1374196 - Fatigue Failure Criterion: Comparing Maximum Phase Angle with Minimum Curvature of Stiffness Curve

This study compares two fatigue failure criteria using uniaxial constant crosshead cyclic fatigue testing results of 12 specimens. To guarantee that the failure occurred in the middle of the specimen, where the strain measurements were carried out, the tests were performed with an Asphalt Mixture Performance Tester (AMPT) on flat rectangular specimens trimmed into a dog-bone shape. Furthermore, the tests were conducted at 40°C to address the potential of top-down fatigue cracking in a high temperature environment such as that of Hawaii, where most of the heavy loading occurs during the day, and where the most commonly testing temperature of 21°C occurs in actual pavements mostly at night with very little traffic.
The first criterion used was the number of repetitions where the phase angle of the mix becomes a maximum, which is one of the criteria commonly proposed in the literature. The second criterion was developed in this study and consists of estimating the number of repetitions where the stiffness vs. loading cycles curve reaches a minimum (most negative) curvature. Both criteria were estimated after fitting harmonics to the stress and strain curves within a small neighborhood (50 cycles) of a point. These fittings make the estimation of both the phase angle and the stiffness of the samples relatively simple. Furthermore, for the proposed criterion, smooth quadratic functions were fitted to the stiffness vs number of cycles curve in a small neighborhood of each of the points at which the stiffness was computed. The fitting coefficients of these polynomials were then used directly to estimate the curvature of the stiffness curve.
Both criteria provide estimates of the number of cycles where a rapid loss in stiffness due to macrocracking starts. For three of the twelve specimens evaluated in this study, the phase angle vs. number of cycles curve was very flat. The flatness of the curve combined with some noise in the data resulted in a substantial underestimation of the location of the failure point using the maximum phase angle criterion. These three specimens corresponded to a mix with small nominal maximum aggregate size and high binder content that was being considered as a candidate for a thin lift overlay. On the other hand, the minimum curvature criterion provided results consistent with what an analysis would visually deduce from the stiffness plot.
For the other nine specimens, the values determined with both criteria were extremely consistent. Considering only these 9 specimens, the R2 of the linear relationship between the two criteria was 0.997. Hence, the minimum curvature of the stiffness curve appears to provide very consistent results with the accepted existing maximum phase angle criterion when that criterion resulted in reasonable results and provided more realistic failure estimates when that criterion failed. Hence, the minimum curvature criterion is recommended as a potential alternative for robust estimation of the fatigue failure point.