You are on page 1of 20

In-place Hot-Mix Asphalt Density Estimation Using Ground

Penetrating Radar

I. L. Al-Qadi, Z. Leng, S. Lahouar, and J. Baek

Paper Number 10-1735

Submitted to:
Transportation Research Board
89th Annual Meeting
January 10-14, 2010
Washington, D.C.

Duplication for publication or sale is strictly prohibited without prior written permission of the Transportation
Research Board
1
2 In-place Hot-Mix Asphalt Density Estimation Using Ground Penetrating
3 Radar
4
5
6 Imad L. Al-Qadi1
7 Founder Professor of Engineering
8 Illinois Center for Transportation, Director
9 Department of Civil and Environmental Engineering
10 University of Illinois at Urbana-Champaign
11 205 N. Mathews Ave., Urbana, IL 61801
12 Phone: (217) 265-0427
13 Fax: (217) 893-0601
14 alqadi@illinois.edu
15
16 Zhen Leng
17 Graduate Research Assistant
18 Department of Civil and Environmental Engineering
19 University of Illinois at Urbana-Champaign
20 205 N. Mathews Ave., MC-250, Urbana, IL 61801
21 zleng2@illinois.edu
22
23 Samer Lahouar
24 Assistant Professor
25 Department of Electronics
26 Institut Supérieur des Sciences Appliquées et de Technologie de Sousse
27 Cité Taffala, Ibn Khaldoun, Sousse 4003, Tunisia
28 Samer.Lahouar@issatso.rnu.tn
29
30 Jongeun Baek
31 Graduate Research Assistant
32 Department of Civil and Environmental Engineering
33 University of Illinois at Urbana-Champaign
34 205 N. Mathews Ave., MC-250, Urbana, IL 61801
35 baek2@illinois.edu
36
37
38
39
40
41
42
43
44 Number of Words: 4449 + 3000 (4 Tables and 8 Figures) = 7449

1
Corresponding Author
Al-Qadi, Leng, Lahouar, and Baek 2

1 ABSTRACT
2 This research proposes the innovative use of ground penetrating radar (GPR) to effectively,
3 continuously, and rapidly estimate in-place hot-mix asphalt (HMA) density. Based on
4 electromagnetic (EM) mixing theories, three candidate models were developed to determine the
5 dielectric constant of HMA considering dielectric and volumetric properties of the three major
6 components of HMA, i.e., air, binder, and aggregate. Laboratory tests were conducted on mid-
7 sized HMA slabs (60cm×60cm×7.5cm) to evaluate the models. After evaluating and comparing
8 the three models, the authors determined that the prediction model based on the Rayleigh mixing
9 theory was the most accurate. The selected model was calibrated with a field core and then
10 validated using field GPR measurements of a composite pavement with an HMA surface. The
11 selected model provided accurate HMA density within a reasonable range.
12
13 KEYWORDS: Hot-mix asphalt, density, ground penetrating radar, electromagnetic mixing
14 theory
15
16
Al-Qadi, Leng, Lahouar, and Baek 3

1 INTRODUCTION
2 During the construction and life of hot-mix asphalt (HMA) pavements, the pavements’ in-place
3 characteristics need to be assessed for quality control (QC) and quality assurance (QA). The
4 density of HMA is one of the most important parameters to monitor. Typically, the percentage of
5 in-place HMA density to maximum theoretical density ranges from 91% to 98% (1). Insufficient
6 HMA density is the most frequent construction-related performance problem. Hot-mix asphalt
7 density that is either too high or too low causes premature pavement failure (1). Low in-place air
8 voids (high density) can result in bleeding and shoving. On the other hand, high air voids (low
9 density) can lead to water damage, oxidation, raveling, and cracking.
10 Traditionally, two techniques have been used to estimate in-place HMA density: a
11 laboratory test on pavement cores and an in-situ nuclear gauge method. The laboratory technique
12 is a destructive test, during which field cores are extracted from a pavement to directly measure
13 the thicknesses and the volumetric properties of the HMA. Although this method provides
14 accurate measurements, it only provides a very limited number of measurements. Also, it is time
15 consuming due to the relatively slow procedures for coring and testing. On the other hand, the
16 nuclear gauge is a nondestructive technique that provides reasonably accurate estimates of the
17 HMA mat density. However, the nuclear gauge technique also has some drawbacks. First, it
18 provides limited information about the layer density since nuclear measurements are usually
19 taken with high spatial spacing. Second, a nuclear gauge operation uses radioactive materials and
20 therefore should only be conducted by authorized personnel with special licensing.
21 Recently, electromagnetic (EM) density gauges have entered the market as an alternative
22 to the nuclear density gauges and the coring process. These non-nuclear devices use EM signals
23 to measure in-place HMA density. These EM density gauges eliminate the need for licenses,
24 training, and specialized storage, as well as the risks associated with devices that use a
25 radioactive source (2). However, similar to the traditional methods, the non-nuclear density
26 gauges do not provide continuous information from the entire pavement area.
27 Determining pavement quality requires a specified number of samples regardless of the
28 density measurement method used. Because the random samples tested are only a small fraction
29 of the entire material, an uncertainty or risk exists in the measurement as well as the decision. To
30 enhance the level of confidence in the evaluation, a rapid and reliable test method that covers a
31 relatively larger surface area is necessary. Therefore, this study proposes mathematical methods
32 to estimate in-place HMA density using ground penetrating radar (GPR), which is fast,
33 continuous, and reliable. In addition, the GPR tests provide additional information about internal
34 pavement structures such as lift thickness and subsurface distress.
35
36 GROUND PENETRATING RADAR APPLICATIONS ON PAVEMENTS
37 Ground penetrating radar is a special type of radar designed to evaluate internal inhomogeneosity
38 and predict layer thickness of structures by penetrating the surface with electromagnetic waves.
39 Among the various types of GPR systems, a pulsed (or impulse) GPR system is the most
40 commercially available and commonly used to evaluate transportation infrastructure (3). The
41 principle of the pulsed systems is that a narrow EM pulse transmitted into the ground is reflected
42 back from the interfaces of materials that have distinct dielectric properties. The most common
43 uses of GPR in pavement engineering are to measure pavement layer thicknesses, to identify
44 large voids, to detect the presence of excess water in a structure, to locate underground utilities,
45 and to investigate significant delamination between pavement layers (4).
Al-Qadi, Leng, Lahouar, and Baek 4

