Nondestructive quality assessment of asphalt pavements based on dynamic modulus

Construction and Building Materials 112 (2016) 836–847

Contents lists available at ScienceDirect

Construction and Building Materials journal homepage: www.elsevier.com/locate/conbuildmat

Nondestructive quality assessment of asphalt pavements based on dynamic modulus Shibin Lin, Jeramy C. Ashlock ⇑, R. Christopher Williams Department of Civil, Construction & Environmental Engineering, Iowa State University, Ames, IA 50010, USA

h i g h l i g h t s

g r a p h i c a l a b s t r a c t


A novel nondestructive QC/QA

procedure is proposed based on in situ dynamic modulus.
A method is proposed to correct in situ moduli to a reference asphalt temperature.
In situ shear-wave velocity is very sensitive to pavement type and temperature.
In situ density is much less sensitive than velocity to pavement type and temperature.
Modulus is a critical, objective, and sensitive property of asphalt pavement systems.

a r t i c l e

i n f o

Article history: Received 1 September 2015 Received in revised form 9 February 2016 Accepted 25 February 2016

Keywords: Asphalt pavement Quality control Quality assurance Dynamic modulus Shear-wave velocity Density

a b s t r a c t Dynamic modulus has been recognized as an objective and sensitive material property for designing and evaluating pavement systems. To accurately measure the in situ elastic modulus (E = 2(1 + m)qVs2) for nondestructive quality assessment of asphalt pavements, field measurements of density (q) via an electromagnetic gauge and shear-wave velocity (Vs) via surface-wave testing were examined for four paving projects covering a range of mixes and traffic loads. A quality control/quality assurance (QC/QA) procedure was developed to correct the in situ moduli at different field temperatures to a common reference temperature using a fitting function from experimental data for QC and using master curves from laboratory dynamic modulus tests for QA. The corrected in situ moduli can then be compared against the maximum moduli for an assessment of the actual pavement performance. Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction Asphalt pavements suffer various modes of failure after construction, and commonly require rehabilitation to enable them to reach their design life-spans. The empirical design of asphalt pavements has been identified as one of the major reasons for cases of ⇑ Corresponding author. E-mail addresses: [email protected] (S. Lin), [email protected] (J.C. Ashlock), [email protected] (R.C. Williams). http://dx.doi.org/10.1016/j.conbuildmat.2016.02.189 0950-0618/Ó 2016 Elsevier Ltd. All rights reserved.

inadequate performance. The mechanistic-empirical pavement design guide (MEPDG) was therefore developed to enable quantitative performance prediction for the design of new and rehabilitated pavement structures [1]. In addition to traffic and climate data as inputs, mechanistic-empirical (M-E) design procedures require measurement of fundamental pavement material properties rather than use of empirical relationships. Quality control and quality assurance (QC/QA) procedures based on measured fundamental properties are thus necessary for enabling quantitative evaluations of pavement condition and performance.

S. Lin et al. / Construction and Building Materials 112 (2016) 836–847

Traditional QC/QA procedures are based on properties of asphalt mixtures (e.g., density, air voids, permeability) measured in one of two ways: (1) using in situ measurements taken on the surface of an asphalt layer (e.g. [2–5]), or (2) using laboratory tests on field cores (e.g. [5–7]). Although these traditional QC/QA procedures play an important role in ensuring high-quality pavements, they employ only the volumetric properties, thickness, and roughness (i.e., present serviceability index), but not actual mechanical properties such as modulus. However, the mechanical properties are more sensitive than volumetric properties to variations in quality, and are also required for performance-based M-E pavement design. Additionally, the quality and remaining life of a pavement can be evaluated based on modeling it as an elastic and/or viscoelastic multilayered system, and predicting the strains and stresses at interfaces of different layers [8]. The modulus of an asphalt layer is also of critical importance for estimating fatigue cracking [9]. The use of modulus for QC/QA has been examined by Celaya et al. [8], Li and Nazarian [10], Nazarian et al. [11], Celaya and Nazarian [12], Jiang [13], Barnes and Trottier [14,15], and Icenogle and Kabir [16]. A variety of nondestructive surface-wave testing equipment for in situ modulus measurement has been developed by Nazarian [17], Stokoe et al. [18], Park et al. [19], Ryden [20], and Lin and Ashlock [21,22], among others. Overall, good correlations have been demonstrated between in situ modulus and laboratory modulus (e.g. [23,24]). Additionally, QC/QA methods based on measured moduli have been demonstrated to be more objective for characterization of asphalt layers by accounting for effects of temperature and loading frequency [8,11,25]. The advancement of QC/QA from methods based on volumetric and geometrical properties to those based on mechanical properties has been driven by the evolution from empirical to M-E design procedures [24]. Significant progress has been made in previous studies on the use of modulus for QC/QA of asphalt pavements (e.g. [8,11,25]), but several challenges remain, such as: (1) statistical variability and uncertainty of in situ shear-wave velocity measurements, (2) a systematic method for correction of in situ moduli measured at different field temperatures to a common reference temperature, and (3) a straightforward and practical QC/QA procedure. To overcome these challenges, the present study is aimed at developing the following: (1) a more robust and less delicate seismic testing system to provide consistent, reliable, and objective in situ velocity measurements, (2) a systematic method to correct in situ moduli measured at different field temperatures to moduli at a common reference temperature based on a fitting function from experimental data for QC and master curves from laboratory dynamic modulus tests for QA, and (3) a straightforward and practical QC/QA procedure to determine quantitative measures of pavement quality from the in situ dynamic modulus measurements.

2. Nondestructive measurement of dynamic modulus Surface-wave methods (SWM) and falling weight deflectometer (FWD) tests have been widely used for determination of asphalt modulus (e.g. [16,26]). However, it is difficult to accurately determine the modulus of individual pavement layers from FWD tests because of the relatively larger receiver spacing, lack of highfrequency content in the impact load, and lower sampling rates compared to SWM. On the other hand, SWM test equipment can be used with relatively smaller receiver spacing to capture shorter wavelengths for the thin layers of interest, and smaller hammers to generate much greater high-frequency content (typically 10– 30 kHz) than FWD tests (typically 1 kHz) [16]. Typical surface wave methods used for pavements include the Spectral Analysis of Surface Waves (SASW, e.g. [17]), Multichannel Analysis of Surface

837

Waves (MASW, e.g. [19,22]), and Multichannel Simulation with One Receiver (MSOR) methods [20]. MSOR and SASW methods involve measurement of surface motion for sequential impacts over an array of source-to-receiver offsets, using a single receiver for (MSOR) or a pair of receivers (SASW). MASW involves simultaneous measurement of surface motion by an array of sensors for a single impact. Surface wave data are processed to obtain the shear-wave phase-velocity spectra in the form of dispersion images or dispersion curves. Surface wave phase velocities at low frequencies correspond to material properties at greater depths, and high frequencies correspond to shallower depths. To more accurately measure the properties of the pavement surface layer, the measurement depth can be reduced by decreasing the receiver spacing, increasing the high-frequency content of the impact, and increasing the sampling rate. For the small strains involved in surfacewave testing, the asphalt behaves viscoelastically, and the Young’s modulus (or stiffness, E) can be obtained from shear wave velocity (VS) as

