University of California Pavement Research Center

An experimental study was conducted to determine the effect of deflection waveform on four-point flexural fatigue test results for hot mix asphalt. Seven asphalt mixtures, comprising a wide variety of gradations, binder types, and binder contents, were selected for this study. Four of the mixes were tested at the University of California Pavement Research Center (UCPRC) and three were tested in Australia by the ARRB Group. The mixes were tested at different strain levels under both haversine and sinusoidal deflection-controlled modes without introducing rest periods between load cycles. The haversine and sinusoidal testing modes were compared to each other from different perspectives. This comparison showed no indications of differences in damage as measured by stiffness reduction between the haversine and sinusoidal waveforms for six of the seven mixtures. This outcome was attributed to the viscoelastic nature of asphalt mixes. Because of this viscoelasticity, it is believed that the beam at-rest position in haversine testing will move to halfway between zero and maximum deflection, and so the same stress is produced by haversine and sinusoidal deflection waveforms as soon as the peak-topeak deflection amplitudes are equal. For one of the seven mixes, sinusoidal testing produced considerably longer fatigue lives. No reason was found to explain why the results for this mix differed considerably from the results with the other mixes. Based on these results, no compelling reason was found to recommend that UCPRC change from its use of a haversine to a sinusoidal mode of testing, but similarly no reason was found not to make this change. One reason to move to the sinusoidal wave is that in terms of stress, haversine testing becomes sinusoidal after a small number of load repetitions. It is therefore recommended that the UCPRC and Caltrans support the standardization of waveform requirements in the ASTM and AASHTO test methods to use the sinusoidal waveform because the same fatigue life and stiffness degradation curves are expected from the two testing modes when rest periods are not included between loading cycles. It is anticipated that none of the historical fatigue model parameters calibrated by UCPRC would need to be adjusted. Also examined in the study was the use of alternative failure criteria compared with the currently used criterion of 50 percent loss of stiffness.