1 Depending on how the antennae are deployed, GPR systems are classified into air-
2 coupled (or launched) or ground-coupled systems (5, 6). Air-coupled antennae are typically
3 mounted 15cm to 50cm above the surface (Figure 1a). These systems produce a clean radar
4 signal at the pavement surface and allow for high-speed surveys of up to 100km/h. The
5 drawbacks of air-coupled systems are their relative low penetration depth (although this could be
6 overcome by the recent introduction of the 500MHz air-coupled antenna) and a portion of the
7 EM energy is reflected back by the pavement surface. In contrast, a ground-coupled GPR
8 antenna is in full contact with the ground (Figure 1b), which gives a deeper penetration at the
9 same frequency but limits the speed of the survey to less than 8km/h. For HMA thickness and
10 density measurements, the air-coupled system is preferred because of its high survey speed and
11 better accuracy.
12

13
14 (a) (b)
15 FIGURE 1 Typical types of GPR antennae: (a) air-coupled and (b) ground-coupled (6)
16
17 The primary material property obtained from GPR surveys is the dielectric constant. The
18 dielectric constant, also known as the relative permittivity, (εr), of a homogeneous media relates
19 the relative EM velocity in a material to the speed of light in free space, c (Eq. 1).
20
2
⎛c⎞
21 εr = ⎜ ⎟ (1)
⎝v⎠
22
23 where, εr is the dielectric constant; c is the speed of light in free space of 3.0×108m/s; and v is
24 EM velocity in the material.
25
26 For a layered structure, such as a pavement, the dielectric constant of an HMA surface
27 layer, εr,HMA, can be estimated from the amplitudes of the reflected pulses (Figure 2) as follows
28 (5):
29
2
⎛ 1 + Ao / A p ⎞
30 ε r , HMA =⎜ ⎟ (2)
⎜ 1 − Ao / A p ⎟
⎝ ⎠
31
Al-Qadi, Leng, Lahouar, and Baek 5

1 where, εr,HMA is the dielectric constant of HMA; Ap is the amplitude the incident GPR wave
2 obtained by collecting data over a copper plate placed on the surface of the pavement; and Ao is
3 the amplitude of the surface reflection.
4
5 Then, the thickness of the HMA layer can be calculated as follows (5):
6
ct HMA
7 d HMA = (3)
2 ε r , HMA
8
9 where, dHMA is the HMA layer thickness; and tHMA is the two-way travel time of the GPR signal
10 within the HMA layer.
11
A0 Ap
HMA
tHMA
(εr,HMA, dHMA)

Reflection from the copper place


Base

Subgrade
12
13 FIGURE 2 Typical GPR signal for an HMA pavement system
14
15 The use of GPR for HMA density measurements has been progressively studied, but the
16 total research devoted to this method is minimal compared to the research devoted to other
17 applications for pavement systems. Lytton (7, 8) developed a computer program to determine the
18 density and water content of various pavement layers using conventional GPR (U.S. Patent No.
19 5384715). In Europe, Saarenketo (9) established a method for using GPR to measure HMA
20 pavement density. Based on the long-term studies on applying GPR for pavement QC, an
21 exponential relationship was developed between the surface dielectric constant and the void
22 content. Silvast (10) used GPR to measure air void contents of a HMA runway at the Helsinki-
23 Vantaa airport in Finland. He calculated the dielectric constant of the HMA using a surface
24 reflection technique and concluded that GPR technology presents a functioning QC method for
25 runway pavements. Silvast confirmed that rapidity and the ability to cover a large area are the
26 main advantages of GPR surveys compared to traditional methods. In addition, GPR enables the
27 monitoring of changes in pavement quality and structure over time. Two non-destructive test
28 methods, infrared imaging and GPR, were applied in Texas to evaluate the uniformity of density
29 in HMA overlays (11). By using the exponential equation proposed by the Finnish researchers,
30 the relationship between the surface dielectric constants and air voids was achieved through a
31 regression analysis. The GPR was concluded to be a much better tool for this investigation than
32 the infrared devices if density changes are the primary heterogeneities in the new HMA surface.
Al-Qadi, Leng, Lahouar, and Baek 6

