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.

71 CHAPTER SEVEN PAVEMENT ENGINEERING APPLICATIONS AND ISSUES INTRODUCTION Thus far, the relationship between APT and constituent elements of the pavement engineering system has been re- viewed. In this chapter, a number of APT-related issues are covered. However, before considering the various issues, it is appropriate to discuss the wide range of APT applica- tions that have been recorded. In NCHRP Synthesis of Highway Practice 235 (Metcalf 1996), the applications of APT to practice were discussed in some detail. The fields of application in this report were expanded in the current study to cover some additional as- pects of pavement engineering. The applications cited by Metcalf (1996) were then categorized according to the de- fined fields. The list was expanded to include applications that had been reported at the recent International APT Con- ference in Reno, Nevada, together with related publications issued since then. The results of this analysis are contained in Appendix G indicating the respective fields and the re- lated objectives of the tests. This summary is indicative of the potential of APT as a core element of pavement engi- neering. The same applies to the annotated bibliography that was assembled from the responses to the question- naires. The aforementioned information should be of value when exploring documented case studies that have been reported by APT users. In the remainder of the chapter, the discussion is fo- cused on several important and related pavement engineer- ing issues and two special APT applications. • The relationship between APT and in-service pave- ments, • Failure criteria – Rutting and – Fatigue and cracking, • The relationship between APT and LTPP studies, • APT applications to block pavers, and • APT applications to airport pavements. RELATIONSHIP OF ACCELERATED PAVEMENT TESTING TO IN-SERVICE PAVEMENTS WITH CONVENTIONAL TRAFFICKING The first question that needs to be asked is: Why is there a difference between APT and conventional trafficking? Ac- cording to Metcalf (1996) there are essentially two rea- sons. • Environmental effects, especially long-term aging, are difficult to capture in APT. (The combined effect of environment and time difference is not simulated.) • A full spectrum of in-service wheel loads are not ap- plied in APT. In reality, all APT tests have to consider these two fac- tors when developing methodologies for using APT results optimally. Therefore, conventional in-service highways and/or formal LTPP studies feature both in such development. This is an important aspect of APT that warrants detailed discussion. Very little has been reported on this issue. However, the approach of some well-advanced APT pro- grams provided valuable input. In this section, the relation- ship between APT and in-service highways in general is discussed. Later in the chapter, the focus will be on formal LTPP studies. LCPC Approach The LCPC has an effective method of addressing the problem (Gramsammer et al. 1999). The procedure is as follows: • Laboratory fatigue tests are conducted on materials that are to be used in a pavement structure. Strains within the pavement structure are then determined theoretically by analytical or numerical modeling ac- cording to their rational pavement design method. • The calculated strains are then used to evaluate per- formance life in terms of the laboratory findings. • These findings are then compared to the performance life of the pavement under trafficking by the carousel (LCPC–APT device). The two results are used to es- tablish a “coefficient of correspondence.” • By relating the theoretical laboratory performance life to conventional field traffic performance, a “coef- ficient of adjustment” is determined. • The interrelationship between laboratory, APT, and field traffic is then established by comparing the co- efficient of correspondence to the coefficient of ad- justment. The extensive knowledge that they have on long-term performance of in-service pavements serves as benchmarks for establishing the relation- ships.

72 On the basis of the established relationships, the LCPC researchers are able to evaluate the performance of new materials using APT. This methodology has been used ex- tensively by LCPC and their industrial partners to establish the performance of new materials (many of which are pro- prietary) and/or structural design systems. The coefficient of correspondence, Kc, for the Circular Test Track is given by s ubb calc ΚΝΕf,Κ δ6 6 0110θεε −⋅〉〈⋅〉〈= where NE = number of axle load passages leading to the rupture of the structure, εcal = strain calculated in the modelized equivalent structure, ε6(θ, f) = strain leading the rupture during the laboratory fatigue test for 1 million cy- cles at the temperature θ and the fre- quency f, b = slope of the material fatigue curve, δ = standard deviation depending on actual thicknesses and fatigue test results, u = random variable related to the risk con- sidered of the normal law, and Ks = correction coefficient of soil homogene- ity. Corté et al. (1997) stressed that to compare performance of different materials to a reference mix it is necessary that • The rut depths are sufficiently large after 100,000 loadings to generate significant differences in per- formance, and • Speeds of load application are the same. In this way, LCPC is able to interrelate the carousel and an in-service pavement that has a proven record of per- formance. This also allows them to use their LTPP results in conjunction with APT results. ARRB Approach In 2001, K.G. Sharp provided a summary on the procedures that are followed by the ARRB in their transformation of ALF trafficking into conventional traffic on in-service highways. Because of the generic nature of the outline, it is quoted below with minor editorial amendments: The method of handling loading obviously depends on the type of experiment/test. In many experiments we are only concerned with the relative performance of one type of “ex- perimental” pavement with another type of “control” pave- ment. The idea is that we have a good knowledge of how the “control” performs in service (because it is a standard pave- ment widely used in practice), so, if we can demonstrate that the performance of the experimental pavement is similar to the control pavement, then there is increased confidence in its likely performance in service. These “experimental” pave- ments might incorporate non-standard materials, recycling techniques, etc. In those cases, the Equivalent Single Axle [load] (ESAL) concept is as good as any, i.e., cycles of load of X kN converted to the Standard Axle load assuming the 4th power law. Similarly, we might be comparing a range of maintenance options. In those cases, we will traffic the trial pavements with the