude of top tensile stresses need to be predicted. For the no joint load transfer case the full gears of the A-380 and MD-11 had top to bottom tensile stress ratios of 1.0 and 0.86, respectively. For the 85% LTE cases, the common critical aircraft position was when one gear straddles the adjacent slab at the corner and the other gear is located on the main slab producing the critical top tensile stress. Critical top tensile stresses occurred mostly at the transverse joint for almost all aircraft analyzed, promoting propagation of longitu- dinal or corner cracks. It was observed that the joint LTE and multiple gear load configuration relative to the slab geometry were the most significant factors affecting the critical top tensile stresses and top and bottom tensile stress ratios. dowel Bars California HVS testing indicated that dowel bar retrofit sig- nificantly improved LTE and reduced deflections from loads and environment compared with un-doweled pavements (Harvey et al. 2003; Bian and Harvey 2006; Bian et al. 2006). Primary performance criteria for dowel joints are LTE and vertical deflection of the joints. Larger vertical joint deflec- tions and lower LTE are strongly correlated with increased rate of faulting and roughness development. Three dowels per wheel path were found to have significantly lower LTE than four dowels per wheel path, whereas fiber reinforced polymer (FRP) dowels and grout filled hollow stainless steel dowels had similar performance to epoxy-coated steel dowels. The use of de-icing salts in cold climates results in sig- nificant deterioration of steel dowels. Khazanovich et al. (2006) conducted f-sAPT (Minnesota) on two jointed test pavement specimens that incorporated stainless steel hol- low tube dowel and epoxy-coated solid steel dowels. Deflec- tions in the loaded test specimens yielded similar LTE when compared with numerical analyses of the doweled joints, indicating that the Minnesota Accelerated Loading Facility (MinneALF-2) provided good approximation of the in situ behavior of jointed pavements in terms of LTE and differ- ential deflections. Test results illustrated that the LTE for the stainless steel dowel tubes was lower than the LTE for the epoxy-coated dowels and that the stainless steel tubes were capable of long-term performance above 70% LTE (the minimum acceptable LTE value for MnDOT). Based on the f-sAPT results, the stainless steel tubes are capable of pro- viding sufficiently high long-term load transfer efficiency for concrete pavement joints. Rohne and Burnham (2010) evaluated concrete pave- ment joint performance and observed a unique distress phe- nomenon in the transverse joints. It occurred where jointed concrete pavements were constructed on undrained gravel bases and presented through cores with a significant amount of concrete material missing from the middle section of the joint. The area of greatest distress was located just below the saw cut, approximately at mid-depth. From the investigation it was evident that sections with base layers that adequately drained water within the joints performed significantly bet- ter. The distress was identified as a likely combination of freeze/thaw damage and erosion resulting from fast-travel- ling trucks. curling of concrete slabs Daiutolo (2008) evaluated curling through f-sAPT at NAPTF to avoid premature corner cracking of slabs in future testing using a single concrete slab with high fly ash content (added to reduce the flexural strength of the concrete) on a rigid support. When slab curl becomes excessive, premature top down cracking is likely to occur when the pavement is sub- jected to test loading. Keeping the pavement continually wet was found to be an effective procedure at the NAPTF where daily variations in concrete temperatures, and consequently curl, are not as severe in the indoor environment as would be expected outdoors. repair of concrete slabs Priddy et al. (2009, 2010) examined nine rapid-setting materi- als for repair of PCC runway pavements through laboratory characterization and f-sAPT (USACE, Mississippi). Seven of the nine materials demonstrated the ability to sustain air- craft traffic within 3 hours of repair completion. Based on the f-sAPT, a wide variety of rapid setting material types were suitable for rapidly repairing small, full-depth sections of PCC pavement. Full repairs including removal and replacement of subgrade materials followed by placement of a rapid-setting cementitious cap and foam injection were tested comparing repairs (three different foam volumes and a control repair) under normally distributed simulated F-15E aircraft traffic. Foam-injected repairs provided more passes before failure and were generally less expensive and time-consuming than pourable foam repairs. Foam-backfilled repairs were not cost- effective. The use of debris backfill eliminated the requirement for compaction compared with the clay gravel repair, reducing the overall repair time, and also provided similar pass-to-failure numbers as compared with traditional backfill materials such