1 However, in most of the previous studies, the prediction of HMA density was based on
2 pure statistical analysis of test data without any electromagnetic theory support. This study
3 applies electromagnetic mixing theory to develop the density prediction models.
4
5 RESEARCH OBJECTIVE AND APPROACH
6 Hot-mix asphalt is a composite material that contains asphalt binder, aggregates, air, and
7 possibly water. The density of HMA depends on the specific gravities and volumetric fractions
8 of its components. In a similar way, the dielectric constant of the HMA is a function of the
9 dielectric and volumetric properties of its components. The effective dielectric constant of HMA
10 is usually within the range of 3 to 12 (5). Various EM mixing models are available to predict a
11 mixture’s dielectric constant based on the dielectric constants and volume fractions of its
12 components (12). Most of these models hypothesize that a mixture is composed of a background
13 material with inclusions that are different sizes and shapes.
14 The objective of this study is to develop mathematical models that can be used to predict
15 HMA bulk specific gravity, which is equivalent to HMA density, based on its dielectric constant
16 using these EM mixing formulae. The following steps were taken to achieve this objective:
17 (1) Development of HMA bulk specific gravity models. Three existing EM mixing
18 models were selected as candidate models to determine the dielectric constant of a homogeneous
19 mixture (12-15): the complex refractive index model (CRIM), the Rayleigh mixing model, and
20 the Böttcher mixing model. The dielectric constant of the mixture is calculated as a function of
21 the dielectric and volumetric properties of its components. Then, the specific gravity models
22 were developed based on the EM mixing models, the known dielectric constant, and other known
23 volumetric properties, such as asphalt binder content.
24 (2) Evaluation of the HMA bulk specific gravity models using laboratory prepared HMA
25 specimens. HMA slabs (60cm×60cm×7.5cm) were prepared with the same job mix formulae but
26 various densities. The dielectric constants of these HMA slabs were measured using GPR, and
27 their bulk specific gravities were measured according to the AASHTO standard test method (16).
28 Based on the collected data, the best HMA bulk specific gravity model was selected from the
29 three candidate models developed in the previous step.
30 (3) Validation of the HMA bulk specific gravity model through in-situ GPR tests. Ground
31 penetrating radar tests were conducted on a test section to calibrate and validate the specific
32 gravity model. The test section was a 50-mm-thick HMA overlay on a concrete pavement. The
33 bulk specific gravity profile of the HMA overlay was created using the model selected in the
34 previous step. Five cores were extracted from the field, and their bulk specific gravities were
35 measured to validate the accuracy of the prediction model.
36
37 SPECIFIC GRAVITY MODEL DERIVATION
38 The three major components of HMA are aggregates, asphalt binder, and air, as shown in Figure
39 3. The volumetric and mass contribution of each component to the entire mixture is represented
40 by V and M, respectively; the specific gravity and dielectric constants of each component are G
41 and ε, respectively. The three EM models assume that the background material is asphalt binder
42 and all inclusions have spherical shapes.
43
Al-Qadi, Leng, Lahouar, and Baek 7

Gmb, Gmm MT

Va Air Gmb = bulk specific gravity of HMA


εa Gmm = maximum specific gravity of HMA
Asphalt binder VT = total volume
Va = volume of air
Gb Vb = total volume of binder
Vb Pb(%)
εb Mb Vsb = bulk volume of aggregate
Vse = effective volume of aggregate
VT MT = total mass
Absorbed Asphalt binder Mb = total mass of binder
=1 Ms = mass of aggregate
Gb = specific gravity of binder
Aggregate Gsb = bulk specific gravity of aggregate
Vsb Gsb Gse = effective specific gravity of aggregate
Vse Gse Ms εa = dielectric constant of air = 1.0
εs εb = dielectric constant of binder
εs = dielectric constant of aggregate
Pb = binder content
1
2 FIGURE 3 HMA volume and mass composition
3
4 Complex Refractive Index Model (CRIM)
5 Based on the CRIM mixture theory, the dielectric constant of a homogenous mixture can be
6 estimated as follow (12, 13):
7
8 (ε HMA )1/ α = Va (ε a )1 α + Vsb (ε s )1 α + Vb (εb )1 α , with εa=1 (4)
9
10 where α is a constant dependent on the mixture’s composition and is usually assumed to be
11 `1(13). Assuming the total volume of HMA, VT = 1, the following relations are concluded from
12 the volumetric properties of HMA:
13
Gmb
14 Va = 1 − (5)
Gmm
15
M b Gmb Pb
16 Vb = = (6)
Gb Gb
17
M s Gmb (1 − Pb )
18 Vsb = = (7)
Gsb Gsb
19
20 Substituting Eqs. 5, 6, and 7 in Eq. 4 gives:
21

