Skip to contentUnited States Department of Transportation - Federal Highway AdministrationSearch FHWAFeedback


<< PreviousContentsNext >>

An Investigation of the Endurance Limit of Hot-Mix Asphalt Concrete Using a New Uniaxial Fatigue Test Protocol
Final Report


Fatigue cracking is one of the major distresses in hot-mix asphalt (HMA) concrete pavements, and a fatigue distressed pavement is very costly to repair. Fatigue cracking can originate from the bottom of the pavement layer propagating to the top, or it can originate from the pavement surface propagating to the bottom. The magnitude and frequency of loads, environmental conditions, engineering properties of the hot-mix asphalt concrete, condition of underlying layers, and pavement structure are all contributing factors to fatigue cracking.

A major controlling factor for bottom-up fatigue cracking is the magnitude of the tensile strain at the bottom of the HMA layer. It is a well-established concept in pavement design that decreasing this tensile strain results in an increase in pavement fatigue life. It is also believed by some asphalt and pavement design experts that an endurance limit exists for HMA. If the tensile strain is maintained at levels below the endurance limit, the pavement will have infinite fatigue life. National Cooperative Highway Research Project NCHRP 9-38 is an ongoing study investigating the validity of the endurance limit hypothesis and the effect of HMA materials, mixture factors and testing conditions on the endurance limit.


The research study presented in this report was conducted to investigate existence of an endurance limit for a limited number of hot-mix asphalt mixtures under uniaxial sinusoidal tension-compression loading.

Literature Review

The presence of an endurance limit is a well-recognized behavior for ferrous materials such as steel. However, limited research has been conducted for HMA and the existence of an endurance limit, and factors affecting it, if it does exist, are not well known at this point. The possible existence of an endurance limit for hot-mix asphalt concrete (HMAC) has been ignored by most researchers studying the fatigue behavior of asphalt mixtures in the laboratory. However, as pointed out by Carpenter et al., in 1970 Monismith et al. suggested that an endurance limit exists for HMAC and suggested 70 µ-strain as a likely value( 1 ). Carpenter et al. recently addressed the question of the existence of an endurance limit in HMAC, concluding that an endurance limit does exist and that it is in the range of 70 to 90 µ-strain at 20°C for a loading frequency of 10 Hz( 2 ). Based on a damage analysis of laboratory test results obtained from a uniaxial tension-compression test, Soltani postulated the presence of an endurance limit for two mixtures tested at 10°C and 10 Hz. He estimated endurance limits of 30 and 80 µ-strains for the two mixtures that contained unmodified binder and modified binder, respectively( 3 ).

Research Plan

The testing configuration and protocol used in this study to establish the presence of an endurance limit for asphalt mixtures is very similar to that used previously for FHWA/PennDOT-sponsored fatigue research( 4 ). Major deviations from the previous study are in the preparation of test specimens, some improvements in the control, data acquisition and analysis software, and the use of a more appropriate load cell with smaller capacity. For the research presented here, cylindrical test specimens were cored and saw cut from Superpave Gyratory Compactor specimens. A photograph of the test specimen, fixtures, transducers and thermocouples is shown in Figure 1.

Figure 1. Photograph of test fixtures, specimen, transducers and thermocouples.

Figure 1. Photograph of test fixtures, specimen, transducers and thermocouples.

The testing was conducted in the controlled strain (displacement) mode using cylindrical test specimens. In this type of loading the state of stress and strain inside the middle portion of the test specimen can be considered uniform, as long as the specimen is homogeneous, the ends of the specimen are restrained from rotation, and the applied load is concentric with the axis of the test specimen. The following test conditions were selected for this study:

  • Test loading: sinusoidal centered at zero (push-pull configuration);
  • Test frequency: 10 Hz;
  • Test temperature: 10°C;
  • Strain: 20 to 30 µ-strain;
  • Specimen: cylindrical, 75.5-mm diameter by 120-mm height;
  • Gauge length: 75 mm; and
  • Test mode: constant strain (controlled displacement by one transducer).

Test Equipment

Details of the test equipment are reported elsewhere( 4 ). Details regarding the testing equipment and data acquisition relevant to this study are presented below.