same number of cycles of the same load and compare per- formance. This, of course, is also the classic way of conduct- ing an “axle load equivalency” trial—either test all pavements as above or, alternatively, compare the number of cycles of differing loads required to generate the same performance (say a 20-mm rut). Such an approach was also appropriate during the series of asphalt deformation trials we conducted. The aim was to com- pare the performance of the various mixes/binders, so we placed the test mixes onto a very strong cemented base to en- sure that all deformation would be induced in the mixes. We also used the pavement heating system to ensure that all test- ing was conducted at the same temperature (50οC). We then tested all mixes with the heaviest load (80 kN) and compared the deformation—leading to a “ranking” of the mixes in terms of deformation resistance. Meanwhile, we conducted labora- tory testing on the same mixes (both cores taken from the pavement and laboratory manufactured samples) and also “ranked” them according to the results. We then compared the field and laboratory results. We have also done this type of testing of unbound materials using ALF and Repeated Load Testing (RLT) testing. It is when we are looking at specifically structural design issues, that it gets a bit more complicated. I am referring here mainly to fatigue tests of bound materials where you are trying to either develop, or confirm, fatigue life prediction models. At the moment in Australia the damage exponent for cemented materials is 12 (based on “experience” and limited testing), while for asphalt it is 5 (Shell). Using these relationships we can predict the “allowable life” for a given test pavement by modeling the pavement using CIRCLY (Wardle 1977), calculating the hori- zontal strain at the base of the bound layer, and hence fatigue life using these performance relationships. These are in the Aus- troads Pavement Design Guide (Austroads 1992), which is currently being updated for release in 2002. We then turn ALF on and monitor the number of loading cycles to “failure” (whatever that is) and compare it with the predicted life. We also do some laboratory testing and com- pare the three predictions. Defining “failure” is tricky enough: we know that surface cracking may not be the failure point. The layer may have cracked at the bottom and the crack worked its way up and we also know that the layer, even though cracked, still has a strength (cemented layers can survive for years after cracking—you simply apply a surface seal every 5 years or so when the cracks reflect through to the surface!). Another way of defining “failure” of bound materials is in terms of back-calculated modulus. We conduct FWD tests at the start of trafficking and use EFROMD2 (Vuong 1991) to predict the modulus of the bound layer. We continue to con- duct FWD testing during trafficking and note the decrease, if any, of the modulus of the bound material (the theory being that the moduli of the other layers (e.g., subgrade) will stay the same because test conditions are controlled). We have found that the modulus of asphalt can decrease to about half its ini- tial value before surface cracking is observed whilst, for ce- mented materials, the value can reduce to as much as 1/10 of the initial value before cracking is observed. Fatigue testing of bound (particularly cemented) materials is tricky at the best of times because of the variations in parent

73 material, binder type, binder content, etc. We never have had a good agreement with laboratory beam testing. This is not a problem unique to us! (K.G. Sharp, ARRB, personal commu- nication, December 3, 2001). LINTRACK Approach Bhairo et al. (1998a,b) debated this issue in some detail. Their LINTRACK program is strongly focused on model- ing. Coupled with that is the question of whether or not a pavement has failed. The problem lies in that stiffness loss invariably occurs well before cracking manifests. The Dutch researchers assumed that a section has failed when the stiffness has dropped to a level of 50% of the virgin as- phalt. The procedure takes account of factors such as lat- eral wander and healing because of rest periods of nonop- erational time. Allowance is also made for crack propagation, which depends on the layer thickness. However, when all of these factors are accounted for, there is still a 2- to 4-fold difference between LINTRACK life and the laboratory-determined value. The authors do not discuss the issue of extending the performance relationship to include a regular highway. South African Approach The South African approach does not have conversion fac- tors between APT structural pavement behavior and real- life structural pavement performance. However, this is not believed to be crucial to the success of the process (De Beer et al., personal communication, November 26, 2001). They consider the number of variables that affect the real- life performance and life expectancy of a road to be so high that they state: It is extremely difficult to directly relate real-life to APT unless you have a huge APT database where you have varied (and tested) all these impacting factors. The approach on the implementation side is therefore rather an “adaptive manage- ment process” than a “predictive control process.” Real-life dictates and we have to manage our activities (maintenance and rehab) to ensure we get the performance that we want from the road. The South African process is shown in Figure 16. In es- sence, there is no attempt to convert HVS-initiated pave- ment structural behavior to real field pavement perform- ance. The approach is to study pavement response and performance in great detail by means of APT and then de- velop practical pavement designs by applying the knowl- edge gained. The South African mechanistic–empirical pavement designs are calibrated from time to time in col- laboration with pavement engineering consultants and the road authorities. This approach is underpinned by the work that was done during the early phases of the SA–HVS program. In 1984, Freeme summed up the approach as follows: • Primary indicators of performance such as rutting and cracking were monitored. In the same vein, sec- ondary indicators, which are responses in terms of Actual pavement performance and life Life-cycle strategy Maintenance Traffic Construction REAL LIFE IMPLIMENTATION APT, ANALYSIS AND DESIGN ADAPTIVE, MANAGED PROCESS Analysis and Design: Accurate modelling and unbiased design APT: Equal testing ground A n a l y s e C a l i b r a t e Improved understanding of pavement behavior TRANSFER CONTROLLED PROCESS Design Environment FIGURE 16 Schematic outline depicting the interaction between SA–HVS APT results and real-life implementation and pavement performance (De Beer et al. 2001, personal communication).

