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.

82 IntroductIon This chapter provides detailed findings on specific interactions between the loading device and the pavement structure with a focus on the linkages to real traffic loading, incorporating the effects of load equivalence, tire types, real-life environ- mental effects, etc. F-sAPT aims to evaluate pavement sec- tions under a range of loading and environmental conditions to improve the knowledge of the potential performance of the pavement layers and structure under a full range of operational conditions. Using this philosophy it is standard f-sAPT practice to select a range of vehicular loading condi- tions (i.e., load levels and tire inflation pressures) as well as environmental conditions (i.e., temperatures and moisture conditions) for different tests and obtain the response of the pavement under the specific selected conditions. The outputs of these tests are combined to develop a model of pavement response under expected field conditions. Twenty respondents indicated that environmental aspects such as noise and dust are not applicable to f-sAPT, although seven respondents do include evaluation of the effects of road noise in their f-sAPT. The majority of respondents related their f-sAPT data to pavement temperature and ambient air tem- perature (Figure 46). Most respondents control the pavement and ambient temperature during tests, with moisture control being a secondary parameter that is controlled or monitored (Figure 47). Load Parameters Increased Load Levels Alabaster et al. (2004a, b) compared the potential impact of increasing the standard legal limit for a dual tired axle of 8.2 tons to 10 and 12 tons using the CAPTIF f-sAPT facility and thin pavements containing a strong dry subgrade. Data indicated that the use of subgrade strain is a poor predictor of pavement life as the basecourse aggregates significantly influ- enced rutting resistance. Using Vertical Surface Deformation (VSD) proved to be the most useful measure for monitoring pavement wear with 60 kN wheel loads resulting in nearly twice the VSD obtained with the 40 kN wheel load in all the pavement segments. The VSD data were modeled using a power law relationship and exponents varied between 2 and 9, depending on the pavement-type and end-of-pavement-life definition. The analysis illustrated that weaker sections in the road network that are adequate at the lower legal axle load may fail quickly under the higher legal axle load. tire contact stress effects De Beer et al. (2004) demonstrated the effect of tire load and inflation pressure on the shape of the tire–pavement contact stress pattern and potential implications for f-sAPT response analysis. High edge stresses develop at high tire loads (referred to as m-shape distributions). Maximum vertical contact stresses appear to be as much as twice the inflation pressure (at extremely high levels of loading). For normal loading conditions the maximum vertical stress exceeds the infla- tion pressure by approximately 30%. Results also show that the length of the tire contact patch increases with increased loading, whereas the tire width remains constant. Quantifi- cation of the horizontal interface shear forces (or stresses) between the tire and the surface of the road could assist in the understanding of the horizontal stress regime in the tire patch of a moving tire on a coarse surface. Lateral and longitudinal stresses could be as high as 36 psi (250 kPa) and 24 psi (165 kPa), respectively. Park et al. (2005) compared predicted pavement response from 3-D FEM and layered elastic programs with the objec- tive of establishing guidelines on a better approximation of pavement response parameters for pavement design and eval- uation purposes. Tire–pavement contact stress data obtained from f-sAPT tests were used to predict pavement response. The horizontal strain at the bottom of the asphalt layer, com- pressive strain at the top of the subgrade, and the principal stresses at different depths were predicted. The decimated 3-D tire contact stresses gave computed tire loads similar to the original measured data. Incorporation of the 3-D tire con- tact stresses mainly influenced the pavement responses in the surface layer. Various researchers evaluated the pavement responses of typical pavement structures under the combined actions of variable wheel loads and tire pressures using multilayer lin- ear-elastic theory to estimate the pavement responses under uniform constant stress and actual contact stress distributions obtained from f-sAPT-based SIM measurements (Prozzi and Luo 2005; Machemehl et al. 2005; Wang and Machemehl 2006a, b). Statistical analysis found that tire inflation pres- sure was significantly related to tensile strains at the bottom chapter four VehIcLe–PaVement–enVIronment InteractIon

83 of the HMA layer, and stresses near the pavement surface for both the thick and thin pavement structures. However, tire pressure effects on vertical strain at the top of the subgrade were minor, especially in the thick pavement. Steyn (2009a) reported on the effect of nonuniform tire– pavement contact stresses as measured through application of two distinct types of tire–pavement contact stresses onto the HMA pavement using a HVS. The rutting response of the pave- ment mirrored the contact stress shapes, indicating the direct effect that the tire–pavement stress has on rutting development in HMA layers. Ozawa et al. (2010) derived closed form governing equa- tions with the assumption that a rectangular load moves at a constant speed on a surface of a pavement system com- posed of Voigt-model-type layers using Fourier transforms to derive theoretical solutions. The theoretical results indicated that pavement response decreases in magnitude with increas- ing speed of the moving load. For the same speed of the mov- ing load, a decrease was observed in the pavement response magnitude with increasing damping ratio of the materials. The effect of material density was found to be insignificant to the pavement responses. Wide-Base tires Various researchers evaluated the status of wide-base tire technology specifically regarding its effects on pavement response (Elseifi et al. 2005; Timm and Priest 2006; Yeo and 0 5 10 Pa ve me nt tem pe rat ure Am bie nt air te mp era tur e Wa ter ta ble Ra inf all Dr ain ag e Ag ing Re lat ive hu mi dit y De pth to be dro ck 15 20 25 30 N um be r o f r es po nd en ts FIGURE 46 Environmental data to which f-sAPT data are related during analysis. FIGURE 47 Environmental parameters controlled during f-sAPT. 0 2 4 6 8 10 Pa ve me nt tem pe rat ure Am bie nt air te mp era tur e Su bg rad e m ois tur e Wa ter ap plic ati on Co ntr olle d a gin g Dr ain ag e 12 14 16 18 N um be r o f r es po nd en ts