E ¼ 2ð1 þ mÞqV 2S

ð1Þ

where m is Poisson’s ratio which has a relatively minor influence and can be assumed constant for each layer, and q is mass density which can be measured in situ using an electromagnetic gauge. Because the near-surface velocity (stiffness) decreases with depth for typical pavement profiles consisting of pavement, base, subbase, and subgrade layers, the corresponding phase-velocity spectra from surface-wave tests primarily show an increase in phase velocity with frequency. However, the phase-velocity spectrum of a layered pavement system actually consists of several branches that can be approximated as multiple modes of antisymmetric and symmetric Lamb waves for a free plate corresponding to the material properties of the pavement layer (Ryden et al. [27]). The correspondence to Lamb waves is approximate, because the pavement layer is not truly free but interacts with the underlying base and subgrade layers to create partial branches of leaky quasi-Lamb waves in the low-frequency regime. At high frequencies (typically above 10 kHz), the experimental phase velocities approach those of the fundamental anti-symmetric (A0) and symmetric (S0) modes of dispersive Lamb waves, which themselves asymptotically approach the pavement layer’s Rayleigh-wave velocity [28]. To obtain accurate properties of the base and subgrade layers, inversion of the phase-velocity dispersion data may therefore require matching forward-modeled theoretical dispersion curves to the low-frequency branches generated by interaction of the leaky quasi-Lamb waves. Alternatively, if only the properties (modulus and thickness) of the stiff top pavement layer are desired, inversion can possibly be avoided by using a simplified analysis in which the experimental surface-wave phase velocity is matched to the fundamental A0 Lamb-wave dispersion curve of a free plate (as well as segments of the S0 mode if detected), as described by Ryden et al. ([27,29]). If the experimental dispersion data can be measured to sufficiently high frequencies such that a horizontal asymptote is actually observed, then the Rayleigh-wave velocity of the pavement layer may simply be read as the asymptotic value of the dispersion curve. For the bandwidth-limited MSOR tests in this study, a more general approach was taken in which a numerical model of the pavement, base, subbase, and soil layers was used in a multilayer inversion procedure to solve for the thickness and phase velocity of each layer. Because the solutions are non-unique, the inversion procedures employ optimization methods to minimize the misfit between the experimental dispersion curves and the theoretical dispersion curves of randomly generated multilayer models [30].

838

S. Lin et al. / Construction and Building Materials 112 (2016) 836–847

Whether employing a true multilayer inversion or the free-plate Lamb wave approximation, resolution of the pavement layer properties requires accurate experimental measurement of the phasevelocity spectrum at high frequencies. A high-resolution testing setup and delicate operation are required, because of the high wave speeds, short wavelengths, and small motions involved. To reliably measure high-frequency dispersion characteristics with results comparable to MASW data, the MSOR method requires a highly repeatable impact source that can generate surface waves with consistent timing and triggering [31], with minimal deviation from the intended impact locations. This is because all sensors in an MASW test measure the same seismic waves from a single triggered impact, whereas generation of the equivalent data by MSOR involves measurement of different sets of seismic waves from an array of triggered impact locations. Based on the results of several MSOR and MASW tests performed in this research, the MASW method can more easily provide reliable high-frequency measurements owing to the fixed receiver locations and less-stringent requirements on impact repeatability. Although not examined in this research, it has been reported that a coupling spike for applying impacts at precise locations can extend the high-frequency range of MSOR data [29]. The primary drawbacks of MASW testing for pavements are the costs of multiple accelerometers and a multichannel signal analyzer, as well as the time required to couple and decouple multiple accelerometers. For more than a decade, the portable seismic pavement analyzer (PSPA) has been the state-of-the-practice equipment for surface-wave testing for engineers and Departments of Transportation (DOTs) [8,16,32]. However, PSPA users have experienced occasional difficulties in measuring high-quality data due to deterioration of the rubber feet, and rough asphalt surfaces [16]. Moreover, the PSPA averages a wide range of discrete dispersion data at short wavelengths (below 6 in.) and high frequencies (above 10 kHz) [8,12], which can produce significant uncertainty in the shear-wave velocity of asphalt pavement surface layers. To overcome several of the challenges associated with surface-wave testing identified above, Lin and Ashlock [22] demonstrated that the custom-built MASW testing system and data acquisition (DAQ) program developed in this study can reduce the uncertainty and extend the frequency range of dispersion data with great efficiency at relatively low cost. 3. In situ tests 3.1. Asphalt test sections A total of eight different representative test sections from four asphalt paving projects were selected to cover a range of pavement types including high-volume versus low-volume, hot-mix asphalt

(HMA) versus warm-mix asphalt (WMA), and newly-constructed versus resurfaced. The following projects were selected: (1) newly constructed pavement sections for a research project at the Boone Central Iowa Expo site featuring low-volume roads with HMA and WMA, with various base and pavement treatments and construction techniques; (2) US 69, a medium-volume road with HMA resurfacing; (3) IA 93, a low-volume road with cold-in-place recycling, thin overlay, and full-depth-reclamation resurfacing methods; and (4) US 6, a high-volume road with sections of HMA and WMA containing steel slag. Details for each paving project and designations for the test locations and core samples are given in Table 1. For each project site, at least six test locations in the paving lane were randomly selected. Surface-wave tests were performed along straight lines centered on the test locations and parallel to the paving direction. The in situ density was measured at each test location immediately before or after the surface-wave tests, as described in the following section. Field pavement temperatures were recorded by inserting a thermocouple probe into freshly placed asphalt pavements, or by holding the probe against the surface of cured asphalt pavements. Core samples were then extracted at the exact test locations where the SWM, density, and temperature measurements were performed. ‘‘Hot” tests were performed up to a few hours after paving, and ‘‘cold” tests were performed after cooling the pavement surface with dry ice. Additional ambient-temperature tests were performed on selected projects one or more days after paving.

3.2. Density measurements The Troxler Model 2701-B PaveTracker (PT) Plus non-nuclear electromagnetic (EM) gauge was employed to measure in situ density immediately before or after surface wave tests at the core locations detailed in Table 1. For each test location, the gauge was calibrated to a reference standard, then five readings were recorded with the device rotated 90° between readings and the values were averaged. For the Boone base courses, hot tests were performed a few hours after paving and ambient-temperature tests were performed the following day. For the US 69 project, hot tests were performed a few hours after paving, then cold tests were performed after applying crushed dry ice for approximately 2 min. Upon analysis of the 44 tests performed at the Boone and US 69 projects, it was concluded that EM density measurements on the cold and ambient-temperature pavements were of limited value, while measurements on hot pavements exhibited consistent trends with small statistical variation. Therefore, EM density measurements were not taken for the cold and ambient conditions on the IA 93 and US 6 projects.

Table 1 Description of paving projects in this study. Test section No.

