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.
126 APPENDIX B: EVALUATION AND SELECTION OF MIXTURE FATIGUE TESTS FOR USE IN NCHRP 9-59 INTRODUCTION Fatigue cracking occurs in asphalt pavements as the result of repeated bending due to traffic loads. This bending results in tensile strains in the asphalt layer which result in the formation of cracks, either at the bottom of the layer for thin pavements (classical bottom-up fatigue) or at the surface in thicker pavements (top-down fatigue). Top-down cracking is also influenced by tire- pavement interactions and by the oxidative aging of the asphalt binder used in the surface layer (Zhou, 2014). Bottom-up fatigue is considered to consist of two phases. The first, crack initiation, involves the development of micro-cracks that grow from a microscopic size to a critical length of approximately 7.5-mm. During the crack propagation phase, the cracks spread from the bottom of the pavement layer to the surface (Zhou, 2014). This literature review looks at three tests that have been developed to measure fatigue cracking potential: the Bending Beam Fatigue (BBF), the Overlay Test (OT), and uniaxial fatigue. While the studies found during the literature search provided much information about the fatigue behavior of asphalt mixtures, the focus of this literature review is placed on the relationships between these three tests, field cracking performance, and laboratory asphalt binder fatigue cracking tests. BENDING BEAM FATIGUE The Bending Beam Fatigue (BBF) test has been used to characterize the fatigue behavior of asphalt concrete since the 1960s but was not standardized until the early 1990s as part of the Strategic Highway Research Program (SHRP) (Tayebali et al., 1994). The test is performed by subjecting a 380-mm by 50-mm by 63-mm beam of compacted HMA (cut from either a laboratory compacted specimen or a field sample) to four point bending in a fixture like the one shown in Figure B-1.
127 Figure B-1. Bending Beam Fatigue Fixture. The beam is clamped into the fixture and placed into a temperature chamber at 20Â°C to simulate an intermediate pavement temperature. The test is performed by moving an actuator attached to the middle two clamps to apply a repeated load using either a sinusoidal or haversine wave form. Loading frequency ranges from 5-10 Hz, with 10 Hz being the most commonly used frequency. The BBF test may be performed using either a controlled stress or controlled strain mode with controlled strain being the most common. Strain levels may be set between 200 and 800 microstrain, which allows the user to account for pavement thickness (higher strains = thinner pavement). Test results include number of cycles to failure and cumulative Dissipated Energy (AASHTO T321, ASTM D7460). The fatigue life determined from the laboratory BBF test may be used with a shift factor to predict the fatigue life of an asphalt pavement constructed using the same mix. The failure point for the BBF was initially defined as the number of cycles for the stiffness of the beam to fall to 50 % of the original stiffness value. This value was rather arbitrary and did not always correlate well to the development of macro-cracks in the test specimen. There are two methods of determining the number of cycles to failure currently being used for beam fatigue testing. Both methods use the maximum tensile stress and strain per loading cycle to determine the flexural stiffness of the beam at each cycle. The stiffness calculated at the 50th loading cycle is typically defined as the initial stiffness of the beam. AASHTO T321 uses an exponential model to determine the relationship between loading cycle and calculated beam stiffness and sets the failure point at a 50% reduction in the initial stiffness from the model. ASTM D7460 uses an indicator called the âNormalized Stiffness X Cyclesâ (NSC) to define the failure point. For each cycle, NSC is defined as a ratio of the product of stiffness and number of cycles to the product of the initial stiffness and initial cycle. NSC is plotted versus cycle and the point at which the maximum NSC is reached is defined as the failure point (Rowe and Bouldin, 2000).