84 Sharp 2006, 2007; Al-Qadi and Elseifi 2007; Dessouky et al. 2007; Wang and Al-Qadi 2008; Yeo et al. 2008; Greene et al. 2010; Wang and Morian 2010; Xue and Weaver 2011). Tire widths can be divided into three groups. The first group is traditional tires used as dual sets on trucks, the second group is the first generation of wide-base tires (385/65R22.5 and 425/65R22.5), and the third group is the second generation wide-base tires (445/50R22.5 and 455/55R22.5). The second generation of wide-base tires provides substantial benefits in fuel efficiency, hauling capacity, tire cost and repair, ride, comfort, and vehicle stability (Al-Qadi and Elseifi 2007). The first generations of wide-base tires were found to cause a significant increase in pavement damage compared with dual tires. Results indicated that both dual and second gener- ation wide-base tire configurations produced similar pavement responses at the bottom of 6.7-in. (170-mm)-thick HMA, which indicates that both tire assemblies would produce sim- ilar fatigue damage. Both tire configurations also produced equal vertical stress on top of the subgrade, which indicates that the secondary rutting performance of both tire configura- tions would be the same. Studies also supported the notion that layered elastic theory (LET) may not be used to compare pavement response with different loading configurations as it assumes a uniform pressure distribution that is only a func- tion of the load and a circular contact area while improve- ments in the second generation wide-base tires cannot be quantified using this simple method. Up to six times greater rutting and roughness damage were observed on thin HMA pavements when using first generation wide-based tires com- pared with dual tires, while damage resulting from second generation wide-based tires was less severe than 385/65R22.5 tires. For thick pavements and primary roads the overall effect of the second generation of wide-base tires is expected to be equivalent to dual-tire assemblies, in that the frequency of top-down cracking in the wheel path is clearly reduced. Given the relatively thick HMA layer used in these pavement classes, the probability of fatigue failure is usually low. By increasing the tire contact area, pavement damage is gener- ally decreased. aircraft tire Loading It appears that more airfield-related work has been conducted since the previous synthesis. This may be linked to the initia- tion of testing using facilities with dedicated f-sAPT devices for airfield testing (i.e., NAPTF and WES). The new facility at Toulouse–Blagnac airport in France (Fabre et al. 2009) is another example of current developments of these facilities. Differences in tire and bogey configuration and tire loads necessitate a focused evaluation of these facilities. The NAPTF was constructed to generate f-sAPT data to study the performance of airport pavements subjected to complex gear loading configurations of new generation air- craft. Comparative effects of four- and six-wheel aircraft gear loads can be evaluated and failure criteria developed for mechanistic-based airport pavement design procedures. The first construction cycle consisted of three rigid pave- ment test items and six flexible pavement test items, while the second construction cycle consisted of a rigid pavement test strip, a free-standing test slab, and three rigid pave- ment test items. The third construction cycle consisted of four flexible pavement test items. Test data from the first tests involving Boeing 777 and Boeing 747 loading gears were reported by various researchers (Garg and Marsey 2004; Hayhoe 2004; Hayhoe and Garg 2006; Gopalakrish- nan and Thompson 2006a, b). FWD and HWD evaluation of the pavements before and during trafficking indicated that, among the deflections and the Deflection Basin Parameters considered, the deflection ratio (D1/D3) showed the stron- gest correlation with the number of load repetitions to reach functional rutting failure. Results from the f-sAPT have been incorporated in the FAA layered elastic flexible airport pavement thickness design procedure and for establishing new alpha factors for the California Bearing Ratio airport pavement thickness design procedure. Kim and Tutumluer (2005) presented findings on predicted performance and field validations of granular base/subbase layer permanent deformation models using the NAPTF f-sAPT facility combined with a comprehensive set of repeated load triaxial tests. Both constant and variable confining pressure conditions were evaluated on a granular base and subbase material. A comparison of measured and predicted rutting indicated that a good match for the measured rutting magni- tudes and the accumulation rates could only be achieved when the magnitudes and variations of stress states in the granular layers, number of load applications, gear load wander patterns, previous loading stress history effects, loading rate effects, and principal stress rotation effects owing to moving wheel loads were properly accounted for in the laboratory testing and per- manent deformation model development. Load repetition factors (Alpha factors) were calculated for NAPTF test data through a least squares quadratic curve fit procedure for four- and six-wheel configurations. Alpha factors at 10,000 coverages were compared with the International Civil Aviation Organization standard for computing Aircraft Classification Number, showing con- sistent relationships with the existing alpha factor of 0.825 for four-wheel gears. However, the six-wheel alpha fac- tor at 10,000 coverages should be changed from 0.788 to a value approximately equal to the interim value of 0.72 (Hayhoe 2006). Improvements in the Department of Defense’s flexible pavement design procedure put a renewed emphasis on the design and construction of contingency pavements prompt- ing concerns regarding the design and evaluation of thin HMA pavements with minimum HMA thickness require- ments. Bell (2008) evaluated the effects of operating cargo and fighter aircraft on representative thin HMA pavements using a C-17 and an F-15E tire in an f-sAPT evaluation. The principle failure mechanisms included rutting, polished