Project site/pavement type

# of test locations

Test and core location designators

Design traffic

1

Boone/HMA base

19

NA

2

Boone/WMA base

16

3

US 69/HMA

9

4 5 6 7 8

IA 93/FDR IA 93/CIP IA 93/OL US 6/HMA US 6/WMA

2 2 2 6 4

HB1-1, HB1-3, HB1-5, HB1-7, HB2-1, HB2-2, HB2-5, HB2-7, HB5-1, HB5-3, HB5-6, HB5-7, HB6-3, HB6-4, HB6-5, HB6-7, HB7-2, HB7-4, HB7-7 WB3-2, WB3-4, WB3-5, WB3-7, WB4-1, WB4-2, WB4-3, WB4-7, WB8-4, WB8-5, WB8-7, WB8-8, WB9-2, WB9-3, WB9-6, WB9-7 US 69–1, US 69–2, US 69–3, US 69–4, US 69–5, US 69–6, US 69–7, US 69– 8, US 69–9 IA 93 FDR-1, IA 93 FDR-2 IA 93 CIP-1, IA 93 CIP-2 IA 93 OL-1, IA 93 OL-2 US 6 H20-2, US 6 H20-3, US 6 H25-1, US 6 H25-2,US 6 H30-2, US 6 H30-3 US 6 W15-1, US 6 W15-2, US 6 W30-2, US 6 W30-3

NA 2011 AADT = 3400 VPD, 5% Trucks, 600,000 ESALs 2011 AADT = 1630 VPD, 475,960 ESALs 2013 AADT = 22,000 VPD, 4% Trucks, 3,700,000 ESALs

Notes: Cores in bold were selected for laboratory dynamic modulus tests. NA = Not Applicable, VPD = Vehicles per Day, AADT = Average Annual Daily Traffic, ESALs = Equivalent Single Axle Loads.

839

Average hot density (kg/m 3)

S. Lin et al. / Construction and Building Materials 112 (2016) 836–847

the US 6 project. The IA 93-OL, IA 93-FDR, and IA 93-CIP cores have the lowest densities, which are approximately 90%, 83%, and 78%, respectively, of the average US 6 density. The somewhat lower densities for IA 93-FDR and IA 93-CIP may be due to these specimens crumbling apart after coring. The correlation between in situ density of hot pavements and laboratory core densities for all 60 test locations is relatively low, with an R2 of 0.2397 (Fig. 3a). Taking the average densities for each of the 8 test sections only slightly improves the correlation coefficient to 0.2983 (Fig. 3b).

2,800 2,600 2,400 2,200 2,000 1,800

3.3. Shear-wave velocity measurement

2,800 2,600 2,400 2,200 2,000 1,800

Fig. 2. Comparison of average laboratory densities for the 8 test sections.

In situ denstity (kg/m3)

The average EM densities measured on hot pavements for each of the 8 test sections are shown in Fig. 1. Compared to the SWM measurements discussed in Section 3.3, the EM densities show less variation with pavement type. The US 6 pavements with the highest traffic volume have the highest densities, as expected, but are followed closely by IA 93-FDR, while the remaining 5 test sections have similar densities within a range of 3% of their average value of 2230 kg/m3. The CoreLok method [33] was employed to measure the laboratory density of all cores, because this method could be used on the loose core specimens which tended to crumble, particularly the FDR and CIP specimens from IA 93. A comparison of the average CoreLok densities for each of the 11 test sections is shown in Fig. 2. Similar to the field test results, the US 6 HMA and WMA cores have the highest laboratory densities, but the order of the remaining sections is not consistent with the field densities measured by the EM gauge. The Boone and US 69 pavements have similar densities, which are approximately 94% of the average density from

(a) 2,600

R² = 0.2397

2,400 2,200 2,000 1,800 1800

2000

2200

2400

Core density (kg/m 3 )

The surface wave testing equipment and DAQ program developed in this research were used for MSOR testing at the Central Iowa Expo project site in Boone County. Several different types of pavements, base courses, and stabilization technologies were used at this site for the Iowa DOT research project ‘‘Boone County Expo Research Phase I-Granular Road Compaction and Stabilization.” MSOR tests were performed on the base courses at 19 HMA and 16 WMA locations, as detailed in Table 1. For the first eight test locations, MSOR data were recorded using 24 impacts located 0.1–2.4 m from the receiver in 0.1 m increments. After examining the data for the first eight locations, the testing spread was reduced to 0.05–1.2 m with smaller 0.05 m increments for the remaining Boone and IA 93-FDR test sections, to obtain better resolution for the asphalt surface layer which is of interest for this research. The testing spread was further reduced to 0.03–0.72 m with 0.03 m increments for the US 69, IA 93-CIP and IA 93-OL test sections. The MSOR testing approach was replaced by MASW for the US 6 project test sections. Field time-domain data and corresponding dispersion trends are shown in Fig. 4 for hot and ambient-temperature MSOR surface-wave tests on the Boone HMA base course test location HB1-1. The experimental dispersion images were obtained using the MASW phase-velocity scanning method [34] with enhancements by Lin [30]. The experimental dispersion image is an intensity plot, for which the maximum amplitude at a given frequency gives the surface-wave phase velocity at that frequency. The maximum amplitudes were picked algorithmically to give an experimental dispersion curve, which was then used in the inversion procedure to determine the material properties (modulus and thickness) of the pavement-system layers. The tests on hot asphalts consistently resulted in experimental dispersion data with relatively low maximum measurable frequencies and corresponding low phase velocities. As the pavements cooled, their modulus increased and damping decreased, resulting in increased phase velocities and significantly greater maximum measurable frequencies. For example, the hot test with an asphalt layer temperature of 43.4 °C shown in Fig. 4b has a maximum measurable frequency of only 350 Hz, corresponding to a relatively low maximum phase

Average in situ denstity (kg/m3)

Average density (kg/m3)

Fig. 1. Comparison of average EM densities measured on hot pavements for the 8 test sections.

(b) 2,600

R² = 0.2983

2,400 2,200 2,000 1,800 1800

2000

2200

2400

Average core density (kg/m3)

Fig. 3. Correlation between in situ EM hot density and laboratory CoreLok density: (a) cores from all 8 test sections (n = 60), (b) average densities for each test section (n = 8).

840

S. Lin et al. / Construction and Building Materials 112 (2016) 836–847

Fig. 4. MSOR surface-wave test results for Boone HMA base course location HB1-1: (a) normalized time-domain data for hot test, (b) experimental dispersion image for hot test, (c) normalized time-domain data for ambient-temperature test, (d) experimental dispersion image for ambient-temperature test.

velocity of 170 m/s. The ambient-temperature test at 26.6 °C at the same location has a much greater maximum frequency of 4 kHz and corresponding phase velocity of 1000 m/s (Fig. 4d), whereas the phase velocity at 350 Hz also increased to 250 m/s. For the dispersion data from the ambient-temperature and cold tests, the Genetic and Simulated Annealing (GSA) inversion program developed by Lin [30] was used to back-calculate the shear-wave velocities of the multiple-layer pavement profile using assumed layer thicknesses from the design plans, and assumed values of Poisson’s ratio (0.3 for pavement and base layers, 0.35 for subbase layer, and 0.4 for subgrade layer) and density for each layer. Experimental dispersion curves were input into the inversion program for the MSOR tests on the Boone, US 69 and IA 93OL cores selected for laboratory dynamic modulus testing as indicated in Table 1. The inversion program searches over layered profile parameters that are randomly perturbed within specified ranges, and performs a nonlinear optimization to match theoretical dispersion curves to the experimental dispersion curves. The output is the final inverted shear-wave velocity profile whose theoretical dispersion curve best fits the experimental one. The solutions are non-unique, however, because each time the Monte-Carlobased GSA program is run, it will converge to one of several similar final inverted profiles that are close to the true profile. To enhance measurement precision in the high-frequency regime which provides information on the pavement surface layer, the MSOR equipment and DAQ program were modified to enable MASW testing using a linear array of 9 accelerometers. MASW tests were then performed at six HMA and four WMA surface course locations on US 6, with receiver stations spaced 0.05 to 0.45 m from the impact point in 0.05 m increments. Hot tests and ambient-temperature tests were performed several hours and several days after paving, respectively. Representative field data and corresponding dispersion trends are shown in Fig. 5 for hot and ambient-temperature MASW tests on WMA at location US 6 W30-2. Similar to the MSOR test in Fig. 4, the maximum useable frequency in the MASW dispersion data increased significantly as the asphalt temperature decreased, from a frequency of 9 kHz in the hot test at 44.0 °C to 25 kHz in the ambient-temperature test at 22.4 °C. More importantly, the tests on US 6 revealed that the MASW testing approach enabled measurement of dispersion data

