Contents lists available at ScienceDirect NDT and E International journal homepage: www.elsevier.com/locate/ndteint Continuous real-time monitoring of flexible pavement layer density and thickness using ground penetrating radar Siqi Wang * , Shan Zhao, Imad L. Al-Qadi Department of Civil and Environmental Engineering, University of Illinois at Urbana-Champaign, 205N. Mathews Ave, IL, 61801, Urbana, USA ARTICLE INFO Keywords: Continuous real-time monitoring Thickness and density estimation Ground penetrating radar Asphalt concrete ABSTRACT Static or discrete time prediction of flexible pavement asphalt concrete (AC) layer density and thickness utilizing the ground penetrating radar (GPR) has proved to be relatively accurate compared to other in-situ methods. However, the accuracy during continuous monitoring of construction has not been validated by real construction projects. This study investigated the continuous estimation of AC layer density and thickness using GPR at a construction site in Macomb, Illinois. The first test compared AC layer density and thickness for pavement segments built in two consecutive days. The second test evaluated the construction quality of pavement segments built utilizing two types of material transfer vehicles (MTV). The GPRan software and the proposed truncation algorithm were used to calculate AC layer thickness continuously. The AC layer density and thickness estimation results were compared to core data using the ALL model. The GPR estimation values were found to be reliable. Errors of thickness results between continuous GPR data and core data from two tests are approximate 3% and 6%, respectively. The density of the new pavement segments was larger than the old ones while the thickness was almost the same. The average error between continuous density predictions and core density was 3%. The two MTVs performed similarly based on layer thickness and density results. 1. Introduction The implementation of non-destructive testing (NDT) has been in- creasing in civil engineering in recent years. As for pavement en- gineering, NDT is useful for quality control and quality assurance (QC/ QA) by means of thickness and density monitoring [1–10]. Traditional in-situ thickness and density testing methods such as coring have sig- nificant limitations, including being time-consuming and destructive to pavement. Moreover, cores can be drilled only at a very small percen- tage of the surveyed area, which limits the information available for asphalt concrete (AC) layer density and thickness. Besides destructive testing methods, NDT methods such as in-situ nuclear gauge measure- ment have been used for compaction monitoring [1,3,11,12]. However, this method has limitations such as large variations and requirements of special licensing operators [7,11]. Ground penetrating radar is a widely-used NDT method for esti- mating AC layer thickness and density. The major advantages of GPR are non-destructive and time efficient. The GPR transmits electron- magnetic (EM) waves and receive reflections at interfaces in pavement structures. These reflected signals are processed through digital signal processing to calculate layer thickness and density. Meanwhile, the dielectric constant of the layer can be back-calculated when layer thickness is known [1,3]. The GPR has proved to provide accurate es- timation of layer dielectric constant, thickness and density for new and in-service pavements [5,13–20]. Prediction errors in thickness ranges from 2.8% to 27.6% using raw data and ranges from 0.2% to 9.4% using deconvolution methods [2,4,21]. However, the deconvolution method requires longer time for data processing than using the raw data. This impedes the GPR application for continuous testing. Prediction models as well as advanced digital signal processing methods have been de- veloped to improve the accuracy of AC layer thickness prediction [4,5,20–22]. For the density monitoring, static testing using GPR shows high accuracy. The errors range from 1% to 13.4% using various pre- diction models [23]. Meanwhile, continuous and real-time layer density and thickness prediction is important for GPR implementation. Furthermore, accurate layer thickness and density estimation results in achieving desired compaction through real-time monitoring during construction. Although step-frequency antennas were used to address the effect of antenna vibration during testing; common mid-point method was used to improve the dielectric constant estimation [23–25]. In this study, two sets of continuous real-time tests of AC layer density and thickness using air-coupled horn antennas were conducted at a construction site located in Macomb, Illinois, US. The first test https://doi.org/10.1016/j.ndteint.2018.08.005 Received 15 January 2018; Received in revised form 2 August 2018; Accepted 21 August 2018 * Corresponding author. E-mail addresses: siqiw2@illinois.edu (S. Wang), szhao28@illinois.edu (S. Zhao), alqadi@illinois.edu (I.L. Al-Qadi). NDT and E International 100 (2018) 48–54 Available online 23 August 2018 0963-8695/ © 2018 Elsevier Ltd. All rights reserved. T