22 (ε HMA )1 / α = (1 − Gmb ) ⋅ 1 + Gmb (1 − Pb ) (ε s )1 α + Gmb Pb (ε b )1 α (8)


Gmm Gsb Gb
Al-Qadi, Leng, Lahouar, and Baek 8

1
2 Reorganizing Eq. 8 gives the following:
3

4 Gmb =
(ε HMA )1/α − 1 (9)
Pb
(ε b )1 α + (1 − Pb ) (ε s )1 α − 1
Gb Gsb Gmm
5
6 For α=2, Eq. 9 is rewritten as follows:
7
ε HMA − 1
8 Gmb = (10)
Pb
εb +
(1 − Pb ) ε − 1
s
Gb Gsb Gmm
9
10 Rayleigh Mixing Model
11 With the Rayleigh mixing model, the effective dielectric constant, εeff, of a mixture composed of
12 a background material (dielectric constant εb) with N inclusions of different dielectric constants
13 is given by the following equation (12, 13, 15):
14
ε eff − ε b N

15 =∑ i i (11)
ε eff + 2ε b i =1 3ε b
16
17 where ni and αi are the number of inclusions and the polarizability factor of material i,
18 respectively.
19
20 The polarizability αi of a spherical inclusion of radius ai is given by:
21
εi − εb
22 α i = 4πai3ε b (12)
ε i + 2ε b
23
24 where εi is the dielectric constant of inclusion i.
25
26 Thus, the dielectric constant of the mixture is given by the following:
27
ε eff − ε b N
4πai3 ni ε i − ε b
28 =∑ (13)
ε eff + 2ε b i =1 3 ε i + 2ε b
29
30 Since the volume of ni spheres of radius ai is:
31
4πai3 ni
32 Vi = , (14)
3
33
34 Eq. 13 is rewritten as:
Al-Qadi, Leng, Lahouar, and Baek 9

1
ε eff − ε b N
ε −ε
2 = ∑ Vi i b (15)
ε eff + 2ε b i =1 ε i + 2ε b
3
4 where Vi is the fractional volume of inclusions i if the total material volume VT is considered
5 equal to one.
6
7 To apply the Rayleigh mixing model, the authors assumed that the HMA is composed of
8 an asphalt binder (dielectric constant εb) as the background material and that it includes
9 spherical-shaped aggregates and air particles. From Eq. 15, the HMA dielectric constant is given
10 by the following equation:
11
ε HMA − ε b ε −ε ε −ε
12 = Vsb s b + Va a b (16)
ε HMA + 2ε b ε s + 2ε b ε a + 2ε b
13
14 where all the parameters are as defined in Figure 4 and εa=1.
15
16 After solving the Eq. 16, the following equation can be developed:
17
ε HMA − ε b 1−εs

ε HMA + 2ε b 1 + 2ε b
18 G mb = (17)
⎛ ε s − ε b ⎞⎛ 1 − Pb ⎞ ⎛ 1 − ε s ⎞⎛ 1 ⎞
⎜⎜ ⎟⎟⎜⎜ ⎟⎟ − ⎜⎜ ⎟⎟⎜⎜ ⎟⎟
⎝ ε s + 2ε b ⎠⎝ G sb ⎠ ⎝ 1 + 2ε b ⎠⎝ G mm ⎠
19
20 Böttcher Mixing Model
21 The Böttcher mixing model is derived in the same way as the Rayleigh mixing model. The
22 dielectric constant of HMA in this case is given (12, 14):
23
ε HMA − ε b εs − εb εa − εb
24 = Vsb + Va (18)
3ε HMA ε s + 2ε HMA ε a + 2ε HMA
25
26 where all the parameters are as defined in Figure 4 and εa=1.
27
ε HMA − ε b 1−εb

3ε HMA 1 + 2ε HMA
28 G mb = (19)
⎛ ε s − ε b ⎞⎛ 1 − Pb ⎞ ⎛ 1 − ε b ⎞⎛ 1 ⎞
⎜⎜ ⎟⎟⎜⎜ ⎟⎟ − ⎜⎜ ⎟⎟⎜⎜ ⎟⎟
ε
⎝ s + 2ε HMA ⎠⎝ G sb ⎠ ⎝ 1 + 2ε HMA ⎠⎝ G mm ⎠
29
30 Model Sensitivity Analysis
31 Figure 4 shows, respectively, the variations of Gmb and air voids (Eq. 5), AV, as a function of
32 εHMA for the three mixture models using the following assumed parameters: εs = 6, εb = 3, Pb =
33 5%, Gb = 1.015, Gsb = 2.705, Gmm = 2.521, and α = 2. According to Figure 4a, the three models
Al-Qadi, Leng, Lahouar, and Baek 10