to significantly higher frequencies (around 25 kHz) than the MSOR approach (around 4 kHz) using the authors’ testing system and procedures. This is due to several factors, including much less stringent requirements on impact repeatability and trigger accuracy in MASW testing, as discussed in Lin and Ashlock [22]. With its improved high-frequency measurement capabilities, the MASW testing system enabled the velocity of the quasiRayleigh waves in the first layer ðV R Þ to be measured as the highfrequency horizontal asymptote of the phase velocity (Fig. 5d). The shear-wave velocity of the surface pavement layer ðV S1 Þ can therefore be obtained directly from V R without the need for inversion, using the following approximate relation from [35]:

 V S1 ¼

 1þv VR 0:87 þ 1:12v

ð2Þ

which greatly simplifies data analysis and enables the shear-wave velocity to be determined immediately in the field. For each of the core locations in Table 1, the in situ elastic modulus of the surface pavement layer was calculated by substituting the shear wave velocity V S1 from SWM testing into Eq. (1) along with the density measured by the EM gauge. Based on the discussion in this section and Section 3.2, the EM-gauge density data were taken from hot tests, while the SWM data were from ambient-temperature (for US 6, Boone, IA 93-FDR, and IA 93-CIP) and cold tests via dry ice (for US 69 and IA 93-OL). For each of the eight test sections, the average V S1 and modulus values are shown in Fig. 6. The average field temperatures ranged from 12.9 to 33.2 °C, as indicated in the figure. The HMA and WMA pavements from US 6 have the highest in situ moduli of approximately 18,000 MPa. The Boone, IA 93OL, and US 69 pavements have intermediate moduli that are approximately half of those from US 6. The IA 93-CIP and IA 93FDR pavements have the lowest moduli, which are approximately 5% of those from US 6. One would typically expect the modulus of HMA to be greater than that of WMA, and the modulus of a highvolume mix to be greater than that of a low-volume mix. However, the in situ HMA tests from US 6 had a lower average modulus than the WMA tests. As will be shown in the next section, when all in situ moduli at different temperatures are corrected to a common

841

S. Lin et al. / Construction and Building Materials 112 (2016) 836–847

Average Vs (m/s)

(a)

2,000 1,500

4

2.0E+04 1.5E+04

4

6

6 2

6

6 6

1,000 500

(b) Average E (GPa)

Fig. 5. MASW surface-wave test results on WMA for location US 6 W30-2: (a) normalized time-domain data for hot test, (b) experimental dispersion trend for hot test, (c) normalized time-domain data for ambient-temperature test, (d) experimental dispersion trend for ambient-temperature test.

2

2

0

1.0E+04

6

2

6 6

5.0E+03

2

2

0.0E+00

Fig. 6. Comparison of surface-wave tests on eight test sections: (a) average shear-wave velocity, (b) average dynamic modulus. Average testing temperatures indicated in horizontal axis labels. Standard error and number of data points are shown at the top of each bar.

reference temperature, the corrected moduli follow the expected trends with mix type. Therefore, a correction method is developed in the next section using laboratory dynamic modulus measurements on the field cores to account for the effect of temperature on the in situ measured modulus.

4. Correction of in situ moduli at different temperatures to a common reference temperature 4.1. Laboratory visco-elastic modulus measurement using indirect tension (IDT) testing method To assess the accuracy of the modulus values obtained from the in situ nondestructive surface-wave tests, the visco-elastic dynamic moduli of the field cores identified in Table 1 were measured by the indirect tensile test method [36,37]. Specimens were tested at six frequencies: 25, 10, 5, 1, 0.5, and 0.1 Hz, and three temperatures: 4, 21, and either 37 or 32 °C. The decreased temperature of 32 °C was used for samples that exhibited permanent deformations above the limit recommended in the proposed stan-

dard [37]. Before performing IDT dynamic modulus tests on core samples from each test section, tuning of the testing system was performed to minimize the loading error and attain the target strain range recommended by the proposed standard. The average moduli of the cores from each test section were used to construct master curves with a reference temperature of 21 °C (Fig. 7). The overall rank of the master curves in Fig. 7 in decreasing order of modulus is as follows: US 6-HMA, US 6-WMA, Boone HMA base, Boone WMA base, US 69, and IA 93-OL. For the in situ moduli from SWM testing shown in Fig. 6b, however, US 6 WMA and IA 93 OL have higher ranks than in Fig. 7, because the field temperature significantly affected the measured in situ modulus. Therefore, when using surface wave methods for QC/QA of asphalt pavements, it is necessary to account for the effect of pavement temperature on in situ modulus. 4.2. Field Modulus Correction Procedure A procedure is proposed herein to correct the in situ modulus measured at any given field temperature to the modulus at a reference temperature (e.g., 21 °C), using master curves from laboratory

842

S. Lin et al. / Construction and Building Materials 112 (2016) 836–847

(a)

(a) 1,00,000 |E*|, MPa

US 6 HMA US 6 WMA Boone HMA base Boone WMA base US 69 IA 93 OL

1,000

100 0.001

0.1

10

10

|E*|, MPa

(b)

1000

Frequency, Hz Fig. 7. Dynamic visco-elastic modulus master curves from field cores for reference temperature of 21 °C: (a) experimental master curves, (b) fitted master curves.

dynamic modulus tests. First, three master curves are obtained from laboratory dynamic modulus tests performed at three reference temperatures. For example, the three fitted master curves from the US 69 cores tested at 4, 21, and 32 °C are shown in Fig. 8a. A new master curve corresponding to any given field-test temperature within this temperature range can then be interpolated based on the three measured master curves, as described below. Five different methods were examined for interpolating between the three measured reference master curves to obtain a new master curve for a given field-test temperature. The methods include linear interpolation of modulus (E), linear interpolation of log10(E), quadratic interpolation of log10(E), quadratic interpolation of the master-curve fitting coefficients, and the standard quadratic fit of the temperature shift factor (e.g. [39,40]). To identify the most accurate interpolation method, an actual laboratory master curve at 12.5 °C was measured using the core from a field test on US 69 at the same temperature (solid line in Fig. 8a), and then compared against the five interpolated curves (symbols in Fig. 8a). The relative error between the measured and interpolated master curves (Fig. 8b) reveals that (1) quadratic interpolation of log10(E) results in the lowest error over the entire frequency range (5.0% average error compared with 28%, 8.7%, 61%, and 5.4% using the other methods); and (2) linear interpolation of log10(E) results in the lowest error over the high-frequency range of interest above 0.1 Hz, in which the design frequency typically falls (0.9% average error compared with 9.2%, 2.3%, 29%, and 3.6% using the other methods). Therefore, linear interpolation of log10(E) can be employed to interpolate between the measured master curves to obtain a master curve at the field temperature (TF) for each test using Eq. (3),