74 deflection with depth, and in situ strength and strain, were monitored. With this process, changes in secon- dary indicators were linked to primary indicators and used to calibrate mechanistic models of pavement struc- tures. Confidence was built when measured deflection profiles matched calculated deflections. • There has been reasonable correlation between HVS performance and actual in-service pavement per- formance. This has enabled general performance of pavement structures to be clearly illustrated and un- derstood. In turn, this has improved understanding of factors that influence performance of pavement struc- tures. This knowledge has enabled performance of rehabilitation strategies to be predicted. From the foregoing discussion, it is apparent that APT findings are transformed in various ways to enable the re- sults to be used by the different APT users for their specific purposes. This process is not without constraints. Some of the problems that have arisen will be illustrated by means of selected case studies indicating how the ultimate per- formance of APT sections compare with the original APT findings and the related performance prediction. CONSIDERING SOME CONSTRAINTS IN THE PROCESS OF TRANSFORMATION OF TEST FINDINGS BETWEEN TRAFFICKING SYSTEMS It is necessary to review some of the constraints that need to be considered in transforming test findings between traf- ficking systems. Typically, the constraints relate to issues that have been discussed elsewhere in the report, such as vehicle–pavement–environment interaction over time (see chapter three) and construction and maintenance (see chap- ter six). The matter is further exacerbated by differences in distress criteria that apply to APT and in-service highways. These constraints are particularly relevant in the case of performance prediction of conventionally trafficked in- service highways, as will be seen from the following case studies. Three cases, each extending over almost two dec- ades, were selected to illustrate the complexity of the prob- lem. Case Study One: A composite pavement with thin HMA and a crushed-stone base on cement-treated subbase— Route P157-1 South Africa. Jooste et al. (1997) presented a comparison of HVS-predicted performance with the actual performance of a road in South Africa. The road (Route P157-1) had been tested 16 years earlier with the HVS. Characteristics such as deflection, Dynamic Cone Pene- trometer shear strength, moisture, and density were found to correlate reasonably well with the initial measurements. Environmental effects did not appear to significantly influ- ence the performance of the pavement. However, the re- searchers point out that a 13-mm surface seal coat had been placed some time earlier and this had probably pre- vented the ingress of surface water. The average rutting was similar to that measured during APT at the equivalent number of standard axles. However, in both cases, the ruts were very small while the variance differed. This was the first comparison of its kind that they had undertaken and, although good agreement was found in the study, they identified a number of important issues that needed to be considered in such comparisons. In particular, accurate traffic statistics in terms of axle loads are necessary. They also suggest that statistics on ride quality would be useful. A greater number of similar tests would also be necessary to improve the level of confidence. Case Study Two: HMA surfacing and base—Natal, South Africa. Wolff et al. (1997) investigated the perform- ance of a pavement with a HMA surfacing and base struc- ture that had been tested with the HVS in 1984. Their goal was to compare the actual performance of the pavement with that predicted after the HVS testing. The original test was to evaluate the performance of a structure consisting of bituminous surfacing and base on a stabilized granite subbase in a wet environment. In the original study, the pavement was subjected to the equivalent of 20.3 million ESALs with the HVS in a dry condition. At that time, the pavement showed a 10-mm rut that was primarily the result of deformation in the HMA. When the section was then artificially wetted, it failed after application of a further 300,000 ESALs with a rut of 20 mm (10 mm in the asphalt and 10 mm in the stabilized subbase). The failure mechanism was reported as being • A longitudinal crack in the wheel path owing to fa- tigue in the asphalt layer initiated by a stabilization crack in the subbase layer or a crack resulting from the settlement of the fill, • Pumping the result of trafficking following ingress of water into the pavement system, • Erosion of the stabilized subbase layers and subse- quent pumping into layers above because of traffick- ing, and • Permanent deformation of the carbonated stabilized subbase. The artificial wetting was accomplished through surface wetting during trafficking and filtering water into the layer system from the side under a head of 1 to 2 m of water. The moisture content in the subbase layer, after artificial wetting, was still slightly less than the actual pavement at the time of failure; that is, after 25 years of service life. The actual pavement that was investigated in close proximity to the original test section exhibited distress in the form of pumping after 1.4 million ESALs of conven- tional trafficking under the natural environment.