1 give approximately the same results for the specific gravity Gmb (e.g. εHMA = 5.2, Gmb = 2.334
2 from the CRIM, Gmb = 2.396 from the Rayleigh model, and Gmb = 2.363 from the Böttcher
3 model). The air voids are also comparable for the three models as shown in Figure 4b (e.g. εHMA
4 = 5.2, and AV = 7% from the CRIM model, AV = 5% from the Rayleigh model, and AV = 6%
5 from the Böttcher model).
6
2.6

2.5

2.4
Gmb

2.3

2.2 Gmb(CRIM)
Gmb(Rayleigh)
2.1 Gmb(Bottcher)

2
4.6 4.7 4.8 4.9 5 5.1 5.2 5.3 5.4 5.5 5.6
HMA Dielectric Constant

(a)
20
AV(CRIM)
AV(Rayleigh)
15 AV(Bottcher)
Air Void (%)

10

0
4.6 4.7 4.8 4.9 5 5.1 5.2 5.3 5.4 5.5 5.6
HMA Dielectric Constant

7 (b)
8 FIGURE 4 (a) Gmb and (b) air void variations as a function of εHMA
9
10 Figure 5 shows the specific gravity sensitivity of the three models with respect to
11 dielectric constant errors. For example, an εHMA error of 10% gives a Gmb error of 8.8% for the
12 CRIM model, 7.6% for the Rayleigh model, and 8.7% for the Böttcher model, respectively.
13 Hence, the error on the specific gravity is at most equal to the error on the dielectric constant. Of
14 the three models, the Rayleigh model has the lowest sensitivity to dielectric constant errors.
15
Al-Qadi, Leng, Lahouar, and Baek 11

20

15

10
Relative Gmb Error (%)

Gmb(CRIM), ε =4
-5
Gmb(Rayleigh), ε =4
Gmb(Bottcher), ε =4
-10 Gmb(CRIM), ε =5
Gmb(Rayleigh), ε =5
-15 Gmb(Bottcher), ε =5

-20
-20 -15 -10 -5 0 5 10 15 20
Dielectric Constant Error (%)
1
2 FIGURE 5 Comparison of relative Gmb error and εHMA error
3
4 MODEL EVALUATION USING LABORATORY TESTING
5 Laboratory tests were conducted on HMA slabs to evaluate the three specific gravity models. In
6 these tests, four 60cm×60cm×7.5cm HMA slabs were prepared with various compaction levels
7 and were tested using GPR to estimate their dielectric constants. The slab sizes were chosen
8 based on the footprint size of the GPR antenna used in this study. The specific gravity of these
9 slabs was estimated by taking direct measurements on cores and was later correlated to the GPR
10 results.
11
12 Slab Preparation
13 The HMA slabs were prepared of limestone aggregates and PG64-22 asphalt binder. The
14 gradation of this mixture is presented in Table 1. The asphalt binder content is 5% by weight.
15 Preheated HMA was placed in a 60cm×60cm wooden mold and then compacted with a small
16 vibratory roller. To achieve various density levels, specified quantities of HMA were poured in
17 the mold. Then, all slabs were compacted to a final design thickness of 76mm. After cooling
18 overnight, the slabs were removed from the mold for GPR testing.
19
Al-Qadi, Leng, Lahouar, and Baek 12

1 TABLE 1 Aggregate Gradation of the Slab and Field


Passing Ratio (%)
Sieve Size (mm)
Slab Field
25.0 100.0 100.0
19.0 100.0 98.0
12.5 100.0 76.0
9.5 97.0 66.0
4.75 56.3 39.0
2.36 34.5 23.0
1.18 22.7 16.0
0.6 15.0 10.0
0.3 8.9 7.0
0.15 6.3 6.0
0.075 5.1 4.5
2
3 Ground Penetrating Radar Data Collection
4 Ground penetrating radar data were collected from various slabs using a 2.0GHz air-coupled
5 antenna and the SIR20 GPR system manufactured by GSSI. As shown in Figure 6a, Styrofoam
6 sheets were placed between the antenna and the slab to ensure the air-coupling of the antenna;
7 Styrofoam has a dielectric constant close to that of air (εa = 1.0). To detect the reflection at the
8 bottom of the slab, an aluminum plate, which is a perfect EM reflector, was placed underneath it.
9 Figure 6b shows a typical GPR scan collected from one of the HMA slabs. The
10 reflections from the surface and bottom of the slab were easily detected. Besides, the other peaks
11 resulted from coupling between the transmitter and the receiver at 1.0ns and multiple ringing
12 from the aluminum plate, e.g., at 5.0ns. From the GPR data, the dielectric constant of each slab
13 was estimated from the reflection amplitude at the surface using Eq. 2. The dielectric constants
14 of the different slabs are 5.22, 5.12, 6.26, and 5.62.
15
Al-Qadi, Leng, Lahouar, and Baek 13

