Use of Fine Aggregate Matrix Experimental Data in Improving Reliability of Fatigue Life Prediction of Asphalt Concrete Sensitivity of This Approach to Variation in Input Parameters

Asphalt concrete (AC) material performance has been assessed by numerous mechanistic models over the years. Often these models are purported to enable more accurate prediction of the pavement service life than existing empirical models. Most of these models use fundamental material properties, which are obtained by performing experiments on the materials, as input variables. However, by introducing more variables, these models create the potential for greater uncertainty because the variables have inherent variability. Variations observed in these input parameters affect the reliability of any resulting performance predictions. In an effort to improve the reliability of fatigue life predictions, experimental data from the fine aggregate matrix (FAM) phase was used in this study for predicting fatigue life. From the comparative assessment, it was observed that the reliability of fatigue life predictions was improved by more than 50% when data from the FAM phase rather than AC data were used. An upscaling procedure was used in predicting the AC material fundamental properties and then in performing a reliability analysis with the predicted data. More reliable fatigue prediction results were also observed when the AC predicted data were used; however, this improvement was not as good as that in the FAM phase. A parametric sensitivity analysis was performed to determine whether variation in any one parameter resulted in a greater impact on the resultant reliability than did variation of other parameters. From the analysis, it was observed that the variation of the modulus parameter affected the reliability predictions more than did the variation of the other input parameters considered in this study, regardless of the model failure criteria used.