75 From an examination of the long-term performance of the highway pavement, the researchers concluded that the failure mechanism was similar to that which had been found with HVS testing. However, failure had occurred af- ter a significantly different number of ESALs. In the case of the HVS tests, the failure had occurred after approxi- mately 300,000 E 80s of wet trafficking, whereas it took a total of 1.4 million ESALs under the natural, but wet cli- matic conditions to reach similar failure on the highway before rehabilitation was undertaken. This difference was attributed to differences in the ingress of water into the sys- tem. Apparently there had been intermittent phases of wet- ting and drying of the layers in the long-term pavement test. The researchers suggested that it would be better to wet the pavement intermittently throughout the HVS test rather than only at the end of the dry test phase or, alterna- tively, better to wet at an early stage rather than only at the end of the dry test phase. A further interesting point from the study is that a second rehabilitation was initiated in 1996 when the PMS signaled the development of the same type of distress as before. At that time approximately 1.5 million ESALs had been added after the rehabilitation. This was almost the same as the traffic carried during the first phase (1.4 million). It is also worth noting the high HVS traffic volume that the section had carried in the dry state (20.3 million). This case study demon- strates the importance of integration of APT and studies of the performance of in-service pavements and the need to take into account the environmental impact in a manner that simulates the actual field conditions during trafficking. Case Study Three: Overlay on an alkali–aggregate re- action distressed concrete pavement. In chapter six, the re- habilitation of a jointed concrete pavement in South Africa was discussed (Strauss et al. 1988). The pavement had suf- fered severe distress because of the reaction between the high alkali cement and the concrete aggregate. The reha- bilitation was undertaken after completion of HVS testing of alternative options that were constructed on the in- service pavement (Viljoen et al. 1987; Strauss and Van der Walt 1990). The HVS trials showed little rutting with the alternative options. Cracking was the predominant mode of failure. The following options were investigated: • 150 mm crushed stone with 50 mm asphalt wearing coarse, • 30 mm open-graded asphalt with a bitumen–rubber binder wearing coarse, • 80 mm asphalt with bitumen binder, • 80 mm asphalt with low-modulus nonwoven geofab- ric interlayer, • 80 mm asphalt with bitumen–rubber stress-relieving interlayer, and • 80 mm asphalt with high-modulus woven geofabric interlayer. Testing took place primarily during the wet winter sea- son. Water was also sprayed onto the surface when there was no rain. Conditions were therefore particularly harsh during APT. Cracks initiated where the underlying old concrete pavement exhibited movement. Several rehabili- tation options were considered feasible, but the most eco- nomical one for a short term (7-year life cycle) was found to be bitumen–rubber asphalt. The HVS study predicted 3 years of life for the bitumen–rubber asphalt. The rehabilitated pavement has been in service for al- most 19 years with only limited maintenance. The mainte- nance was in the form of a seal coat and joint repair through removal of failed material and reinstatement of concrete where the material had deteriorated owing to the ingress of water and air (J.C. Van der Walt, personal communication, December 3, 2001). The performance of the road has been intermittently monitored since the origi- nal construction. A report was compiled in 1997 (NDOT– SA 1997). It was apparent that the performance was far better than found with HVS testing. The theoretical analysis by Strauss et al. (1988) concluded that the actual performance could be as much as four times longer than the HVS per- formance, if the mean condition of the underlying concrete pavement is considered instead of the most severely dis- tressed sections. In addition, the harsh conditions during APT trafficking would have led to a conservative estimate of performance life. The information gathered from the di- agnostic performance monitoring has provided noteworthy insight on the relationship between APT and in-service per- formance. Some lessons learned were • The actual long-term performance of the original test sections was as found with the HVS testing in 1984. • The concrete overlays (jointed and CRCP) have per- formed well, with little or no visible distress. • The bitumen–rubber asphalt section that had been constructed with extender oil added has performed well. • The stress-absorbing interlayers have played an im- portant role in the long-term performance. This re- lates to the reduction of the ingress of water into the pavement structure and reduction of shear and tensile stresses in the asphalt. The three case studies demonstrate how vehicle– pavement–environment interactions affect LTPP. Clearly, the transfer of APT results to in-service pavements has to be done with caution and specific consideration of all fac- tors that can affect the results, particularly the environ- mental factors. Also to be considered are the frequency of maintenance and possible alternative failure mechanisms, especially when the pavement structure is altered.