2.0GHz air-coupled GPR antenna

Styrofoam

HMA Slab
Aluminum foil

(a)
25

20
Surface reflection Bottom reflection
15
Amplitude (x1000)

10

-5

-10

-15
0 1 2 3 4 5 6 7
Time (ns)
(b)
1
2 FIGURE 6 GPR test on HMA slab: (a) GPR test setup and (b) typical GPR data collected
3 from the slab
4
5 Hot-Mix Asphalt Density Measurement from Cores
6 According to the AASHTO standards (16), the maximum theoretical specific gravity (Gmm) of
7 HMA was measured on loose mixture. After collecting GPR data from all the slabs, two cores
8 were taken from each slab to directly measure the specific gravity (Gmb) and the air-void content
9 (16). The results of these tests are presented in Table 2. Due to laboratory limitations in
10 preparing relatively large HMA slabs, the air-void content of the slabs was greater than 9%.
11 Hence, field validation will be conducted on a full-scale test site of HMA sections having a wide
12 range of air-void content.
13
Al-Qadi, Leng, Lahouar, and Baek 14

1 TABLE 2 Dielectric Constant and Specific Gravity for the Tested HMA
Dielectric Dry Sample Submerged SSD Air Voids
No. Gmb Gmm
Constant Weight (g) Weight (g) Weight (g) (%)
1 5.22 912.2 518.7 948.4 2.122 2.526 16.0
2 5.12 1041.1 597.9 1090.9 2.111 2.526 16.4
3 6.26 1013.7 579 1024.6 2.274 2.526 9.9
4 5.62 833.6 476.2 856 2.194 2.526 13.1
2
3 Hot-Mix Asphalt Specific Gravity Model Selection
4 Using the results of Table 2, the HMA specific gravity variations as a function of the change in
5 dielectric constants are presented in Figure 7. This figure also shows the fitting results of the
6 three models to the measured data using a non-linear least squares fitting algorithm due to the
7 complexity of the models. In this case, two variables were assumed constants (Gsb = 2.610 and
8 Gb = 1.015), and the remaining four parameters were determined from the curve fitting.
9

2.6
Measured Gmb
CRIM (R2=0.7481)
Rayleigh (R2=0.8001)
2.4 Bottcher (R2=0.7483)
Gmb

2.2

2.0

1.8
5.0 5.4 5.8 6.2 6.6
HMA Dielectric Constant
10
11 FIGURE 7 Measured and fitted Gmb variation as a function of εHMA
12
13 The results of the curve fitting procedure for the three models are presented in Table 3.
14 Based on these results, the dielectric constant estimation by the Rayleigh model gives more
15 accurate Gmb results (higher R2 value and lower root mean square error) than the other two
16 models for this particular mix. Therefore, the Rayleigh model’s performance was further
17 validated by using the GPR data collected in the field. The preliminary results show that the
18 specific gravity and air voids content of the HMA can be predicted from its dielectric constant
19 estimated from GPR data if a proper model is used.
Al-Qadi, Leng, Lahouar, and Baek 15

1
2 TABLE 3 Statistical Analysis Results for the Proposed Models
HMA Parameters
Model R2* RMSE*
εs εb Pb (%) Gmm
CRIM 6.38 3.21 5.1 2.602 0.7481 0.0857
Rayleigh 7.44 2.00 4.1 2.276 0.8001 0.0443
Böttcher 8.33 7.00 6.0 3.026 0.7483 0.0727
* 2
3 R is the coefficient of determination and RMSE is root mean square error.
4
5 MODEL VALIDATION IN A TESTING SITE
6 To further evaluate the performance of Rayleigh model for predicting in-place HMA density, a
7 field GPR survey was conducted on a composite pavement located at the University of Illinois.
8 This 75-m-long composite pavement consists of a 57-mm-thick HMA overlay on a continuously
9 reinforced concrete pavement (CRCP). The aggregate used in the HMA is limestone, and the
10 asphalt binder is PG64-22 at 4.6%. The gradation of the HMA is shown in Table 1. Although the
11 same aggregate type and binder grade were used, the job mix formulae of HMA overlay and the
12 laboratory slabs are different. The same 2.0GHz air-coupled antenna was used to collect GPR
13 data (Figure 8a). The survey speed was approximately 30km/h, which means it only took
14 approximately 10s to survey the entire 75-m-long pavement section. Figure 8b shows a GPR B-
15 scan image for the surveyed composite pavement. Based on the GPR data, the dielectric constant
16 profile of the HMA surface was calculated using Eq. 2 as shown in Figure 8c. The dielectric
17 constant values of the HMA surface are relatively constant throughout the entire pavement
18 section. All dielectric constant values range from 7.0 to 7.5. This indicates that the HMA density
19 is also relatively uniform for the entire section.
20 After the dielectric constant profile of the HMA surface was obtained, the bulk specific
21 gravity profile of the HMA surface was calculated using the specific gravity model based on the
22 Rayleigh model presented in Eq. 17. In this model, the following parameters are either constant
23 or measurable: εb = 3.0 (as a typical value), Pb = 4.6% (as measured), Gb = 1.015 (as a typical
24 value), Gsb = 2.70 (as a typical value), Gmm = 2.488 (as measured). Only one parameter, εs, is
25 unknown on the right hand side of Eq. 16. Since the dielectric constant of limestone, εs, is within
26 a relatively wide range of 7.5 to 9.2 (17), one HMA core was extracted from the field and its
27 bulk specific gravity was measured in the lab to obtain a good estimate of this value. Based on
28 the measured Gmb from the HMA core, εs was determined as 8.78. The bulk specific gravity
29 profile of the HMA surface was calculated and presented in Figure 8c.
30
31
Al-Qadi, Leng, Lahouar, and Baek 16