128 Dissipated energy is a measure of the energy used to cause damage in the BBF test specimen during loading. It is defined as the area under the stress strain curve. Changes in dissipated energy can be used to define the failure point and to indicate specimen behavior independent of loading rate. In 2000, Ghuzlan and Carpenter theorized that a sudden change in dissipated energy is an effective indication of damage accumulation and failure in BBF tests. A ratio of the change in dissipated energy between cycles (RDEC) and the dissipated energy of the previous cycle was defined. Changes in the relationship between the RDEC and the load during the BBF test can be used to indicate the point where failure occurs (Ghuzlan, 2000). The BBF can also be used to determine an endurance limit for fatigue in asphalt pavements. The endurance limit is the theoretical strain below which the mix will not crack and can be used as a tool for designing the thickness of asphalt pavements. In the laboratory, the endurance limit is defined as the strain level corresponding to a fatigue life of 50 million cycles in the BBF. This concept was verified by several studies (Carpenter et al. 2003, Peterson et al. 2004, Prowell et al. 2010). As part of NCHRP 9-38, Validating the Fatigue Endurance Limit for Hot Mix Asphalt Pavements, Prowell et al. verified the endurance limit concept and made recommendations for decreasing the amount of testing time required to determine the endurance limit. A single-stage Weibull Survivor function was recommended for estimating the fatigue life for BBF tests conducted as strain levels near the endurance limit (Prowell et al. 2010). CORRELATION OF BENDING BEAM FATIGUE RESULTS TO FIELD PERFORMANCE AND ASPHALT BINDER PROPERTIES. Many studies have been carried out to evaluate the use of the BBF test and to correlate it with field performance and with asphalt binder properties. Some of the studies are briefly described in this section. The BBF has been used to study the benefits of using a rich bottom layer to mitigate fatigue cracking. These layers are placed between pavement layers and provide two elements believed to extend fatigue life: increased asphalt content and decreased air voids. In 1996, Harvey and Tsai confirmed that for a typical Caltrans mix, using a rich bottom layer could increase the fatigue life 25-45% based on BBF results. Their results also showed an unexpected increase in fatigue life with increasing initial stiffness in the BBF. Because of the possible effect of other variables on the initial stiffness, it was suggested that initial stiffness not be included in the fatigue life models unless these relationships are clearly understood. Instead, relative changes in predicted pavement life should be used to evaluate the effect of mix variables on fatigue life (Harvey, 1996). A study by Deacon et al. in 1997 studied the new binder loss modulus value (G*sin(Î´)) proposed by SHRP to limit fatigue cracking and its effect on laboratory fatigue performance. While it was not clear what laboratory fatigue test was performed for this study, the procedure used was described as a controlled strain, flexural beam fatigue type test run on a similar size specimen as the BBF procedure. Thirty-two asphalt mixtures were evaluated, using eight different asphalt binders, two aggregates, and two air void contents. The results of the study showed that asphalt binder G*sin(Î´) correlated well with laboratory fatigue resistance for mixtures and could identify changes in asphalt binder properties in mixes that were otherwise identical. The researchers cautioned, however; that there was not a good enough correlation between the two properties to use G*sin(Î´) by itself as an indicator of fatigue resistance of asphalt mixtures (Deacon et al., 1997)
129 In 2001, Bahia et al. (Bahia et al., 2001) used the BBF as part of NCHRP 9-10, Characterization of Modified Asphalt Binders in Superpave Mix Design, to evaluate the effectiveness of the asphalt binder fatigue parameter G*sin(Î´) (AASHTO M-320, Performance- Graded Asphalt Binder). For this study, four aggregate gradations were used: coarse limestone, coarse granite, fine limestone, and fine granite. Mixtures were made using each of the gradations blended with nine different modified asphalt binders ranging in high temperature PG grade from 58 to 82. After mixing, the mixtures were short-term oven aged for 4 hours at 135Â°C then tested in the BBF at a temperature corresponding to the temperature where the asphalt binder used in the mix met the G*sin(Î´) requirement of 5,000 kPa. A loading frequency of 10Hz was used for the test and controlled strains ranging from 250 to 750 microstrain. The fatigue failure point for this testing was defined as a 50% reduction in the initial beam stiffness. Figures B-2 and B-3 show a summary of the results from the study. Figure B-2. Fatigue Life of NCHRP 9-10 Mixes with 9 Modified Asphalt Binders (Bahia et al., 2001)
130 Figure B-3. Correlation between Asphalt Binder G*sin(Î´) and Mixture Fatigue Life (Bahia et al., 2001) Figure B-2 shows that for this testing, the BBF appears to be sensitive to both binder modification type and aggregate properties, with binder type having a larger effect than aggregate type. The effect of the binder also appears to be independent of the aggregate effect. Given that the BBF test was run at temperatures that should have corresponded to an equal binder stiffness value, these results raised concerns about the validity of the asphalt binder G*sin(Î´) parameter. Figure B-3 shows the correlations between asphalt binder G*sin(Î´) and mixture fatigue life. For this comparison, the binder tests were run using RTFO-aged binder at the same temperatures that gave G*sin(Î´) = 5,000 kPa for the PAV-aged binders. The RTFO-aged binder was used for this comparison instead of the PAV-aged binder specified in AASHTO M-320 because the researchers felt that the RTFO-aging procedure might be a better match for the short-term aging used for the mixes. These plots show that there is very little correlation between the mixture fatigue life and the binder G*sin(Î´) parameter (Bahia et al., 2001). In 2001, Stuart et al. (2001) presented the results of a study conducted using the FHWA- ALF. Four lanes of the FHWA-ALF experiment were used to validate the BBF and the Dynamic Shear Rheometer (DSR) intermediate temperature parameter, G*sin(Î´). Lanes 1 and 2 were constructed with a 100-mm layer of HMA over a 560-mm unbound crushed aggregate base and a
