Significant Findings from Full-Scale Accelerated Pavement Testing (2004)

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.

19 CHAPTER THREE VEHICLE–PAVEMENT–ENVIRONMENT INTERACTION INTRODUCTION The topic of vehicle–pavement–environment interaction is complex and probably one of the most controversial as- pects of APT. Croney and Croney (1991) noted that age strengthening appears to be a major factor in the perform- ance of well-designed pavements. Nunn (1997) also re- ported on this phenomenon. These effects were found to be the result of changes in the asphalt over time, which lent support to the statement by Croney and Croney (1991) that the environmental limitations placed APT in jeopardy. Dif- ferences between loading used during APT studies and conventional traffic have also led to questioning of the ap- plicability of APT findings to the performance of conven- tional in-service highways. However, during the last decade, advances have been made toward a better understanding of these phenomena. The results from research by a number of APT programs will provide evidence of this. Naturally, there are differ- ences in the extent to which the various programs include these issues in their test plans. In subsequent chapters some case studies will be presented to illustrate how this affects design and construction, which should help APT users at- large and DOTs to gain confidence in using and applying the various findings from APT programs. Before discussing the wide variety of applications and relevant issues that were found in the literature, the results from a synthesis of the questionnaire survey is presented here. QUESTIONNAIRE SURVEY The responses to Questions 3.1 to 3.4 on loading and envi- ronment are shown in Figures C27 to C30 in Appendix C. These responses were synthesized, and the results are con- tained in the following list: • The primary load characteristic to which perform- ance is related is the wheel load and whether it is ap- plied with or without lateral wander (Figure C27). • Of almost equal importance to load is tire pressure. • Temperature (both pavement and air) is the primary environmental element that has been related to per- formance (Figure C28). • The environmental condition that is controlled most is pavement and air temperature (Figure C29). • Of almost equal frequency of control is the subgrade moisture with related factors such as the water table and drainage. • Most tests were conducted at moderate temperatures (>10°C and <40°C) (Figure C30). Views of respondents to the survey on loading and envi- ronment are presented in Table D3 in Appendix D. ELEMENTS OF THE VEHICLE–PAVEMENT–ENVIRONMENT SYSTEM The structural configuration of the pavement system is normally fixed by design or policy, including the materials that are to be used. The structural system is then subjected to the impact of traffic loading under the prevailing envi- ronmental conditions that affect its performance. The response and performance of the pavement system is therefore subject to an array of influential factors that have variable levels of control and are time dependent to a greater or lesser degree. The following need to be consid- ered: • Pavement materials; • Trafficking comprising – wheel loads that can be single axle or multiple axle, – wheel load(s) that can be static or dynamic, – wheel loads that wander laterally, – suspension systems, – tire pressure/contact stress, – tire type, and – speed; • Environmental impact of – wind and radiation, – temperature, and – water in a variety of forms. Clearly, the performance of the pavement is dependent on the interaction of these factors. APT programs have proven to be invaluable because of the ability to partially or fully control these factors. The synthesis of the findings from the various APT programs provides valuable insight and some useful applications, which will be considered in terms of the respective factors. The effect of materials will be considered in conjunction with the respective factors per- taining to traffic and the environment, where appropriate.

20 Chapter four discusses the relationship between APT and materials and tests comprehensively. TRAFFICKING APT programs have used both conventional trucks and a variety of vehicles for simulating conventional trafficking of pavements. The wheel loads are applied at selected static levels that may be varied according to the needs of the experiment, with or without some type of suspension. In some cases the suspension or even the undercarriage is the same as that used in conventional trucks. In addition, the devices have used wheel loads that vary from the con- ventional truck wheel loading to aircraft loading. The tire pressures have also been varied to accommodate both con- ventional and extraordinary tire pressures to explore the impact of this variable. Because of the range of variables, the findings from the different programs need to be care- fully scrutinized to determine to what extent the results are comparable. Some protocols have been established and these have provided a means of limited comparison of the respective test programs. Details of the respective APT trafficking devices in use through the mid-1990s can be found in NCHRP Synthesis of Highway Practice 235 (Metcalf 1996). It provides an excellent overview of the trafficking devices that had been used up to that time. Some details of the latest additions to the APT scene, such as trafficking systems, mechanical functioning, loading characteristics, and test plans, are given in chapters one and eight and in Appendix E of this synthesis. In this section, the focus will be on aspects of APT trafficking that affected the performance of the pave- ments in experimental and related analytical studies. Wheel Load Intensity and Load Equivalency Wheel load intensity has by far the most profound effect on pavement performance and yet it is, to a large extent, an un- controlled variable in real traffic. This explains the interest in the well-known “fourth power law” that dates back to the AASHO road test (AASHO 1961). Understandably, APT pro- grams are increasingly focusing on this aspect as test capabili- ties improve in their sophistication. It had been reported that the relationship of load with performance was neither con- stant nor linear. The so-called fourth power law was found to be exponential and highly dependent on the thickness and configuration of the layer(s). In addition, it depends on whether the axle has a suspension, on the type of suspen- sion, and on the degree of smoothness of the pavement sur- face. Some relevant studies are discussed here. In 1989, the Organization for Economic Cooperation and Development (OECD) undertook a full-scale pave- ment test in France at the circular fatigue test track of the LCPC in Nantes. The primary purpose of the test was to assess the relative damage of the maximum legal axle loads of 15 European countries, in comparison with the 11.5-t axle load that was to be the future standard axle load in the European Union (Gramsammer et al. 1999). Because of the many factors that affect performance, Gramsammer et al. (1999) are of the opinion that APT was the only means of coming to a quantitative solution that would be acceptable to all parties concerned. A report on the joint test program that became known internationally as the FORCE project was published in 1991 (OECD 1991a,b). The findings were also discussed at a 1991 conference in La Baule, France (OECD 1992). For the analy- sis of the comparison of the 10- and 11.5-t axle loads, relative damaging effects were calculated in terms of cracking and rut- ting performance, and on the basis of response measure- ments. The damage exponent was determined separately for rutting and cracking. The report concluded that the fourth power law constitutes only a general description and approximation of relative damage owing to axle loads. For the rutting, the thin asphalt (67 mm) flexible pave- ment had an exponent of 5.74. The value for the thick as- phalt (140 mm) pavement was 2.88 and the semi-rigid pavement (170-mm cement-treated gravel base with a 63- mm surface layer of asphalt) had a value of 1.47. The ex- ponent decreased with increasing stiffness and the thick- ness of the bituminous layer. The cracking evaluation was limited to the flexible pavements because of the lack of deterioration in the ce- ment-treated pavement structure. The exponent varied be- tween 1.8 and 9 depending on the degree of deterioration, the criterion used for comparison, and the condition of the pavement at the time of comparison based on cracking of the thin asphalt section. The exponent increased linearly with the degree of cracking up to a level of 400,000 load applications. Thereafter, the rate of increase gradually re- duced. Crack distress was defined as crack length per 100 m (linear) of pavement. The report points out that 25% to 30% of cracking at the FORCE experiment yields an expo- nent varying between 2.7 and 4.3 on the basis of 60–100 m/100 linear meters of an in-service pavement. The neces- sity of qualifying the degree of cracking in the pavement is apparent. It should be noted that there was little deteriora- tion in the cement-treated pavement structure for the num- ber of axle load applications that had been applied at the time. Nor was there any significant cracking in the thick asphalt. The exponent applicable to the semi-rigid pave- ment structure was expected to change as the layer deterio- rates if and when further trafficking is applied. A number of participating countries with APT facilities such as the United States, Spain, Switzerland, and Finland,

