Incorporation of GPR and FWD into pavement Mechanistic-Empirical design

Construction and Building Materials xxx (2017) xxx–xxx

Contents lists available at ScienceDirect

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

Incorporation of GPR and FWD into pavement Mechanistic-Empirical design Mesbah U. Ahmed ⇑, Rafiqul A. Tarefder Department of Civil Engineering, University of New Mexico, Albuquerque, NM 87131, United States

h i g h l i g h t s
Pavement layer thickness and modulus prediction using GPR and FWD tests.
Input generation for pavement M-E analysis from the nondestructive tests.
Pavement performance evaluation through the M-E analysis.

a r t i c l e

i n f o

a b s t r a c t

Article history: Received 21 November 2016 Received in revised form 22 May 2017 Accepted 18 June 2017 Available online xxxx

A Mechanistic-Empirical (M-E) design based pavement quality or performance prediction methodology incorporating nondestructive test such as Ground Penetrating Radar (GPR) and Falling Weight Deflectometer (FWD) tests is demonstrated in this study. To facilitate, three pavement sites, namely, US285, US54, and I-40 in New Mexico are selected for nondestructive tests such as GPR and FWD. It is observed that the GPR predicted AC layer thickness is more consistent than base layer thickness. The FWD test was also conducted on the same locations, and like GPR, backcalculated layer moduli also shows varying level of inconsistency in different pavement sections. Based on the GPR and FWD tests, each of these three sites is divided into two major areas: one with the mean and the other with the minimum layer thicknesses and modulus. Layer thicknesses as well as moduli are incorporated to the AASHTOWare-ME software to determine load related distresses such as bottom-up crack and rut. Based on the amount of bottom-up cracks, the US54 pavement section is predicted to fail during 15th year of its service life, and the time of failure can be very early in the area with minimum layer thickness and modulus. Based on the rut prediction, pavement sections in US54 and I-40 is predicted to fail early. Finally, this methodology is recommended to implement to perform pavement quality assessment incorporating GPR and FWD tests in both project and network levels. Ó 2017 Published by Elsevier Ltd.

Keywords: GPR FWD Pavement M-E Thickness Modulus AASHTOWare-ME

1. Introduction Nondestructive tests such as Ground Penetrating Radar (GPR) and Falling Weight Deflectometer (FWD) are currently used in pavement evaluation [1–4]. A GPR is mainly used for subsurface evaluation, and the most common application is pavement layer identification and thickness prediction in case of asphalt pavement using electromagnetic signal [5,6]. A FWD, on the other hand, is mainly used for evaluation of structural capacity through applying load on pavement and measured deflections [7,8]. These two tests are combined during the interpretation of layer moduli using the FWD test data and GPR predicted thicknesses [9]. Based on these ⇑ Corresponding author. E-mail addresses: (R.A. Tarefder).

[email protected]

(M.U.

Ahmed),

[email protected]

tests in network and/or project selection level, a pavement condition ranking can be made which is later used in developing Maintenance and Rehabilitation (M&R) strategy [10]. Limitation of this strategy is that this ranking system is developed based on the current pavement condition, and the future condition cannot be predicted. There may be some empirical relationships available to predict the future pavement condition [11]. However, those relationships are only valid for regions with specific materials, climate and so on. During a pavement design, stiffness or modulus is typically determined from volumetric properties of layer materials [12,13]. After the construction, pavement quality is assessed only investigating whether volumetric properties meet the design criteria based on the materials collected field coring and boring. Therefore, it is difficult to assess the as-built material stiffness. In addition to uncertainty in attaining expected stiffness, layer thickness is

http://dx.doi.org/10.1016/j.conbuildmat.2017.06.105 0950-0618/Ó 2017 Published by Elsevier Ltd.

Please cite this article in press as: M.U. Ahmed, R.A. Tarefder, Incorporation of GPR and FWD into pavement Mechanistic-Empirical design, Constr. Build. Mater. (2017), http://dx.doi.org/10.1016/j.conbuildmat.2017.06.105

2

M.U. Ahmed, R.A. Tarefder / Construction and Building Materials xxx (2017) xxx–xxx

another important issue. Layer stiffness and thicknesses over a pavement segment are not consistent [1,6,9]. Due to these issues, a pavement section may show excessive distresses than expected, and thereby, it may fail before the specified service life. Load related pavement distresses, such as cracks and rut or permanent deformation of layers due to repeated traffic load, are typically evaluated manually or automated method at different times over service life [14,15]. Priority and timing of M&R are then setup based on the evaluation at the time of survey and future condition can only be predicted whenever distress data of several years in past are available. However, this type of prediction is still empirical which is suitable for any specific pavement site and/or region. Recently, several researches have been performed to evaluate the structural capacity of asphalt pavement incorporating nondestructive tests such as GPR and FWD. Plati and Loizos [16] demonstrated the use of GPR to predict the overlay thickness required for an asphalt pavement. Hu et al. [17] used Dynamic Cone Penetrometer (DCP) and GPR to assess the in-situ pavement foundation as well as predict layer thicknesses. Later, Smith et al. [18] proposed a methodology to incorporate field damaged asphalt modulus into Mechanistic- Empirical (M-E) analysis of pavement. It is known that the most updated pavement M-E design program requires temperature and frequency varying dynamic modulus of asphalt layer, unbound layer modulus at a stress-state corresponding to wheel load, and as-built layer thicknesses [19]. In addition, analysis incorporating pavement layer thickness and material stiffness at multiple locations is necessary to evaluate the overall pavement performance. The current practice does not meet all the requirements in a single approach. There is a need to develop a methodology that will integrate the use of GPR and FWD to address the earlier mentioned research need. Therefore, this study is motivated to demonstrate a pavement M-E analysis based pavement quality assessment methodology incorporating GPR and FWD tests.