Data Acquisition System and Testing Machine

The electronics included an Instron Model 8800 System equipped with FastTrack™ software and hardware. The FastTrack™ 8800 unit, which contains two general purpose interface bus (GPIB) boards each with four channels of data acquisition, is used to control the servo valve that controls the pressure in the hydraulic actuator. The first GPIB board communicates with the position, load, strain 1, and strain 2 channels. The second GPIB board communicates with the strain 3 channel.

The FastTrack™ 8800 unit, which is itself a computer, receives commands from either a hardware source or from another computer that generates commands through software. Instron provides a separate Man-Machine Interface (MMI) that can be used as a hardware source. The Instron multi-axial library (which is appropriate when more than one GPIB board is in use) was used in developing the LabVIEW™ data acquisition and control program for this study.

Specimen displacement was measured with three Epsilon transducers that were interfaced with the FastTrack™ 8800 hardware. Data were collected from the actuator LVDT, the load cell and each of the three displacement transducers at the rate of 100 points/cycle/channel (1 KHz/channel). Data were collected and analyzed using LabVIEW™. Excel™ files of reduced data were generated by LabVIEW™ for plotting and further analysis. The approach presented by Chapra and Canale( 5 ) was used to apply sinusoidal signal curve fitting to the data obtained from all channels. The curve fitting was used to generate the stress and strain amplitude and phase angle data.

An MTS 5-KN load cell was used in this study to measure the load magnitude. This lower-capacity load cell was utilized to improve the resolution in measuring the load since very low load levels were needed to induce the targeted strain levels. This is a change from the previous study( 4 ) where a 100-kN load cell was utilized.

The rate of data logging was adjusted in concordance with the rate of change in the modulus or to the importance of the loading cycles. Consequently, at the beginning of each stage when the modulus is either dropping (Stages I and II) or recovering (Stage III) the data for the first 100 cycles and the last 100 cycles of the stage were acquired. For other cycles of each stage, the cycles used for data collection can be found in Table 1.

Table 1. Data acquisition schedule for each stage.
Range in CyclesNumber of Cycles between Measurements
First 100 cycles of Stage1
Cycles 101 to 1,000 of Stage20
Cycles 1,001 to 1,000 of Stage100
Cycles 1,001 to 100 cycles prior to end of Stage1,000
Last 100 Cycles of Stage1

Temperature Measurements

Eighteen thermocouples (TC) were used for the measurement of temperature at various points. The specimen surface temperature was measured using nine thermocouples (Figure 2). One TC was placed in the middle of each pair of transducer contact points (thermocouples 2, 5, and 8). Above and below each transducer, three thermocouples were placed at the very top of the specimen (thermocouples 1, 4, and 7), and three thermocouples were placed at the very bottom of the specimen (thermocouples 3, 6, and 9). TC 10 was placed in the water bath and TC 11 was placed next to the RTD that controls the test chamber temperature. Thermocouples 12 and 13 were placed on the upper and lower aluminum heads and TC 14 was placed on the load cell. The air temperature of the walk-in chamber was monitored with the built-in thermocouple of the multiplexer. TC 15 was placed at the mid center of a dummy specimen and TC 16 at its center. The laboratory ambient air temperature was monitored with the built-in thermocouple of the Micrologger.

An AM25T multiplexer and a Campbell Scientific CR23x Micrologger were used to collect and store temperature data. Temperature measurements were obtained once per minute during the fatigue testing and synchronized in the time domain with the dynamic measurements obtained with the FastTrack™ software.

Data Synchronization

The complexity of the data acquisition system requires caution with respect to the synchronization of the data. If the computer in the Fast Track 8800 cannot acquire and process data fast enough to prevent buffer overruns, data from the various channels will not be properly synchronized in time. The results of this problem can be seen as excessive variability in the plots of modulus versus loading cycles as shown in Figure 3.

Figure 2. Location of thermocouples on the test specimen.

Figure 2. Location of thermocouples on the test specimen.

As discussed above, the data are collected from two separate GPIB boards (number 3 and 4) within the Fast Track 8800. Except for transducer 3 all the channels are located on GPIB 3. The Fast Track 8800 system is in charge of keeping GPIB 3 and 4 synchronized so all the collected data are obtained at the same time. Data synchronization is well maintained by Fast Track 8800 as long as the buffer is not full. If for any reason the computer does not read the data fast enough from the buffer, the buffer becomes full and Fast Track 8800 loses its ability to keep the two GPIB boards synchronized. If such an event occurs, data from GPIB 4 (where transducer 3 is connected) are no longer synchronized with GPIB 3 (where the rest of channels are connected). Unfortunately from the moment Fast Track 8800 loses its ability to maintain synchronization, all the collected data are desynchronized, even after the buffer is fully read and empty.