85 aggregate, and surface cracking, with most of the failure gradu- ally appearing in the subgrade. The outcome indicated that the Department of Defense’s minimum HMA thickness cri- terion was more than adequate with sections trafficked to an equivalent of a C-17 aircraft movement on an airfield every 3 days for 20 years or six movements of an F-15E aircraft on an airfield every day for 20 years. Leahy et al. (2008) evaluated the use of M-E analyses to predict the performance of the air-side HMA pavements at the New Doha International Airport. Data from the RSST- CH, AASHTO T-320, and mechanistic pavement analyses were used to calculate the expected rutting performance. These rutting depth estimation procedures were developed during the WesTrack f-sAPT program. Results suggest that the proposed mixture design should provide adequate rut- ting resistance and that the overall pavement thickness is adequate. The concentrated effects of high tire pressure in the sur- face layers of the pavement structure, specifically on HMA surfaces at elevated temperatures, are being evaluated by the FAA to develop HMA mixes that can withstand these aggres- sive loading conditions (APWGM proceedings 2009). Load equivalence Factor development Load equivalence factors (LEFs) represent the ratio of the number of repetitions of any axle load and configuration to the number of applications of the standard 80-kN single-axle load necessary to cause a specified reduction in serviceability. The ratio between the AASHTO LEF for any two-axle loads of the same configuration is thought to be approximated by the 4th power law. When dealing with f-sAPT, it should be appreciated that there exists a difference between f-sAPT and in-service performance not only because of load differences but also because of the effects of the environment, age, and traffic mix. Most of the references on development of LEFs based on f-sAPT indicated concern about the range of values and the effect of various pavement and material parameters on the actual value for specific cases. Dawson (2008) found that the use of the 4th Power Law cannot be recommended for predicting rutting development as a function of the number of load applications and the load magnitude for many low-volume pavements in different cli- mates. Further, if the behavior of pavement materials is well understood, it makes power law response unlikely as materials often have a steady-state response at lower stress levels. Real pavement response cannot be modeled using the power rela- tionship as there often appears to be a threshold stress above which large deformations result rapidly that cannot be related to the responses at lower stress levels. It is found internation- ally that even in controlled experiments the power law expo- nent varies widely as a function of material properties and pavement layer structure. Chen and Shiah (2001) compared data from two test roads with f-sAPT data to evaluate the effect of heavy traf- fic loading on pavement distress. Based on Present Ser- viceability Index loss, a power law exponent of 8 existed for the ALF, in contrast to the 4th Power Law in full-scale test roads. This could explain why pavements tested by the ALF failed much faster. Critical loads appeared to be pres- ent for pavements tested by the ALF with pavements tested beyond the critical load failing predominantly because of traffic loading. The rut-based LEFs for the full-scale test road were closely correlated with AASHTO equivalencies, while the rut-based ALF LEF is much higher, which may be attributed to the slower speed at which the ALF oper- ates. Crack-based LEFs were closely correlated, most prob- ably because the terminal level of 20% cracking is reached after the terminal levels for rutting and serviceability are reached. Chen et al. (2004) conducted a study to validate and improve the VESYS 5 pavement performance model using f-sAPT results, as well as AASHO data in the calibration. Close agreement was found among the predictions from VESYS 5, ABAQUS, and field rutting measurements from the f-sAPT results. Overload damage relationships were sub- sequently generated from the calibrated VESYS 5 model. The AASHO 4th Power Law relationship was shown to be valid for Pavement Serviceability Index as a pavement per- formance measure but not for rutting, where power values between 1.5 and 2 were found to be more appropriate. It appears as if load equivalency factors for thicker pavements do not follow the 4th Power Law. Yeo et al. (2004, 2006) evaluated the effects of increased axle loads on the performance of thin-surfaced unbound granu- lar pavements. The results of f-sAPT showed the lower quality material to experience much higher deformation and roughness progression compared with the high-quality materials. Load damage exponents (LDEs) were derived for each test pavement based on load cycles to equivalent surface rutting. These values ranged between 2.0 and 4.4 depending on the material quality. Chen et al. (2006) calculated LDEs as a function of structural number, load cycles, and load levels for f-sAPT data on pavement structures with four subgrade types and three moisture conditions originating from USACE-CRREL. These effects were determined in terms of LDEs as a func- tion of structural numbers, cycles of load, and load levels. Four test pavements had LDEs greater than 7.8, while three other pavements had LDEs greater than 4. The high LDEs were attributed to low structural capacity (structural num- bers ranged between 1 and 2.6) of the test pavements. The 4th power rule may thus underestimate rutting damage for weak pavement structures. It was also observed that, based on available data, the LDE increased for an AASHTO clas- sification of A-2-4 and A-4 soils with increasing moisture content.