Field Test o Selection of pavement sites o GPR survey o FWD test o Coring & boring

GPR survey 1. Length of segment 2. Antenna frequency

FWD test 1. Length of segment 2. Test interval

1. Number of layers 2. Layer thicknesses

1. Load 2. Deflections

Layer moduli from backcalculation of FWD test data incorporating GPR predicted thicknesses

Determination of material modulus/stiffness such as dynamic modulus and resilient modulus

Perform pavement ME analysis in the AASHTOWare-ME program

Pavement quality/performance assessment based on distress summary

2. Objectives Fig. 1. Outline of the test and analysis.

Main goal of this study is to demonstrate a methodology of pavement M-E analysis incorporating GPR and FWD tests for pavement quality assessment in network or project selection level. Specific objectives are as below:
Conduct nondestructive tests such as GPR and FWD for pavement thickness and stiffness evaluation. Later, these are used to generate inputs for the pavement M-E analysis.
Perform pavement M-E analysis in the AASHTOWare-ME incorporating the inputs generated by the nondestructive tests to predict pavement distresses over its service life. Pavement quality or performance is assessed based on the summary of pavement distresses. 3. Methodology The methodology in this study is outlined in Fig. 1. Prior to field tests, pavement sites were selected. Field tests included both GPR survey for sub-surface and FWD tests for strength evaluation respectively. GPR survey was conducted over a specific length of segment, i.e., 1-mile, on different pavement sections. In addition, signal frequencies were selected based on the requirement, i.e., layer identification and thickness prediction of asphalt and aggregate layers at both shallow and deeper depths. FWD test was also conducted on the same segment of the pavement site right after GPR survey. Layer number and thicknesses were interpreted from the GPR survey data. In addition, as-built thicknesses were measured from coring and boring to use these data as ‘Ground-Truth’

during the GPR data interpretation. Later, pavement surface deflections were collected in response to 9 kip load during FWD testing. Layer stiffnesses, i.e., modulus, were backcalculated from the FWD load and deflections incorporating GPR predicted thicknesses. Once the stiffness or modulus was determined, these were converted to laboratory modulus, such as temperature and frequency varying dynamic modulus for an Asphalt Concrete (AC) layer and resilient modulus for unbound layers, using a set of correlation models for inputs to the pavement M-E analysis. The pavement M-E analysis was performed incorporating the GPR predicted thicknesses and converted laboratory dynamic modulus of AC as well as resilient modulus of unbound layers. Some other major inputs such as traffic and climate data were also collected and used in the analysis. Quality or performance of the selected pavement sections was assessed based on the summary of a number of simulation results such as rutting or permanent deformation and bottom-up crack due to traffic load repetition over a specific year.

4. Description of site A total of three pavement sections in New Mexico were selected for this study. These are US285 near Clines Corner, US54 near Carrizozo, and I-40 near Albuquerque. Fig. 2 shows the cross-section of these pavements. The first two sections have three layers, i.e., AC, base, and subgrade. The base and subgrade of these sections comprise granular aggregate and fine soil respectively. The pavement

Please cite this article in press as: M.U. Ahmed, R.A. Tarefder, Incorporation of GPR and FWD into pavement Mechanistic-Empirical design, Constr. Build. Mater. (2017), http://dx.doi.org/10.1016/j.conbuildmat.2017.06.105

M.U. Ahmed, R.A. Tarefder / Construction and Building Materials xxx (2017) xxx–xxx

OGFC: 0.6 in

OGFC: 0.6 in

OGFC: 0.6 in

AC: 6 in

AC: 6 in

AC: 10.5 in

Base: 6 in

Base: 6 in

Base: 6 in

3

or more- axles single unit truck; (h) Class 8: Four or fewer-axle single-trailer trucks; (i) Class 9: Five-axle single-trailer trucks; (j) Class 10: Six or more-axle single-trailer trucks; (k) Class 11: Five or fewer-axle multi-trailer trucks; (l) Class 12: Six-axle multitrailer trucks; and (m) Class 13: Seven- or more axle multi-trailer trucks. Distributions of traffic classes are also collected from the selected sites and plotted in Fig. 3. In case of US285 and US54, Class 9 traffic is dominant. On instrumented section (I-40), Class 9 is 18% of the total traffic. 5. Field tests

Subbase: 8 in Subgrade

Subgrade

The nondestructive tests such as GPR and FWD were conducted on the three pavement sites. The tests conducted on a selected one mile long segment. In addition to the field tests, coring and boring were conducted so that the layer thicknesses can be used later for velocity calibration of the GPR data.