21 undertook studies that would serve as “cross-tests” for the tests that were conducted at the LCPC track in France. The purpose of the cross-tests was to promote international co- operation toward more effective use of large and expensive facilities. The intent was, among others, to conduct the tests on the basis of the specifications and procedures used for the tests in France. Axle loads of 10 t and 11.5 t were used with single and dual tires. The results of the various cross-tests did not correspond directly with the findings from the LCPC test track. This was ascribed to differences in pavement configuration, material characteristics, sub- grade strength, climatic conditions, and the type of test. Some examples are discussed here. Kenis and Lord (1992) reported on FHWA participation in a cross-test. The performance of the FHWA ALF test dif- fered significantly from that in France in terms of both rut- ting and cracking. Furthermore, the findings were not con- sistent. In the case of the thin asphalt, the OECD pavement performed poorly when compared with the FHWA test, whereas the reverse was found in the case of the thick as- phalt. In a cross-test conducted in Finland (Pihlajamäki 1992), the thick asphalt and semi-rigid pavements were tested to 7 million applications of a 25-kN single-wheel load without serious distress. Thereafter, subgrade condi- tions were artificially changed to accelerate distress by ad- justing the water table. The first cracks manifested after 11 million applications. Distress was primarily in the form of rutting. The performance was better than the test at the LCPC track, which was terminated after 13.5 million load applications. In Switzerland (Scazziga et al. 1992), the thin asphalt pavement performed as well as the thick asphalt pavement at the LCPC track. One reason for this was thought to be the difference in subgrade bearing capacity. Romero et al. (1992) reported on a cross-test in Spain. The cracking and rutting damage exponents from this cross-test once again differed from the LCPC test results for the thick asphalt pavement test. For example, in the case of the 11.5- t versus the 13-t axle loads, the exponents were greater than 12. In the light of these findings it is understandable why it was concluded in the report that cross-tests were misnomers and that they were more correctly comparative tests on a qualitative basis (OECD 1991). Núñez et al. (1999) reported on APT studies of sound, weathered-basalt gravel base pavements with thin surfac- ings in Brazil. The Load Damage Equivalency exponent was found to be four. Soil moisture was monitored with jet-fill tensiometers. This provided valuable information on the variance of subgrade modulus. This topic is discussed at greater length later in this report. Hugo et al. (1997) reported on tests with the TxMLS in which the damage exponent in terms of rutting of a pave- ment comprising 75 mm of asphalt on a 150-mm cement- treated siliceous gravel base course was found to be as high as 7. Freeme (1984) reported damage exponents that had been determined for rutting in the SA–HVS program for different pavement structures. For pavements with granular layers, the values varied between 2 and 4. For thick asphalt layers, the exponent was 4. Overall, the load- ing exponent varied between 2 and 6, depending on the relative stiffnesses of the pavement layers and the respec- tive pavement layer thicknesses. The findings lead to the identification and subsequent validation by HVS testing of the concept of strength-balance of a pavement structure in terms of depth (Kleyn et al. 1985, 1989). This enhanced the understanding of pavement behavior relative to struc- tural composition and provided a basis for selecting layer composition and configuration when designing pavements. Vuong et al. (1994) reported on Australian studies per- taining to performance of subgrade and base course layers using fine-grained marginal material under accelerated loading. They concluded that excessive deformation of poor sandstone base course caused the loading exponent with respect to differing wheel loads to increase to a value of 7 to 8. As a result, they noted that there was a need to revise the Austroads subgrade critical strain criterion to improve the prediction of pavement life. Suspension/Dynamic Load The fourth power law of load equivalence was derived from experiments in which the applied wheel forces were dynamic (AASHO 1961). However, performance models for asphalt pavement generally included the effects of traf- fic loading on the basis of static wheel loads and numbers of applications; vehicle dynamic effects are not explicitly taken into account. Pidwerbesky et al. (1997b) pointed out that Sweatman (1983) had deduced that there was a degree of spatial repeatability in the dynamic loads that occur un- der wheel loads. As a result, the damage at peak load loca- tions would be the critical factor in pavement performance. This was validated through a study of the influence of dy- namic axle loads on pavement response and deterioration at the Canterbury Accelerated Pavement Testing Indoor Facility (CAPTIF) (Pidwerbesky et al. 1997b). In this study, the effect of three types of suspension, steel multi- leaf, twin parabolic spring, and air suspension, on pave- ment deterioration was investigated in two tests. The wheel loads were identical statically, but quite different dynami- cally. Analysis ranked the suspensions from worst to best as steel, parabolic, and air. In the first test it was found that there was a good corre- lation between the dynamic wheel forces and pavement distress. The results also showed that using a suspension system that is not designed for the load it is carrying could have significant impact on the distress in the pavement. The second test was part of the OECD DIVINE (Dynamic

22 Interaction between Vehicles and Infrastructure Experi- ment) project (OECD 1998). Two suspensions, a steel multi-leaf and a dual air suspension, were compared. The modes and level of pavement distress proved to be depend- ent on the particular suspension characteristics. The air suspension generated dynamic loads less than half those generated by the steel suspension. Surface rutting and lon- gitudinal profiles, as well as surface distress, were moni- tored and analyzed. The difference in pavement condition led the researchers to conclude that the steel spring suspen- sion caused an increase in deterioration rate compared with the airbag suspension. The mean vehicle speed was 45 km/h with the inner and outer Simulated Loading and Vehicle Emulator (SLAVE) units traveling at 43.0 km/h and 47.0 km/h, respectively, on the inner and outer circles. This slight difference between the speeds was not considered to be of concern. Single wide-base tires were used, and the amount of lat- eral vehicle wander was limited to ±100 mm to maximize the separation between the vehicles. Pavement failure was de- fined as maximum rut depth exceeding 25 mm or surface cracking exceeding 5 m/m2 over 50% of the trafficked area. Steven et al. (1999) reported that the two different types of vehicle suspensions (dual air suspension and multi-leaf steel spring) in the CAPTIF tests produced no significant difference in the mean level of pavement wear in terms of surface roughness. However, the variability in the wear is quite different, with higher dynamic wheel loads producing greater variability. Similarly, the average wheel load was the same, but the response of the suspensions differed con- siderably. The authors noted that the findings have impor- tant implications for pavement maintenance requirements. In the same vein, the benefit of a smooth, uniform pave- ment structure is apparent. As stated by Steven and Pidwerbesky (1997) This research provided the first real insights into the effect of suspensions and dynamic loading on pavement life and per- formance; only an accelerated loading facility could provide an affordable means of obtaining the results in controlled ex- perimental conditions within a reasonable time frame. The pavement condition and response changed with increased numbers of applied loads, and varied between the different suspensions. The effect of dynamic loading was clearly dem- onstrated by severe depressions in the pavement at intervals relating to the dynamic characteristics of the different suspen- sion types. Kenis and Wang (1999) focused on one important as- pect of the OECD’s DIVINE program in New Zealand. The goal was to investigate whether it was possible to dis- tinguish between the development of pavement distresses resulting from initial variations in material properties and layer thicknesses, and those resulting from variations in the dynamic wheel forces imposed on pavements due to tire-suspension dynamics. Research was conducted to determine if such differences in the level of these two phenomena, that is, the effects of pavement structural variability and dynamic wheel force on pavement per- formance, were detectable. They concluded that it was. They found that the change in mean Vertical Surface Deformation in both wheel paths with trafficking was similar, thus confirming theoretical assumptions that dynamic loading had no effect on the accumulation of mean surface deformation. However, they did find that there was about a 27% greater standard deviation of the Vertical Surface Deformation in the outer wheel path than the inner wheel path. This was probably because of the increase in the dynamic load in the outer wheel path. Cross-correlation analysis showed that the initial profile had little influence on the final profile of the pavement regardless of the type of suspension. The reliability analysis by Kenis and Wang (1999) re- vealed that reducing structural variability of the pavement because of variability of pavement materials or nonuni- formity of pavement construction increased the reliability of pavement service life. It is apparent that this improved understanding of the vehicle pavement has important pavement economic consequences, because it affects such factors as construction quality control and the long-term performance of the pavement structure. In contrast to the CAPTIF facility, the LCPC carousel maintains a constant load to minimize the dynamic effect. This is the case for most facilities. Recently the HVS IV was put into operation in Florida, with some limited capa- bility to simulate dynamic loads; however, test results have not yet been reported (Steyn et al. 1999). Gautrans Prov- ince in South Africa also acquired an HVS IV+ with en- hanced capacity for dynamic loading. It can apply a sinu- soidal load pattern along the length of the test section with a varying wavelength. The load amplitude can vary ±20%. The TxMLS (Hugo et al. 1991) was also designed to have some dynamic load effect, but thus far this aspect has not been used except for trial purposes. Test facilities using regular truck trafficking also apply dynamic loads and the performance of the pavements reflects this. For example, in the case of WesTrack (Sime and Ashmore 2000), it was possible to gain insight into the effect on user cost of measurable changes in surface smoothness. This is dis- cussed in more detail in chapter eight. Unidirectional/Bidirectional Loading Contradictory findings have been reported on the effects of unidirectional versus bidirectional trafficking. Tests in the United Kingdom (Brown and Brodrick 1999) reported that the bidirectional trafficking system is more severe in terms