86 uni- and Bi-directional Loading Instability or near-surface rutting of HMA surfacings is a costly form of distress. Novak et al. (2004) evaluated the nature of instability rutting observed under the HVS and spe- cifically the response of various mixtures to rutting under one-way and two-way directional loading conditions. It was demonstrated that there are differences in rutting depth pro- gression between these two modes of operation. A reversal of shear stress pattern occurs in two-way directional loading not seen in one-way directional loading. Although transverse shear stress patterns have no stress reversal (and thus do not differ between one-way and two-way directional loading), longitudinal shear stress patterns do differ between one-way and two-way directional loading. The lack of shear stress pattern reversal (one-way loading) produces greater strain in a visco-elastic material, even with a greater relaxation time compared with one with shear pattern reversal and less relax- ation time (two-way loading). Instability rutting is manifested through lateral deformation in the transverse plane. The observed degree of rutting between one-way and two-way HVS directional loading suggest that nontransverse stresses play a role in instability rutting propagation. A realistic f-sAPT simulation of actual in-service load- ing is essential to obtain accurate pavement responses. Uni- directional loading without any wheel wander is generally used in rutting evaluation using the HVS. Although this load- ing configuration is thought to be more efficient, its effective- ness and appropriateness to simulate actual in-service truck loading remains unclear. Choubane et al. (2004) evaluated the different possible loading combinations for f-sAPT using an HVS with the intent of determining a more realistic APT simulation of actual in-service loading. When wheel wan- der was not considered, the rutting developed approximately 65% faster in the uni-directional than in the bi-directional mode. However, when the wheel wander was not considered, the uni-directional mode appeared to place considerable wear- ing forces on both the tire and the pavement with as much as 25% of the tread depth worn away at localized locations on the tire. Each of the various wander increments considered affected the tire–pavement contact differently across the test track width when wheel wander was considered. The findings illustrated the importance of using wheel wander for more realistic and meaningful results in rutting testing. Wheel Wander FDOT evaluated various loading conditions under an f-sAPT device and found that when wheel wander was not consid- ered, significantly faster rutting development was observed in the uni-directional than in the bi-directional loading mode. When wheel wander was not considered, considerable stresses were placed on both the tire and the pavement. The tire tread pattern had an impact on the pavement deformation patterns with uni-directional loading forming a pattern on the Chen et al. (2008) used a data mining technique to com- pute rut-based LDEs through analysis of data from seven tests conducted using the CRREL HVS and one test from the Texas MLS. A rutting prediction equation using wheel load, load repetitions, and pavement structural number was devel- oped and the LDEs computed. They were found to decrease with increasing structural number values, demonstrating that overload has a more pronounced effect on the rutting devel- opment of weaker pavements than stronger pavements. Guler and Madanat (2011) developed a hazard rate func- tion with data from the AASHO road test, showing that the damage exponent for fatigue crack initiation is significantly higher than the power typically assumed. Studies on marginal cost pricing typically rely on the assumption that the mar- ginal cost prices should be based on axle loads raised to the power of 4. This power is appropriate if the agency bases its maintenance and rehabilitation decisions on serviceability. However, if the agency uses cracking as a basis, the power should be between 8 and 8.5. Erroneous selection of the load damage exponent (too low in this case) would lead to subsidies of heavier axle vehicles by the lighter ones. Long-term monitoring Various LTM studies are conducted outside the U.S. SHRP LTPP program. It has often been stated that ideally the data obtained from f-sAPT should be calibrated for real traffic and environment through LTM data. Of the respondents to the questionnaire, 36% indicated that they incorporate LTPP/ LTM data in some form in their f-sAPT analyses. No explicit references focusing on this aspect could be identified. This may be attributed to the long-term nature of LTM data collec- tion and the duration of time required before adequate data have been collected to analyze and publish. Jones et al. (2004) developed a protocol to standardize the methodology used for establishing and monitoring LTM sec- tions in conjunction with APT sections. The protocol focuses on issues such as the justification for the LTM sections, sus- tainable funding, section location layout and marking, data collection and instrument installation, laboratory and field testing requirements, and monitoring standards. A monitor- ing plan is discussed together with reporting criteria. Typical data requirements include visual assessments, profile mea- surements, density and moisture measurements, DCP logs, general environmental data, and deflection data. A dedicated traffic monitoring plan is also required. F-saPt Load modes The load mode employed during f-sAPT affects the outcome of the specific test. In this section, specific attention is given to the direction, wander, speed, and temperature effects from various f-sAPT devices.