HMA Surface

HMA Bottom

(a) (b)
8.0 2.6

7.5 2.5

HMA Bulk Specific Gravity


HMA Dielectric Constant

7.0 2.4

6.5 2.3

6.0
2.2
Predicted by model
5.5 Measured on cores 2.1
Calibration value
5.0 2.0
0 15 30 45 60 75

Distance (m)
(c)
1
2 FIGURE 8 HMA pavement GPR survey using a 2.0GHz air-coupled antenna: (a) GPR
3 field survey; (b) GPR data image; and (c) dielectric constant and bulk specific gravity
4 profile of the HMA surface
5
6 To validate the accuracy of the predicted specific gravity values, four more field cores
7 were taken from the same location where the GPR test was conducted. The measured bulk
8 specific gravities of these HMA cores are also shown in Figure 8c (marked as square shapes) and
9 Table 4. The predicted HMA bulk specific gravity values show reasonable agreements with the
10 measured values from the cores. The maximum relative error of the bulk specific gravity
11 prediction is 5.1% (average 4.3%), which indicates that the estimations by the bulk specific
12 model based on the Rayleigh formula are reasonably accurate after calibration with the pavement
13 core. In order to improve the methodology and the prediction accuracy of the developed models,
14 more field validations are being conducted currently and will be presented in the future.
15
16
17
Al-Qadi, Leng, Lahouar, and Baek 17

1 TABLE 4 Relative Errors in Bulk Specific Gravity Prediction


Core # 1 2 3 4 5*
Gmb measured on cores 2.398 2.395 2.380 2.368 2.350
Gmb predicted by the Rayleigh model 2.388 2.375 2.384 2.362 2.350
Relative error (%) 4.3 5.1 3.6 4.6 0.0
2 * Core #5 was used for calibration
3
4 SUMMARY AND DISCUSSION
5 Based on the EM mixing theory, three HMA bulk specific gravity models were developed in this
6 study. The best model was selected through laboratory testing and validated by field testing. The
7 ability of GPR to predict in-place HMA density was investigated. The findings are summarized
8 below:
9
10 • Estimated HMA dielectric constant from GPR data can yield its density and air void
11 content when an appropriate model is used. For the case presented in this paper, the
12 relative errors of the predicted HMA bulk specific gravities based on the Rayleigh mixing
13 model are less than 6% (average 4.3%) after the model was calibrated with a field core.
14 • When GPR is used for in-place HMA pavement density prediction, the density profile of
15 the entire pavement can be obtained. This significantly improves the reliability of the in-
16 situ pavement density prediction considering that all the other in-place density
17 measurement methods can only provide density data at discrete sampling locations.
18 • The proposed method to predict in-place HMA density using GPR is relatively faster than
19 other conventional HMA density measurement methods. For example, for the 75-m-long
20 pavement section surveyed in this study, it only took approximately 10 seconds to collect
21 the GPR data and several seconds to analyze it.
22
23 However, it should be noted that the method developed in this study is only valid for one
24 type of limestone aggregate and PG64-22 asphalt binder, and all materials were dry. In future
25 research, the model should include more variables, such as different aggregate types and
26 moisture effects. The accuracy of the model should also be validated with more data collected
27 from in-service airport and highway pavements. The proposed steps can assure the accuracy of
28 the HMA specific gravity models, so this new approach for in-place HMA density prediction can
29 be implemented in practice.
30
31 ACKNOWLEDGMENTS
32 The authors would like to acknowledge the input of Al Larkin of the Federal Aviation
33 Administration (FAA).This publication is based on the preliminary results of an FAA project,
34 Non-Destructive Testing and Evaluation (NDTE) Technologies for Airport Pavement
35 Acceptance and Quality Assurance Activities. The FAA project is conducted in cooperation with
36 the Center of Excellence for Airport Technology (CEAT) and the University of Illinois at
37 Urbana-Champaign. The contents of this study reflect the views of the authors, who are
38 responsible for the facts and the accuracy of the data presented herein. The contents do not
39 necessarily reflect the official views or policies of the CEAT or FAA. The research was
Al-Qadi, Leng, Lahouar, and Baek 18