76 FAILURE CRITERIA This issue relates to the previous topic. It is the basis for defining benchmarks to ensure comparable APT perform- ance relative to real pavement performance under conven- tional traffic. Several factors have to be taken into account, such as the limited size of test sections, the difference in the nature of trafficking, the difference in time scale in- cluding the effect of aging, the limited number of experi- ments, and the limited ability to determine the integrity of the pavement nondestructively. Two forms of load-related dis- tress are generally considered, rutting and fatigue cracking. Rutting This form of distress is the easier one to monitor and evaluate. The rut profile measurement is taken to be repre- sentative of the performance of a section of in-service pavement that it simulates. In general, the average maxi- mum rut at a specific cross section is reported. A number of impact factors are then taken into account to enable per- formance of a pavement under different trafficking systems or conditions to be compared. In the case of APT testing of an in-service highway, it is taken to be on a one-to-one ba- sis; that is, the rutting performance should be the same af- ter accounting for climatic effects, load amplitude, and fre- quency of load applications. Epps et al. (2001) followed a systematic procedure for doing this in relating the per- formance of the MMLS3 to the full-scale truck trafficking at WesTrack. This allowed for differences in aging, load frequency, temperature, and lateral wander. Epps et al. (2001) also took into account the difference in the extent of the stress profiles owing to the difference in load ampli- tude. Once these factors had been accounted for, they found that the rutting performance was comparable on a one-to-one basis with a small margin of error. It is of interest to note that the method of recording rut- ting used in the Spanish APT program (Romero et al. 1992) is able to capture changes in the transverse rut pro- file that are the result of shoving of the asphalt owing to lateral wander. In their transverse profile measurements they differentiate between the maximum rut depth and the rut depth on the center line of the wheel path. Maccarrone et al. (1997) reported that permanent de- formation results obtained with ALF testing in the field (in Australia) and wheel tracking in the laboratory gave a good correlation. A similar magnitude of deformation level and de- formation rate was obtained with ALF field testing and labo- ratory wheel tracking when the latter was done at 60οC. The problem becomes more difficult where the defor- mation occurs throughout the pavement structure including the nonbound, stress-dependent materials. Theyse (1997) presented a conceptual model for the permanent deforma- tion of pavement layers. In the process, he developed per- manent deformation design transfer functions for a number of unbound material quality groups, following the princi- ples of a basic model. The main source of data for develop- ing the design models was HVS test data collected over a long period of time in South Africa. This covered dense- graded crushed stone and natural gravel with due regard to variations in quality, as well as a range of pavement foun- dation material quality groups. These were subsequently incorporated into the design recommendations for high- ways (Structural Design of Interurban . . . 1996). The ARRB followed a similar procedure (Sharp et al. 1999a). Fatigue and Cracking Fatigue is a more complicated form of distress because it relates to cracking and stiffness loss. In general, it is ac- cepted that failure has occurred once the in situ stiffness has dropped to a level of 50% of the original untrafficked pavement (Bhairo 1998a). Various methods of monitoring the stiffness loss have been reported (Lee et al. 1997; Bhairo 1998a; Harvey et al. 2000). With full-scale APT it is necessary to monitor the cracking as it develops and several procedures have been developed to do this. Scheffy et al. (1999) explored a number of digital and manual crack detection and mapping methods and developed a reliable and consistent method of crack detection and measurement for fatigue cracks. The method employs off-the-shelf soft- ware and hardware, making it inexpensive and easy to im- plement. Hugo et al. (1997) used transparent mylar sheets to log cracking of a 12-m test section as trafficking pro- gressed. Prints were then made of the sheets and their size reduced to enable photocopies to be made. The electronic images were vectorized with the aid of a commercial soft- ware package. The cracks could then be analyzed to cate- gorize them according to orientation. Groenendijk et al. (1997) also present details for capturing and recording cracks in the pavement using Mylar sheets. The methodol- ogy they used for categorizing the crack orientation is the one that was followed by Hugo et al. (1997). Other APT programs each have their own methodology for monitoring and recording cracks. RELATIONSHIP BETWEEN ACCELERATED PAVEMENT TESTING AND LONG-TERM PAVEMENT PERFORMANCE STUDIES There have been, and still are, a wide range of formal experi- ments that are designed to monitor the performance of desig- nated or specially constructed in-service pavement sections, over time. Such formally structured experiments are gener- ally known as LTPP studies. The relationship between APT and formal LTPP studies is discussed in this section.