74 as clay gravel. The foam injection methods were deemed more suitable in scenarios where quantities of high-quality debris backfill, similar to that used, are not available. airfields Brill and Guo (2009) evaluated concrete slab performance under 4-wheel and 6-wheel simulated f-sAPT (FAA NAPTF) airplane loads using visual surveys, crack mapping, destruc- tive and nondestructive testing, and observed top-down and bottom-up cracks. Analysis indicated that the bottom-up cracking initiated soon after starting trafficking at full load and only later progressed to the surface. The presence of a sig- nificant percentage of top-down longitudinal and transverse cracks in airfield pavements has implications for the future development of FAA rigid pavement design procedures as this failure mode is currently not considered. The evaluation demonstrated the importance of frequent and detailed visual observations of pavements. Granular materials rutting Rasoulian et al. (2000) compared rutting performance under f-sAPT (Louisiana ALF) of an inverted (stone interlayer) pavement consisting of a 4 in. (102 mm) layer of stone base on top of 6 in. (152 mm) of in-place cement-stabilized base, with a control section consisting of 8.5 in. (216 mm) of cement- stabilized base layer on top of prepared subgrade. Both sec- tions were surfaced with a 4 in. (99 mm) layer of HMA. The inverted design experienced only 16% of the reflective crack- ing experienced by the control section and had almost five times the performance life of the control section. The struc- tural capacity, ride characteristics, and rutting of the inverted pavement was similar to that of the control pavement. Bejarano (2004) conducted an analysis of rutting in unbound aggregate base, subbase, and subgrade layers using f-sAPT (California HVS). Pavement responses were related to rutting using a recursive rutting model that considers stress level, material condition, and number of load applications. The calculated model coefficients were dependent on stress level and material condition. Evaluation of the subgrade strain criteria indicated that subgrade plastic deformation was prevented. Odermatt et al. (2004) examined the differences in rutting development of crushed and uncrushed layers subjected to f-sAPT (Sweden HVS) and laboratory tests. Indications are that for natural gravel the f-sAPT may cause less rutting than in real roads owing to better side support than that achieved near the edge of the roads, while the same natural gravel rut- ted more than the crushed rock in tri-axial laboratory tests because of improved aggregate interlock in the crushed rock samples. In f-sAPT crushed rock showed greater rutting, probably the result of insufficient compaction. The results indicated that the required degree of compaction for crushed materials in pavements may need to be increased. Korkiala-Tanttu (2004) studied the rutting development of unbound materials in low-volume roads in Finland, focus- ing on the effects of different subgrades, layer thicknesses, water contents, and pavement materials. It was shown that FEM can model the stress distribution of a pavement reliably and that the selected material model drastically affects the stress distribution and therefore rutting development in the model. A conventional linear-elastic material model is not adequate for this modeling as tensile stresses may develop in the modeled unbound pavement layers, causing unreal- istic stress concentrations and erroneous rutting develop- ment data. Use of the Mohr–Coulomb material model was effective for unbound materials and the results of the f-sAPT could be modeled reliably. Data confirmed the dependency of rutting development on pavement geometry, material stiffness, and stresses. Wu et al. (2009a) developed a simplified numerical model for simulation of rutting development in granular base materi- als based on a conventional elastic-plastic model with linear strain hardening with empirical simplifications. The model directly uses different secant moduli for loading and unload- ing behavior and predicts permanent strain for each load rep- etition. The secant modulus is a function of load repetition, stress level, temperature, moisture content, and other material properties. FEM analyses simulating f-sAPT data (Louisiana ALF) showed satisfactory results using the model. Hussain et al. (2011) investigated the geology, nature of fines, effect on the rutting development, and changes in gradation and moisture of unbound flexible pavement grey- wacke base course materials using f-sAPT at CAPTIF. Three similar sourced greywacke aggregate materials with differ- ent gradations were subjected to repeated load cycles during which it was found that the geologic properties of the aggre- gate and minerals present in the aggregate are important to improve the understanding of the layer performance. The aggregate geology showed that the three materials were geo- logically sound and therefore the rutting cannot be affected by the coarse aggregate fractions. The factor affecting the rutting behavior was the gradation. Materials with a Talbot’s gradation value constant of 0.5 performed well, while a value of 0.37 indicated poor performance. structural analysis Kim and Tutumluer (2005, 2006) analyzed the effects of principle stress rotations resulting from moving wheel loads on rutting development in granular pavement granular lay- ers, causing varying vertical, horizontal, and shear stresses on pavements. Laboratory testing was combined with f-sAPT on low-strength subgrade flexible pavement test sections at