Subgrade 5.1. Ground Penetrating Radar (GPR)

(a) US285

(b) US54

(c) I-40

Fig. 2. Cross-section of pavement sites.

section on I-40 has four major layers and these are AC, base with a mix of granular aggregate and Recycled Asphalt Pavement (RAP) material, subbase with Process in place and Compacted (PPC) aggregates, and subgrade with compacted fine soil. The AC layers of all these pavement sections are constructed of compacted Hot Mix Asphalt (HMA). In addition, there is a thin layer of Open Graded Friction Course (OGFC) layer on top of the AC layer. Design layer thicknesses of these pavement sections are shown in Fig. 2. Asphalt mix design information was collected during the construction of the pavement sections, i.e., US285, US54, and I-40 respectively, as summarized in Table 1. Granular aggregates and fine soil were collected from the unbound layers for sieve analysis. Particle gradation of the unbound materials is summarized in Table 2. Traffic data are also collected from this specific pavement sites which include Annual Average Daily Traffic (AADT), (%) truck, and growth rate. The AADT of US285, US54, and I-40 are 5978, 4992, and 20712 with the growth rate of 2.5, 4, and 2.6 respectively. The (%) trucks of these sites are 17.43, 40.41, and 49.0 respectively. Volume of traffic is the maximum on the I-40 which is about four times the traffic on the US54. During pavement M-E analysis in AASHTOWare-ME software, it is necessary to input the distribution of different traffic class/vehicle types. There is total of thirteen traffic classes or vehicle types that range from small car to heavy truck [19]. These are: (a) Class 1: Motorcycles; (b) Class 2: Passenger cars; (c) Class 3: Pick-ups and vans; (d) Class 4: Buses; (e) Class 5: Two-axle trucks; (f) Class 6: Three-axle trucks; (g) Class 7: Four Table 1 Volumetric properties of asphalt mix. Parameter

h¼ US285

US54 S

I40

4 4.2 53 1

4 4.4 55 1

4 4.4 55 1

Percentage of aggregate retained in 3/8-inch sieve, q3=8

26

25

25

Percentage of aggregate passing #200 sieve,

5.7

5

5

Binder type, PG

58–28

64–28

70–22

Percentage of Percentage of Percentage of Percentage of sieve, q3=4

air voids, V a effective binder, V beff aggregate retained #4 sieve, q4 aggregate retained in 3/4-inch

q200

Pavement sections in this study have layers with different thicknesses. During a GPR survey, it is important to select antennas with specific frequencies. It is known that a layer with smaller thickness can be identified by a higher frequency signal and vice versa [9]. In addition, depth of penetration is smaller at higher frequency signal since it attenuates rapidly as it moves deeper into a pavement. However, the lower frequency signal can penetrate deeper in the pavement. It indicates that antennas with small to high frequencies are necessary to identify all the layers. The existing GPR system of the NMDOT has both air-coupled and groundcoupled antennas with different central frequencies such as 2.0 GHz, 400 and 900 MHz respectively. The first two central frequencies, i.e., 2 GHz and 900 MHz, have been used to predict the thickness for asphalt layer. To predict the thicknesses of layers underneath an asphalt layer, i.e., base and sub-base layers, 400 and 900 MHz frequencies have been used. Fig. 4(a) shows that two 2.0 GHz air-launched horn antennas were mounted in front of the NMDOT truck with a clear gap of 20 in. (50 cm) between antenna bottom and pavement surface. A transmitted signal travels into pavement and reflects at different layer interfaces. Signal reflection is strong at the interfaces whenever two layer materials have different dielectric property. After reception of reflected signals, a time vs. amplitude plot is generated, and later, layer interfaces are identified based on polarity change of peak amplitudes. The similar type of time vs. amplitude plot for ground-coupled antenna is shown in Fig. 4(b). In this plot, there is no direct coupling since the antenna is placed on the surface. The time vs. signal amplitude plots at different scans (captured by air-coupled and ground-coupled antennas) are stacked together to generate 2D images as shown in Fig. 4(c) and (d). Layer interfaces are identified from this image, and 2-way travel times at these interfaces are determined to calculate layer thicknesses using the following equations [6]:

vt 2 c

v ¼ pffiffiffi e

ð1Þ ð2Þ

where h = depth of layer, t = two-travel time of a signal, v = velocity of signal in a specific media/pavement layer, c = velocity of signal in air, and e = dielectric constant of a media. 5.1.1. Air-coupled antenna AC layer thickness is determined from the image as captured by the 2.0 GHz air-coupled antenna using the velocity calibration

Please cite this article in press as: M.U. Ahmed, R.A. Tarefder, Incorporation of GPR and FWD into pavement Mechanistic-Empirical design, Constr. Build. Mater. (2017), http://dx.doi.org/10.1016/j.conbuildmat.2017.06.105

4

M.U. Ahmed, R.A. Tarefder / Construction and Building Materials xxx (2017) xxx–xxx

Table 2 Summary of particle gradation of unbound layers. Site

#4

#10

#40

#200

Base

US285 US54 I-40

43.5 39.5 42.0

27.5 26.9 30.0

13.1 13.5 15.0

4.8 2.1 4.0

Subgrade

US285 US54 I-40*

97.8 99.3 44.0

97 98.2 25.0

93 93 15.0

55.7 34 3.0

% Passing

Subbase.