23 of relative performance. In one case (50 mm of asphalt on a 160-mm granular base), the rutting more than doubled for two-way traffic. In contrast, Huhtala and Pihlajamaki (2000) reported that Finnish researchers found no differ- ence in performance. Theoretical analyses have clearly in- dicated that a difference in performance should be ex- pected because of residual stresses (Yandell and Behzadi 1999). Galal and White (1999) have also commented on this aspect. In general, trafficking with the Indiana DOT/Purdue–APT was done unidirectionally. In a single comparative test, White (T.D. White, personal communica- tion, 2002) found that unidirectional trafficking exhibited less deformation than bidirectional trafficking. No follow- up tests have yet been conducted. The reason(s) for this apparent anomaly in the findings is not clear, and further research is needed to address this issue. Lateral Wander During Traffic Loading This topic has a bearing on the number of load applications that cause the damage affecting performance. Most devices have the capability to apply lateral wander, but in the inter- est of increased rates of load application on the center of the wheel path, it is often not applied. The effect of such an approach needs to be considered because it affects the rela- tionship between distress and load applications at trans- verse positions in the wheel path. Epps et al. (2001) reported on the effect of lateral wan- der on a number of the test sections at WesTrack. The re- searchers found that the HMA had moved laterally during trafficking and there was evidence that the material could be shoved upwards when the mix was shear susceptible at high temperature. This phenomenon was also observed by Hand (1999). As a result, the rut depth could be reduced at a specific lateral position with an increase in the number of load applications. Accordingly, Epps et al. (2001) con- cluded that the effect of this phenomenon had to be taken into account when quantifying the rutting performance un- der APT trafficking. Chen et al. (1999) and Chen and Lin (1999) compared the typical wheel path rut profiles on a number of highways in Texas to the wheel path profiles generated by the TxMLS dur- ing testing to relate the performance under APT to actual traf- fic on the highways. They determined an adjustment factor based on the respective wheel path widths. White et al. (1999) also reported on the effect of lateral wander. Their extensive tests showed significant reduction in rut depth and upward heave when the load applications were applied with lateral wander. For example, in a spe- cific case, the rut depth and upward heave were respec- tively 10 mm and 5 mm with wander, compared with 15 mm and 14 mm without wander. In general, they reported reductions of between 30% and 40% in rut depth when lat- eral wander normally distributed over a width of 250 mm was introduced. They also showed that FE analysis and modeling could be used to quantitatively predict the APT rutting with wander from single wheel path rutting. This has an advantage over fitting regression curves to data and extrapolating. The latter is purely empirical and relates only to the data collected. Extrapolation may be inappro- priate On the basis of these reports, it is apparent that lateral wander has an important influence on the rutting perform- ance of pavements. Tire Characteristics and Related Contact Stresses Recent developments in the field of vehicle–pavement in- teraction have provided users of APT programs with im- portant insights into the performance of pavements. It has become clear that the tire–pavement interface has a major impact on the performance of the upper layer(s), especially for flexible pavements. De Beer et al. (1997) described the quantification of three-dimensional (3-D) tire/pavement contact stresses for vehicle tires using the Vehicle–Road Surface Pressure Transducer Array (VRSPTA) system (Figure 2). The sys- tem was developed to measure actual contact stresses un- der a moving tire; that is, Stress-In-Motion as an ancillary device for use in conjunction with the South African APT facilities. The contact stresses were found to be nonuni- form and were considered to be a primary cause of surface cracking. Researchers have shown conclusively that the contact pressure is not uniform over the contact surface. Contract pressure varies between the center of the contact area and the outer edge, depending on the tire pressure and the nature of the tire; that is, whether it is a radial ply tire or not. Measured peak contact stresses were up to 100% higher than the inflation pressure (De Beer et al. 1997). Blab (1999) also found that tire inflation pressure, load, FIGURE 2 The VRSPTA on the pavement under the HVS with a 315/80 R22.5 tire (De Beer et al. 1997).

