Prediction of asphalt concrete pavement density from processing ground-penetrating radar compaction data

Wang, Siqi

Permalink

https://hdl.handle.net/2142/110836 Copy

Description

Title

Prediction of asphalt concrete pavement density from processing ground-penetrating radar compaction data

Author(s)

Wang, Siqi

Issue Date

2021-04-20

Director of Research (if dissertation) or Advisor (if thesis)

Al-Qadi, Imad L

Doctoral Committee Chair(s)

Al-Qadi, Imad L

Committee Member(s)

• Leng, Zhen
• Popovics, John S
• Ozer, Hasan
• Tutumluer, Erol

Department of Study

Civil & Environmental Eng

Discipline

Civil Engineering

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

Ph.D.

Degree Level

Dissertation

Date of Ingest

2021-09-17T04:04:39Z

Keyword(s)

• Ground-penetrating radar
• Digital signal processing
• Asphalt concrete pavement
• Density prediction
• Transportation

Abstract

Achieving desired asphalt concrete (AC) density during compaction is essential to ensure flexible pavement performance. Approaches such as lab-testing of field cores, in-situ nuclear gauge measurements, and intelligent compaction prediction have shortcomings, such as limited area coverage, labor- and time-intensiveness, use of hazardous materials, and unreliability. Ground-penetrating radar (GPR) technique is proposed to monitor the density change during AC pavement compaction using air-coupled antennas. However, issues, including surface moisture effect due to water spray during compaction, vibration effect during continuous GPR survey, and signal processing complexity, may hinder AC density prediction accuracy. This study addressed the aforementioned issues and developed signal processing algorithms to provide near real-time AC pavement density monitoring during compaction. Signal pre-processing techniques were proposed to clean GPR signals during a continuous survey. The truncation algorithm was used to remove the disguised second-largest amplitude by time-gating the AC layer bottom reflection. This ensures reliable thickness predictions during high-speed GPR survey. A non-linear optimization approach, with a three-layer model, was proposed to address reflected signal overlap and surface moisture effect simultaneously in the time domain. Results from field tests performed in Urbana, IL showed that AC layer thickness and dielectric constant could be reconstructed with errors less than 9.2% and 2%, respectively. The AC dielectric constant and density predictions, using the Al-Qadi-Lahouar-Leng (ALL) model, may be over-estimated due to the increased surface reflection amplitude because of surface water. Surface water is sprayed during compaction to prevent AC mixtures from sticking to the roller drum/wheel. A mean reflection coefficient algorithm was proposed to remove this effect. The input was the time domain pavement surface reflection, truncated using a 1-D Sobel edge detector during a continuous GPR survey. Simulations using finite-difference time-domain (FDTD) method were performed to identify low and high bounds of frequency-select bandwidth. Static lab-controlled GPR tests with water sprays were conducted for validation. Results showed that the dielectric constant was reconstructed with errors of 2.6% and 3.1% using 1 GHz and 2 GHz antennas, respectively. The algorithm is computationally-efficient, automatic, and independent of antenna central frequency. During high-speed continuous GPR survey, vibrations of transmitter and receiver, coaxial cables, and mounting equipment may affect density prediction accuracy. Vibration impact on AC density was identified as low-frequency components considering GPR antenna resolution and spatial sampling rate. Hamming, Hann, Bartlett, and Blackman windows were applied as low-pass filters and found to have comparable smoothing performances. The matched filter and cumulative profile difference approaches were proposed to identify low-pass filter inputs when survey speed is constant. For the inconsistent speed survey, wavelet transform was applied for profile smoothing. The mother wavelet and decomposition level were identified based on profile smoothing requirements. A data-driven empirical mode decomposition (EMD) approach was applied to address noise due to sudden accelerating/braking. GPR was installed on the roller for data collection during compaction. Air void prediction showed a density progression curve with respect to roller passes. The air void values at the final roller pass were 5.1% and 4.4% using wavelet transform and EMD, respectively, which were considered reasonably accurate compared to ground-truth core value (4.5%). However, these results are based on one field test and further tests are needed. The aggregate dielectric constant is an input in the ALL density prediction model. Based on sensitivity analysis, a linear correlation was found between the AC dielectric constant and bulk specific gravity. A calibration algorithm was proposed based on this observation. A user-friendly and time-efficient tool, including the developed algorithms, was introduced for signal processing for continuous GPR survey. Ultimately, the tool may be installed on GPR-equipped compactor for compaction monitoring.

Graduation Semester

2021-05

Type of Resource

Thesis

Permalink

http://hdl.handle.net/2142/110836

Copyright and License Information

Copyright 2021 Siqi Wang

Owning Collections