131 prepared subgrade. Lanes 3 and 4 utilized a 200-mm layer of HMA over a 460-mm unbound crushed aggregate base and a prepared subgrade. Lanes 1 and 3 used a PG 58-34 asphalt binder, while lanes 2 and 4 used a PG 64-22 binder. ALF testing was performed at three temperatures: 10, 19, and 28Â°C. BBF tests were performed on laboratory compacted ALF mixes using a strain controlled sinusoidal load at levels of 300, 600, 900, and 1100 microstrain at temperatures corresponding to those used in the ALF. All the mixes were tested to 40% of their initial stiffness. The researchers directly compared BBF fatigue life to the ALF fatigue life without using shift factors. Results are summarized in Table B-1. Figure B-4 shows a graphical summary of the number of ALF passes to a crack length of 50-mm compared to the BBF fatigue life. Table B-1. Results of FHWA-ALF BBF Comparison
132 Figure B-4. Comparison of FHWA-ALF and BBF fatigue life Both Table B-1 and Figure B-4 show a good correlation between FHWA-ALF fatigue life and BBF fatigue life, with the best comparison occurring at the 28Â°C test temperature. The worst correlation was at 10Â°C. At all temperatures, the BBF did a reasonable job of quantifying the increase in fatigue life due to increased pavement thickness. The BBF was also able to identify, in most cases, an increase in fatigue life due to the softer PG 58-34 asphalt binder and due to the use of a thicker pavement. Figure B-5 shows the correlation between binder G*sin(Î´) and FHWA-ALF fatigue life. For the thinner pavements, the G*sin(Î´) parameter ranked relatively well compared to the ALF fatigue life and provided the expected relationship of decreasing mix fatigue life with increasing asphalt binder stiffness. The opposite effect was seen for the thicker pavement, with fatigue life increasing with increasing binder stiffness (Stuart, 2001).
133 Figure B-5. Comparison of Binder G*sin(Î´) and FHWA-ALF Fatigue Life In 2005 and 2007, Tsai and Johnson presented the results of two studies intended to expand on Deaconâs 1997 study. These studies, performed for the Pacific Coast Conference on Asphalt Specification (PCCAS), investigated the effects of nine asphalt binders on the fatigue life of asphalt pavements. Laboratory binder characterization was done to obtain, among other parameters, the G*sin(Î´) at an intermediate temperature corresponding to the PG grade of the binder. G*sin(Î´) was tested 20Â°C, a temperature corresponding to the BBF test temperature. PAV aged asphalt binders were used for this determination. Other asphalt binder testing was done to determine the relaxation spectra for the binders and their molecular weight distributions. Beam fatigue testing was performed at three temperatures (10, 20, and 30Â°C) and two strain levels (200 and 400 microstrain). Testing was conducted until failure and the results used to obtain the ratio of stiffness at each cycle to the initial stiffness. This value was plotted against the number of cycles and the number of cycles to failure defined as the cycle where the ratio was equal to 0.5. Figure B-6 shows a summary of the asphalt binder fatigue parameter results from
134 this study compared to the laboratory fatigue results. Very little correlation was found between the laboratory BBF results and the asphalt binder G*sin(Î´) values (Tsai et al, 2005). Figure B-6. Comparison of PCAS Asphalt Binder G*sin(Î´) and Mix Nf Johnson et al. (2007) used the same mixture and binders from Tsaiâs study to further investigate the use of a different asphalt binder fatigue test. The proposed test was a strain- controlled, cyclical test performed using the DSR. Repeated cycles of shear loading were applied to a test specimen, and the decrease in complex shear modulus with time was recorded. DSR tests were performed at 10, 20, and 30Â°C to match the mixture fatigue tests from the earlier study. Strain levels of 1 and 2% were evaluated at a testing frequency of 10 Hz. Figure B-7 shows the results of the initial DSR testing compared to the mixture fatigue results from the BBF. The results in Figure B-7 show a good correlation between the BBF results and the DSR cyclical fatigue test. The researchers were concerned, however, about the possible effects of healing in the sample during the test. The first set of DSR tests used a âprecision-samplingâ method in the DSR software to maintain the strain level in the test. This method pauses the test between cycles in order to calculate the necessary stress to apply to the sample to get the desired strain. A second set of DSR tests was performed using a strain-control method that did not pause between cycles. Figure B-8 shows the results from the second set of DSR tests compared to the mixture BBF results; the second set of test results did not provide a good correlation between binder and mixture fatigue results. The researchers stated that more work was needed to fully understand fatigue behavior of asphalt binders tested in the DSR (Johnson, et al., 2007).