24 and tire type are the dominant factors affecting the 3-D ve- hicle–pavement contact stresses. He concluded that high tire edge vertical contact stresses are the principal factors responsible for surface rutting and cracking. Bonaquist (1992a,b) also reported on work that had been done to assess the impact of wide-based single tires. The comparison between dual and wide-based single tires was done in terms of response and performance relative to fatigue and rutting damage potential. Two thicknesses of AC, 89 and 178 mm, over a 305-mm crushed aggregate base, were trafficked. There was a near linear increase in gross contact area for increasing loads for both tire types. For compression within the AC, tire pressure was found more important than load. However, the tensile strain at the bottom of the asphalt layer and the compressive strain at the base and subgrade level were affected more by the load. This was a good example of Saint Venant’s principle, which states that the stress in the depth of the pavement is solely dependent on the size of the load and not the contact stress. Overall it was found that the wide-based single tire generated 1.0 to 2.4 times more rutting than the dual tires with the majority of the rutting occurring in the crushed aggregate base layer. The observed damage in terms of fa- tigue cracking was approximately four times greater with the single tire compared with the dual tire in terms of measured crack lengths. These values compared well with theoretical predictions based on the following damage models: Fatigue Damage bt f a N 1 )(ε= (1) Layer Rutting drp cN∆=∆ (2) where Nf = fatigue life, εt = tensile strain at the bottom of the AC layer, ∆p = layer permanent deformation, ∆r = resilient layer compression, and a,b,c,d = material coefficients. The expected relative damage according to the models was between 3.5 and 4.3 times greater for fatigue damage, and between 1.1 and 1.5 times greater for rutting damage. Corté et al. (1997) reported on results from experiments on the LCPC test track in France to determine the aggres- siveness of wide-based single tires. It was found that the aggressiveness of the axles with wide-based single wheels is greater than that of the dual wheels. For the sensitive mixes, the magnitude of the ruts formed by the wide-based single tires was 50% more than those of the standard dual tires. In subsequent tests with rut resisting mixes, this difference in rutting performance was reduced to 20%. It was therefore concluded that this effect depended on the nature of the asphalt. The more sensitive the asphalt mix was to rutting, the more pronounced was the effect. The relative effect of a tandem axle with wide-based single tires does not appear significantly different to two isolated wide-based single wheels. However, this observation was made for rather low test temperatures and was therefore not favorable for determining differences in rutting per- formance. Gramsammer et al. (1999) reported that LCPC had also tested new 5.3-t extra-wide single tires currently being de- veloped. As always, the tests were related back to reference materials with known performance to serve as a bench- mark. The tests involved sponsors from the industry, which resulted in fast-tracking the findings into field application by the industrial partners, thus expediting economic bene- fits. Increase in tire pressure affects rutting in APT. Both Gramsammer et al. (1999) and Hugo (2000) found the in- crease in rutting of the asphalt layers to be proportional to the increase in tire pressure. De Beer et al. (1997) reported that tire inflation pressure predominantly controls the vertical contact stresses on the pavement at the tire center, whereas the tire load controls those at the tire edges. Their analysis indicated that during instantaneous overloading/underinflated conditions, the maximum strain energy of distortion (SED) in the asphalt surfacing occurs close to tire edges. Under instantaneous uniform vertical stress conditions, the SED is within the asphalt surfacing at the tire center. In a recent Australian study by Foley and Sharp (2001), pavement deformation beneath wide-based radial and con- ventional radial tires was compared under ALF trafficking. A nominal 75-mm-thick asphalt mix of 25-mm maximum aggregate size, exhibited five times more rutting under the wide-based radial tire than under the dual radial tire at an equal vertical loading. It was concluded that the transverse tensile contact stress developed across individual tread ribs causes a reduction in shear strength because of a reduction in confinement of the material. As a result, the deformation was adversely affected. It was concluded that the contact stresses beneath wide-based radial tires needed to be inves- tigated further. Similarly, the ability of models to simulate these conditions should also be explored further. FE methods were considered to be the most useful for these investigations. Sideways shear or shear forces (stresses) that may de- velop under a condition of cornering were investigated. In South African experiments, the HVS test tire was sheared

25 over the surface of the VRSPTA by moving the tire during testing toward one side of the VRSPTA at an angle of 7.5 to 8 degrees. The test results indicated that the transverse stress increases by approximately 30 kPa/degree for con- ventional Type IV tires and approximately 50 kPa/degree for the wide-based tires. In terms of the ratio of maximum stresses, the transverse ratio increases to a range of be- tween 2.6 and 3.61 (De Beer et al. 1997 and Figure 2). Woodside et al. (1999) also reported on a study of tire– pavement interaction in Northern Ireland. They used a multifunctional wheel tracking device known as TRACKER. Test specimens were tracked by a truck (lorry) tire, at a speed of 2.7 km/h. The stress induced by the tire was varied by altering the tire inflation pressure and the loading. A test platform was used to measure dynamic con- tact stress by means of 12 specially designed, inverted T- shaped transducers positioned across the width of the con- tact patch. Surface texture was simulated by varying the extent to which the T-shaped transducers protrude above the cover plate. The researchers reported that contact stresses increased with tire load, inflation pressure, and surface roughness (texture depth); contact stress distribu- tion is not uniform, but rather increases toward the tire edge. In dynamic stress models, vertical stresses are of higher magnitude than lateral or horizontal stresses. They also showed that mathematical models were ideal for analysis of dynamic contact stress. A quantitative model was developed by Groenendijk et al. (1997) to describe the contact stress distribution meas- ured under tires of LINTRACK in tests in South Africa. This model distinguishes between the edge (2 times 20%) and center (60%) zones of the tire width. The vertical stress level in the center zone is mainly determined by the tire pressure and the edge zone level is mainly determined by the wheel load. In follow-up work, Groenendijk (1998) concluded that the measurement of tire/pavement contact stresses with the VRSPTA had improved the estimation of the vertical, transverse, and longitudinal forces; that is, stresses under slow-moving pneumatic tires. He also pointed out that the surface stresses primarily affect the upper 100 mm of the pavement contributing to surface dis- tress. De Beer et al. (1997) reported on an investigation into some effects of the actual tire pressures on thin asphalt sur- facings. They concluded that the design and analysis of flexible pavements with thin surfacing (<50 mm) were af- fected by surface stresses. The results are significantly dif- ferent compared with those used in conventional design and analysis methods. The ability to evaluate the tire pavement interface in such detail under accelerated pavement test devices could be an important reason why APT programs have become more popular for determining how pavements respond and ultimately perform under traffic. In this regard, the com- puter analysis of Stress-in-Motion using the VRSPTA sys- tem is a powerful tool for enhancing pavement load model- ing for both design and analysis. Trafficking Speed (Speed of Load Application) The effect of trafficking speed has not received as much at- tention as other pavement interaction factors. This is probably because most of the APT systems are either not able to test over a wide range of speeds or are not designed to measure dynamic deflections during trafficking. Nevertheless, some findings have been reported. Corté et al. (1997) noted that, as a first approximation, the trend is that in the range of speeds of 40 to 50 km/h, an increase in speed of 15% reduces rutting by 20% to 35%. They mention that there may have been a veering effect of the wheels during circu- lar trafficking. This value becomes more variable with stiff asphalt (small deformation). Also, when the ruts are small (2–4 mm), it is difficult to quantify the influence of speed. Steven et al. (1999) reported that the dynamic load coeffi- cient, as measured with the CAPTIF device, doubled when a parabolic spring suspension was used and the speed was increased from 20 km/h to 45 km/h. The effect was mini- mal with a multi-leaf spring suspension. Lourens (1995) measured both static and dynamic de- flections under wheel loads on in-service highways to re- late these to deflections measured under the HVS. Figure 3 shows the results on a pavement consisting of 60-mm as- phalt surfacing, 210 mm high-quality crushed-stone aggre- gate base, and 150 mm of cement-treated subbase. He ob- tained similar results with a pavement that had 100-mm asphalt surfacing, but a slightly weaker support structure. The results show how the deflection reduces dramatically within the speed range of 0 to 20 km/h. Because of the re- lationship between elastic and plastic strain this would lead to a similar reduction in deformation of the asphalt. The experiments were done with different axle loads and on a range of pavement structures. The range of strain measurements taken in five different pavement structures under 11.5-t axle loads during the FORCE project (OECD 1991) were very similar to those shown in Figure 3. White et al. (1999) demonstrated the effect of speed on rutting with the FE analysis that had been developed in their APT program. They found that the rutting would be halved when the speed is increased to 100 km/h from creep speed. These findings demonstrate why frequency of load applica- tion is important when comparing deformation under dif- ferent trafficking patterns as was done by Epps et al. (2001). It is clear that speed of loading has to be carefully con- sidered in judging results from APT trafficking, especially