87 3.4, and 5 mph (0.3, 0.5, 0.8, 1.1, 2.2, 5.4, and 7.9 km/h)], with dual-wheel configurations at wheel loads of 24, 30, and 36 kip (106.8, 133.5, and 160.2 kN) and tire pressures of 200 psi (1 378 kPa) at the NAPTF. The HMA was located on top of a stabilized base. Measurements were made at asphalt temperatures of 52°F (11.1°C) and 72°F (22.2°C). Significant permanent deformations were found in the mea- surements, and the strains varied strongly with temperature and test speed (between 300 and 2,000 microstrains). Slower load application resulted in reduced asphalt stiffness, and increases in the amount of viscous flow and total strain within the HMA mix. Powell (2008a) described a new method of characterizing traffic (load-temperature spectra) used for weighting traffic as a function of pavement temperature. Axle loads are banded in accordance with high temperatures in the performance grading system for binders and a weight factor is developed for each band, reflecting increased rutting potential at higher temperatures. A distinct model is devel- oped for various laboratory tests to predict field performance at an empirically determined single pavement age resulting from the application of banded, weighted traffic. Finally, a time-dependent shift factor is developed to change the model output to predict rutting performance at all other ages. Pow- ell (2008b) applied the model when comparing rutting per- formance findings from HVS tests with similar findings from NCAT. A direct comparison of the NCAT and HVS data, using the load-temperature spectra method, showed gener- ally good agreement between the different experiments with each pass on the NCAT track equivalent to 1.1 HVS wheel passes. The reduced speed of the HVS load wheel [6.2 mph (10 km/h) versus 43.5 mph (70 km/h) for the NCAT trucks] supports the finding. Robbins and Timm (2008) evaluated methods to improve pavement response accuracy through quantification of the effects of vehicle speed and pavement temperatures on pave- ment response. Increasing the rate of loading (vehicle speed) caused a substantial reduction in strain levels with strain rate reduction being more sensitive to vehicle speeds at warmer temperatures. The tensile strain at the bottom of the HMA layer was found to be proportional to the natural logarithm of the vehicle speed, with the mid-depth pavement temperature correlating best with the induced tensile strain. Increasing the mid-depth temperature resulted in exponential eleva- tions in the tensile strain induced. Regression equations were developed for the pavement test section to predict strains for various temperatures and speeds, enabling a comparison to a laboratory threshold level of 100 µε. It indicated that the crit- ical mid-depth pavement temperature for vehicle speeds of 87 mph (104 km/h) or less was approximately 79°F (26°C). The application of f-sAPT is sometimes deemed limited as factors such as the loading speed are not widely repro- duced. Traffic speed on real roads is higher [generally 50 mph (80 km/h)] than that used to perform f-sAPT [less than 7.5 mph (12 km/h)] and owing to the visco-elasticity of HMA mixes, fatigue cracks appear quite rapidly in f-sAPT because of low speeds. To apply f-sAPT results on HMA to real roads, a formulation to estimate the influence of loading speed is surfacing matching the general tire tread pattern, both under channelized and wandering traffic (Tia et al. 2003). Garcia and Thompson (2008) evaluated the effect of load- ing speed on longitudinal and transverse tensile strain pulses measured in a HMA section of the University of Illinois ATLAS facility. The speed of loading was between 1.9 and 10 mph (3 and 16 km/h) and it is known that frequency sig- nificantly influences HMA performance. For realistic evalu- ations it is necessary to simulate the real loading frequency. The CalME relationship for estimating vertical stress pulse durations was successfully used to estimate the pulse dura- tions in the f-sAPT, and it was found that the transverse strain pulse durations were about three times those in the longitu- dinal direction. Subsequently, the transverse tensile strains were approximately 1.5 times greater than those in the lon- gitudinal direction and the transverse pulse was also more sensitive to the lateral position of the moving load. If con- siderable moving load wander is expected, the longitudinal strain pulse should be considered the most critical, because the probability that the transverse strain pulse is greater is low, and vice versa for channelized conditions. F-sAPT using HVS to evaluate rutting performance of different HMA mixtures are typically conducted at elevated temperatures and with channelized traffic, although actual highway traffic always has a certain amount of wander. Under- standing the effect of traffic wander on rutting performance of HMA under HVS loading has been evaluated using FEM of rutting development on a flexible pavement (Wu and Harvey 2008). An elasto-plastic constitutive model based on a bounding surface concept was developed to describe plastic deformation of HMA under shearing. During calibra- tion of the model with RSST-CH test data it was found that the model can describe the accumulation of plastic strain of HMA in RSST-CH tests with high accuracy. A FEM analysis was used to simulate rutting development in flexible pave- ments with HMA surfacing. The effect of traffic wander on rutting performance was evaluated by comparing simulations of a HVS rutting test under both wandering and channelized traffic. The FEM simulation results were found to agree very well with typical observations under the HVS. Discrepan- cies between the two can be explained by the lack of plastic volumetric compression in the material model, and the actual values of the calculated total rutting were found to agree very well with measured data from HVS tests. Allowing HVS traffic to wander decreases the amount of total rutting consis- tently by approximately 56% (for the structure analyzed) and it is recommended that channelized traffic should be used for conducting HVS tests because it allows the same qualitative answer to be obtained much faster. speed Garg and Hayhoe (2001) described asphalt strain data mea- sured at the bottom of a 5-in. (125-mm)-thick HMA surface layer obtained at a range of speeds [0.2, 0.3, 0.5, 0.7, 1.4,