77 As discussed earlier, a number of APT programs have monitored and kept detailed records on the performance of pavements tested in an accelerated mode, either directly or indirectly. Other programs, such as several in South Africa, have only recently included such procedures in their test plan (Jooste et al. 1997). This ongoing process provides an important comparative base for relating actual performance of formal LTPP sections under conventional trafficking to performance under accelerated trafficking. The benefits of such an approach are highlighted in an overview on the topic submitted by Sharp and Clayton (personal communi- cation, November 30, 2001) in which they give details of the relationship between APT and long-term performance trials of the ARRB. This is of a generic nature and consid- ered to be of value to the APT community at-large. There- fore, it has been included as Appendix H. Aspects of their approach are briefly discussed later. It appears as if this in- teraction between APT and LTPP was also contemplated in the original formulation of the SHRP–LTPP experiment according to the following extract from the TRB SHRP Research Plan report (1986): The LTPP program embraces the total range of pavement in- formation needs. It draws on technical knowledge of pave- ments presently available and seeks to develop models that will better explain how pavements perform. It also seeks to gain knowledge of the specific effects on pavement perform- ance of various design features, traffic and environment, use of various materials, construction quality, and maintenance practices. In the years to come, as sufficient data become available, “analysis” is expected to be conducted by numerous interested agencies. The formal LTPP products are expected to include better predictive models for use in design and pavement man- agement, better understanding of the effect of many variables on pavement performance, and new techniques for design and construction. These improved “tools” are then expected to re- sult in improved pavement management. The LTPP study program has three potential types of stud- ies. These include General Pavement Studies (GPS), Specific Pavement Studies (SPS), and APT. The GPS involves a very large experiment that embraces an array of site selection fac- tors expected to produce a broad range of products and results. The SPS will have their own set of more limited goals, con- struction needs, and experimental approaches; and are gener- ally aimed at more intensive studies of a few independent variables for each of a number of study topics. The other cate- gory of studies, which is to be considered for future adoption, is the Accelerated Pavement Testing Program. It is apparent that APT and formal LTPP programs share many goals. This is significant given the growth in the APT programs in the United States. Furthermore, both sys- tems are attempting to build pavement knowledge in areas such as • Validation of laboratory performance-related materi- als tests; • The linkage between laboratory, APT, LTPP, and normal in-service highways in terms of distress and performance; • Construction-related pavement defects; • Quantification of risks in warranty projects; • The efficacy of maintenance procedures; • Vehicle–pavement–environment interaction; and • Development of durable, low-maintenance pave- ments with improved performance prediction. Because of the commonality, APT records contain in- formation compatible with, and in many respects similar to, that already being collected as part of formal LTPP studies or programs, such as the SHRP–LTPP experiment. This provides a basis for efficiently and effectively com- paring APT performance with that of related or comparable LTPP sections. It is therefore understandable why research entities are in general keen and willing to interface their APT programs with formal LTPP programs. The Australian ARRB approach to this matter is presented in Appendix H as outlined by Sharp and Clayton (2001). The following are salient features of their approach. An Austroads-funded project was established to address APT and LTPP issues and to take advantage of the oppor- tunity to be directly involved in the SHRP program. This had as its primary aim the monitoring of the performance of a range of Australian test sections as a complimentary project to the U.S. SHRP–LTPP experiment. The goal was to improve performance prediction models for the benefit of LTPP. The overall objectives of the Austroads LTPP study are to • Enhance asset management strategies through the use of improved pavement performance models based on an improved understanding of the behavior of pave- ment structures (the SHRP–LTPP program), and • Compare the results of accelerated pavement test studies with actual road pavement performance (the ALF–LTPP program). Nineteen Australian test sections have been monitored continuously for 5 years. Some of these were specifically established in tandem with ALF trials. The preliminary analyses conducted to date have already produced signifi- cant findings, many of which are discussed elsewhere in this report. With the expansion of the APT programs in the United States, the benefits to be gained from closer linkage be- tween the performance of in-service highways, formal LTPP programs, and APT studies are apparent. As was evi- dent from earlier discussions in this report, there is already an extensive array of significant findings that can serve as a basis for such a collaborative effort. The foregoing discussions demonstrated that the in- teractive use of results from the performance of in-service pavements, together with formal LTPP monitoring and