26 FIGURE 3 Effect of speed on pavement deflection (Lourens 1995). when a substantial layer of asphalt is present in the structure. On the other hand, it is apparent that the wide range of vari- ants that have to be considered in pavement performance can most probably only be investigated with any degree of success by using APT. In this regard, Gramsammer et al. (1999) re- ported the following in their presentation on research at the LCPC test facility during the 1999 Reno conference: The Carousel made it possible to optimize the inverted pave- ment structure quickly. . . Could the hierarchical organization of anti-rutting qualities of the materials have been worked out as easily (without the use of APT)? How could the traffic, the speed, and the temperature conditions have been as rigorously guaranteed with another testing facility? LOAD COMPOSITION AND CONFIGURATION (SINGLE OR MULTIPLE AXLES) According to LCPC reports, the effect of multiple axles has not been excessively detrimental (Gramsammer et al. 1999). The phenomenon is however receiving specific at- tention at the Mn/ROAD project (Newcomb et al. 1999). At this facility, a range of vehicle loads, axle configura- tions, and vehicle speeds are being applied to instrumented pavement sections. The results were successfully used to develop a mechanistic–empirical design process that takes actual load repetitions into account (see the summary at the end of this chapter). By reviewing HVS test data from actual constructed pavements, Wolff (1992) was able to determine the effect of molding (post-construction consolidation) of a pave- ment structure due to wheel load applications. This had previously been postulated and verified by Kleyn et al. (1985). Wolff found that the application of different load sequences on the pavement structure affected the perform- ance. It appeared, from the limited data that were available, that the stress path influenced the final performance. The extensive nature of the impact of tire–pavement in- teraction explains why more APT programs are incorporat- ing this factor in their test plans. The same applies to the attempts at simulating dynamic effects of truck trafficking through APT devices. ENVIRONMENTAL IMPACT As mentioned earlier, environmental impact is essentially caused by physical and chemical action related to tempera- ture, water, wind, and radiation. It was found that the im- pact of temperature and water on pavement structures had been investigated by a number of APT programs, and a wide variety of pavement structures and materials had been tested. As far as could be ascertained, wind and radiation have not been featured to any extent in APT studies. In considering the effects of temperature and water it was apparent from the literature and the responses to the questionnaire that it was necessary to differentiate between short-term and long-term conditions. In addition, non- traffic effects had to be considered, especially in the case of temperature. This portion of the synthesis, on applica- tions and significant findings relative to environmental im- pacts, was therefore compiled with due regard to these in- fluence factors. During the course of events, test protocols have been developed, and these have provided a means of comparison between the findings of the various test programs.

27 Environmental Impact Factors Both non-traffic and traffic-related findings will be dis- cussed in this section. Aging of asphaltic materials, caused by time-related ex- posure to heat, oxygen, radiation, and wind, is one of the primary non-traffic-related effects that have to be consid- ered. Surrogate tests have provided useful laboratory in- sight into the effect of such aging on pavement perform- ance. However, to capture all aspects of this phenomenon, the actual effect of aging on pavement performance, a long-term study at a dedicated site(s) is needed. Thus far, no formal study has been undertaken specifically for this purpose by any APT program. Epps et al. (2001) did con- sider the effect of aging in their analysis of the findings from the MMLS3 (one-third scale mobile load simulator) study at WesTrack. It has also been recognized by Hand (1999) as a factor that needs to be considered in the analy- sis of APT performance. As expected, rutting was reduced because of the hardening of the aged binder. In addition, artificial accelerated aging has been used to simulate natu- ral aging of asphalt in APT studies (Hugo et al. 1987; Van der Merwe et al. 1992). Van der Merwe et al. (1992) aged two HMA test sec- tions through heating, at 100°C for 7 days and 28 days, re- spectively. They found a dramatic increase in the rut resis- tance after the artificial aging. However, at the same time, the resistance to cracking had decreased, because surface cracking was experienced much earlier. Shift factors relat- ing the respective performance parameters (rutting and cracking) before and after aging were 1:1.5 (rutting) and 5:1 (cracking). It was clear that aging had affected per- formance both positively and negatively. Unfortunately, the study was not taken any further, but it was apparent that the entire issue of aging needed additional research. In the second study, Hugo et al. (1987) reported on testing of similarly aged material at low temperature (–10°C). Aging again resulted in more extensive cracking distress. It was found that cracking depends on the interrelationship be- tween the increase in modulus and tensile strength. A pro- posal for investigating the effect of the environment was put forward by Hugo (1999). It entails a method for quanti- fying the effect of the environment over a period of several years with the aid of a structured APT study. The primary goal would be to develop procedures for determining the possible changes in the intrinsic life of a pavement due to environmental impact. The remainder of the discussion in this section fo- cuses on the various findings that have come about from APT programs that have addressed the influence of temperature and water in their testing of pavement struc- tures without specific attention being paid to the effect of aging. Impact of Temperature on Performance Performance of pavements is affected by the full tempera- ture range that occurs during its life cycle. During earlier APT studies tests were simply conducted at ambient tem- perature, while monitoring and recording the actual condi- tions during testing. However, recently, many facilities have provided some means to control the temperature of the pavement. These programs are identified in Appendix F. Test temperature capabilities vary, with the upper limit reported as high as 76°C (Harvey et al. 2000). The LCPC system was adapted to heat one of the sections artificially (Corté et al. 1997). A total of 36 tungsten–halogen projec- tors, of 1,000 W unit output, were installed on the inside of the track, 0.5 m from the wheel path at a height of 1.5 m above the ground over a length of 25 m. This enabled the pavement temperature to be increased by 10°C to 14°C above the ambient air temperature. The power needed for this was a little over 2 kW/m2 for testing speeds limited to 40 km/h. It is of interest to note that the power required to maintain a test temperature of 50°C under the MMLS3 on in-service pavements was the same, but was dependent on the ambient conditions (Epps et al. 2001). The difference with the MMLS3 was possible because the device is well insulated, thus minimizing loss of heat. For the initial heat- ing prior to starting an experiment, 6 kW/m2 or more was needed to bring the pavement to a test temperature of 60°C when the ambient temperature was near freezing. The heaters used with the ARRB–ALF system have a maxi- mum capacity of 2.8 kW/m2 (Johnson-Clarke and Fossey 1996). A number of facilities have also acquired the capa- bility to cool the pavement system and test at a controlled low temperature level. In some cases this can be below 0ºC. However, this requires specially equipped facilities such as HVS–CRREL and a limited number of others iden- tified in Appendix F. Because the reports on the application of APT under dif- ferent temperature ranges focus on dissimilar phenomena, the related discussions are considered in separate subsec- tions by means of case studies. APT Tests at Ambient and Elevated Temperatures Corté et al. (1997) discussed in detail the effect of tempera- ture on rutting. They reported that there is a threshold tem- perature above which asphalts show susceptibility to rut- ting. For the reference asphalt mix in the LCPC facility, this threshold temperature was 40°C to 45°C. When the temperature was below this, there was a marked reduction in the rate of rutting, as well as the extent of rutting, with deformation very low or even indeed nonexistent. This threshold value appears to be close to the Ring and Ball