88 the MnROAD facility under crack mouth opening displace- ment control using Semi-Circular Bend specimens obtained from Superpave Gyratory Compactor cylindrical samples. The results showed that neither the binder type nor the speci- men location were significant for the stiffness calculated as the slope of the load-displacement curve, whereas the binder type was significant for the fracture toughness and fracture energy. The Ministry of Transport Quebec (MTQ) and LCPC under- took a project with the objective of optimizing their frost- thaw pavement design method (Savard et al. 2004). Four test sections were constructed, representing an HMA base and a cement-treated base. The thermally un-insulated sections were designed to obtain damage within three years, while the insulated sections disassociate the effect of traffic from the thaw period loss of bearing capacity. A good correlation was obtained between the in situ f-sAPT and laboratory tests as well as the direct and indirect load strain measurements at the bottom of the pavements. FWD data showed a smaller varia- tion in the thaw period subgrade modulus than the quasi- static tests, while the performance prediction models were validated against the measured deformation. Kubo et al. (2006) and Kawakami and Kubo (2008) evalu- ated the urban heat island effect caused by street pavement temperatures and measures to decrease the effect through cool pavements that reflect solar radiation and retain water within the pavement materials. F-sAPT of the pavement structure was performed to evaluate the durability of the mixture-type heat-shield pavement. Temperature monitor- ing showed that the water retention pavements and the heat shield pavements reduced their surface temperatures by up to 68°F (20°C). Measured rutting was lower than that of drain- age and dense-graded pavements, as the HMA temperature was suppressed because of the technology incorporated into the HMA design. Limited cracking occurred during traffick- ing and the in situ permeability remained good. The project proved that the mixture-type heat-shield pavement, when applied to actual roadways, is durable enough to withstand typical traffic loads for the application. Marasteanu et al. (2007) reported on a pooled fund study on low temperature cracking in HMA pavements in the northern part of the United States and Canada. The pre- dominant failure mode observed on HMA in these areas is cracking as a result of high thermal stresses developing at low temperatures. Both traditional and novel experimental protocols and analyses were applied to a set of laboratory prepared and field samples obtained from pavements with well-documented performance to determine the best options for improving low temperature fracture resistance of HMA pavements. Tests used included creep, strength for asphalt binders and mixtures, disk compact tension test, single-edge notched beam test, and semicircular bend test. The coeffi- cients of thermal contraction of HMA samples were measured using dilatometric measurements and discrete fracture and required. Theisen et al. (2009) developed a method to do this by employing visco-elasticity theory and Schapery’s work potential theory using experimental data from a HVS traffic simulator. Adaptation of Schapery’s work potential theory formulation was developed and calibrated using experimental data and the effect of loading speed was estimated to speeds ranging from 2.5 to 50 mph (4 to 80 km/h). This showed results qualitatively similar to those observed in real roads. F-sAPT devices accelerate pavement distress by applying high axle loads at high loading frequency, while the effects of high axle loads are generally computed using load factors. However, linear traffic simulators apply loads to pavements at very low speeds [generally lower than 6 mph (10 km/h)], which accelerate pavement degradation, especially on thick HMA layers. Núñez et al. (2008) developed a simplified approach to apply f-sAPT results to real pavements where trucks apply loads at speeds ten times higher than the f-sAPT device. Using Van der Poel’s nomograph, the HMA binder stiffness moduli corresponding to both loading times may be estimated. Once the HMA binder stiffness moduli at both loading speeds are estimated, appropriate models can be used to estimate the corresponding HMA mixture stiffness moduli. The tensile strains at the bottom of the HMA layer may be computed. Finally, the HMA mixture fatigue life may be esti- mated using appropriate transfer functions. It was shown that reducing the loading speed from 50 mph (80 km/h) (trucks) to 5 mph (8 km/h) (f-sAPT) can accelerate fatigue cracking by 1.86 times. The distress acceleration resulting from high axle loads must be taken into account using load equivalence fac- tors. A 36 kip (160 kN) axle load applied by the f-sAPT device is equivalent to 18.2 equivalent single-axle loads (using AAS- HTO load factors). Therefore the combined effect of high axle loads and low loading speed results in a global acceleration factor of approximately 34, and the HMA layer fatigue life, when submitted to real-time loading, may be computed as: Nreal traffic = Loading speed × Load equivalence factor × Number of f-sAPT load applications. Evaluation of a test on an instrumented f-sAPT section at the German Federal Highway Research Institute [Bundes- anstalt für Straßenwesen (BASt)] (Rabe 2008) provides infor- mation on the basic mechanical behavior of a representative selection of pavements of different strength constructed with different materials. It was observed that vehicle speed has a decisive influence on the stiffness of HMA layers evaluated and therefore on the stress/strain level inside both the HMA and unbound materials. enVIronmentaL Parameters temperature Low temperature cracking is the main distress in HMA pave- ments in the northern United States and Canada. Li et al. (2004) described the evaluation of three HMA mixes used in

