Investigation of Asphalt Pavement Analyzer Testing Program in Nebraska

The asphalt pavement analyzer (APA) has been widely used to evaluate hot-mix asphalt (HMA) rutting potential in mix design and quality control – quality assurance (QC-QA) applications, because the APA testing and its data analyses are relatively simple, rapid, and easy. However, as demonstrated in many studies and also experienced by the state of Nebraska, the APA testing is in question due to its high testing variability and a lack of sufficient correlation with actual filed performance. The primary objective of this research was to find critical materials and/or mixture design factors affecting APA test results so as to eventually improve current APA testing program in Nebraska. In addition to that, development of models to predict APA rut performance with given properties of HMA mixture ingredients and mixture design characteristics were also attempted. To find variables affecting APA rut results and the extent of these variables, SP-4 mixture data from Nebraska and HMA mixture data from Kentucky were statistically analyzed using the multiple linear regression method considering six factors (binder PG, aggregate gradation, nominal maximum aggregate size, aggregate angularity, air voids in mixture, and asphalt content in mixture) as probable candidates significantly affecting APA rut results. For a detailed characterization of gradation effects, three indicators (gradation density, fineness modulus, and restricted zone) were considered and each of them was used for each statistical analysis. Analyses results demonstrated that the binder PG was the only variable that always shows significant impact on APA rut results, which is in a good agreement with other studies. Predicting models developed through the results of multiple linear regression analysis and the artificial neural network technique presented a relatively low level of model adequacy which can be observed by the coefficients of determination and crossplots between predicted APA rut values and the measured APA rut data. More data would be helpful to confirm the findings from this research and also to develop a better prediction model