28 FIGURE 4 Effects of wheel load and temperature on permanent deformation. softening point temperature of the binder. [The Ring and Ball test is used to determine the temperature at which bi- tumen (asphalt) reaches a certain degree of softness. This is arbitrarily defined in terms of a test to determine when bi- tumen changes from solid to liquid. The test is carried out as follows: The bitumen is melted and poured into a stan- dard brass ring placed on a plate; when cool, a standard- sized steel ball is placed on the bitumen and the ring sus- pended in a water bath, the temperature of which is raised at the rate of 5°C per minute. The temperature at which the bitumen softens sufficiently to allow the ball to pass through the ring and touch the lower plate—a distance of 25 mm—is called the softening point.] For this reason, the thermal history of a pavement plays an important role in the development of rutting and the related performance. Lister reported this in 1972 (Lister 1972 and Metcalf 1996) in one of the very early APT studies in the United King- dom (see Figure 4). From this figure, the dramatic effect of a 10°C or 20°C change in temperature on deformation can be seen. The results reported by Lister show a similar trend toward a threshold temperature at 40°C, as that reported by Corté et al. (1997). This explains why temperature is a major factor in the per- formance of asphalt pavements. The same is true of concrete pavements, albeit for different reasons, such as curling of slabs. The temperature effect is both short and long term. Maccarrone et al. (1997) found that the sensitivity of the deformation rate with increase in temperature is markedly reduced with each of the two ethylene vinyl acetate (EVA)- modified binders that were evaluated. They compared their performance to that of a conventional AC-30 binder (Class 320 Australian binder classification). In all cases, the com- parative mixes were continuously graded asphalt with 10 mm nominal aggregate and 5.1% binder. The rate of de- formation approximately doubled for every 4°C increase in temperature for the Class 320 mix, compared with every 7°C for the two modified binders. Modified bind- ers were found to significantly reduce permanent de- formation in asphalt pavements at elevated tempera- tures. At pavement temperatures of 60°C, it was found that the two tested, modified binders could reduce the rate of deformation by factors of approximately five and eight times, respectively, when compared with conventional AC- 30 binder. Galal and White (1999) investigated the effects of dif- ferent constituents of asphalt mixtures on permanent pave- ment deformation using APT. The investigation was also used to study the effect of lateral wander on rutting, which is discussed elsewhere in detail in this chapter. The re- searchers considered aggregate type, percentage of crushed gravel, and percentage of natural versus crushed sand and binder. Thirty-two mixtures were evaluated at a tempera- ture of 38°C (±1°C) and a trafficking speed of 8 km/h. The elevated temperature and the slow speed enabled definitive conclusions to be drawn. The researchers found that the aggregate type has a significant effect on rutting; slag and limestone mixtures rutted 50% less than gravel mixtures. They also found that an increase of 0.25% in the binder content resulted in as much as a 40% increase in rutting of the gravel mixtures. Harvey et al. (1999) reported on two pavements that were constructed in 1995. One pavement was a Caltrans “undrained” structure consisting of prepared subgrade, an aggregate subbase and base layers, and an AC surface. The second pavement was a Caltrans “drained” structure, in which a 75-mm-thick layer of ATPB was placed beneath the AC, replacing a portion of aggregate base. They re- ported partial verification that rutting of HMA primarily

29 occurs in the upper 100 mm of pavements. The speed of loading during the rutting tests averaged 7.6 km/h, while the nominal surface temperature was 55°C. Once again the slow speed and the elevated temperature yielded definitive results. It was found that 48% to 68% of the observed surface rutting was the result of the deformation of the asphalt layers and that the plastic deformation in the subgrade was minimal. The aggregate base and subbase layers contributed 22% to 50% of the total permanent vertical deformation. Gramsammer et al. (1999) pointed out that rutting experi- ments require fewer load cycles. Testing with the carousel at LCPC requires applying 100,000 to 200,000 load applica- tions on four radii of gyration at speeds of 38 to 48 km/h depending on the radius. Because rut tests require high temperatures, such tests are systematically conducted in midsummer (July and August) when the temperature reaches 30°C within the wearing courses (between 9 h and 18 h solar time). The rutting tests are conducted on a spe- cific site equipped with a subbase course resisting defor- mation, so that rutting measured on the surface is really the result of the plastic deformation of the wearing courses and not of the subsurface layers. Sharp et al. (1999b) also reported that deformation at high temperature was confined primarily to the upper as- phalt layer, indicating that only the upper layer needed to be replaced with a more rut-resistant product to effectively repair temperature-related deformed pavement surfaces. They concluded that the current Austroads design method needed to be revised to reflect the plastic-related deforma- tion in the asphalt layers together with limiting subgrade strain. In that study, a test section incorporating a high bi- tumen content mix in the layer base course showed only limited rutting. This was similar to findings reported by Harvey et al. (1999) in support of the rich-bottom design. Sharp et al. (1999b) concluded that testing at an elevated temperature at 40°C was too low for typical Australian in- service pavements, and suggested 50°C instead. APT Tests at Low Temperatures Low temperature cracking is a phenomenon that has been addressed primarily through improved binder specifica- tions. APT studies in regions with low winter temperatures and no environmental control presented the opportunity to evaluate the success of this approach. In the case of WesTrack, no distress of this nature occurred, thus validating the design strategy (Epps et al. 1999). At Mn/ROAD, the occurrence of low temperature cracking owing to the natural environment is also being monitored (Newcomb et al. 1999). When the pavement is subjected to temperatures below 0°C water becomes an important factor, because it freezes and the freeze–thaw phenomenon has to be considered. It becomes an interactive process between temperature and water. Few of the facilities take this into consideration. The Mn/ROAD is one site that was especially designed to in- vestigate the problem. The proposed design methodology that was implemented to address this is reportedly success- ful (Newcomb et al. 1999). The HVS–CRREL facility is also equipped to investigate this, and Janoo et al. (1999) have re- ported on a study that was conducted with their HVS. Zhang and Macdonald (1999) reported on full-scale APT testing of a pavement structure at low temperature with the Road Testing Machine in Denmark. The purpose was to investigate the effect of freeze–thaw cycles on the response and performance of a pavement that had already been subjected to 150,000 load applications at 25°C with only very limited deterioration. Trafficking included 50,000 load repetitions at dual-wheel loads of 40 kN, 50 kN, and 60 kN each. The International Roughness Index (IRI) had increased to 1.50 m/km from zero with a rut depth of 10 mm. In the freeze–thaw tests, the pavement was frozen to a depth of 1.2 m, which included the top three of nine clayey silty sand subgrade layers. The freezing was achieved by maintaining –10°C in the climate chamber surrounding the pavement section. Once this condition had stabilized, freezing was stopped and profiles measured. During freez- ing, the surface heaved upwards from 6 to 26 mm, in dif- ferent locations, because of ice formation. The temperature was slowly raised to 25°C and the pavement was allowed to thaw. Pore pressures (soil suction) were measured with tensiometers during thawing using a blend of water and anti-freeze. Temperatures were also monitored. A dual 60- kN wheel load was applied at a rate of 150 load cycles per day. The pavement response and pavement profiles were measured at regular intervals. After 1,800 load repetitions, freezing was resumed and a second cycle started. The frost heave was again recorded. During the second thawing period, 3,000 dual-load repeti- tions were applied. The IRI increased to 4.4 m/km, whereas the rut increased to 22 mm. A stress–strain model was developed that closely predicted the permanent strain in the subgrade for a subsequent test. The comparison between measured and calculated pavement responses proved satisfactory. Models for esti- mating plastic strains at the top of the subgrade and the permanent deformation of the subgrade were also devel- oped and found to be satisfactory. It was observed that • Some 60% to 70% of the total increase in plastic strain occurred during the early stages of thaw loading. • Dynamic transient (elastic) stresses and strains in- creased by up to 40% during thaw loading over those measured before the freeze–thaw.