75 NAPTF. By using stationary repeated wheel loading and moving wheel loading-type models, granular subbase rutting development was predicted. As these models do not account for stress rotations, neither the magnitudes nor the rutting accumulation rates could be predicted accurately. The predic- tion models were adapted through a simple laboratory-based approach (catering for different load pulse durations) for the actual NAPTF loading stress history. After the corrections were made, both the coal combustion products (CCPs) and VCP models showed improved predictions, with the variable confining pressure (VCP) model giving better predictions of the rutting magnitude and accumulation rate. The flexible pavement test sections at NAPTF were traf- ficked while load-induced vertical subgrade stress were mon- itored, using a simulated six-wheel Boeing 777 landing gear and a simulated four-wheel Boeing 747 gear simultaneously until the sections failed (Gopalakrishnan and Thompson 2006b). Analysis of the vertical subgrade stresses shows that the subgrade stresses induced by the B777 and B747 did not differ significantly. González et al. (2007) validated 3-D FEM models using data collected at CAPTIF demonstrating the feasibility of using advanced structural analysis tools such as 3D-FEM software and nonlinear-elastic models for the analysis of thinly surfaced pavement structures. Analysis results showed an accurate pre- diction of the behavior of the modeled pavement. Wu et al. (2007) evaluated three f-sAPT sections (Louisiana ALF) with different base and subbase materials and observed that the cement-treated subbase possessed higher load-induced structural capacity than a lime-treated subbase (higher resilient modulus, greater load carrying capability, and lower rutting). Heavier loads caused a higher percent increase of vertical stresses on top of the subgrade than on top of the subbase and thus subgrade rutting owing to overloaded conditions should be less in real applications. Rutting analysis using MEPDG software generally overestimated the rutting development for the f-sAPT sections, and a simple rutting prediction model was proposed, relating flexible pavement rutting development to in situ surface deflections. Kim and Tutumluer (2008) investigated the effect of multiple wheel loading scenarios on granular airport pave- ment structures. The traditional approach of single wheel load response superposition may not be adequate to analyze pavements subjected to such multiple wheel load cases, as these complex loading conditions require advanced pavement layered solutions to realistically consider nonliner behavior of the materials. Analysis of data from NAPTF yielded two peaks directly under the wheels for thinner sections, leading to the highest subgrade vertical deflections and computed stresses. As the granular base gets substantially thicker the two peaks merge toward the middle of the two wheels. Critical pavement responses and locations in the pavement structure were computed and shown to be significantly influenced by multiple wheel loads from single-, tandem-, and tridem-axle- type wheel arrangements and configurations. Load spreading and nonlinear modulus distributions of the granular base lay- ers were found to significantly impact the maximum surface deflections, whereas the subgrade vertical stresses and deflec- tions were considerably influenced by both loading and the base layer thickness. Xu et al. (2008) presented data from a comparison of the performance of various semi-rigid (cement-stabilized crushed stone) base asphalt pavements and flexible (graded crushed stone) base HMA pavements obtained through f-sAPT in China. Although the flexible base structure was thicker than the semi-rigid base structure, it was less temperature-sensitive than similar HMA base pavements and less prone to rutting compared with the flexible base pavement structure. No sur- face cracks were observed during the tests. The deflection index was shown to be related to the bearing capacity of the full pavement structures and not affected by deterioration of the bearing capacity of individual pavement layers. Chen and Abu-Farsakh (2010) evaluated the in situ perfor- mance of raw blended calcium sulfate (BCS), stabilized BCS, stabilized RAP, and stabilized soil as base–subbase materials through cyclic plate load tests and f-sAPT (Louisiana ALF). It was shown that the use of a conventional formula calculat- ing the equivalent elastic modulus with only the individual layer thickness considered is not a good performance indi- cator for multilayer systems and proposed a modified for- mula using a position factor on the basis of Boussinesq stress distributions in soils. Successful application of the model to the tested pavement sections demonstrated that the position factor can turn the conventional equivalent elastic modulus into a good performance indicator. The f-sAPT measured rut- ting depth was higher than the cyclic plate load rutting depth (three to seven times larger) indicating that the rolling wheel load is much more damaging than the cyclic plate load. This phenomenon is probably linked to principal stress rotation, friction-induced tangential forces, and lateral wander during the f-sAPT. The extension–compression–extension multiple stress path-type test with principal stress rotation causes much higher rutting than the single compression stress path- type test with no principal stress rotation, whereas lateral wander most likely decreases the stability of unbound and weakly bound granular base materials. The ground granulated blast furnace slag (GGBFS)-stabilized BCS tested as a supe- rior base material compared with the conventional stone base, while the two fly ash-stabilized RAP base materials did not perform well compared with the other sections. A considerable amount of research focused on the devel- opment of a repeated load triaxial (RLT) test procedure for evaluating natural and marginal unbound base and subbase materials (Jameson et al. 2010). A major Australian research project was undertaken comparing the rutting of four gran- ular bases under f-sAPT with various laboratory perfor- mance tests. Base layers were tested for rutting resistance in