135 Figure B-7. Initial DSR Fatigue Results Compared to Mixture Fatigue From BBF Figure B-8. DSR Fatigue Results Using Continuous Loading The NCAT Test Track structural sections from the 2003 testing cycle were used by Willis and Timm to validate the use of strain thresholds in the design of asphalt pavements. This research focused on setting a horizontal strain limit that would prevent fatigue cracking in an asphalt pavement. Laboratory testing for this study included the use of BBF data to determine a fatigue endurance limit based on the results from the BBF test 20Â°C and multiple strain levels. Phase 1 of the study compared the laboratory endurance limit to a field strain value calculated using strain-temperature equations developed in an earlier part of the study. Phase 2 of the study compared the magnitude of the endurance limit from the laboratory to the measured field strain
136 distributions in the test track sections. Phase 3 of the study compared the laboratory endurance limit based on the BBF and the developed field strain distribution using a fatigue ratio between the measured strain values and the laboratory endurance limit. The first two phases had difficulty finding a clear relationship between laboratory results, field strains, and pavement performance. The third phase showed that there was a distinct difference in the fatigue ratios for test track sections that failed due to cracking and sections that did not, with the cracked sections having higher ratios (Willis et al., 2010). Way, et al. presented the results of a study reviewing the use of the BBF test in the design of several Arizona paving projects dating from 1993 to 2012. The projects used were mostly overlay projects and represented various asphalt mix types and grades of binders. Comparisons between mixture properties showed similar trends to other studies already discussed, with the BBR showing sensitivity to mixture and binder properties. Annual field cracking surveys for the projects studied showed that the BBF results correlated reasonably well with actual field performance of the projects (Way et al., 2012) A study presented in the International Journal of Fatigue in 2015 (Mannan, 2015) compared laboratory mixture fatigue results to asphalt binder fatigue results measured using two provisional test procedures. Two asphalt mixtures, one with 35% recycled asphalt pavement (RAP) and one with no RAP, were evaluated. BBF was performed on the mixes at 20Â°C using four strain levels (400, 600, 800, and 1000 microstrain) and three loading frequencies (1, 5, and 10HZ). Asphalt binder was extracted and recovered from the mix and tested for fatigue resistance using the Linear Amplitude Sweep Procedure (LAS) described in AASHTO TP 101. A time sweep procedure similar to the one used by Johnson (2007) was also used to measure the fatigue resistance of the asphalt binder. Both asphalt binder procedures used for this study were performed using a Dynamic Shear Rheometer (DSR). The time sweep procedure is a cyclical shear loading that looks at the degradation of the complex shear modulus (G*) of the binder over time. For this study, the time sweep was performed at 20Â°C and at three frequencies (1, 5, and 10 HZ) and four strain levels (4, 6, 8, and 10%). The LAS procedure damages the DSR sample by applying a linearly increasing strain to the sample. For this study, the LAS was also performed at 20Â°C and used strains ranging from 1 to 30%. Figures B-9 and B-10 show the results of the comparison between laboratory mixture fatigue and the asphalt binder fatigue tests used in this study. The results of this study showed that there were strong correlations between mixture fatigue life and both binder tests evaluated. The study also showed that the mixture fatigue life from the BBF test decreased with the addition of RAP to the mix, as would be expected due to the increased stiffness from the RAP binder (Mannan, 2015).
