Effect of testing geometry on measuring fatigue of asphalt binders and mastics

Fatigue damage is a distress observed in asphalt pavement at moderate to low temperatures due to cyclic loading of moving traffic. Recent developments in binder fatigue research have been made using parallel plate geometry in a dynamic shear rheometer (DSR) to measure total dissipated energy and its change as a function of test cycles. Recent studies, however, have raised questions regarding the temperature level and stiffness level at which true binder fatigue can be measured with the parallel plate geometry in the DSR. It has been claimed that the binder's stiffness of less than 15 MPa will result in instability failure known as "edge effect," instead of true fatigue. Edge effect, better known as edge fracture in polymer science, is a normal-force dependent phenomenon which is characterized by the inability of material to maintain its geometry at its boundary with air/atmosphere. This study attempts to investigate further the relationship between binder's stiffness and true fatigue using parallel plate (PP) geometry and torsion cylinder (TC) geometry. TC samples are a cylindrical-shaped mix of asphalt binder and Ottawa sand (one size, pass No. 30 sieve, and retained on No. 50 sieve), 12 mm in diameter and 30 mm in height, produced using metal molds, to simulate the effect of traffic load on thin asphalt film in asphalt mixtures. Torsion cylinder high stiffness (range) ensures that the sample will fail in true fatigue. This study hypothesized that PP and TC test results are related, and should PP and TC behave in a similar way, edge effect can be considered not significant enough to effect parallel plate fatigue testing. Testing was done by three different laboratories to evaluate repeatability between these two geometries. A simple shift parameter based on testing geometries allows direct comparison between PP and TC test results. Binder stiffness role in similarities and differences in both geometries is presented. Regression analysis using Minitab showed that test geometry was not sufficiently significant to affect fatigue test results, thus the edge effect was not significant in PP fatigue testing. The same analysis indicated that repeatability is significant in the PP regression model, but not in the TC regression model. This indicated that TC has better repeatability than PP, and showed that sample preparation and adhering of asphalt film to parallel plates can be dealt with by using torsion cylinder geometry.