  T F  T iþ1 log10 ETF ðf Þ ¼ log10 ET iþ1 ðf Þ þ ðlog10 ET i ðf Þ  log10 ET iþ1 ðf ÞÞ T i  T iþ1

ð3Þ

ETF ðf Þ ¼

ET i ðf Þ

ðT F T iþ1 Þ=ðT i T iþ1 Þ

ET iþ1 ðf Þ

ðT F T i Þ=ðT i T iþ1 Þ

1.E+01

US 69-4deg-fit US 69-21deg-fit Linear interpolation of E Quadratic interpolation of log10E Quadratic fit of log of temperature shift factor

US 6 HMA fit US 6 WMA fit Boone HMA base fit Boone WMA base fit US 69 fit IA 93 OL fit

which can be written as

1.E-01

1.E+03

1.E+05

Frequency, Hz

10,000

0.1

1.E-03

1000

(b)

100 0.001

1,000

100 1.E-05

Frequency, Hz

1,000

10,000

(E_interpolated-E_actual)/ E_actual*100%

|E*|, MPa

10,000

180 160 140 120 100 80 60 40 20 0 -20 1.E-05

US 69-12.5deg-fit US 69-32deg-fit Linear interpolation of log10E Quadratic interpolation of coefficients

Linear interpolation of E Linear interpolation of log10E Quadratic interpolation of log10E Quadratic interpolation of coefficients Quadratic fit of log of temperature shift factor

1.E-02

1.E+01

1.E+04

1.E+07

Frequency, Hz Fig. 8. Determination of optimum master curve interpolation method: (a) measured master curves at 4, 12.5, 21, and 32 °C (solid lines), and master curves at 12.5 °C by five interpolation methods (symbols); (b) relative error between interpolated and measured master curves at 12.5 °C.

where i = 1 when TF < T2, i = 2 when TF > T2, f is frequency, T is temperature, and E is modulus. The interpolated field-temperature master curve can then be used to find the reduced frequency corresponding to the measured in situ modulus, as illustrated in Fig. 9. Finally, the temperature-corrected modulus at the same reduced frequency can be found for any of the three master curves at the reference temperatures. To avoid extrapolation error, the maximum and minimum reference temperatures should encompass the range of field temperatures. Using this procedure, the in situ moduli measured under various field temperatures can all be corrected to the same reference temperature. This procedure enables more meaningful comparisons of in situ moduli to moduli from different temperatures, from different field test sites, or from laboratory tests. This procedure was employed to correct the in situ moduli from six of the field test sections in Table 1 to the moduli at a common reference temperature of 21 °C. The results are shown in Fig. 10 and Table 2. Upon correcting the field moduli to the common reference temperature of 21 °C, the rankings of the corrected field moduli in Table 2 were found to agree with those of the laboratory dynamic modulus master curves in Fig. 7, illustrating the importance of correcting the field NDE modulus values for temperature. 5. Quality control and quality assurance procedure A procedure for quality control and quality assurance of asphalt pavements based on a quantitative mechanical property (modulus) from rapid in situ NDT measurements is presented in this section. The procedure is somewhat similar to the basic concept proposed in [25], but employs field density measurements, MASW tests, and the field modulus temperature-correction procedure introduced in the previous section. The in situ modulus is calculated using Eq. (1), and depends strongly upon shear-wave velocity, which is

S. Lin et al. / Construction and Building Materials 112 (2016) 836–847

T1 In-situ modulus ETF

T2

Modulus

Corrected modulus

T3 TF In-situ modulus Corrected modulus

Frequency

Reduced frequency, fr

Fig. 9. Procedure for converting an in situ modulus at a measured field temperature TF to a corrected modulus at a reference temperature T2.

obtained from the MASW tests and is quite sensitive to temperature. The modulus also depends less strongly on density and Poisson’s ratio. Density is much less variable with temperature and is measured by an EM gauge such as the PT, whereas Poisson’s ratio varies over a small range and can be assumed using typical values for pavement. For implementation of the QC/QA procedures into practice, it would be useful for state departments of transportation to develop master curve databases, to correct the in situ moduli measured at various field temperatures to the moduli at a common reference temperature. As described in the following sections, the variation of temperature-corrected in situ modulus measured across a given project site can then be used for quality control. Additionally, the dimensionless ratio of the achieved field modulus to the maximum modulus can be used as a quality assurance metric.

843

5.1. Primary factors affecting asphalt pavement quality Asphalt pavement density has been widely employed as a quantitative indicator of quality, measured either in situ by EM or nuclear gauges [2–5] or using cores in the laboratory by Saturated Surface Dry (SSD) or CoreLok test methods [7]. Pavement quality is conventionally assumed to increase as density increases (or air voids decrease). However, as can be seen by comparing Figs. 1 and 2 to Fig. 6, density is not as sensitive as modulus for distinguishing between quality differences among different pavement types. Furthermore, density fails to capture the temperature dependence of asphalt pavement mechanistic properties such as modulus. In contrast, shear-wave velocity is very sensitive to pavement type as well as variations in temperature, as demonstrated above. Therefore the small strain seismic (elastic) modulus of Eq. (1), being a combination of density and shear-wave velocity, is a sensitive and objective measure of pavement quality. Moreover, the measured in situ modulus can be directly used to estimate the fatigue damage of an asphalt layer [38], by enabling calculation of tensile strain at the bottom of the layer [9]. 5.2. Correlations between density and shear-wave velocity The field and laboratory data from Sections 3 and 4 are used in this section to study the main factors affecting an accurate estimation of in situ asphalt pavement modulus. As described in Section 3.3, shear-wave velocity data presented in this section are for SWM tests on cold and ambient-temperature pavements, and in situ densities are from EM gauge tests on hot pavements. Additionally, the shear wave velocities were determined from the highfrequency horizontal asymptotes of the dispersion trends for the

Fig. 10. Correction of in situ moduli to a reference temperature of 21 °C: (a) Boone HMA base (TF = 28.4 °C, T3 = 32 °C), (b) Boone WMA base (TF = 33.2 °C, T3 = 32 °C), (c) US 6 HMA (TF = 26.6 °C, T3 = 37 °C), (d) US 6 WMA (TF = 22.4 °C, T3 = 37 °C), (e) US 69 (TF = 17.9 °C, T3 = 32 °C), (f) IA 93 OL (TF = 12.9 °C, T3 = 32 °C).