78 APT programs, has improved understanding of the per- formance of pavements and its prediction. In turn, this has led to cost savings and it has enhanced construction and rehabilitation practices. As a result it has reduced risk in performance prediction. These findings are being used in warranty projects and are proving of value in the im- provement of pavement management. In the next sections the discussion will focus on APT studies conducted on nonconventional pavements such as concrete block pavers and airport pavements. APPLICATION OF ACCELERATED PAVEMENT TESTING TO BLOCK PAVERS Concrete block pavements have been tested using APT in a number of facilities worldwide. Shackel (1990) provided an overview of developments going back as far as 1967 in Rotterdam. At the time, the test facility comprised a 20-m- diameter circular test track. Subsequently, tests were car- ried out in Australia, New Zealand, South Africa, and Ja- pan. Studies have focused on • Strengths, • Shape of pavements, • Thickness of pavements, and • Layout pattern. In studies using the HVS, Shackel (1980, 1982) ex- plored the failure mechanisms and the factors that affect performance. Performance has been found to be somewhat similar to flexible pavements, with some specific differences such as • Ability to tolerate higher resilient deflection, and • Stiffening of the pavement structure after the initial 500 load cycles were needed to develop full interlock. It was also found that the pavements have a prominent structural capacity as reflected by a sharp reduction in ver- tical stress at the level below the bedding sand layer. This brief overview shows how the state of knowledge on pavers has been enhanced through APT and it serves as an excellent example of how applications followed initial research studies to enhance the use of the product. In 2001, Sharp provided an interesting case study, per- taining to the Sydney Opera House, on concrete pavers that dates back 18 years. In 1998 and 1999, the Sydney Opera House Trust commissioned an ALF study to investigate failure of granite sets (pavers) on a heavily trafficked ac- cess road to the Opera House. The sets were paved on a mortar bed and the waterproof membrane on top of a rein- forced concrete slab. The sets had failed through displace- ment by passing traffic, after being loosened during traf- ficking. The ingress of water followed with subsequent secondary effects on the underlying structure. The study focused on determining the optimum laying pattern as well as bedding and jointing methods. Five panels were tested, with alternative set patterns and bedding methods applied to the respective layouts (K.G. Sharp, personal communica- tion, December 3, 2001). In this instance, the goal was to rank the relative per- formance in terms of the extent of cracking and deforma- tion. Another goal for the test was to determine the effec- tiveness of the waterproof membrane. Dual-wheel load tests (40 kN) were conducted, and it was apparent that dif- ferential failure had occurred among the alternative test plans (Figure 17). A second experiment was conducted with 40-, 60-, and 80-kN ALF loads and trafficking was taken to almost 1 million equivalent standard axles. In an attempt to promote premature failure, areas of mortar joint- ing were saw cut to duplicate cracking and allow ingress of water into the pavement. Both wet trafficking and heated trafficking were undertaken. The latter was to investigate the effects of heat on the rubber membrane. Some minor cracking was observed in both panels that were tested after 36,000 cycles. No further distress developed. The saw cuts did not appear to affect the structural performance. Only minor cracking was observed in both panels after 36,000 cycles; no other distress was observed at that time. There appeared to be no evidence of the application of heat af- fecting the performance of the water membrane underlay, and the underlay did not appear to affect the performance of the granite set pavement. FIGURE 17 Failure of block pavers after 9,000 cycles of 40-kN loading. Note the amount of sand that has been pumped from under the granite blocks (Sharp 2001). APPLICATION OF ACCELERATED PAVEMENT TESTING TO AIRPORT PAVEMENTS APT already has an impressive record of applications to airports, so much so that special equipment has been de- veloped for this purpose (Hayhoe et al. 2001). Some of the cases will be used to demonstrate the applications.

79 The HVS–SA program successfully completed APT studies on airport pavements in South Africa in the early 1980s (Clifford et al. 1982; Clifford and Opperman 1983). The first was at the Cape Town International Airport and the second at the Johannesburg International Airport. Guo and Marsey (2001) investigated the effect of temperature variation on the performance of concrete slabs in the FAA’s NAPTF. Load transfer between adjacent slabs was found to be poor during winter (low temperature). They also investigated joint behavior and concluded that load transfer may be sensitive to traffic direction. An inter- esting finding was that the sum of the deflection in two op- posite directions across a joint varied very little. This may serve as a guide to the sensitivity of slabs to curvature. In the Cape Town International Airport case study, HVS testing was conducted to evaluate the status of the taxi- ways. Three sites were selected and trafficking done with 40- and 80-kN wheel loads (using regular truck wheels). This was followed by trafficking with a 200-kN wheel load, using a single wheel of a Boeing 747 at full load. In the latter case, bleeding occurred and cracks developed. Trafficking was terminated when the rut depth reached 40 mm. From the analysis of the results, it was concluded that deep-seated distress was occurring because of insufficient coverage over the submerged sandy subgrade. It was pro- posed that full-depth asphalt rehabilitation be done. This was undertaken 2 years later and as of 2001 the pavements were being rehabilitated again with milling and the re- placement of the asphalt layers. The historical traffic over the taxiway amounted to 78,000 aircraft wheel-load appli- cations before the recent rehabilitation (F.J. Pretorius, personal communication, December 2001). In the HVS testing in 1982, trafficking was terminated at 31,000 load applications, implying that the HVS