(%) Distribution

*

Layer

50 45 40 35 30 25 20 15 10 5 0

US285 US54S I-40

Class 4

Class 5

Class 6

Class 8

Class 9

Class 10

Class 11

Class 12

Class 13

Traffic classification Fig. 3. Distribution of traffic classes.

method using a 0.25 in. thick 4 feet by 4 feet metal plate. The entire procedure is discussed in Ahmed et al. [9]. This velocity calibration is implemented using the following equation:

en

2 3  2 P n2 Ai A0 An1 1  þ c þ i i¼1 Am Am Am 7 6 ¼ en1 4 5  2 P n2 Ai A0 An1 1  Am þ i¼1 ci Am  Am

ð3Þ

where en = dielectric constant of n-th layer, en-1 = dielectric constant of (n  1)-th layer, amplitude of signal reflected from i–th layer interface, and ci = reflection coefficient. The reflection coefficient can be determined as follows:

pffiffiffiffi

pffiffiffiffiffiffiffiffi

e  ei1 ci ¼ pffiffiffiiffi pffiffiffiffiffiffiffiffi ei þ eiþ1

ð4Þ

where ei = dielectric constant of i-th layer, ei  1 = dielectric constant of (i  1)-th layer, and ei + 1 = dielectric constant of (i + 1)-th layer. Once the dielectric constant was determined, layer thickness was determined using the Eqs. (1) and (2). Later, the predicted thicknesses were calibrated/adjusted using the measured thickness. 5.1.2. Ground-coupled antenna The ground-coupled antennas cannot use the above mentioned velocity calibration method, i.e., Eqs. (3) and (4), to determine thickness of unbound layer at deeper depth. To date, coring and boring were conducted on the selected locations to measure the as-built unbound layer thicknesses. Velocity of GPR signal through unbound layer is determined from the known thickness and 2-way travel time using Eq. (1). Later, this velocity is incorporated to determine unbound layer thicknesses at every other survey locations.

quency distributions are plotted to determine the mean and minimum thickness as a input for the pavement ME design. Fig. 5 (a) and (b) show the frequency distributions of AC and base layer thicknesses respectively in US285. It is observed that the frequency distribution of the AC layer is narrower than that of base layer since the AC thicknesses are more consistent over the 1-mile segment. In addition, it is observed that the frequency distribution of the predicted thickness is a normal distribution. In case of a normally distributed dataset, maximum likelihood value as determined from the maximum value of a frequency distribution can be considered as the mean of the dataset [20]. Based on this consideration, the mean AC and base layer thicknesses are 8 and 6 in. respectively. The minimum thickness is determined by subtracting twice the standard deviation of the thicknesses from the mean. The resulting minimum AC and base thicknesses are 6.9 and 3.8 in. respectively The mean and minimum AC and base layer thicknesses of pavement sites are documented in Table 3. Based on the summary, mean thickness of both AC and base layer is slightly above the design thickness. However, the mean thicknesses in US54 are smaller than the design thickness in case of both AC and base layer. There is a thin Open Graded Friction Course (OGFC) with thickness of 0.6 in. (15 mm) on the top of an asphalt layer as shown in Fig. 2. A GPR signal with the 2 GHz frequency cannot distinguish this very thin layer. The entire asphalt layer with this OGFC is identified as a single layer and thereby, thickness of the AC layer is more than the base layer. It is also observed that the GPR predicted AC layer thickness is more consistent than the base layer thickness in all the sections. Reliability of asphalt and base layer thickness is above 90% in all the pavement sites. 5.2. Falling Weight Deflectometer (FWD)

5.1.3. Predicted thickness During the GPR survey, data were collected at 2 scans per foot and 512 samples per scan using a 4-channel GSSI SIR-30 controller. After the data collection, layer thicknesses are determined, and fre-

The FWD test was conducted on the same 1-mile long segment at 250 feet regular interval according to ASTM [21]. Fig. 6(a) through (b) shows a FWD test on pavement where a load is applied

Please cite this article in press as: M.U. Ahmed, R.A. Tarefder, Incorporation of GPR and FWD into pavement Mechanistic-Empirical design, Constr. Build. Mater. (2017), http://dx.doi.org/10.1016/j.conbuildmat.2017.06.105

5

M.U. Ahmed, R.A. Tarefder / Construction and Building Materials xxx (2017) xxx–xxx

de Amplitud -

+

A

0

A

1

A

2

me (ns) Tim

((a) 2.0 GHz aiir-coupled horrn antenna Amplitudde -

+ A

0

A1 A2 Tiime (ns)

(b) 400 & 900 M MHz ground-cooupled antennaa Distannce (feet)

Disstance (feet)

Layer: 1 Layer: 2 Layer: 3

Time (ns)

Time (ns)

D Direct coupling

Pavvement surface

Botttom of AC layer

(c) Air--coupled antennna

(d) Ground--coupled antennna

100

100

80

80

(%) Frequency

(%) Frequency

Fig. 4. GPR survey on pavement sites.

