Method for Assessment of Modeling Quality for Asphalt Dielectric Constant to Density Calibration

Abstract

Traditional measures of asphalt compaction rely primarily on random cores that only measure a small fraction of the pavement. Recently, the use of ground penetrating radar was indicated to be usable as a nondestructive means for the continuous assessment of asphalt compaction. A proposed Hoegh-Dai (HD) model has been successful in predicting air void content within typically achieved field compaction levels but has reduced accuracy at the extremes. This paper proposes an enhanced Minnesota DOT (MnDOT) model to address this issue. A method for assessing modeling quality is proposed to quantify the improvement of the MnDOT model. The procedure is based on the accuracy of fits when run through a Monte Carlo simulation. The developed procedure indicates that the MnDOT model has improved accuracy—with 0.74% air void variation at a dielectric of 4 compared with 3.83% for the HD fit. Additionally, the MnDOT model is more stable for replicate days of the same mix design and falls within the uncertainty of more of the field cores across several projects than the HD model.