844

S. Lin et al. / Construction and Building Materials 112 (2016) 836–847

Table 2 Rankings of field moduli before and after temperature correction. After temperature correction to 21 °C

Before temperature correction Project/pavement type

Avg. field temp. (°C)

Avg. field modulus (GPa)

Project/pavement type

Corrected field modulus (GPa)

US 6 WMA US 6 HMA Boone HMA base IA 93 OL Boone WMA base US 69

22.4 26.6 28.4 12.9 33.2 17.9

18.5 18.3 10.1 9.62 9.43 7.58

US 6 HMA US 6 WMA Boone HMA Base Boone WMA Base US 69 IA 93 OL

19.5 18.6 15.0 14.0 6.27 6.08

Shear-wave velocity (m/s)

1 2 3 4 5 6

(a) 2,000 1,500

y = 1.3246x - 1960.3 R² = 0.233

1,000 500 0 1500

2000

2500

3000

Shear-wave velocity (m/s)

Rank

(b) 2,000 1,500 1,000 500 0 1500

US 6 HMA

Boone HMA base

2000

2500

3000

Core density (kg/m3)

In situ hot density (kg/m3) US 6 WMA

y = 2.5321x - 4511.5 R² = 0.8926

Boone WMA base

US 69

IA 93 OL

IA 93 FDR

IA 93 CIP

Fig. 11. Correlation between average in situ shear-wave velocity and average density for seven test sections: (a) in situ EM density from hot pavements, (b) laboratory CoreLok density.

20,000

y=x R2 = 0.99

ECL (MPa)

15,000

US 6 WMA US 6 HMA Boone HMA base Boone WMA base US 69 IA 93 OL IA 93 FDR IA 93 CIP

10,000

5,000

0 0

5,000

10,000

15,000

20,000

EEM (MPa) Fig. 12. Correlation between in situ modulus calculated using in situ EM density (EEM) vs. laboratory CoreLok density (ECL) for 8 different pavement test sections.

MASW tests on US 6, and from the multilayer inversion analyses for the MSOR tests at the other project sites. Shear-wave velocity was found to have a very low correlation to in situ density, with an R2 of 0.233 (Fig. 11a), but a high correlation to laboratory CoreLok density, with an R2 of 0.893 (Fig. 11b). If the laboratory density is considered to be objective and accurate, then the low correlation between the in situ and laboratory densities in Fig. 3 indicates that the accuracy of the EM measurement is low, whereas the high correlation between shear-wave velocity and laboratory density in Fig. 11b indicates that the accuracy of the surface-wave measurement is relatively high. Despite the low correlation shown in Fig. 3, the density is still needed for calculation of the in situ modulus by Eq. (1). Furthermore, for the modulus to be a viable QC measure, the density should be obtainable by rapid NDT techniques rather than slower laboratory methods. To assess the error incurred by using the in situ density rather than laboratory density to calculate the in situ modulus, both densities were used in Eq. (1) along with the shear-wave velocity from SWM tests. The correlation between the resulting in situ moduli calculated using the EM density (EEM) versus the CoreLok density

IA 93 CIP IA 93 FDR IA 93 OL US 69 Boone WMA base Boone HMA base US 6 HMA US 6 WMA 0%

5%

10%

15%

20%

|EEM - ECL| / ECL (%) Fig. 13. Relative difference between moduli EEM and ECL calculated with in situ EM gauge vs. laboratory CoreLok densities.

(ECL) is very good, with an R2 of 0.99 (Fig. 12). The EM density therefore causes negligible error when used to calculate the in situ modulus. This is because Vs, with its exponent of 2 and high sensitivity to temperature and pavement type, has a significantly greater influence on modulus in Eq. (1) than does density, which varies over a limited range. The relative difference between EEM and ECL is small; below 10% for the denser pavements, and no more than 20% for the lower density CIP and FDR test sections (Fig. 13). Thus, the in situ modulus EEM calculated from rapid NDT EM density measurements and SWM tests appears to be an acceptable measure for QC/QA of asphalt pavements. 5.3. Calculation of corrected modulus based on in situ modulus and master curve database As demonstrated in Section 4, asphalt pavement modulus is very sensitive to temperature, and a poor-quality pavement tested in situ at a low temperature might therefore have a higher modulus than a good-quality pavement tested at a higher temperature. For meaningful comparisons across different testing locations, project sites, and pavement types, the in situ moduli measured at different field temperatures using SWM and EM tests should be

845

S. Lin et al. / Construction and Building Materials 112 (2016) 836–847 Table 3 Parameters and maximum moduli for fitted 21 °C master curves of Fig. 7b.

Quality rating Ec21(QA) /Emax21

d

a

d+a

Emax (GPa)

US 6 HMA US 6 WMA Boone HMA Boone WMA US 69 IA 93 OL

1.62 1.50 0.87 1.82 1.58 1.20

2.72 2.78 3.55 2.47 2.68 3.12

4.34 4.28 4.42 4.29 4.26 4.32

21.9 19.2 26.7 19.5 18.0 20.5

4.32

21.0

Average

Modulus

Project/pavement type

T1 T2 T3 TF In-situ modulus EC21(QA) Emax21

Very high

>80%

High

60−80%

Moderate

40−60%

Low

20−40%

Very low

<20%

Frequency

ETF -ET21 (MPa)

4,000

Fig. 15. Quality assurance procedure based on NDT measurements.

2,000 Boone HMA Boone WMA US 6 HMA US 6 WMA US 69 IA 93 OL

0 y = -420x R2 = 0.90

-2,000 -4,000 -6,000 -10

-5

0

5

10

Table 4 Proposed quality ratings for QC/QA of asphalt pavements based on modulus ratio. EC21/Emax21 (%)

>80%

60–80%

40–60%

20–40%

<20%

Quality rating

Very high

High

Moderate

Low

Very low

15

TF -21 (°C) Fig. 14. Correlation between modulus difference and temperature difference of field and reference values across six test sections (average values used for each test section).

corrected to a common reference temperature using the correction procedure detailed in Section 4.2 and shown schematically in Fig. 9. For general implementation, however, it may not be practical to measure the required master curves in the laboratory on each new project. Alternatively, a database of master curves can be generated from which a set of three master curves at three reference temperatures (e.g., T1 = 4 °C, T2 = 21 °C, T3 = 37 °C) can be selected according to the asphalt pavement type. The commonly used master curve equation [39–41] is

logðE Þ ¼ d þ

a

1 þ ebc log f r

ð4Þ

where d is the common logarithm of the minimum modulus, d + a is the common logarithm of the maximum modulus, b and c are the shape coefficients, and fr is the reduced frequency. For example, the values of d, a, d + a, and the maximum modulus (Emax), which would be cataloged for different pavement mixes in such a database, are listed in Table 3 for each of the fitted master curves in Fig. 7b. Upon selecting three reference master curves from the database for the appropriate asphalt mix used on the project, linear interpolation of log10(E) can employed as described in Section 4.2 to obtain master curves corresponding to the field temperature (TF) of each individual NDT test location. For each test, the reduced frequency of the in situ modulus can then be found on the interpolated field-temperature master curve, and the corrected modulus having the same reduced frequency on the 21 °C reference master curve can be found. 5.4. Quality control procedure For quality control, a quicker procedure is needed to convert the in situ measured density and shear-wave velocity to a quality metric such as the temperature-corrected modulus for comparison to a quantitative design-related value such as the maximum modulus. One possible method is to employ a correlation between the modulus difference (field modulus minus corrected field modulus at 21 °C in Table 2) and the temperature difference (field temperature minus 21 °C), as shown in Fig. 14 for the average values from six different test sections. Regression of this data gives an estimate of the corrected in situ modulus at 21 °C as

