Using Test Track Data to Validate Mechanistic Pavement Design Models
Asphalt pavement thickness has historically been designed based on vehicle type, standardized axle loads and material properties based on results from the AASHO Road Test in the late 1950s. In recent years, however, pavement design has begun to shift toward a mechanistic-empirical framework that uses engineering principles to design pavement structures that will resist specific distresses, including fatigue cracking and rutting, over the required performance period.
Mechanistic-empirical (M-E) design incorporates material properties and environmental data, and uses mechanical analysis to more accurately model a pavement structure. Pavement response, which is calculated based on expected traffic loading, can then be used to predict pavement performance through empirical correlations.
M-E design is slated to become the new AASHTO standard, with DarWIN M-E software scheduled for release in early 2011. As M-E design is implemented, there is an ongoing need for local calibration/validation of the empirically derived pavement performance models, which are dependent on both materials and climate. Data from structural sections at NCAT’s Test Track provide a perfect opportunity to compare actual pavement performance with that predicted by M-E design models.
Comparing Actual and Predicted Performance
Figure 1. Comparison of measured and MEPDG-predicted IRI values for all sections
A variety of inputs are needed in the Mechanistic-Empirical Pavement Design Guide (MEPDG) analysis procedure, including the following:
- detailed traffic data, referred to as load spectra
- detailed climate data, such as air temperature, rainfall, wind speed, relative humidity and percent cloud cover
- mechanistic material properties—dynamic modulus (E*) for hot-mix asphalt (HMA) and resilient modulus (Mr) for granular base and subgrade materials
- HMA properties, such as asphalt content, lift thickness and density
These inputs are readily available for the structural sections of the completed 2003 and 2006 research cycles at the test track. These sections were instrumented with strain gauges to measure pavement response under loading and were subjected to 10 million equivalent single-axle loads (ESALs) during each two-year testing cycle. Design variables included total hot-mix asphalt (HMA) thickness, HMA mix type, base material type and subgrade material type. All structural sections were assessed on a weekly basis for surface performance (rut depth, fatigue cracking and international roughness index, or IRI) as well as structural response (strain and pressure).
Figure 2. Comparison of measured and MEPDG-predicted rut depths for all sections
A comparison of measured and predicted IRI values for all sections (Figure 1, previous page) shows that the MEPDG model gives fairly reasonable results. Several sections, most notably N1 2003 and N2 2003, exhibited severe fatigue cracking—well beyond what is typically considered failure—and the resulting roughness is reflected in the measured IRI data but not in the MEPDG-predicted values for those sections, since the MEPDG model is not based on such severe distress. When comparing actual and predicted rut depths for all sections (Figure 2), it can be seen that the MEPDG rutting model overpredicts permanent deformation, but this can be calibrated or corrected with a simple offset. However, a comparison of measured and predicted fatigue cracking for all sections (Figure 3) reveals mixed results. In several cases, the MEPDG fatigue model does not match actual performance.
NCAT plans to continue MEPDG validation/calibration efforts as data becomes available for structural sections currently under loading at the test track.
Test Track Data Used in TexME Calibration
Recent research conducted by the Texas Transportation Institute (TTI) also employed data from previous test track structural sections in calibrating and validating distress prediction models within the proposed TexME design system.
Figure 3. Comparison of measured and MEPDG-predicted fatigue cracking for all sections
To predict rutting, TTI selected the VESYS layer rutting model, which accounts for the permanent deformation properties of each layer. Material properties were determined using dynamic modulus and repeated load testing. Calibration of the proposed rutting prediction model involved three correction factors—pavement temperature, modulus and HMA thickness—that were determined using field rutting data. Eight structural sections from the 2006 NCAT Test Track cycle, representing a range of rut depths from low to very high, were used to determine the pavement temperature and modulus correction factors. In addition, the LTPP SPS-5 sections on US175 in Texas were used to determine the thickness correction factor. The correction factors were established by minimizing the difference between predicted and measured rutting. The accuracy of the calibrated rutting model was verified using data from three sections of the 2000 test track cycle, with predicted rutting generally matching the rutting observed in the field.
Figure 4. Comparison of measured and predicted fatigue cracking for section N7 using the TTI-developed fatigue model
The proposed TexME fatigue cracking model, which considers crack initiation and crack propagation, uses an enhanced two-step Overlay Test to determine HMA fracture properties. Preliminary calibration included data from seven 2006 test track sections, some with no fatigue cracking and others exhibiting severe fatigue damage. The differences between measured and predicted fatigue cracking were minimized in order to develop the necessary calibration factors. Figure 4 illustrates the measured and predicted fatigue cracking for section N7. Data from two sections of the 2003 test track cycle were used to validate the fatigue cracking model.
Performance data from the current test track cycle will allow further validation/calibration of the TexME distress prediction models.