In normal situations the speed of reading the data by the Fast Track 8800 computer is sufficient to prevent buffer overruns. This keeps the buffer empty almost all the time, ensuring that no data are lost by overwriting. In the event that the computer processor is busy with some other tasks, then there is a risk that the buffer will overflow. To avoid this situation all the possible actions were taken to be certain that the computer was fully and exclusively devoted to the task of data acquisition. All possible services were turned off; the computer was disconnected from the network, the antivirus was disabled, all the services from the operating system were deactivated, and temperature data acquisition was loaded on a separate computer. Although these actions reduced the number of times that this problem occurred, only at the end of this project was it realized that some unknown application(s) was (were) occasionally taking processor time and causing the data not to be synchronized.

Figure 3. Example of the effect of data synchronization on variability in the measured modulus.

Figure 3. Example of the effect of data synchronization on variability in the measured modulus.

At the beginning of the project, when the data synchronization problem was first realized, the only solution immediately available was to stop and restart the loading when the buffer overflow occurred. This could add some other risks to the testing in addition to the fact that any discontinuity in the loading applied to the specimen affected its modulus. As a consequence a new module was programmed and added to the control and data acquisition software. This module allowed the operator to reset the buffer and to synchronize the data without any interruption in the loading. The only downside of this procedure was that it was not automatic and required that the operator be continuously present during the test and intervene manually. This meant that from the moment the problem occurred until the moment the operator noticed the problem and reset the buffer, the collected data were not synchronized.

During the testing in this project the difference in time between the first appearance of unsynchronized data collected from transducer 3 and the two others was limited to a maximum of 2 seconds (or 20 cycles). During those time periods in the testing where the variation of modulus was very small, the effect of unsynchronized data on the analysis was negligible. This is especially true for the calculation of the slope of modulus at the end of the stages where there were a large number of data points.

For future testing either a system upgrade will be needed to solve this problem or a module must be added to the control and data acquisition system to periodically check whether the data are synchronized or not. The added module will automatically reset the buffer in the event that the data are not synchronized.

Environmental Chamber

A dual chamber scheme was used to control the testing temperature (Figure 4). The test chamber was mounted inside a walk-in environmental chamber. The temperature inside the test chamber was controlled by connecting a water bath to a heat exchanger mounted in a plenum adjacent to the test chamber. Air was exchanged between the plenum and the test chamber resulting in variations of less than ± 0.015 °C in the test chamber. Two 100-W light bulbs were used as a heat source to make small adjustments in the temperature of the air passing between the plenum and the test chamber.

Figure 4. Schematic of temperature chamber and controller.

Figure 4. Schematic of temperature chamber and controller.

Test Fixtures and Specimen Mounting

A schematic of the test fixture including a test specimen and the transducers is given in Figure 5. A jig was used to align the mounting heads and test specimen during the gluing process. The jig ensures that the ends of the mounting heads as they are attached to the testing frame are parallel and that the axes of the test specimen and mounting heads are concentric.

Figure 5. Schematic of test fixture.

Figure 5. Schematic of test fixture.

In order to prepare a specimen for testing, the test specimen is placed in the jig along with the epoxy-coated upper and lower mounting heads. After the epoxy has cured, the specimens are maintained in the temperature chamber for 24 hours at the test temperature for conditioning and achieving temperature equilibrium. The lower mounting head and the mounting flange are screwed together, and then the upper mounting head is screwed into the load cell. The hydraulic ram is lowered until the mounting flange seats on the lower platen extension. The three bolts connecting the mounting flange and the lower platen extension are then tightened. With this configuration, the ends of the test specimen are held rigidly in the testing machine. By following this procedure damage to the test specimen caused by traction, compression or torque is minimized.

Materials and Specimen Preparation

The original experiment design included four different asphalt mixtures with three replicates for each. However, because a number of problems were incurred during the course of the project, changes were made to the original experiment design. Table 2 presents the mixtures and specimens prepared for this study.