89 with virgin binder and aggregate materials sampled during construction at NCAT. The Hirsch model had the tendency to underestimate the binder G* at higher test temperatures [70°F and 115°F (21.1°C and 46.1°C)]. However, the Hirsch model was calibrated based on the E* data measured using the laboratory prepared mixtures, and the mixtures used in this valuation were plant-produced. Both the un-aged and Rolling Thin Film Oven Test-aged binder physical properties are required to determine the binder critical high temperature. The Hirsch model then needs to be calibrated to backcalculate the un-aged binder physical properties. For the RAP mixture design, the binder critical high temperature can be deter- mined solely based on the Rolling Thin Film Oven Test-aged binder physical properties. Evaluating the model showed that the procedure could be used to backcalculate the binder criti- cal high temperature. The Hirsch model may however need to be calibrated for local materials to reduce the backcalcu- lation errors. Leiva-Villacorta and Timm (2011) validated theoretical strain-temperature curves developed with LET with f-sAPT measured strain-temperature relationships from NCAT using LET. The results confirmed that relationships between predicted strains and mid-depth temperature can be expressed with an exponential function. Comparing differ- ent structural sections confirmed that thicker sections had lower strain levels, while mixes with stiffer binder grades had lower strain levels. Relatively soft mixes were found to be more thermally susceptible. Final analysis indicated that in the some cases LET typically overestimated pavement response at intermediate to high temperatures and underestimated it at low to intermediate temperatures. HMA cooling rates provide important information that can help with planning operations and field decisions made during construction. MultiCool is a program that predicts cooling rates in an HMA mat during construction based on information related to start time of the paving operation, envi- ronmental conditions, existing surface, and mixture specifi- cations. Vargas-Nordcbeck and Timm (2011) validated the cooling rates calculated by MultiCool for nonconventional mixes such as WMA, mixtures containing high percentages of RAP, and mixtures containing modified and alternative binders based on NCAT data. No evidence was found that the difference between the measured and predicted cooling curves over the entire pavement structure exceeded the 50°F (10°C) tolerance allowed. Factors such as mixture tempera- ture, RAP content, binder type, and pavement lift were shown to have a significant effect on model fit. No need exists to adjust the existing cooling curve model in MultiCool for the nonconventional mixtures. moisture Caltrans required inclusion of a 3 in. (75 mm) layer of asphalt-treated permeable base (ATPB) between the HMA and aggregate base layers of all new flexible pavements with the purpose of intercepting water entering the pave- ment resulting from high permeability, and transport it away damage tools utilized to model crack initiation and propaga- tion in the pavement systems. FEM and the TCMODEL were used to predict performance of the laboratory samples and compare it to the field performance data. Robbins and Timm (2008) evaluated the temperature effects of a perpetual pavement at NCAT. Temperature probes captured the full temperature profile of the structure. A gen- eral increase in strain resulted from increasing pavement temperatures, while the mid-depth pavement temperature correlated the best with the induced tensile strain. Increasing the mid-depth temperature resulted in drastic elevations in the tensile strain induced. Control and management of the temperature of a HMA layer in a f-sAPT test section is important to ensure that correct inferences are drawn from the response of the layer during testing (Steyn and Denneman 2008). The effect of temperature on the loading conditions of the test (specifi- cally through the changes in tire inflation pressure and resul- tant contact stresses), as well as during the measurement of pavement response parameters is important. When conduct- ing temperature-controlled f-sAPT, actual field conditions should be incorporated into the planning of the loading con- ditions, HMA layer condition, and test plan to ensure that outputs are applicable to the desired application. Velasquez et al. (2008) developed regression models to predict flexible pavement temperature profiles based on mea- sured values for air temperature, humidity, wind speed, and calculated solar radiation at the MnROAD facility. Binder type did not affect the prediction of the maximum and mini- mum pavement temperature strongly. Verification of the models with a finite difference heat transfer program and data from five locations in the United States indicated that the predicted temperatures of the numerical model agreed reasonably well with the predictions from both maximum and minimum temperature regression models. Wind speed was an important factor, especially at shallow depths. Herb et al. (2009) analyzed the HMA pavement tempera- ture from MnROAD data and conducted simulations using a 1-dimensional finite difference heat transfer model, char- acterizing the diurnal and seasonal changes in pavement temperature. Information was obtained on temperature gra- dients inside HMA layers that assist in the characterization and analysis of temperature-related phenomena in the HMA sections. AASHTO M 323 indicates that a blending chart should be used to determine either how much RAP can be added or which virgin binder to use when working with high RAP contents. For this process, the critical temperatures of the recovered RAP binder need to be determined. Tran et al. (2010) presented a new approach of using predictive mod- els to backcalculate the critical temperatures of the blended binder from mixture properties, based on HMA mixtures

90 changes in the moisture content as well as the potential out- come of the changes in moisture content to be modeled were stressed. During the tests moisture changes need to be moni- tored and logged to ensure accurate data analysis. The use of appropriate control sections without moisture changes are required to ensure comparison of behavior. Erlingsson (2008, 2010a, b) investigated the response behavior and performance of a commonly used pavement structure in Sweden through f-sAPT. It was shown that rais- ing the groundwater table during a test increased the rutting rate in all unbound layers. It further increased both the resil- ient and the permanent strain of all unbound layers above the groundwater table, probably because of increased moisture content in the unbound layers. NCAT compared the construction and performance of per- meable surface mixes containing two different aggregate sources that were placed with conventional and dual layer paving equipment on perpetual foundations (Powell 2009). One section contained cubical aggregates and the other more flat and elongated aggregates. The third section was built using the same slightly flat and elongated aggregates and a surface consisting of a thin 0.4 in. (9.5 mm) drainable mix- ture over a thicker 0.5 in. (12.5 mm) drainable mix. The field drainability exhibited by the section constructed with slightly flat and