76 a RLT test using constant confining pressure based on the Austroads test method, the Department for Transport, Energy and Infrastructure South Australia test method, and the Tran- sit New Zealand (NZ) method. The crushed hornfels rutted much more rapidly than the crushed rhyolite, and there was concern that there was no indication from any of the exist- ing processes of this behavior with both the Department for Transport, Energy and Infrastructure and Transit NZ methods indicating allowable performance for all four tested granular bases. The findings questioned the usefulness of the perma- nent strains measured in the Austroads test method, as the RLT permanent strain results did not identify the hornfels as a material unsuitable for base course because of its low resis- tance to lateral shoving. The inability of the permanent strain test to identify the poor performance of the hornfels supports the contention that incorporation of aggregate interlock under shear stress reversals occurring under rolling wheel loads is vital for accurate rutting models. The laboratory wheel track- ing results correlated well with the f-sAPT data indicating that the wheel tracking test is probably the most suitable test for the characterization of the rutting potential of unbound bases under thin bitumen-treated surfacings. Navneet et al. (2010) described f-sAPT using the NAPTF and B777 and B747 loading on four flexible pavement test sections incorporating 5 in. (125 mm) HMA surfacings on top of a crushed stone base, varying thickness subbases, and a silty clay subgrade. A fixed wander pattern was used. Results showed nonlinear behavior (stress-softening) of the subgrade with a final surface rutting of approximately 3 in. (75 mm). Post-traffic tests showed no signs of shear failure or rutting in the subgrade, with most of the rutting contributed by the subbase layer. stabilization and modification of soils Romanoschi et al. (2006) found cement and lime to be the most effective stabilizers for a nonsulfate-bearing clayey soil tested in f-sAPT (Kansas), resulting in the lowest vertical compressive stresses at the top of the unbound clayey sub- grade. Fly ash-treated subgrade soil generated higher verti- cal compressive stresses at the top of the subgrade as well as higher rutting depths than both the cement- and lime-treated subgrade soils. De Vos et al. (2007) reported on the development of guidelines for an M-E pavement design model for cement- stabilized sand bases in Mozambique based on f-sAPT, scaled APT, and laboratory tests. The maximum tensile stress at the bottom of these bases for both the scaled APT and f-sAPT were calculated and related using the 4th Power Law (expo- nent of 4.2). Similar performance was observed between the laboratory and field pavements in terms of distress and num- ber of axle load applications. The dynamic deflection of the pavement progressively increased relative to the initial value and the PSPA stiffness decreased simultaneously with the stiffness ratio reaching 50% of the initial value when signifi- cant distress manifested. Since 2005, pavement design in Denmark has been based on a three-tiered design guideline (1st level—design accord- ing to a catalog, 2nd level—automated M-E design, and 3rd level—based on pavement performance simulation and incre- mental-recursive methodology) (Busch 2008). The recursive- incremental models for cement-stabilized materials are based on f-sAPT data. These models should still be verified using in-service pavement performance to improve reliability, although they currently do provide credible outputs. Yeo (2008) investigated parameters influencing the per- formance of unbound and cemented pavements with thin bitumen-treated surfacings to improve the understanding of fatigue performance and derive fatigue-consistent-terms rela- tionships for cemented materials based on f-sAPT and labora- tory strength, modulus, and fatigue characteristics. The most appropriate means to assess the fatigue performance was backcalculation of material moduli from FWD data. Mea- sured fatigue life was compared with initial tensile strain in the cemented layer to develop a relationship for each mate- rial. Based on the f-sAPT and laboratory data, a load damage exponent of between 5 and 8 was calculated that does not support the current Austroads load damage exponent of 12. Wu et al. (2009b) evaluated BCS as a pavement base layer using laboratory and f-sAPT, compared with a crushed stone base. FWD results indicated higher in situ stiffness for the BCS/GGBFS layer than for an HMA layer, which resulted in an inverted pavement structure for one test section. It was found that the multilayer elastic theory overpredicted the ver- tical compressive stresses developed in an unbound aggre- gate layer and underpredicted the stresses for other bounded base materials, because the elastic theory predicted an unre- alistic tension zone inside the base layer. Most of the rutting occurred in the base layers. The 15% fly ash-stabilized BCS base performed better than the crushed stone base, probably because of a shear flow failure related to water susceptibility. It was demonstrated that the thickness of the HMA layer can be reduced significantly by using a GGBFS-stabilized BCS base instead of a crushed stone base. In the current MEPDG, cement-stabilized layers are assumed to have no contribution to the total rutting develop- ment of the pavement. Wu et al. (2011) developed a unified rutting model to simulate the rutting behavior of a cement- stabilized pavement layer. The model was used successfully in a FEM analysis to evaluate the performance of f-sAPT sections, and a shift factor of 1.13 was obtained from calibra- tion to account for the condition differences between labora- tory permanent deformation tests and pavements under traf- fic load. It was demonstrated that rutting contribution from cement-stabilized layers may account for a considerable por- tion of surface rutting.

