Factors Impacting Monitoring Asphalt Pavement Density by Ground Penetrating Radar

摘要

Real-time asphalt concrete (AC) pavement density monitoring is important for quality control (QC) and quality assurance (QA) purposes, because it allows correction during the compaction process. Ground penetrating radar (GPR) is capable of providing real-time AC mixture density prediction using the Al-Qadi, Lahouar, and Leng (ALL) density prediction model. However, noise sources, such as surface moisture and vibrations, may jeopardize the AC density prediction accuracy. This study proposes a mean reflection coefficient algorithm and digital filter design method to remove the surface moisture and smooth the density profile. In the mean reflection algorithm, the frequency-select bandwidth was selected as 40-70% of the actual peak frequency in the magnitude spectrum through the simulation studies. White Gaussian noise was added in the models for robustness testing. In the digital filter design method, the magnitude spectrum of the GPR predicted density profile was analyzed to decide filter types and corresponding parameters. Thresholding method was used to remove abnormal values, and window-based finite impulse response (FIR) filters were used to smooth the density profile. Lab-controlled and field tests were performed for both algorithms. Estimated aggregate dielectric constant was used to predict pavement density. A sensitivity analysis was performed to evaluate the effect of different aggregate dielectric constant on density (or air void). For surface moisture effect removal, mean reflection coefficient algorithm may be utilized to reconstruct dielectric constant values at an error less than 4%. This algorithm is independent of the antenna central frequency. For the density profile smoothing during continuous GPR survey, results show that various filter types have comparable smoothing performances. For the effect of aggregate dielectric constant on density prediction, sensitivity analysis shows that when aggregate dielectric constant values changes from 6.5 to 7, the predicted air void increases from 2.5% to 6.3%. This indicates the importance of an accurate aggregate dielectric constant estimate when applying ALL model for pavement density predictions; hence, aggregate dielectric constant estimate must be utilized.