Use of image analysis for the evaluation of rolling bottle tests results

The adhesion between bitumen and aggregates is of paramount importance for asphalt mixtures, because it is confirmed that a weak bond strength results in a premature failure of the pavement. Methods for determining the affinity or the adhesion between components are made both on loose and compacted samples. Among the first category the rolling bottle method, which is standardized in EN 12697-11 part a, is very common. It represents a simple, rapid and low cost test for an indication of the affinity between aggregate and bitumen and its influence on the susceptibility of the mixture to stripping. This paper proposes the use of 2D image analysis to evaluate the rolling bottle test results, overcoming the limits and shortcomings of the visual analysis prescribed by the reference standard. In order to demonstrate its applicability to a broad range of materials, this procedure was applied to both light and dark aggregates, mixed with a wax modified binder. The mixing temperature was varied so that the influence of the binder viscosity on the adhesion was assessed. A comparison between visual and semi-automatic estimation is presented, demonstrating that the latter brings to far better results. The accuracies were determined through confusion matrixes that permit to identify the errors made during the process of classification.