EC21 ¼ ETF þ 420ðT F  21Þ

ð5Þ

where modulus is in MPa, temperature is in °C, TF is the in situ temperature, and ETF is the in situ modulus measured at temperature TF. A quick QC procedure can thus be summarized in four steps: (1) measure in situ moduli and temperatures of several randomly selected locations in one construction section; (2) correct the average in situ modulus to account for temperature effects using Eq. (5) or a similar regression from local construction projects; (3) find the maximum modulus (Emax) of the constructed pavement section from a master-curve database based on the pavement type and reference temperature of 21 °C; and (4) calculate a dimensionless index as the ratio of corrected modulus to maximum modulus (EC21(QC)/Emax  100%) to quantify the achieved quality. In step 2, the average modulus can be replaced with the individual moduli to assess variability across a given project site for QC purposes. For example, any test location having a corrected modulus more than two standard deviations below the average value could be flagged as a potential problem area. 5.5. Quality assurance procedure After the pavement has been constructed as guided by the QC procedure, a more precise QA procedure is needed to convert in situ measurements to quality ratings using a master curve database. The following QA procedure is proposed, which is shown schematically in Fig. 15: (1) Measure the field density and shear-wave velocity using the NDT techniques described herein. (2) Substitute the measured density and shear-wave velocity into Eq. (1) to calculate the in situ modulus (ETF) at the field temperature TF. (3) Construct a set of three master curves at three reference temperatures (T1, T2, T3) according to the asphalt pavement type from a master-curve database. (4) Determine the maximum modulus (Emax21) for the 21 °C reference-temperature master curve from the database. (5) Employ linear interpolation of log10(E) to obtain a master curve corresponding to the measured field temperature TF. (6) Determine the reduced frequency (fr) of the in situ modulus on the interpolated master curve for the temperature TF. (7) Use the reduced frequency to find the corrected modulus EC21(QA) on the 21 °C reference-temperature master curve.

846

S. Lin et al. / Construction and Building Materials 112 (2016) 836–847

Table 5 QC/QA assessment for six pavement test sections in this study. Project/pavement type

US 6 HMA US 6 WMA Boone HMA Boone WMA US 69 IA 93 OL

QC

QA

EC21(QC) (GPa)

EC21(QC)/Emax (%)

Quality rating

EC21(QA) (GPa)

EC21(QA)/Emax (%)

Quality rating

20.7 19.1 13.2 14.6 6.28 6.21

94.5% 99.6% 49.3% 74.9% 34.9% 30.2%

Very high Very high Moderate High Low Low

19.5 18.6 15.0 14.0 6.27 6.08

89.3% 97.0% 56.2% 71.7% 34.9% 29.6%

Very high Very high Moderate High Low Low

(8) Calculate the ratio of corrected modulus to maximum modulus to quantify the achieved quality factor, Q:

Q ¼ EC21ðQAÞ =Emax21  100%

ð6Þ

A quality rating system is proposed in Table 4 to classify the resulting achieved quality factor. This rating system could also be used for the QC modulus ratio described in the previous section. It should be noted that moderate or low quality ratings might be acceptable for low-volume roads. 5.6. QC/QA procedures applied to the asphalt projects in this study The QC/QA procedures described above were applied to the six pavement sections in this study for which laboratory dynamic modulus tests were performed. For the QC procedure, corrected moduli were estimated by substituting the field moduli and temperatures from Table 2 into Eq. (5), while the corrected moduli for QA were obtained by interpolation of master curves as previously reported in Table 2. The QC/QA modulus ratios and quality ratings defined above were then determined for all six test sections using the maximum moduli from Table 3. The results are reported in Table 5. In both the QC and QA assessments, US 6 WMA has the highest quality factors of 99.6% and 97.0%, whereas IA 93 OL has the lowest quality factors of 30.2% and 29.6%. One should note that the quality ratings in Table 5 do not take into account the design traffic volumes of each project. Although IA 93 OL has the lowest quality rating among the pavements in Table 5, it has the smallest traffic volume, as shown in Table 1. If the traffic volume is taken into account, the relative quality rating for IA 93 OL might be deemed acceptable. 6. Conclusions Objective nondestructive quality assessment procedures were developed in this study, using an in situ dynamic modulus obtained by two efficient and economical NDT methods for measuring pavement density and surface-wave phase velocity. The in situ density can be measured by non-nuclear electromagnetic devices, while the shear-wave velocity can be obtained by inversion of MSOR dispersion data, or by simply multiplying the highfrequency horizontal asymptote of MASW phase velocity by a constant related to Poisson’s ratio. The in situ shear-wave velocity was demonstrated to be very sensitive to pavement type and temperature, whereas the EM density measurements were much less sensitive. The EM density had a low correlation with laboratory density, but the differences in density were demonstrated to have little effect on the calculated in situ modulus. The shear-wave velocity and therefore in situ modulus is very sensitive to temperature variation, whereas the in situ density is not. To account for effect of temperature on modulus, a procedure was presented to correct the in situ moduli measured at different field temperatures to a common reference temperature for quality assessments.

For the data in this study and the design frequency range of interest (>0.1 Hz), linear interpolation of log10(E) was found to have the greatest accuracy for obtaining the master curve at a given field temperature using data from measured master curves at three different reference temperatures. A quick QC procedure was developed in which the modulus is corrected using a fitting function from experimental data and compared to the maximum modulus for a given pavement type. A more precise QA procedure was developed by correcting the field modulus using the interpolated field-temperature master curve from a database, and comparing the corrected modulus to the maximum modulus. Further studies are needed to validate the proposed QC/QA procedures before they can be implemented in practice. Acknowledgements Support for this work was provided by the Iowa Highway Research Board through the project ‘‘Assessment of NonDestructive Testing Technologies for Quality Control/Quality Assurance of Asphalt Mixtures – TR-653” (Addendum 446). This support is gratefully acknowledged. Any opinions, findings, conclusions, or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the sponsoring organization. References [1] ARA Inc., and ERES Consultants Division, Guide For Mechanistic-Empirical Design of New and Rehabilitated Pavement Structures. Final Report, NCHRP Project 1-37A, Transportation Research Board of the National Academies, Washington, D.C., 2004. [2] D.A. Larsen, J.W. Henault, Quantifying Segregation in HMA Pavements Using Non-nuclear Density Devices: Data Collection Report for Connecticut. Report, Connecticut Department of Transportation, Newington, CT, 2006. [3] S.G. Williams, Non-nuclear Methods for HMA Density Measurement. Report, Mack-Blackwell Transportation Center and Arkansas State Highway and Transportation Department, Fayetteville, AR, 2008. [4] M.E. Mason, R.C. Williams, Investigation of Electromagnetic Gauges for Determination of In-place Density of HMA Pavements. Report, Iowa Department of Transportation, Ames, IA, 2009. [5] R.C. Williams, C. Can, T. Ahmed, H. Lee, Quality Control/Quality Assurance Testing for Joint Density and Segregation of Asphalt Mixtures. Report, Iowa Department of Transportation, Ames, IA, 2013. [6] S.G. Williams, Bulk specific gravity measurements of 25.0-mm and 37.5-mm coarse-graded superpave mixes, Trans. Res. Rec. 2007 (2001) 110–117. [7] M. Buchanan, T. White, Hot mix asphalt mix design evaluation using the Corelok vacuum-sealing device, J. Mater. Civ. Eng. 17 (2005) 137–142. [8] M. Celaya, S. Nazarian, M. Zea, V. Tandon, Use of NDT Equipment for Construction Quality Control of Hot Mix Asphalt Pavements. Report, Arizona Department of Transportation, Phoenix, AZ, 2006. [9] F. Finn, C. Saraf, R. Kulkarni, K. Nair, W. Smith, A. Abdullah, The use of distress prediction subsystems for the design of pavement structures, in: Paper Presented at the 4th International Conference on the Structural Design of Asphalt Pavements, 1977. Ann Arbor, MI. [10] Y. Li, S. Nazarian, Evaluation of aging of hot-mix asphalt using wave propagation techniques, in: Gerald A. Huber, Dale S. Decker (Eds.), Paper presented at Engineering Properties of Asphalt Mixtures and the Relationship to Their Performance, American Society for Testing and Materials, ASTM STP 1265, Philadelphia, 1995. [11] S. Nazarian, V. Tandon, D. Yuan, Mechanistic quality management of hot mix asphalt layers with seismic methods, J. ASTM Int. 2 (2005) 1–12.