137 < Figure B-9. Comparison of Mixture and Asphalt Binder Fatigue for 35% RAP Mixture Figure B-10. Comparison of Mixture and Binder Fatigue Life (Time Sweep) UNIAXIAL FATIGUE TESTING The uniaxial fatigue test is similar to the flexural fatigue test, but specimens are cylindrical in shape and are loaded in simple tension and/or compression rather than in flexure. This has the advantage that stresses and strains are uniform across the central portion of the specimen. Data from uniaxial fatigue tests are often analyzed using concepts from continuum damage theory, although this type of analysis is not a required part of uniaxial fatigue testing (Kim and Little, 1990; Underwood et al., 2012; Christensen and Bonaquist, 2012). Regardless of the type of analysis used, fatigue data from uniaxial tests is most often used to evaluate the way damage accumulates in the asphalt mixture as the number of loading cycles increase. The most common way of evaluating damage growth is by plotting the reduction in modulus against a damage function, usually designed so that damage curves collected over a range of conditions collapse to a single âmaster curveâ for fatigue damage. This is one of the main advantages of this type of testingâthe rate of damage that occurs during fatigue loading can be characterized in a single
138 damage curve with just a few parameters. These parameters can then be correlated to other factors, such as mix composition and various asphalt binder properties. As with flexural fatigue testing, uniaxial testing can be done either in stress or strain control, although properly constructed damage curves should be independent of the manner of testingâanother advantage of uniaxial fatigue tests. However, as with flexural testing, strain controlled uniaxial fatigue testing often results in very gradual damage accumulation with a difficult-to-define failure point. In fact, most of the emphasis in analyzing uniaxial fatigue test data has been in the rate of damage accumulation rather than characterizing and modeling the point of failure, although more emphasis has recently been placed on developing failure criteria for this type of data (Kim and Little, 1990; Underwood et al., 2012). Along with the advantages mentioned above, there are however several disadvantages in uniaxial fatigue testing compared to flexural fatigue testing. One disadvantage is that because uniaxial fatigue testing is relatively new, there is not yet a large body of data linking data from this test to field performance. Another important disadvantage is that the vast majority of fatigue tests on asphalt mixtures to date have been done using flexural fatigue testing, and so there is a large body of knowledge revolving around this type of testing and relatively little using uniaxial fatigue testing. Most methods of flexible pavement design, including the latest AASHTO procedure (AASHTOWare Pavement ME Design) have used data based on flexural fatigue testing, and many have an option for including input from flexural fatigue tests on specific asphalt mixtures. For these reasons, most practicing pavement engineers are more familiar with flexural fatigue tests and are more comfortable interpreting flexural fatigue data; uniaxial fatigue testing is at this point largely a research tool. Although continuum damage theory is potentially a very useful tool in characterizing the fatigue behavior of asphalt mixes and for the design of flexible pavement it is unfortunately also very complex and difficult to understand. The fact that uniaxial fatigue testing of asphalt mixes has largely been used within the context of continuum damage theory has probably increased the amount of confusion many engineers feel concerning this test and the resulting data. Although the disadvantages of uniaxial fatigue testing are significant, it does have significant advantages, including the possibility that this will become the standard method for characterizing the fatigue performance of asphalt concrete mixtures in the future. TEXAS OVERLAY TEST A common method of rehabilitating old asphalt and concrete pavements involves the placement of an asphalt overlay. These overlays, compared to new asphalt pavements, are relatively thin and are placed over already damaged structures. Reflective cracking can occur in asphalt overlays at locations corresponding to existing cracks or joints in the underlying pavement. Movement in the vicinity of these joints or cracks, caused by bending or shearing action due to traffic loading or temperature changes causes high concentrations of stress to form at the bottom of the overlay. Eventually, these stresses will lead to cracks which can propagate to the surface of the pavement. Figure B-11 shows the effect of traffic loading on an asphalt overlay. Figure B-12 shows the common failure modes associated with slab movements (Zhou,
139 2014). Figure B-11. Reflection Cracking Mechanisms Figure B-12. Common Failure Modes Associated with Slab Movements
140 The Texas Overlay Test (OT) was developed in the late 1970s at the Texas Transportation Institute (TTI) by Lytton et al. as a means of simulating the movement of joints or cracks in a pavement, as discussed earlier. The original apparatus, shown in Figure B-13, consists of two steel plates. One plate is fixed, while the other can move horizontally, thus simulating the opening and closing movements believed to cause reflective cracks (Germann et al., 1979). Figure B-13. Original Overlay Tester Concept The original OT device had some limitations, mainly associated with the large test specimen size required. The 375-mm (15-in.) length of the specimen made it difficult to fabricate samples in the lab, and even more difficult to obtain field samples (Zhou et al. 2005). Work by Zhou, et al. in the early 2000s focused on upgrading the existing OT equipment to fully automate the test procedure and to determine a sample size that was easier to work with. This project came about due to concerns that the asphalt mixture design procedure in Texas was focused heavily on rutting resistance, leading to the design of asphalt mixtures that had good rutting properties, but were very stiff and had low asphalt contents. These stiff, dry mixes were prone to cracking. The intent of this study was to develop a procedure that could be used in conjunction with the Hamburg Wheel Tracking test for rutting resistance to ensure that the mixes produced met both rutting and cracking resistance requirements (Zhou et al. 2003). As part of the study conducted by Zhou et al. (2003), a smaller sample size was developed that was easier to obtain using existing Superpave Gyratory Compactor (SGC) molds and standard size drill bits for field cores. The recommended test specimen size was a 150-mm (6- in.) by 75-mm (3-in.) wide by approximately 50-mm (2-in.) tall specimen that could be prepared from either an SGC specimen or a field core. The height of the specimen was chosen to correspond to the average 2-in overlay thickness used in Texas. Figure B-14 shows the upgraded OT equipment.