60 40

1 inch = 25.4 mm

60 40 20

20

0

0 3

4

5

6

7

8

9

10

11

12

3

4

5

6

7

8

9

10

AC thickness (inch)

Base thickness (inch)

(a)

(b)

11

12

Fig. 5. Frequency distributions of layer thicknesses (US285).

on surface using a spring-mass system, and in response, surface deflects vertically to form a deflection basin. In a regular setting, seven geophones resting at different radial offsets, i.e., 0, 8, 12,

18, 24, 36, and 60 in. (0, 200, 457, 609, 914, and 1524 mm) from the center of the loading plate, measure these deflections. The test load varies from 6 to 16 kips (26.7–71.2 kN). Distribution of deflec-

Please cite this article in press as: M.U. Ahmed, R.A. Tarefder, Incorporation of GPR and FWD into pavement Mechanistic-Empirical design, Constr. Build. Mater. (2017), http://dx.doi.org/10.1016/j.conbuildmat.2017.06.105

6

M.U. Ahmed, R.A. Tarefder / Construction and Building Materials xxx (2017) xxx–xxx

Table 3 Summary of layer thicknesses. Layer

AC thickness, inch (mm)

US285 US54 I-40

Base thickness, inch (mm)

Mean

Min.

COV (%)

Mean

Min.

COV (%)

8.0 (203) 5.5 (140) 12.0 (305)

6.9 (175) 4.8 (122) 11.3 (287)

6.0 7.4 3.0

6.0 (150) 5.0 (125) 6.0 (150)

3.8 (96) 4.2 (106) 6.3 (160)

14.8 15.0 7.6

Faalling weight Geeophones

h

5 1

2

3

7

6

4

(b)) Deflection baasin

(a) FWD testt 15 Deflection (mil)

D1

1 mil = 25.4 μm

D D7

12 9 6 3 0

Station (mille)

(c) Deflecction profile: S Sensor 1 & 7 Layer modulus (ksi)

1000

100

10

AC C

Base

Subgradee

1 ksi = 6.89 MPa

1 0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9 9

1

Station (m mile)

(d) Backkcalculated lay yer moduli Fig. 6. Analysis of FWD test data (US285).

tions at the geophone 1 and 7 are plotted in Fig. 6(c). It is observed that deflections vary over a wide range which indicates to the variation in pavement properties. Layer moduli (E-values) are determined from the backcalculation of FWD test data. The backcalculation is performed using the Layered Elastic Analysis technique of the ELMOD, a backcalculation software developed by the Dynatest Inc. The basic analysis requires inputs such as layer number and thicknesses as predicted by GPR, load, and deflections at different radial offsets [7]. Fig. 6(d) shows

the distribution of backcalculated AC, base, and subgrade modulus over the test locations in US285. The mean as well as minimum layer moduli are determined from the frequency distribution of the E-values. These values are summarized in Table 4. Pavement sections in the study are mostly newly constructed pavements, and thereby, asphalt layer was not hardened enough to reach its full strength after it was open to traffic. In addition, volumetrics of asphalt mixed are different due to varying target traffics. For this reason, asphalt moduli of US54

Please cite this article in press as: M.U. Ahmed, R.A. Tarefder, Incorporation of GPR and FWD into pavement Mechanistic-Empirical design, Constr. Build. Mater. (2017), http://dx.doi.org/10.1016/j.conbuildmat.2017.06.105

7

M.U. Ahmed, R.A. Tarefder / Construction and Building Materials xxx (2017) xxx–xxx Table 4 Summary of backcalculated layer moduli. Site

EAC, ksi (MPa)

US285 US54 I-40

Ebase, ksi (MPa)

Esubgrade, ksi (MPa)

Mean

Min.

COV (%)

Mean

Min.

COV (%)

Mean

Min.

COV (%)

600 (4136) 650 (4481) 800 (5515)

468 (3226) 510 (3516) 652 (4495)

11.1 8.7 22.1

55 (379) 60 (413) 60 (413)

43 (296) 48 (331) 47 (324)

10.9 9.6 25.4

25 (172) 35 (241) 35 (241)

14 (96) 29 (200) 27 (186)

26.6 36.3 11.5

and US285 are smaller than that of I-40 [22]. Coefficient of Variations (COV, %) of AC, base, and subgrade modulus are also determined and summarized in this Table to observe the level of consistency of modulus. It is observed that backcalculated modulus of all the layers such as AC, base, and subgrade shows inconsistency since the COV (%) ranges from 8.7 to 36.3. It is observed that the AC and base layer moduli of US 285 and US 54 are more consistent than those of I-40. Reliability of the data, i.e., backcalculated layer moduli, varies from 66 to 96%. The reasons behind such variation in the backcalculated moduli are: (a) variation of layer thicknesses during backcalculation; and (b) variation in surface deflections under FWD test load.