30 A model for the IRI was also developed to reflect the permanent deformation at the surface. A number of conclusions were drawn from the study. • It is important to measure pore pressures in the sub- grade before and during thawing. • The subgrade permanent strain model closely pre- dicted the permanent deformation (rutting), but un- derpredicted the permanent strain in the subgrade. • Freeze–thaw experiments are time consuming. It took 2 months of continuously applied freezing tempera- tures at –10°C for a single cycle. The experiment took 11 months to complete. Impact of Water on Performance The impact of water on pavement performance has been explored in a number of APT programs. In general, the ef- fect of water ingress into the structure is dramatic. The im- portance of this is endorsed in the questionnaire response from CAPTIF, which states that the exclusion of environ- mental factors such as moisture greatly increases pavement life when compared to the expected design life (see Ap- pendix D, Table D3). As much as feasible, the discussion will differentiate be- tween the effect of surface infiltration of water and subsur- face ingress of water. However, as was shown in the case of temperature, water also has an effect that is non-traffic-related, at least until it causes distress that affects interaction with traf- fic. A major factor in this regard is chemical distress, which is primarily the result of chemical disintegration of material(s) or deterioration of layers. It is another form of non-traffic- related impact. The study by Roesler on alkali–silica reac- tion (ASR) attack on PCC in conjunction with the CAL/ APT program in Palmdale, California, is an example of this (Roesler 1998; Roesler et al. 1999). South Africa’s testing of cement-stabilized gravel that is deteriorating because of carbonation and/or crushing (Steyn et al. 1997) is another example of how APT has provided insight into the application of distress mecha- nisms and remedial measures. Similarly, the diagnostic study of the distress that the N2 National Road section in the Western Cape province of South Africa has suffered because of alkaline reaction is a good example of how ex- tensive the impact of this form of distress can be and how APT was used to provide parameters for evaluating opti- mum solutions for the rehabilitation. This is discussed in chapter six of this report (Strauss et al. 1988; Strauss and Van der Walt 1990). Infiltration of Surface Water In his study of the deformation of unbound bases, Maree et al. (1982) found that the ingress of water into the pavement structure had a major impact on the deformation. This is vividly demonstrated in Figure 5, which shows how rain and the subsequent artificial spraying of water onto a 30 mm surface under HVS testing in South Africa led to a dramatic increase in the deformation. The researchers re- port that this does not occur when the pavement is properly maintained, for example, by crack sealing. They also report FIGURE 5 Rutting performance of a pavement with a granular base layer during HVS testing (Maree et al. 1982).

31 FIGURE 6 Comparison of the pavement deformation induced in various crushed-stone base pavements by HVS trafficking (using n = 3) (Walker 1985). that the rate of rutting is reduced significantly when the pavement is allowed to dry out after the wetting phase. Water affects pavement performance when it gains ac- cess to the pavement structure. During trafficking pore wa- ter pressures develop, causing a loss of shear strength and even disintegration, depending on the nature and quality of the asphalt or other pavement materials. Sharp et al. (1999b) reported on APT that had been done on gravel base courses in Australia. Time Domain Reflectometry (TDR) gauges developed by the Council for Scientific and Industrial Research Organisation (CSIRO) were used suc- cessfully to monitor moisture in the subgrade. The lateritic soil was found to be very susceptible to moisture, and field road performance correlated well with the moisture varia- tion in the wheel path. Walker (1985) discussed the impact of moisture on pavement performance and used data that had been ob- tained from various HVS tests on South African field pavements to show how the effect differs depending on the type of pavement and the degree of saturation. Figure 6 (Walker 1985) shows the deformation relative to equiva- lent 80-kN standard axles using n = 3 in the load equiva- lency formula. Highway P 157/1 (see Figure 6) was subse- quently investigated in 1996 (Jooste et al. 1997) to evaluate how it had performed since the initial HVS testing (see chapter seven). Walubita et al. (2000) reported on studies that were conducted on Texas US-281 with the MMLS3 in con- junction with the TxMLS. The goal was to investigate whether moisture damage of the asphalt mixtures would occur as a result of the wet trafficking. The MMLS3 tests were conducted on the pavement adjacent to the TxMLS with a sheet of water flowing over the surface during trafficking. The most important finding was a significant drop in modulus of elasticity measured by spectral analysis of surface waves (SASW) owing to wet trafficking. This was in contrast to an increase in SASW modulus of elasticity because of trafficking under warm conditions (38°C). There was also a significant reduc- tion in the indirect tensile fatigue life of the wet traf- ficked asphalt. This reduction was attributed to mi- crofracturing, which would indicate that the test under the TxMLS that had been trafficked dry was less severe than the test with the scaled device. This meant that a much lighter wheel load was caus- ing more damage to the pavement through wet traffick- ing than the full-scale trafficking could achieve under dry conditions; the apparent remaining life appeared to have been reduced to 20% to 40% of that in the dry traf- ficked areas. Related findings by (Pidwerbersky et al. 1997a) and others indicated that the impact of water during trafficking on remaining life and performance can be significant.

32 Surface and Subsurface Water In an APT study on the performance of subgrade and base course layers of marginal quality, Vuong et al. (1994) re- ported on the importance of moisture content, making the following noteworthy points: • Back calculation of layer moduli from FWD deflec- tion data provides information on the effect of mois- ture content. The variation in back-calculated moduli with time was a useful tool for determining the rela- tive effects of seasonal environmental conditions. • Under dry conditions the sandstone had adequate stiffness (at say 50% of optimum moisture content). The material behaved nonlinearly, with the stiffness increasing as the load increased. However, the nega- tive effect of wetting the sandstone was also clearly demonstrated (see Figures 7 and 8). As a result of this finding it was apparent that the modulus had to be adjusted to account for seasonal moisture variations, because it could significantly impact the performance of low-cost pavements. • The Regression Rut Depth Model was used to analyze the performance data from ALF tests when load and en- vironmental conditions vary. It was also used to derive the power law in the load damage relationship. The model is illustrated in Figure 9. The 40-kN load applica- tions are progressively adjusted to be equivalent to 80 kN by varying the exponent value. The best-fit regres- sion indicates the applicable damage exponent. Vuong et al. (1996) reported on 34 experiment applying 3 million light-load cycles to 10 pavement types that were completed in Australia between February 1992 and June 1993. The trails contributed to the state of the art for the design and construction of stabilized and unstabilized granular pavements. Some of the important findings were as follows: • The effect of heavy rain in a concentrated short pe- riod of time needs to be considered in the adjudica- tion of performance, because it can override the good results over a long period of time. Granular materials that are sensitive to moisture do not dry back quickly and the prevention of ingress of water and the possible increase in moisture content within the material is critical. When the subgrade has ade- quate strength to support traffic, the main function of unbound marginal material is to carry the surface seal rather than support the load. • The relative effect of the environment can be ex- pressed in terms of an environmental equivalency factor or an environmental damage exponent (EDE). The EDE is defined by the ratio between rut depth and the number of load repetitions on a log-log ba- sis. EDE values of 6 to 10 were obtained from the trials indicating that the environmental effects could be more significant than the loading effects. The Australian roads subgrade strain model overpre- dicted the lives of granular pavements because it did not FIGURE 7 Variation in sandstone layer modulus with different loads and moisture conditions (Vuong et al. 1994).

33 FIGURE 8 Comparison of average layer moduli in site C and cumulative rainfall over time (Vuong et al. 1994). FIGURE 9 Comparison of net rut depth with 80-kN load repetitions (site C) (Vuong et al. 1994). take into account the deformation of granular pavement layers that are sensitive to moisture and compaction levels. Freeme (1984) had previously reported a damage expo- nent that varied between 2 and 6 in studies with the HVS.