141 Figure B-14. OT Test Equipment To run the OT, the test specimens are glued to the plates, using tape to prevent the glue from seeping into the opening between the moving and non-moving plate and to provide a consistent gauge length. The OT is then performed by opening the space between the test plates to a width of 0.64-mm (0.025-in), then repeatedly closing it at a rate of 1 cycle per 10 seconds until the specimen reaches the failure limit. A typical test temperature of 25Â°C is used. During the test, tensile load and displacement (measured by a linear variable differential transducer (LVDT) under the fixed plate) are recorded every 0.1 seconds. Figure B-15 shows a typical set of OT test results. Figure B-15. Typical OT Test Results As shown in Figure B-15, the OT results can be divided into 3 distinct phases. Phase 1 consists of the crack initiation and early propagation phase and is distinguished by a rapid drop in test load as cracks develop and move through the test specimen. Phase 2 is the late propagation phase and is indicated by the slow decrease in the maximum load applied to the specimen, with the load staying in phase with the displacement. Phase 3 is sample failure and is distinguished by the small applied load occurring before the displacement reaches a maximum
142 value. The sample is considered to have reached failure when the peak load during a cycle is reduced at least 93 % compared to the peak load of the first cycle applied. If failure is not reached, the test is terminated at 1200 cycles. One proposed benefit of the OT over other fatigue testing methods is its ability to measure not only the crack initiation phase, but also the crack propagation phase (Zhou et al. 2003, TX-248-F-09) A recent upgrade of the OT equipment by IPC Global provided an OT device for use in the Asphalt Mixture Performance Tests (AMPT). In this device, the OT is performed vertically. The top plate is fixed while the bottom plate is opened and closed. Two LVDTs are used, one for measure the actuator displacement and one for measure the opening displacement. Figure B-16 shows the IPC Global device (IPC Global, 2012). Figure B-16. IPC Global AMPT OT Device A study of the sensitivity of the OT showed that it was sensitive to mixture properties, including temperature, asphalt content, asphalt type, aggregate type, and air voids. The OT results were also found to be sensitive to the displacement used to run the test. Studies of repeatability found that the test provided results that had similar standard deviations and coefficient of variation values compared to other mixture cracking tests and that three replicates were enough for running the test. Based on the results of five case studies in Texas, the OT seemed to show good correlation with field reflective cracking performance. The OT was also found to have good correlations with Bending Beam Fatigue results and low temperature performance of the mixtures. Based on this work limiting criteria were proposed for the OT results based on the number of cycles to failure (Nf): Nf > 750 cycles for rich bottom layers and Nf > 300 cycles for all other mixture types (Zhou, 2005). These limits were used by Holdt and Scullion in 2006 as part of a study evaluating different materials used to mitigate joint reflection cracking in Texas and were found to be able to distinguish materials that performed well (Holdt et al. 2006). In 2007, Zhou et al. reported that the 300 cycles limit was also reasonable for evaluating the fatigue cracking resistance of mixtures and proposed a method for using the OT device as a
143 means for predicting the fatigue life of asphalt pavements using a modified version of Parisâ Law. For this analysis, Zhou used a measurement-based method to determine the Paris Law parameters using the normalized OT load results. The normalization of the load allowed a representation of the damage in the specimen during the test relative to its initial state. The traditional fatigue models used in the Mechanistic Empirical Pavement Design Guide (MEPDG) showed poor correlation to fatigue performance measured in the Federal Highway Administrationâs Accelerated Loading Facility (FHWA-ALF). This led to the desire for a better prediction model for fatigue life. The approach developed by Zhou differed from previous methods in that it considered not only the crack initiation stage but also the crack propagation stage. The OT tester results allow for the measurement of cracking characteristics associated with both of these stages and allows a more accurate prediction model (Zhou, et al., 2007). Further studies by TTI have continued to provide improvements in the test method and to recommend its use in the design of asphalt pavements. A major area of focus has been on improving the repeatability of the test. Items of interested have been sample fabrication, testing temperature, gluing methods, and plate opening. Continuous efforts are also being made to improve the proposed fatigue cracking model (Hu, 2008; Zhou, 2011; Zhou, 2013; Walubita, 2012). CORRELATION OF OVERLAY TESTS RESULTS TO FIELD PERFORMANCE Many studies have been done, particularly by TTI, to validate the results of the OT compared to field performance. In 2003, a validation of this work was done by taking cores from several Texas highways. One of the locations had very little cracking after 10 years of service (overlay placed over a cracked, stabilized base) in both the control sections (SPS5 Virgin) and in a section containing recycled asphalt pavement (RAP) (SPS5 Recycled) while the other sections showed significant cracking after only a few months of service. Figure B-17 shows a comparison of the OT results from the pavement sections evaluated. The OT was able to distinguish between the sections with and without cracking. Figure B-17. OT Test Results on TXDOT Highway Sections