77 Geogrids Dawson et al. (2004) assessed the potential of geosynthetics as unsealed road reinforcements in 16 test sections incorpo- rating a variety of geosynthetics, natural reinforcement, or no reinforcement subjected to f-sAPT (United Kingdom), evaluating the influence of nonlinearity, material variability, pavement cross section, and other variables. There was strong evidence that the fundamental mechanism of reinforcement in unsealed pavements is the result of the prevention of out- ward shear on the subgrade. The commonly adopted 4th Power Law load equivalency approach was not valid for the pavements studied. Perkins and Cortez (2004) found that similar rankings in terms of rutting were observed between f-sAPT and labora- tory box test sections under a cyclic load for geosynthetic- reinforced pavement materials. The magnitude of reinforce- ment benefit (defined as Traffic Benefit Ratio) was generally lower in the f-sAPT sections (located at USACE–CRREL) compared with the box test results. A marked difference was observed in the stress and strain response measures for the f-sAPT data, including dynamic vertical stress on the top of the subgrade, dynamic vertical strain in the base and sub- grade, dynamic transverse strain in the bottom of the base and top of the subgrade, and permanent vertical strain in the base and subgrade layers. Differences between reinforced sections were apparent but not significant. Reyes and Kohler (2006) described the performance of three geogrid-reinforced sections and a control section under f-sAPT in Colombia. Failure was observed owing to fatigue on the HMA mixture instead of under compression on the subgrade with best performance evident when the geogrid is placed in the granular base. As long as the geogrid reinforce- ment is placed in the upper layers, the fatigue performance improved. Nine pavement test sections were evaluated using f-sAPT to determine the effectiveness of geogrid in secondary flex- ible pavements (Al-Qadi et al. 2008a, b). All sections were constructed over a weak subgrade in Illinois. The geogrid was effective in reducing the horizontal shear deformation of the aggregate layer, and thus the effectiveness of the geogrid for aggregate base layers with thicknesses ranging from 8 to 18 in. (200 to 457 mm) was demonstrated. The optimal geogrid location (for a thin aggregate layer) was found to be at the unbound aggregate-subgrade interface, while another geogrid at the subgrade-base layer interface may be needed for stability of thicker base layers. Numerical modeling and f-sAPT of geogrid base-reinforced pavement systems (located in Illinois) indicated the stiffen- ing effect around geogrid reinforcement and its effect on improved pavement performance (Kwon and Tutumluer 2009). A stiffened zone was observed around the geogrid reinforcement, primarily the result of the aggregate particle interlock in the apertures of the geogrid, as shown through Discrete Element Modeling. The stiffened zone is located 100 to 150 mm above and below the geogrid. F-sAPT results demonstrated the performance benefits of using geogrids especially in the reduced horizontal base course movements. Pavement structures and systems The validity of material response models is important for reliable evaluation of the structural pavement condition. An instrumented field test pavement was established on top of a sandy material sufficiently thick to be considered a half space to verify predicted pavement response (Hildebrand 2003). FWD data on each layer provided elastic layer moduli, while in situ sensors registered stress and strain in three orthogo- nal directions. Pavement response was predicted using a wide range of response methods from a simple Boussinesq model to FEM and compared with f-sAPT data (Denmark). The comparison showed good agreement for strain response, whereas the agreement was more questionable for stress. It was concluded that although field verification is difficult, it is possible to construct an approximately homogeneous and isotropic test site suitable for pavement response verification. Blab (2004) evaluated an existing semi-rigid pavement in Poland using laboratory testing and f-sAPT program. Design calculations indicated that the thickness of the cement- stabilized base layers could be reduced and still fulfill the performance requirements for the type of design, as the f-sAPT on the in-service pavement showed no fatigue dam- age at the end of the tests. Lenngren (2004) evaluated eight f-sAPT sections (located in Sweden) to determine the influence of moisture on a pave- ment where mica is present in the granular materials. Results showed that the influence of moisture did affect the deterio- ration of the road, whereas the presence of mica had limited influence under dry conditions. There was no clear evidence in the test data that mica contents of between 6% and 34% affected the rutting during dry conditions. The top of sub- grade strain rutting criterion proved to be a very good single indicator of rutting development, but could not predict initial rut. The evaluation demonstrated the importance of a good understanding of subbase and subgrade properties in pave- ment evaluation. Martin (2005) evaluated refining existing road deteriora- tion models for Australian sealed granular pavements based on f-sAPT data. Different rutting rates were observed for seemingly similar pavements with a crushed concrete blend showing 3 to 10 times as much rutting as a high-quality gran- ular material. Relative performance factors and load dam- age equivalence (LDE) factors were developed based on the f-sAPT data for rutting and roughness development based on statistically significant relationships using variables for load cycles, pavement layer strengths, and wheel load. The