1 conducted at the Advanced Transportation Research and Engineering Laboratory (ATREL) at


2 the University of Illinois at Urbana-Champaign. This report does not constitute a standard,
3 specification, or regulation.
4
5 REFERENCES
6 1. Killingsworth, B. M. Quality Characteristics for Use with Performance Related
7 Specifications for Hot Mix Asphalt. NCHRP Project 9-15 Research Results Digest 291,
8 Washington, D.C.: Transportation Research Board, National Research Council, 2004.
9
10 2. Romero, P., and F. Kuhnow. Evaluation of New Non-nuclear Pavement Density Gauges with
11 Data from Field Projects. In Transportation Research Record: Journal of the Transportation
12 Research Board, No. 1813, Transportation Research Board of the National Academies,
13 Washington, D.C., 2002, pp. 47-54.
14
15 3. Al-Qadi, I. L. Radar Testing of Concrete, Where Are We Now. Keynote Presented at the 8th
16 International Conference and Exhibition on Structural Faults and Repair, London, United
17 Kingdom, 1999.
18
19 4. Federal Aviation Administration (FAA) Advisory Circular: Use of Non-destructive Testing in
20 the Evaluation of Airport Pavements. AC No. 150/5370-11A, Washington, D.C.: U.S.
21 Department of Transportation, 2004.
22
23 5. Lahouar, S. Development of Data Analysis Algorithms for Interpretation of Ground
24 Penetrating Radar Data. Ph.D. Dissertation, Virginia Polytechnic Institute and State
25 University, Blacksburg, VA, 2003.
26
27 6. Leng, Z., I. L. Al-Qadi, J. Baek, and S. Lahouar. Development of Data Analysis Algorithms
28 for Interpretation of Ground Penetrating Radar Data. Presented at the 86th Annual Meeting
29 of the Transportation Research Board, CD-ROM, Transportation Research Board of the
30 National Academies, Washington, D.C., 2007.
31
32 7. Lytton, R. L. System Identification and Analysis of Subsurface Radar Signals, U.S. Patent No.
33 5384715, Texas A&M University, Licensed to Lyric Technologies, Inc., Houston, TX, 1995.
34
35 8. Lytton, R. L. Characterizing Asphalt Pavements for Performance, In Transportation
36 Research Record: Journal of the Transportation Research Board, No.1723, Transportation
37 Research Board of the National Academies, Washington, D.C., 2000, pp.5-16.
38
39 9. Saarenketo, T. Using Ground Penetrating Radar and Dielectric Probe Measurements in
40 Pavement Density Quality Control. In Transportation Research Record: Journal of the
41 Transportation Research Board, No.1575, Transportation Research Board of the National
42 Academies, Washington, D.C., 1997, pp. 34-41
43
44 10. Silvast, M. Air Void Content Measurement Using GPR Technology at Helsinki-Vantaa
45 Airport: Runway No. 3, Survey Report, Roadscanners [Electronic version], 2001.
46
Al-Qadi, Leng, Lahouar, and Baek 19

1 11. Sebesta, S., and T. Scullion. Application of Infrared Imaging and Ground-Penetrating Radar
2 for Detecting Segregation in Hot-Mix Asphalt Overlays. In Transportation Research Record:
3 Journal of the Transportation Research Board, No.1861, Transportation Research Board of
4 the National Academies, Washington, D.C., 2003, pp.37-43.
5
6 12. Al-Qadi, I. L., and S. M., Riad. Characterization of Portland Cement Concrete:
7 Electromagnetic and Ultrasonic Measurement Techniques, Report Submitted to the National
8 Science Foundation, Washington, D.C., 1996.
9
10 13. Sihvola, A. Electromagnetic Mixing Formulas and Applications, the Institution of Electrical
11 Engineers (IEE) Publishing, London, United Kingdom, 1999.
12
13 14. Bottcher, C. J. F. Theory of Electric Polarisation, Elsevier Publishing Company, Amsterdam,
14 Netherland, 1952.
15
16 15. Behari, J. Microwave Dielectric Behavior of Wet Soils, Springer, Berlin, Germany, 2005.
17
18 16. American Association of State Highway and Transportation Officials. Standard
19 Specifications for Transportation Materials and Methods of Sampling and Testing. (22nd
20 Ed.). AASHTO, Washington, D.C., 2001
21
22 17. Johnson, D., and K. Pile. Well Logging, Penn Well Publishing Company, Tulsa, OK, 2002.

You might also like