34 FIGURE 10 Impact of water ingress on pavement performance (Rust et al. 1997). Vuong and Sharp (1999) reported that the typical con- struction of 95% of sealed roads in Australia is unbound crushed-stone/gravel bases with sprayed seals or thin HMA. As a result, the ALF program in Australia has fo- cused heavily on this type of pavement structure. This is evidence of the appropriateness of APT as a tool for inves- tigating a wide range of pavement structures and configu- rations. The same is true of several other programs such as SA–HVS (Rust et al. 1997) and CEDEX in Spain (Ruiz and Romero 1999). Meng et al. (1999) reported from their tests on a high- way in China that rainfall is one of the major environmental factors that can accelerate pavement deterioration. They also commented that the damage is primarily the result of pump- ing and erosion at the underside of the asphalt layer. Rust et al. (1997) pointed out that extensive studies have been undertaken to explore the impact of water on the performance of the pavement. Figure 10 shows the defor- mation performance schematically. Odermatt et al. (1999) completed accelerated testing on two subgrade soils at two moisture contents. The structures included 76 mm HMA, 229 mm gravel base, and 3 m sub- grade. They investigated the effect of increasing the mois- ture content in the subgrade and mathematically modeled this in terms of a power function relating accumulated per- manent strain in the subgrade and rut depth (see chapter five). From the study it was found that stresses and strains increased rapidly during the early loading cycles and the rate then tended to slow down. Either the dynamic or per- manent strain could be used to relate to surface rutting. The use of the ε-mu coil system proved to be very successful. This method of measuring strain uses a signal conditioner system for collecting and decoding signals generated be- tween two coils placed in the pavement layer system. It re- quires a good low-pass filter system to remove excessive noise from the dynamic output signal. Further validation was planned to include other soils and moisture contents. Under freeze–thaw conditions, the effect of water is greatly increased because of the development of pore pres- sures. The result was a dramatic reduction in subgrade strength. The topic was discussed earlier in this chapter in the section on the influence of low temperature on the pavement structure. Artificial Wetting of the Pavement Structure Several innovative ways have been used to simulate the in- gress of water into the pavement structure to study the ef- fect of water and/or moisture. These include trafficking

35 with water on the surface (Walubita et al. 2000). In the SA–HVS program the researchers spray water onto the surface to apply wet trafficking. In addition, they inject water through drilled pipes into one or more of the pave- ment layers under 1 to 2 m positive static pressure (Maree et al. 1982; Rust et al. 1997). Kadar and Walter (1989) give details of a procedure that was used to weaken the subgrade to achieve acceleration of distress. In essence, this entailed among other procedures • Using a longitudinal gradient for the test pad at 1% fall, • Constructing the drainage layer above ground (for positive drainage), • Allowing positive drainage flow into the structure, and • Constructing impermeable barriers between layers and between test layers. Moisture in the Pavement Structure During Freeze–Thaw Cycles The effect of moisture in the pavement structure is well known. A number of APT programs have studied this phe- nomenon and some valuable findings have been reported. Newcomb et al. (1999) extensively discussed the impact of moisture in the pavement structure, particularly as it af- fects the pavement during the freeze–thaw period of the year. They noted that at the moment when the soil becomes unfrozen and there is excess water in the base course, the soil weakens to its lowest point. This affects the stiffness of the layers and hence the pavement design and operational requirements. Saarelainen et al. (1999) reported on full-scale APT of a pavement on thawing, frost-susceptible subgrade using the HVS–Nordic with environmentally controlled conditions. The objective was to simulate natural thaw weakening in pavement subgrades. The pavement consisted of an HMA (50 mm), on a crushed-rock base layer (200 mm), and a subbase layer of sand (250 mm). The structure was con- structed on six subgrade layers of lean clay. The tests were conducted by freezing the pavement to a depth of 1.2 m and testing with the HVS–Nordic once the thaw depth had reached 0.9 m, with the temperature kept constant at 10°C. Pressure cells and strain gauges were installed and pore pressures were measured. It was found that the APT cyclic loading increased the state of static pore pressure. The time between two load pulses was approximately 7 s. The main purpose of the test was to determine a reference limit for the deflection of the pavement during thawing to ensure a predetermined rut depth limit (see chapter four for a further discussion). The dramatic impact of thawing on the sub- grade support was evident. Tests to evaluate this have been reported by Zhang and MacDonald (1999). These were dis- cussed in the section on APT at low temperature earlier in this chapter. MINNESOTA ROAD RESEARCH PROJECT—A COMPREHENSIVE CASE STUDY OF VEHICLE– PAVEMENT–ENVIRONMENT INTERACTION Mn/ROAD was intended to provide a full-scale test facility where the complex interaction between climate, materials, and traffic could be studied relative to the design and per- formance of the pavements. Forty pavement sections were constructed consisting of concrete-, asphalt-, and aggre- gate-surfaced structures. Both low- and high-volume traffic were considered. Climatic conditions are such that the frost-susceptible silty clay subgrade would provide insight into the environmental impact on the pavement structure. Provision was made to allow for nonfrost-susceptible ma- terial to allow for comparison. The results have been very good and have provided the basis for developing the mechanistic–empirical design system. In addition, it was possible to determine the effect of the spring thaw on the pavement structure, and this provided the basis for struc- turing guidelines for restricting truck loads in the state dur- ing the critical period of the year. The analytical model and the transfer functions, as well as seasonal changes in mate- rial properties, provided the basis for the systematic devel- opment of the procedure. The system comprises a computerized design proce- dure called ROADENT, which is an interactive, user- friendly, flexible pavement-thickness design program. WESLEA (Waterways Experiment Station Layered Elas- tic Analysis) is the analytical model used in ROADENT for calculating strains in critical locations. Strains at the bottom of the asphalt layer in a variety of pavement struc- tures in six instrumented flexible pavement sections in the Interstate portion of Mn/ROAD were used to confirm the reasonableness of WESLEA calculations (Chadbourn et al. 1997). It is important to note the close collaboration with other entities that provided input into the system, such as the USACE and the state of Washington. Another aspect that re- ceived attention was the determination of vehicle load damage. As could be expected, material characterization and instrumentation formed a major part of the program. A variety of other concerns are being addressed, such as ul- tra-thin white topping, drainage characteristics, and low- volume roads. An integral part of the design system is the use of Monte Carlo simulation. This enables the design to take care of input variability in terms of moduli and thick- nesses of the layers. The computer program Evercalc was employed to accomplish back-calculation. In the analysis a stiff layer is used to accommodate the relatively high water

36 table. Changes in the moduli during the year were moni- tored and used for the design system. Miner’s hypothesis was used to define the state of damage. Also investigated was the use of air temperature as an indicator of thawing following the research done in Washington. Revised refer- ence temperatures were determined for calculating a cumu- lative degree days thawing index. The onset of thawing could be predicted from measurements of actual in situ air temperature. This information was subsequently used to revise Minnesota DOT policy on the application of spring load restrictions. SUMMARY The studies cited show the importance of vehicle– pavement–environment interaction. A wide range of factors that affect this interaction were discussed. Through the ap- plication of APT, it was possible to gain insight that has benefited pavement engineering in general. As an example, it has been found that the damage exponent in terms of rut- ting and cracking varies depending on the degree of dete- rioration, the criterion used for comparison, the pavement structure, material characteristics, and the condition of the pavement at the time of comparison. The range extends from as low as 1.7 to as high as 10 and even more. Simi- larly, the impact of suspension type on the extent of dam- age owing to wheel load is now well understood. However, it was also apparent that there was a need to address several knowledge gaps, particularly in terms of the impact of the environment on pavement performance. A failure to properly incorporate environmental effects was the one factor that could jeopardize the broader application of APT in the field of pavement engineering. Nevertheless, there were several important case studies available to serve as guidelines for the best way to approach this problem. The extent of APT work that is needed became clear through the synthesis of all of the experimental studies. Some of the topics that need further attention are • Lateral wander, • Unidirectional versus bidirectional trafficking, • The effect of surface shear, • The effect of speed, • The impact of wet trafficking, and • The effect of soil moisture. It was apparent that there is a need to formulate and conduct a comprehensive collaborative APT program to develop guidelines to account for environmental effects on vehicle–pavement interaction.