78 rutting LDE of the marginal material was lower than the 4th Power Law and higher than the 4th Power Law for high- quality materials, while the roughness LDE was close to the 4th Power Law for the marginal material and significantly higher than the 4th Power Law for high-quality materials. The use of the relatively short [39 ft (12 m)] ALF test pave- ment to derive relative roughness estimates appears to be inappropriate. Steyn (2008) developed guidelines for design of long-life pavements based on the experiences from mobile f-sAPT (South African HVS) on pavements that performed well after more than 30 years in the field. The basic principles of pavement engineering are fundamental to these designs, and it is thus important to keep the pavement layers dry, sup- port the various pavement layers well, ensure good strength- balance between the various layers in the pavement, select appropriate aggregate gradations to prevent rutting and use bituminous binders that do not age quickly to prevent fatigue. MnROAD performed maintenance treatments in 2003 to improve the riding quality before reconstruction could be scheduled and used the opportunity to evaluate differ- ent maintenance treatments and their effectiveness (Worel and Clyne 2009). The slurry seals used for the maintenance performed well in terms of restoring riding quality and also performed adequately for filling existing ruts. However, the maintenance treatments studied were not successful in pre- venting reflective cracking from underlying transverse or top down cracks. Martin (2010) derived an equation for estimating load- related road wear for uncracked, reasonable, quality-sealed unbound pavements (in Australia) using a component cumu- lative roughness deterioration equation based on observa- tional and experimental data. The equation differs from the HDM4 incremental component roughness model and pro- vides varying estimates of road wear as affected by traffic load, pavement strength, and environmental conditions. Parmeggiani (2009) developed an alternative approach for the design of pavements with infinite structural life and low maintenance using an M-E design method. The concept of long-life pavements defines structures that provide strong, thick HMA layers having the required structural capacity to provide long service lives under very heavy traffic load- ing conditions with only periodic surface renewal required during this period. The individual HMA layers are typically designed to resist specific modes of distress to maintain the critical strains in the HMA layers and subgrade below critical threshold levels. Although traditional perpetual pavements rely on the total HMA layer thickness together with the fatigue-resistant lower HMA layer to provide the required HMA fatigue life, the alternative approach entails provision of enhanced support to the HMA layers to comply with the same strain threshold levels in the HMA and subgrades as specified for the present perpetual pavements. The approach envisages complete elimination of HMA fatigue cracking and subgrade deformation and subsequent design of the HMA layer to maximize its rutting resistance without com- promising durability. F-sAPT experiments (Australia) on maintenance treat- ments were used to refine relative performance factors for cumulative rutting and roughness on uncracked and cracked seals as the performance of these seals are prone to the influ- ence of surface water (Martin 2010). Observational data were obtained for the development of road deterioration models from LTM, LTPP maintenance, and f-sAPT in Australia. The developed road development models are applicable for the gradual rutting, cracking, and roughness deterioration of sealed granular pavements under standard in-service conditions (Martin and Choummanivong 2010). An f-sAPT evaluation of the behavior of full-depth HMA pavements designed using perpetual pavement concepts was conducted in Binzhou, Shandong Province, China (Timm et al. 2011b). PerRoad was used for probabilistic analysis comparison of the pavement responses and it was confirmed that the tensile HMA strains were sufficiently low to prevent bottom-up fatigue, except where full-slip conditions existed between layers. The importance of good bonds between lay- ers was evident from the behavior of the sections. miscellaneous toPics Bridges F-sAPT on an HMA-surfaced steel bridge indicated that HMA behavior depends on proper choices of material as well as the type of waterproof bonding layer used (Shao et al. 2004). Emphasis should be placed on ensuring that the HMA is nondeformable and stable at high temperatures. F-sAPT research into orthotropic steel deck bridge deck behavior with HMA overlays combined with FEM analysis indicated that the assumptions upon which current design techniques are based proved not to be true (Huurman et al. 2004). FEM calculations showed that strain gradients are strongly influenced by the bridge deck geometry, loading, and the behavior of both the membrane interface and the sur- facing material. A constitutive model was developed that is capable of simulating the f-sAPT observed behavior of the surfacing asphalt (Liu et al. 2006). Jun et al. (2008) used circular f-sAPT and FEM to ana- lyze rutting development of steel bridge decks surfaced with HMA. The HMA was modeled using a creep model in the FEM, and the data compared well to the measured tertiary rutting development on the f-sAPT test.