frequency (f), and temperature (T) to develop the following correlations [23]:

log10 ðRÞ ¼ 0:8031 þ 0:2418log10 ðf =6:4Þ  0:5984ðT=21:1Þ; 

> 10 C

T

2

R ¼ 0:98

ð6Þ

log10 ðRÞ ¼ 130:3e0:3298ðf =6:4Þ þ 131:2e0:3283ðf =6:4Þ ; T ¼ 10 C

R2 ¼ 0:57

ð7Þ

Pavement M-E analysis using the AASHTOWare-ME requires the temperature and frequency varying dynamic modulus values. Therefore, it is necessary develop a conversion or adjustment factor which can be used determine dynamic modulus values from a single FWD backcalculated AC modulus as below:

The first set of correlation is used to determine E⁄-values at a temperature of 14 °F whereas the second set is used at temperatures above 14 °F. After development of these correlations, the Rfactors are determined over a range of frequency (0.1, 0.5, 1, 10 and 25 Hz) and temperatures (10, 0, 5.4, 21.1, 37.8 and 54.4 °C). Later, E ðf ; TÞ are determined by multiplying the R-factors with the mean backcalculated AC modulus as documented in Table 4. Fig. 7(a) through (c) show the dynamic modulus plots derived from FWD modulus for US285, US54, and I-40 respectively. It is observed that dynamic modulus at high at low temperature and gradually decreases towards high temperature as expected. In addition, it increases as frequency increases. Dynamic modulus in I-40 is the maximum since the mean FWD AC modulus is greater than that of the two other sites. The reliability coefficient, r, is determined based on the dynamic modulus values predicted for the pavement sites in this study. The r-value is 0.71 which indicates this correlation is reliable. Dynamic modulus is also determined based on minimum FWD modulus for each of the pavement sites, and later, these are also incorporated to the pavement ME analysis.

E ðf ; TÞ ¼ EFWD;T¼70 F  Rðf ; TÞ

6.2. Unbound layer

6. Analysis Pavement ME analysis by the AASHTOWare-ME requires traffic, material properties, section geometry, and climate data as major inputs [17]. Traffic data of the selected pavement sites are discussed earlier. Climate data were input to the software based on weather station data collected from the specific regions, and these data are already available in the software. Material inputs are discussed in detail as follows: 6.1. AC layer

ð5Þ



where E ðf ; TÞ ¼ frequency and temperature varying dynamic modulus, EFWD;T¼70 F ¼ backcalculated AC modulus from FWD test at 70 °F, and Rðf ; TÞ ¼ frequency and temperature varying correlation factor. A set of regression equations for FWD-dynamic modulus (E⁄) correlation factor (R) are developed based on asphalt mixes in New Mexico to determine the frequency and temperature varying dynamic modulus from backcalculated AC modulus. During the development of the correlation equations, FWD tests were conducted on several pavement sections all over New Mexico with known mix design information. The E⁄-values are determined from the dynamic modulus prediction model by Tarefder and Rahman [12]. The R-factor is then determined from the ratio of dynamic modulus and FWD modulus at different frequencies and temperatures. Finally, regression analysis is performed on the R-factor,

The unbound layer moduli are backcalculated from the FWD test data. However, these values cannot be directly incorporated to the AASHTOWare-ME since it requires resilient modulus. Resilient modulus of an unbound material is stress-dependent and to date, a set of conversion factors were developed to determine resilient modulus from the FWD unbound layer modulus. These equations are as follows [24]:

Rbase ¼ 1:1

 1:67  2:89 hFWD soctFWD þ1 pa pa

ð8Þ

 1:03  6:54 hFWD soctFWD þ1 Rsubgrade ¼ 3:47 pa pa

ð9Þ

Table 5 Summary of converted resilient modulus values. Project

Converted resilient modulus, ksi (MPa) Base

US285 US54 I-40

Subbase

Subgrade

Mean

Min

Mean

Min

Mean

Min

20.3 (140) 25.3 (174) 31.2 (215)

13.5 (93) 16.1 (111) 18.2 (125)

– – 12.3 (85)

– – 7.4 (51)

16.8 (116) 21.7 (149) 8.8 (60)

8.9 (61) 10.8 (74) 6.8 (47)

Please cite this article in press as: M.U. Ahmed, R.A. Tarefder, Incorporation of GPR and FWD into pavement Mechanistic-Empirical design, Constr. Build. Mater. (2017), http://dx.doi.org/10.1016/j.conbuildmat.2017.06.105

T: 40 F

T: 100 F

T: 130 F

T: 70 F

0.1

1

T: 40 F

T: 100 F

T: 130 F

10

0.1

1

Frequency (Hz)

(b) US54 T: 14 F

T: 40 F

T: 100 F

T: 130 F

T: 70 F

1.E+4

1.E+5

1.E+6

1.E+7

10

Frequency (Hz)

(a) US285

Dynamic modulus (psi)

T: 70 F

1.E+6

T: 14 F

1.E+5 1.E+4

1.E+6 1.E+5 1.E+4

1.E+7

T: 14 F

Dynamic modulus (psi)

1.E+7

M.U. Ahmed, R.A. Tarefder / Construction and Building Materials xxx (2017) xxx–xxx

Dynamic modulus (psi)

8

1 psi = 6.89 kPa 0.1

1

10

Frequency (Hz) (c) Instrumented section (I-40) Fig. 7. Dynamic modulus based on mean backcalculated AC modulus.