Table 2. Properties of the mixtures prepared for this research.
Mix IDPG GradeTarget AV%Specimen #Actual AV%%AC
M1241APG 64-
M2167PG 76-
M3298PG 64-
M5201PG 58-
I-80PG 76-227.0---6.65.3

With the exception of the I-80 specimen the specimens were compacted using a Pine™ Superpave Gyratory Compactor. The compacted specimens were 150 mm in diameter and 150 mm in height. Attempts were made to achieve air voids in the range of 7.0 ± 0.5 percent. The I-80 specimens were compacted as part of a previous fatigue study using the French rolling wheel compactor as described elsewhere( 4 ).

Properties of Mixtures

The gradation and properties for the mixtures prepared for this study are presented in Appendix A. Three of the mixes were obtained from a PennDOT-sponsored project, Superpave In-Situ Stress-Strain Investigation (SISSI).

The I-80 mix was used in the previous fatigue study( 4 ). Unfortunately, due to time constraints no tests were conducted on the last mix (M5201) and limited testing was conducted on the remaining mixes.

Fatigue Testing

During a previous study( 6 ), a new fatigue testing protocol was proposed in which three stages of continuous loading without any rest period were considered, as shown in Figure 6. Schematic of loading in Stages I, II, and III. The same strain level, not exceeding the endurance limit of the HMA and consequently at a level that does not produce fatigue damage, is applied during Stages I and III. During Stage II, a strain with a magnitude exceeding the endurance limit and consequently causing fatigue damage is applied. The difference between the moduli at the end of Stages I and III (represented by two asymptotes) should directly indicate the level of "true" fatigue damage imposed during Stage II, as shown in Figure 6. Schematic of loading in Stages I, II, and III.

Damage should occur during Stage II only if the strain level during Stage II is greater than the endurance limit. In the study presented here, strain levels were expected to be considerably less than if the endurance limit were applied. For both Stages I and III, strain magnitude is maintained at the same level (σ1). For Stage II, a larger strain level σ2 is applied. At each stage, loading continues until the curve of modulus versus number of cycles reaches a zero slope (i.e., SAB = SCD = SFG = 0, where S indicates the slope of line presenting modulus as a function of loading cycles). The expectation is that the modulus at the end of Stage III will return to the same level as was observed at the end of Stage I if the applied strains are below the endurance limit. Therefore, the following goals will be achieved from such an experiment:

  1. Demonstrate that the drop in modulus during Stage II is not the result of fatigue damage. This is believed to be the case when at the end of Stage III the modulus returns to the original level E1 of Stage I once strain is reduced from σ2 to σ1 (as shown in Figure 7).
  2. Demonstrate that the induced strains σI and σII are indeed within the endurance limit, even though a drop in modulus is observed during Stages I and II; and
  3. Verify the existence of an endurance limit for the mixtures that is equal to or greater than σ2.

Figure 6. Schematic of loading in Stages I, II, and III.

Figure 6. Schematic of loading in Stages I, II, and III.

Figure 7. Testing at specific strain levels to validate existence of the endurance limit.

Figure 7. Testing at specific strain levels to validate existence of the endurance limit.


  1. Monismith, C. L., J. A. Epps, D. A. Kasianchunk, and D. B. McLean, Asphalt Mixture Behavior in Repeated Flexure, Report No. TE 70-5, Institute of Transportation and Traffic Engineering, University of California, Berkeley, 1970.
  2. Carpenter, S. H., K. A. Ghuzlan, and S. Shen, "A Fatigue Endurance Limit for Highway and Airport Pavements." Paper No. 03-3428, presented at the Annual Meeting of the Transportation Research Board, January 2003.
  3. Soltani, A., "Comportement en Fatigue des Enrobes Bitumineux," Doctoral Dissertation, Ecole Nationale des Travaux Publics de l'Etat - INSA, Lyon, France, 1998.
  4. Soltani, A., and D. A. Anderson, Uniaxial Fatigue of Asphalt Mixes: A New Approach, Report No. PTI 2003-13, Pennsylvania Transportation Institute, The Pennsylvania State University, January 2005.
  5. Chapra, S.C., and R.P. Canale, Numerical Methods for Engineers, McGraw-Hill, New York, 1988, Section 13.1.
  6. Soltani, A., and D. A. Anderson, "New Test Protocol to Measure Fatigue Damage in Asphalt Mixtures," Journal of Road Materials and Pavement Design, Vol. 4, 2005.
<< PreviousContentsNext >>

More Information

PDF files can be viewed with the Acrobat® Reader®
Updated: 04/07/2011

United States Department of Transportation - Federal Highway Administration