elongated aggregates appeared to be better than the section built with the more cubical aggregates after prolonged rain, while the field drainability exhibited by the dual layer pavement appeared to be better than the conventional single layer drainable surface. Schaefer et al. (2010) described the construction and per- formance of a pervious concrete overlay constructed at the MnROAD low-volume road. The overlay performed well with only localized surface pavement distress and good per- formance and durability. The pavement provided good miti- gation of splash and spray and reduction of hydroplaning through good flow characteristics. Twelve sets of full-scale flexible pavement test sections were built inside the Frost Effects Research Facility of the U.S. Army Corps of Engineers to develop subgrade failure criteria and performance prediction models that consider the subgrade soil type and moisture condition (Cortez et al. 2007). The pavement materials and layer thickness were kept constant, while the subgrade soil type and moisture condition varied from test section to test section. The experimental data suggest that for subgrade soil type AASHTO A-2-4 and up to 15% moisture content moisture above the conventional opti- mum moisture content was beneficial in reducing permanent deformation. For soil types A-4 and A-6, moisture in excess of optimum had a weakening effect that gradually increased the rutting rate. For soil type A-7-5, the effect of moisture in excess of optimum was initially negligible; however, above certain moisture contents it resulted in sharp increases in rut- ting. With low-plasticity subgrade soils, such as types A-4 from the pavement before it reaches the unbound materials. Bejarano et al. (2003, 2004a, b, c) summarized results from a study using f-sAPT to evaluate the performance of drained and undrained flexible pavements under wet (saturated base) conditions. A drained structure is a pavement section that contains an ATPB layer between the HMA and the aggregate base, whereas an undrained structure does not contain an ATPB layer. It was observed that the ATPB placed between the HMA and base stripped in the presence of water and heavy loading. Further, clogging of the ATPB with fines from the aggregate base was observed in the wheel path area. Surface rutting was the prominent failure mode because of stripping of the ATPB. For the undrained sections, fatigue cracking was the predominant failure mode. Improved compaction of the HMA layer and the use of improved structural section design suggest that consideration should be given to the elimination of the ATPB directly beneath the HMA layer. The detrimental impact of the moisture intrusion into base and subgrade layers of pavements is well-quantified based on laboratory tests and field studies. A series of tests were car- ried out on small-scale specimens [40 in. (1 m) in diameter] to quantify the impact of moisture on one base and two types of subgrade using a new test set-up. The results were com- pared with those from two full-scale test sections made from the same materials (Amiri 2006). It was shown that if the small-scale specimens are carefully constructed to achieve similar densities and moisture contents to the pavement sec- tions, the load-deformation responses are reasonably close, indicating that the small-scale tests can be effectively utilized to calibrate and validate the existing models under different conditions. For clayey subgrades, the moisture in the sub- grade dominates the performance of the pavement, whereas for the sandy subgrades the moisture conditions within the base, as well as the subgrade, influence on the behavior of the pavement. The test pit at LCPC’s f-sAPT facility was filled with clayey-silty sand with an elastic modulus of between 12.3 and 15.9 psi (85 and 110 MPa), depending on moisture con- tent. Four different flexible pavements and one bituminous reference structure were constructed on top of the clayey- silty sand and tested to generate information for improving the understanding of subgrade moisture content effects on flexible pavement design. It was observed that the rutting of the unbound granular subgrade significantly contributed to the overall rutting development. Although the evaluation of the behavior using the current French flexible pavement design method provided satisfactory results, it underesti- mated the rut resistance of the HMA surfacing (Balay and Kerzreho 2008). Steyn and Du Plessis (2008) discussed the objectives, potential effects, and available methods for performing HVS tests on pavements where the moisture condition of the pave- ment is a major parameter in the experimental design test matrix. The importance of clearly defining the objective of

91 exists for evaluating the effects of climate change on pave- ments through the judicious application of artificial tem- perature and moisture changes (based on expected weather conditions) during f-sAPT. The major effects of tire contact stresses and loading conditions on pavement response were highlighted by many researchers and are shown to be incorporated as a factor in many of the test programs evaluated for this synthesis. Improved measurement systems as well as novel analysis techniques for incorporating actual tire–pavement contact stresses into analyses allows for an improved understand- ing and appreciation of this parameter. The loading effects caused by wide-based and also aircraft tires are being evaluated at various f-sAPT facilities. A limited number of facilities incorporate LTM of existing pavements into their research, although it is appreciated that it is an important link between f-sAPT and real life data. and A-6, pumping can occur under repeated loads. It has been recognized that the most important mechanism produc- ing rutting in unbound soil materials is shear distortion and not densification. The relative contribution of the various pavement layers to the total vertical rutting varies accord- ing to subgrade soil type and moisture condition. The data suggest that the percentage deformation in the surfacing and base layers increases with stronger subgrades. chaPter summary This chapter focused on the vehicle–pavement–environment interaction that is vital for a complete understanding of f-sAPT. Respondents view pavement and ambient tem- perature as the most important environmental parameters to relate f-sAPT data with and to control during tests. This is probably related to the high percentage of HMA-type f-sAPT evaluations conducted (chapter three). Potential