S. Lin et al. / Construction and Building Materials 112 (2016) 836–847 [12] M. Celaya, S. Nazarian, Seismic testing to determine quality of hot-mix asphalt, Trans. Res. Rec. 1946 (2006) 113–122. [13] Z. Jiang, Innovative Nondestructive Testing (NDT) for Condition Assessment of Longitudinal Joints in Asphalt Pavements Master’s Thesis, University of Waterloo, Waterloo, Ontario, Canada, 2007. [14] C.L. Barnes, J.F. Trottier, Evaluating in-service asphalt concrete damage using surface waves, Int. J. Pavement Eng. 11 (2010) 449–458. [15] C.L. Barnes, J.F. Trottier, Evaluating laboratory-induced asphalt concrete moisture damage using surface waves, Int. J. Pavement Eng. 11 (2010) 489– 497. [16] P. Icenogle, M.S. Kabir, Evaluation of Non-destructive Technologies for Construction Quality Control of HMA and PCC Pavements in Louisiana. Report, Louisiana Department of Transportation, Baton Rouge, LA, 2013. [17] S. Nazarian, In-situ Determination of Elastic Moduli of Soil Deposits and Pavement Systems by Spectral-analysis-of-surface-waves method PhD Dissertation, University of Texas, Austin, 1984. [18] K.H.I.I. Stokoe, G.W. Wright, A.B. James, M.R. Jose, Characterization of geotechnical sites by SASW method, in: R.D. Woods (Ed.), Geophysical Characterization of Sites, Oxford Pub., 1994, pp. 15–25. [19] C.B. Park, J. Ivanov, R.D. Miller, J. Xia, N. Ryden, Seismic investigation of pavements by MASW method-Geophone approach, Paper Presented at SAGEEP 2001, 2001. Denver, CO. [20] N. Ryden, Surface Wave Testing of Pavements PhD Dissertation, Lund University, Lund, Sweden, 2004. [21] S. Lin, J.C. Ashlock, A study on issues relating to testing of soils and pavements by surface wave methods, in: Paper Presented at the 38th Annual Review of Progress in Quantitative Nondestructive Evaluation (QNDE 2011), 2011. Burlington, VT. [22] S. Lin, J.C. Ashlock, Comparison of MASW and MSOR for surface wave testing of pavements, J. Environ. Eng. Geophys. 20 (4) (2015) 277–285, http://dx.doi.org/ 10.2113/JEEG20.4.277. [23] A. Saeed, J.W. Hall, Comparison of non-destructive testing devices to determine in-situ properties of asphalt concrete pavement layers, in: Paper Presented at the 82nd Transportation Research Board (TRB) Annual Meeting, 2003. Washington, DC. [24] X. Bai, Assessment of Relationship Between Dynamic and Seismic Moduli of Asphalt Concrete Mixtures Master’s Thesis, University of Texas, El Paso, 2004. [25] I. Abdallah, A. Meshkani, D. Yuan, S. Nazarian, Design Modulus Values Using Seismic Moduli. Report, Texas Department of Transportation, Austin, TX, 2005. [26] A. Tertre, G. Cascante, S.L. Tighe, Combining portable falling weight deflectometer and surface wave measurements for evaluation of longitudinal joints in asphalt pavements, Trans. Res. Rec. 2152 (2010) 28–36.

847

[27] N. Ryden, M.J.S. Lowe, P. Cawley, C.B. Park, Evaluation of multilayered pavement structures from measurements of surface waves, in: Paper Presented at the 32nd QNDE, 2006. Portland, OR. [28] N. Ryden, C.B. Park, Surface waves in inversely dispersive media, Near Surf. Geophys. 2 (2004) 187–197. [29] N. Ryden, C.B. Park, P. Ulriksen, R.D. Miller, Multimodal approach to seismic pavement testing, J. Geotech. Geoenviron. Eng. 130 (2004) 636–645. [30] S. Lin, Advancements in Active Surface Wave Methods: Modeling, Testing, and Inversion PhD Dissertation, Iowa State University, 2014. [31] C.B. Park, N. Ryden, R.D. Miller, P. Ulriksen, Time break correction in multichannel simulation with one receiver (MSOR), in: Paper Presented at SAGEEP 2002, 2002. Las Vegas, NV. [32] S. Nazarian, D. Yuan, M.R. Baker, Rapid Determination of Pavement Moduli with Spectral-analysis of Surface Waves Method. Report, Texas Department of Transportation, El Paso, TX, 1995. [33] AASHTO, Standard Method of Test for Bulk Specific Gravity (Gmb) and Density of Compacted Hot Mix Asphalt (HMA) Using Automatic Vacuum Sealing Method, AASHTO T331-13-UL, Washington, DC, 2013. [34] C.B. Park, R.D. Miller, J. Xia, Multichannel analysis of surface waves, Geophysics 64 (1999) 800–808. [35] R.F. Vidale, The Dispersion of Stress Waves in Layered Media Overlaying a Half Space of Lesser Acoustic Rigidity PhD Dissertation, University of Wisconsin, 1964. [36] J.S. Kim, Complex Modulus from Indirect Tension Testing Master’s Thesis, University of Florida, 2002. [37] AASHTO, Proposed standard test method for determining the dynamic modulus for hot-mix asphalt (HMA) using the indirect tension testing method, AASHTO, Washington, DC, 2014. [38] M. Ayres, M.W. Witczak, AYMA: mechanistic probabilistic system to evaluate flexible pavement performance, Trans. Res. Rec. 1629 (1) (1998) 137–148. [39] R. Bonaquist, D.W. Christensen, Practical procedure for developing dynamic modulus master curves for pavement structural design, Trans. Res. Rec. 1929 (1) (2005) 208–217. [40] J. Bari, M.W. Witczak, Development of a new revised version of the Witczak E⁄ predictive model for hot mix asphalt mixtures, J. Assoc. Asphalt Paving Technol. 75 (2006) 381–423. [41] AASHTO Designation PP 62-10, Developing Dynamic Modulus Master Curves for Hot Mix Asphalt (HMA), 2010. Washington, DC.