30 25

Crack (%)

where Rbase and Rsubgrade ¼ R-factor for base and subgrade respectively, hFWD ¼ bulk stress determined from state of stresses under FWD test, soctFWD ¼ octahedral shear stress determined from state of stresses under FWD test, and pa ¼ atmospheric pressure. The unbound layer moduli are backcalculated from deflections at 9 kip load. The hFWD and soctFWD corresponding to this test load is used for the conversion using Eqs. (8) and (9). The converted resilient modulus of both base and subgrade is summarized in Table 5. These values are later incorporated to the pavement M-E analysis.

US285

US54

I-40

20 15 10 5

7. Results & discussion

0 5

Pavement M-E analysis is performed in the AASHTOWare-ME software incorporating all the major inputs. Pavement distresses such as bottom-up crack due to fatigue and rut are determined from the simulations. The outcomes are discussed below:

10

15

20

Age (year) Fig. 8. Bottom-up crack due to fatigue.

7.1. Bottom-up cracks Fig. 8 shows the amount of bottom-up crack (% of lane area) at 5, 10, 15, and 20 years of pavement service life for US285, US54, and I-40 respectively incorporating the mean stiffness. As a normal trend, crack increases with age due to repeated number of traffic load. Comparing the three sites, US54 shows the maximum amount of cracks which is excessively greater than the two other sites. It is observed earlier that layer material stiffness in this site is close to the US285. However, the AC thickness is smaller than the other two pavement sections. For instance, AC thickness in US285 is 8.0 in. whereas that in US54 is 5.5 in. Due to this smaller thickness, comparatively high tensile strain is developed at the

bottom of the AC layer, and thereby, such a high amount of cracks are predicted in US54. It is also observed that after 15 years of pavement service life, amount of crack in the US54 section reaches 25 which is the general maximum limit of crack set in AASHTOWare-ME. In addition, right after 5th year, 15% of lane area will show bottom-up crack 7.2. Rut Fig. 9(a) through (d) shows the rutting accumulation at 5, 10, 15, and 20 years of pavement service life for US285, US54, and I-40 respectively incorporating the mean stiffness. Based on

Please cite this article in press as: M.U. Ahmed, R.A. Tarefder, Incorporation of GPR and FWD into pavement Mechanistic-Empirical design, Constr. Build. Mater. (2017), http://dx.doi.org/10.1016/j.conbuildmat.2017.06.105

9

M.U. Ahmed, R.A. Tarefder / Construction and Building Materials xxx (2017) xxx–xxx

overall observation, rut increases with age due to traffic load repetition. Fig. 9(a) shows the rut accumulation in AC layer. Amount of rut in US285 is smaller than two other sites. AC layer thickness is the minimum, i.e., 5.5 in (140 mm), in US54. In addition, traffic growth rate is the maximum, i.e., 4%. It leads to very high rut in US54 which is even greater than I-40. During 15th year of service life, rut in AC layer reaches 0.25 in (6 mm) and this is the maximum limit of rut in AASHTOWare-ME. Rut in base is very small compared to that in AC layer (see Fig. 9(b)). Subgrade rut becomes higher due to its smaller stiffness value (see Fig. 9(c)). Ruts in all the layers are summed up to determine total rut. It is observed that total ruts in US54 and I-40 are close and greater than US285. All the tree sites show that the total rut is below the maximum limit of 0.75 in (19 mm). The earlier results are based on the mean stiffness as well as layer thicknesses. It is necessary to investigate the regions with minimum thicknesses and stiffness leads to earlier attainment of maximum limit of crack and/or rut. To further investigate, a comparison is made as follows: 7.3. Bottom-up crack (mean vs. minimum) Fig. 10(a) through (c) show the bottom-up crack for three different pavement sites. It is observed that bottom-up crack occurs more rapidly at the locations with the minimum layer thickness and modulus. In US285, this location shows earlier cracks whereas there is not much difference in I-40. In summary, the same pavement may show very small to high amount of bottom-up cracks based on the variation in layer thicknesses and modulus.