79 coal combustion Product Tu et al. (2006) evaluated the performance of pavements constructed with recycled CCPs using f-sAPT. The perfor- mance of the CCP pavements was promising as potential substitutes for conventional base materials. All the pavement sections exhibited similar surface distress performance, with the CCP base/subbase mixes outperforming the control mix. Both CCP sections showed lower deformation, higher over- all stiffness, and lower traffic-induced stress in the subgrade than the conventional pavement. edge damage Olson and Roberson (2003) described research on joint seal- ing activities at MnROAD where pavement sections were instrumented to provide automated measurements of pave- ment moisture and drainage changes resulting from varying climate conditions. The research included measuring changes in edge drain outflow and base moisture content in response to precipitation events. Data were collected before and after seal- ing of edge joints on concrete sections. Although no signifi- cant difference was observed in the volume drained between the control section and test section before sealing the joint on the test section, a significant reduction in the volume drained from the test section was observed after sealing the edge joints. It was shown that sealing the concrete pavement edge joint with a bituminous shoulder reduced the total volume of water entering the pavement system by as much as 85%. Saleh (2006) evaluated the factors affecting the develop- ment of pavement edge failure in New Zealand using 3-D FEM incorporating shoulder width, shoulder stiffness, axle load, tire pressure and pavement thickness, and compared the output with multilayer analysis and f-sAPT. Shoulder stiff- ness was shown to be a significant factor affecting the lateral support and therefore pavement response in the outer wheel path with axle load being a significant factor for all responses affecting the edge damage. The use of stiff shoulders should thus help to reduce edge damage of the pavement. large stone Hot mix asphalt The applicability of using large stone HMA for airport pave- ments surface courses was evaluated using static loading, FWD tests, and f-sAPT (Tsubokawa et al. 2004). The research has shown that application of large stone HMA in the surface and binder courses of airport pavements may reduce rutting by up to 20% compared with conventional HMA. Backcalcula- tion analysis showed the elastic modulus of large stone HMA to be 1.2 to 1.3 times larger than that of conventional HMA. low-volume roads Snyder (2008b) evaluated the design and performance of the five original MnROAD low-volume road concrete sec- tions. Application of MEPDG provided performance esti- mates that closely matched observed pavement roughness trends, although it still underpredicted project performance by about 23%. The low-volume road test cells show ride deterioration rates suggesting that the design parameters that were varied on these test cells (i.e., load transfer, drainage, foundation preparation, etc.) appear to be more significant for thinner concrete pavements. Initial road roughness on concrete pavements has a tremendous impact on the pave- ment life and thus the performance life may be extended sig- nificantly by simply improving construction quality. Panel cracking was mainly attributed to localized loss of support, with panel length changes not strongly linked to changes in panel cracking. The use of dowels and a thick sand-filled layer effectively reduced joint faulting. Pay Factors A procedure was developed to establish pay factors for HMA construction using performance models for fatigue and rutting based on a combination of M-E analyses, Strategic Highway Research Program (SHRP) developed laboratory test data, and f-sAPT data (Popescu and Monismith 2007). The mod- els use means and variances rather than the Percent Within Limits (PWL) approach, with the influence of asphalt content, air-void content, and aggregate gradation considered for rut- ting, and air-void content, asphalt content, and asphalt concrete thickness considered for fatigue. The relative performance of the as-constructed HMA is determined based on its measured mean property and a reasonable standard deviation. Relative performance is defined as the ratio of off-target traffic (ESALs) to target or design traffic (ESALs). Only agency cost conse- quences delaying or accelerating the time to the next rehabilita- tion are included in the cost model. The methodology provides for a full bonus for superior construction and a full penalty for inferior construction. The approach emphasizes the importance of adhering to the target value for a specific pavement charac- teristic and maintaining uniformity to achieve or exceed the desired performance level. Pipes MnDOT conducted research on the performance of large diameter corrugated polyethylene pipes installed under high- way vehicle loading to improve analysis and design methods for these conditions (McGrath 2005). Corrugated polyethylene pipe sections were installed on the MnROAD low-volume loop and instrumented. It was observed that the pipe has performed well with minimal load response under moderate live loads for a period of 3.5 years without any deterioration. Because of the high coefficient of thermal expansion and the temperature extremes in the Minnesota environment, the pipe expansion and contraction caused the pavement surface to become rough and crack the pavement over the pipes. The research lead to a recommendation that a minimum cover depth of 2 ft (600 mm) or 0.5 times the nominal pipe diameter is required.