144 The OT was also able to distinguish between different asphalt binder sources and showed that stiffer asphalt binders had much lower cracking resistance in the OT than softer binders. Figure B-18 shows a comparison of OT results for a set of test specimens made using different binder types in a set of mixes where all other properties were the same. It can be seen from Figure B-18 that the softer asphalt binders, particularly those with polymer modifiers perform better in the OT than stiffer asphalt binders (Zhou, 2003). Figure B-18. Comparison of OT Results for Mix Sample with Different Asphalt Binders Further work by Zhou (Zhou, 2005) analyzed additional roadways in Texas. Figure B-19 shows OT results for three more pavements, all of which showed reflective cracking within a few years of construction. Figure B-19: OT Test Results on Texas Highways
145 In addition to the results shown, a section of the SH3 project was constructed using a crack resistant material called Strata as an interface between the overlay and the existing pavement. OT tests from this section showed a fatigue life of greater than 750 cycles. A final comparison in this study used three sections from MnRoad. Cores were taken from the three sections (15, 18, and 20) and shipped to TTI for overlay testing. Table B-2 shows the comparison between the OT results and the cracking observed in the sections. Table B-2. MnRoad Comparison Between OT and Field Cracking It can be seen from Table B-2 that the OT provided a good ranking of cracking resistance compared to the actual observed cracking on the test sections. Based on the work done in this study, a proposed OT fatigue life limit of greater than 750 cycles was proposed for overlays using rich bottom layers from crack relief and greater than 300 cycles for all other materials (Zhou, 2005). In 2007, cores taken from the FHWA-ALF fatigue test lanes were sent to TTI for testing. The OT results were compared to the measured cracking performance in the accelerated loading facility. OT testing was conducted at 19Â°C to match the test temperature in the FHWA-ALF. The results showed an improved correlation with the actual fatigue performance of the ALF pavements compared to traditional methods of predicting fatigue life. Figure B-20 and Table B- 3 show a summary of the actual cracking performance of the FHWA-ALF sections and the predicted fatigue life calculated using the proposed fatigue cracking model. The results show that the number of cycles to failure correlates well with the actual observed cracking performance of the FHWA-ALF sections (Zhou, 2007).
146 Figure B-20. FHWA-ALF Test Section Cracking Performance Table B-3. FHWA-ALF Cracking Performance Calculated Using Fatigue Model (Zhou, 2005).
147 In 2008, Bennert el al. used the OT as one method of evaluating the effectiveness of an asphalt interlayer type mix (reflective crack relief interlayer (RCRI)) at reducing the reflective cracking in composite pavements. In this study, falling weight deflectometer (FWD) and weigh- in-motion sensors were used to determine the vertical movements at the concrete slab joints in the underlying pavement. Cores of the underlying concrete were tested to determine the coefficient of thermal expansion of the concrete and this value was used to estimate the horizontal movements in the concrete due to environmental changes. The horizontal and vertical deflections where then used in laboratory performance tests to evaluate the performance of HMA mixtures commonly used in New Jersey for overlays. Three test sections were constructed, two consisting of an RCRI overlaid by a 12.5-mm Superpave mix and the other test section consisting of a 12.5-mm Superpave mix overlaid with a 9.5-mm Superpave mix. OT tests showed that the RCRI interlayer should have a fatigue life of over 46,000 cycles while the other two mixes used had fatigue lives in the range of 22-24 cycles. Actual performance, however; showed cracking in all sections after 6 months. Forensic analysis of the test sections attributed the poor performance of the RCRI sections to construction issues which led to poor bonding between the layers and also to inadequate compaction thickness of the RCRI layer (Bennert et al., 2008). A study by Zhou, et al. (2010) used the OT to analyze six different types of overlays placed at the University of California Pavement Research Center. The goal of this research was to further validate the proposed fatigue cracking model. A 90-m test road was built at the University of California, consisting of compacted clay subgrade, a 410-mm aggregate base, and a 90-mm dense HMA surface. A heavy vehicle simulator (HVS) was used to induce fatigue damage to a density of 2.5 m/m2. Six different overlay types (control, plus 5 different rubber modified binders) were placed over the damaged pavement. The HVS was again used to induce fatigue damage of 2.5 m/m2 in the overlays. Cores were taken from each section and submitted to TTI for OT testing. Table B-4 shows a summary of the results from the study. The results of the study showed good correlations between both the OT test results and the fatigue life predicted by the model compared to the actual number of equivalent single axle loads (ESALs) to failure as determined by the HVS. (Zhou, et al., 2010). Mixes from five test sections located on the National Center for Asphalt Technology (NCAT) test track were used by Ma to recommend a new method of determining the failure point of the Overlay Test based on a normalized load cycle procedure. OT results from this study show that the procedure provided similar rankings for the materials from the test sections compared to the Bending Beam Fatigue procedure. The OT results also correlated well with the field performance data of the five test sections seen on the test track (Ma, 2014).