tests were con- servative in application. A somewhat similar result was found with a second test run close by, which was termi- nated after 14,000 load applications with the rut depth at only 14 mm. Hayhoe et al. (2001) investigated the impact of multi- ple-wheel and multiple-truck landing gear configurations and their spacings on six flexible pavements. A total of 522 tests were performed. Various items were also involved in 108 tests on three rigid pavements. The test vehicle com- pleted these slow-rolling tests at 0.15 m/s. The purpose was to study wheel load–pavement interaction in terms of response under the moving wheel loads. Multi-Depth Deflectometers were used in conjunction with the static load tests on the flexible pavements under varying wheel loads. Garg and Hayhoe (2001) reported on tests that were done with various slow-moving wheel loads. The loads were varied and so was the temperature. As expected, the strains varied strongly with temperature and speed; at the upper range, the values were as much as three times higher than those predicted by layered elastic computer programs. They also found significant permanent deformation at high temperature. It should be remembered that trafficking done was bi- directional, which is believed to be more severe than unidirectional (Brown and Brodrick 1999). This could have affected the performance. The primary objective of the HVS–A research program is to validate the 3-D pavement design and evaluation pro- gram being developed by the ERDC. The airfield pave- ment design system was a direct result of a long history of APT/full-scale testing. At the Johannesburg International Airport, HVS–SA tests were conducted on fresh asphalt on a new taxiway ad- jacent to the newly constructed parallel runway. The total asphalt thickness was to be 100 mm, semi-gap graded with 19 mm aggregate. The pavement was very stiff with high- quality crushed-stone base course on cement-treated sub- base layers. With the HVS testing it was found that the de- formation increased sharply when the test temperature was raised from 30°C to 40°C. The laboratory Marshall designs had indicated rutting susceptibility. This was confirmed with the HVS testing when 40 mm rutting was reached af- ter 27,500 repetitions of a 200 kN Boeing 747 wheel. As a result it was recommended that the asphalt layer thickness be reduced because the substructure was sufficiently stiff. As of December 2002 the pavements are still in service. The test plan provides for trafficking with the HVS–A to determine response and performance using multiple- wheel loadings. The first two flexible pavement systems were designed, constructed, and instrumented using the USACE Layered Elastic Design program. The structure in- cluded a clay subgrade with a CBR of 6, a crushed lime- stone base with a CBR of 100, and an asphalt surface. Traf- ficking simulated a 727 aircraft with a modified overall load of 223 kN per tire on the main gear. The pavement configuration allowed for different base course thick- nesses. Local materials were used. The quality of HMA as con- structed did not meet specifications, but the test was com- pleted to explore the effect of noncompliance. Tests were continued until trafficking was no longer possible. The test sections were fully instrumented and included three five– level, multiple-depth deflectometers. The test program is continuing. SUMMARY In essence, APT results have been applied toward

80 • Validation and modification of design procedures, • Pavement configuration comparison in terms of per- formance, • Evaluation of material performance, • Performance prediction of pavements, • Evaluation and improvement of construction prac- tices, and • Evaluation of maintenance and rehabilitation prac- tices. In this chapter, the applications and issues emanating from APT applications were considered. It was evident that the applications that have been reported worldwide offer a vast reservoir of knowledge that should be tapped by all APT programs prior to embarking on new endeavors. The fields that were covered represent not only a wide range of applications but also some very valuable information that provides strong economic incentives for conducting APT. It is apparent that the pavement engineering knowledge base has been enhanced through APT and much of the knowledge is being applied. However, it remains a chal- lenge to integrate the information into the entire pavement engineering design and construction system. A comprehensive statement on key findings is provided in chapter nine. It was clear that all APT programs had al- ready contributed valuable information toward the building of a sound understanding of pavement performance under APT. However, gaps were identified with respect to vehi- cle–pavement–environment interaction and substantial ef- fort will be needed in this regard to create a sound basis for future development. Another matter that was evident was the advantage in establishing a close relationship between LTPP studies and APT. The benefit of progressive review of pavement per- formance under conventional traffic and the possibility of emulating this through intermittent APT trials offers scope for further study, and this should be pursued. Thus far the economic aspects of applications have not been specifically considered and these form the topic of the next chapter. It is clear that work remains in various fields of pave- ment engineering and in this regard the following should be noted: • The need to formalize the approach for extrapolating APT to full-scale trafficking on regular, conventional pavements; • The need to attempt improvement on the quantifica- tion of the impact of the environment on APT results. This is not only necessary to improve the understand- ing of pavement performance, but also to validate the credibility of APT; and • The need to establish repeatability of test findings and related confidence limits. Collaborative APT pro- grams have already demonstrated the benefits in this regard.