80 Gaz de France developed a concept of a thin trench filled with a self-placing concrete in which polyethylene gas pipes are directly placed (Balay et al. 2008). The concept was eval- uated in the LCPC f-sAPT facility through the construction of five trenches with depths ranging between 16 and 32 in. (400 and 800 mm), as well as a conventional trench, which indicated that the concept performed very well with no dete- rioration or damage of the structures or the adjacent pave- ment. Strain measurements indicated that the bending ten- sile loads created by rolling loads remained well below the fatigue damage limits foreseeable with this material, while no subsidence of the trenches was observed. Farrag (2011) performed laboratory and f-sAPT on several excavatable controlled low-strength material (CLSM) mix- tures in accordance with the standard specifications of several state highway agencies to evaluate its use as an alternative to traditional soil backfill materials around buried pipes. Lower settlement was observed than for other compacted soil back- fills; however, CLSM with high fly ash had lower frost-heave and thawing resistance than soils. The long-term compressive strength of the CLSM with fly ash increased over time and exceeded the limits used for manual excavation. Construction quality control of CLSM is a challenge and mostly limited to characterizing the flowability of the mix. Prefabricated Bituminous slabs The Dutch Ministry of Public Works and Water Manage- ment anticipates the need for future mobility evaluating by a prefabricated road surface (Van Dommelen et al. 2004). Four designs proposed by the private sector were constructed and evaluated for functional performance on pilot test sec- tions, while structural evaluation was conducted using the LinTrack APT facility, supported by FEM and laboratory material research. This mix of research approaches provided sufficient evidence to justify decisions about wider scale application without the need for 10 to 15 years of test section evaluation. The evaluations showed that the modular road surface is cost-effective and viable relative to traditional road construction and maintenance efforts. The project demon- strated the benefits of the combination of f-sAPT, in-service test sections, modeling, and laboratory material testing to enable responsible decisions about the implementation of innovative pavement concepts. recycled concrete Reflection cracking is a common distress in asphalt over- lays on cracked concrete pavements. Rubblization, which can minimize reflection cracks, has been used as an effi- cient rehabilitation method for aged concrete pavements in the United States. Garg et al. (2007) conducted f-sAPT on three rubblized rigid airport pavements overlaid with 5 in. (127 mm) of HMA at the NAPTF. MRC (rubblized con- crete on conventional base) was observed to suffer severe structural distress, while MRG (rubblized concrete on grade) suffered severe structural deterioration toward the end of trafficking although it retained sufficient structural capacity to support the applied load. MRS (rubblized concrete over econocrete base) accumulated significant rutting and shear flow in the asphalt layer but did not appear to suffer severe structural deterioration (MRS is the most common of pave- ment structures encountered on commercial airports in the United States). The analyses indicated that the current design assumptions are overly conservative for the conditions eval- uated. The structural performance of MRS suggests that rub- blized concrete pavements with HMA overlay are a viable option for commercial airports. road user charges Martin (2010) used data obtained from f-sAPT research to develop a model for implementation of varying heavy vehi- cle road user charges on Australia’s arterial roads. An equa- tion for estimating load-related road wear was derived for uncracked, sealed, unbound pavements typical of Australia’s sealed road network using a cumulative roughness deteriora- tion equation based on observational and experimental data for the gradual deterioration phase of sealed granular pave- ment life. This equation gave varying estimates of load wear with changes to the traffic load, pavement strength, and the environmental coefficient, and was also found to give dif- ferent estimates of load wear than that provided by an equa- tion based on the HDM4 incremental component roughness deterioration model. snow-melting systems Embedding snow-melting equipment in airport HMA layers is considered an effective way of avoiding disruption from extreme weather events. Hachiya et al. (2008) evaluated the performance of two types of snow-melting equipment: a warm water heating pipe system and an electric heating wire system. Pavements were trafficked in f-sAPT fashion using a wheel configuration identical to that of a B747 aircraft. Pavement surface deflection under the aircraft loading was slightly higher when the snow-melting system was present; however, there is no clear difference arising from the type of snow-melting system and its embedded depth. The rutting development was similar with and without either of the two snow-melting sys- tems present. 3-D FEM analysis demonstrated that the stress imposed on the heating unit when aircraft loading is applied to the pavement is much less than its strength, and it was thus concluded that snow-melting systems could be successfully introduced into airport asphalt pavements. Pavement markings Choubane et al. (2006b) evaluated the viability of testing the deterioration of raised pavement markers using f-sAPT

81 (Florida HVS). This option allowed for the controlled appli- cation of realistic wheel loads to the installed markers, simu- lating long-term, in-service loading conditions. Four types of markers were evaluated and the outcome provided a rela- tive rating between the durability of the four markers and the amount of traffic applied to the markers. Correlation to field measurements was beyond the scope of the project. cHaPter summary This chapter focuses on behavior and performance of specific pavement layer materials evaluated using f-sAPT. The major focus of the chapter is on HMA materials. The questionnaire trend indicating that f-sAPT focuses on materials used closer to the surface was confirmed through the evaluation of the published literature. Apart from the major emphasis on HMA rutting, a developing trend is the evaluation of environmen- tally sensitive materials such as WMA and RAP, as well as the focus on pavement life extension through application of HMA overlays and various types of thin concrete overlays. Although some f-sAPT on granular materials is still being conducted, the applications are limited. A number of miscel- laneous unique applications of f-sAPT were also discussed, including the testing of pipes in trenches, snow melting systems, and the use of f-sAPT data to assist in calculating road user charges and pay factors. Although the focus of this section is on the evaluation of specific materials in specific layers, most of the research incorporates the effects of the supporting layers into the pavement response analyses. The volume of research still being performed on the analysis of specific material response models indicates the importance of this work to ensure that appropriate models are developed for use on pavement design methods.