148 Table B-4. Summary of OT Results Section ID Fracture properties from OT Predicted ESALs to 50% reflective cracking (from model) Measured ESALs to 2.5 m/m2 cracking (from field HVS test) A n 586 2.77Ã10 â6 4.974703 No crack propagation after 91 million ESALs None after 88 million ESALs 587 3.41Ã10 â9 4.003453 50.4 million 60 million 588 6.10Ã10 â10 4.9019 8.3 million 16 million 589 2.44Ã10 â8 5.543798 No crack propagation after 66 million ESALs None after 66 million ESALs 590 6.52Ã10 â8 5.184681 No crack propagation after 37 million ESALs None after 37 million ESALs 591 3.44Ã10 â10 4.763871 None after 91 million ESALs None after 91 million ESALs MIXTURE CRACKING TESTS SELECTED FOR USE IN RELATED PROJECT NCHRP 9-57 The objective of NCHRP 9-57 was to develop an experimental design for field validation of selected laboratory tests to evaluate the cracking potential of asphalt mixturesâclearly a project closely related to NCHRP 9-59. After the Asphalt Cracking Tests Workshop organized as part of NCHRP 9-57 in February 2015, the following mixture cracking tests are recommended for including in NCHRP 9-57âs experimental design to evaluate three types of cracking. 1. Bottom-up fatigue cracking a. Flexural fatigue testing b. Semi-circular bend test at intermediate temperature 2. Top-down fatigue cracking a. Semi-circular bend test at intermediate temperature b. Indirect tension test with dissipated creep strain energy and energy ratio 3. Reflection cracking a. Overlay test b. Semi-circular bend test at intermediate temperature c. Flexural fatigue testing The two mixture flexural tests recommended for use in NCHRP 9-57 through consultation with an expert panel were BBF and the OT. Uniaxial fatigue testing was not recommended. This
149 supports the use of these tests in NCHRP 9-59 for both technical reasons, and because this suggests that findings based upon these tests would have the best chance of being accepted by other pavement researchers and engineers. MIXTURE FATIGUE TEST METHODS SELECTED FOR USE IN NCHRP 9-59 Both the BBF and OT methods have been evaluated to determine their effectiveness at predicting asphalt mixture cracking resistance. The BBF test showed reasonable correlations with field performance and was able to distinguish between different mixture properties. The BBF showed mixed results when compared to the asphalt binder fatigue tests (G*sin(Î´), LAS, cyclical testing). While some research showed a correlation with the binder parameter, other research was unable to show any correlation at all. This lack of correlation is likely not due to an issue with the BBF procedure, but may be an indication of the inability of the binder parameter by itself to account for the effects of other mixture properties such as asphalt content, aggregate type, air voids, etc. on fatigue life (Deacon, 1997). For the OT, the studies discussed in this literature review show that this procedure was also able to correlate reasonably well with field performance and can differentiate between mixture properties. The OT was also able to identify the benefits of reflective crack relief interlayers. The selection of BBF and the OT as preferred test methods in NCHRP 9-57 lends additional support for using these tests in NCHRP 9-59. Correlations between uniaxial fatigue data and field performance are not nearly as extensive as for the BBF and OT, largely because this is a relatively new procedure that has not yet been widely used by pavement engineers and researchers. Because of the rigor of this test method and the related continuum damage theory, significant resources have gone towards its development and implementation, and it will likely become more widely used in the future. After consultation with the panel, it was decided to use both bending beam flexural and uniaxial fatigue testing in NCHRP 9-59. Although limited overlay tests were performed early in the project, modifications of the work plan ultimately lead to discontinuing this test. Because the available time and funding were not adequate to allow testing of all mixes using both fatigue test methods, it was decided to rely more on uniaxial fatigue testing. All 16 NCHRP 9-59 mixtures were tested using uniaxial fatigue testing, while only 9 of 16 mixes were tested using flexural fatigue.