7.4. Rut (mean vs. minimum) Fig. 11(a) through (c) show the rut accumulation in the AC layer considering both mean and minimum layer thicknesses and modulus. In US285, rut in the AC does not increase much at the location of minimum thickness and modulus. However, in US54 and I-40, the maximum AC rut of 0.25 in (6 mm) is attained at 10th year of pavement service life. In case of base layer, rut increases are high in US54 and I-40 (see Fig. 11(d) through (f)). Fig. 11(g) through (i) show the rut accumulation in subgrade. It is observed that rut in this layer also increases in US285. In summary, pavement locations with minimum layer thicknesses and modulus show greater amount of rut accumulation in US 285 and I-40. However, the US285 shows rut increase only in subgrade. 8. Conclusions A methodology is demonstrated in this study to assess pavement quality or performance through pavement M-E analysis incorporating GPR and FWD tests. Based on this study, the following conclusions are made:
The GPR and FWD test were conducted on selected pavement sections and varying level of inconsistency is observed in both GPR predicted layer thicknesses and FWD backcalculated layer moduli. It is also observed that the GPR layer thicknesses are smaller than design thicknesses in case of US54. It indicates that a pavement section may have weaker area with thickness that is smaller than design or recommended thickness and thereby, it may show earlier distress and/or failure.

0.30

0.35

US285

US54

I-40

0.30

Rut in base (inch)

Rut in AC (inch)

0.35

0.25 0.20 0.15 0.10 0.05

US285

10

15

US285

US54

I-40

0.15 0.10

20

5

10

15

Age (year)

Age (year)

(a) AC

(b) Base

US54

20

0.80

I-40 Total rut (inch)

Rut in subgrade (inch)

0.20

0.00 5

0.30

0.25

0.05

0.00

0.35

1 inch = 25.4 mm

0.25 0.20 0.15 0.10

US285

US54

I-40

0.60 0.40 0.20

0.05 0.00

0.00 5

10

15

20

Age (year)

5

10

15

20

Age (year)

(c) Subgrade

(d) Total Fig. 9. Rutting in pavement sections.

Please cite this article in press as: M.U. Ahmed, R.A. Tarefder, Incorporation of GPR and FWD into pavement Mechanistic-Empirical design, Constr. Build. Mater. (2017), http://dx.doi.org/10.1016/j.conbuildmat.2017.06.105

10

M.U. Ahmed, R.A. Tarefder / Construction and Building Materials xxx (2017) xxx–xxx

Mean Min.

60 40 20

80

Crack (%)

80

Crack (%)

Crack (%)

80

60

Mean Min.

40 20

0 10

15

20

40 20

0 5

Mean Min.

60

0 5

10

Age (year) (a) US285

15

20

5

10

15

Age (year)

Age (year)

(b) US54

(c) I-40

20

Mean Min.

0.3 0.2 0.1 0.0

10

15

0.3 0.2

Mean Min.

0.1

20

10

15

20

5

(b) US54

(c) I-40

0.00 10

15

0.08 0.06 0.04

Mean Min.

0.02

20

0.08 0.06 0.04

Mean Min.

0.02 0.00

5

10

15

20

5

10

15

Age (year)

Age (year)

(d) US285

(e) US54

(f) I-40

Rut in subgrade (inch)

Mean Min.

0.25 0.00 10

15

20

0.75 0.50 0.25

Mean Min.

0.00 5

10

15

20

Rut in subgrade (inch)

Age (year)

0.75

20

0.10

Rut in base (inch)

0.10

0.00

5

15

(a) US285

0.02

0.50

10

Age (year)

0.04

5

Mean Min.

0.1

Age (year)

Mean Min.

0.06

0.2

Age (year)

0.10 0.08

0.3

0.0 5

Rut in base (inch)

Rut in base (inch)

0.4

0.0 5

Rut in subgrade (inch)

0.4

Rut in AC (inch)

0.4

Rut in AC (inch)

Rut in AC (inch)

Fig. 10. Bottom-up cracks (Mean vs. Minimum).

20

0.75 1 inch =25.4 mm 0.50 0.25

Mean Min.

0.00 5

10

15

Age (year)

Age (year)

Age (year)

(g) US285

(h) US54

(i) I-40

20

Fig. 11. Rut in pavement layers (Mean vs. Minimum).


AC dynamic modulus and unbound layer resilient modulus are determined from the FWD backcalculated layer moduli which are used as input to the pavement M-E analysis. These moduli are determined based on both mean and minimum FWD modulus so that pavement performance can be predicted in both average and weaker zones.
Bottom-up cracks over pavement service life is predicted, and it is observed that amount of cracks exceeds the maximum threshold at 15th year in US54 due to layer thicknesses smaller than specified. Amount of cracks are enhanced in the region with minimum layer thickness and stiffness, and thereby, earlier failure is predicted.


Rut accumulation in pavement layers over service life is predicted, and it is observed that rut in AC layer of only US285 is below the maximum threshold on areas with both average and minimum layer properties. Total ruts in all the tree sites are below the maximum limit of 0.75 in. (19 mm) at 20th year. However, in the weaker area, total rut in US54 and I-40 exceeds the limit. There is possibility of some error during the thickness and modulus prediction that may affect the final prediction of pavement performance. Due to lack of sufficient data, this effect is not investigated within the current scope of this study. It is recommended

Please cite this article in press as: M.U. Ahmed, R.A. Tarefder, Incorporation of GPR and FWD into pavement Mechanistic-Empirical design, Constr. Build. Mater. (2017), http://dx.doi.org/10.1016/j.conbuildmat.2017.06.105

M.U. Ahmed, R.A. Tarefder / Construction and Building Materials xxx (2017) xxx–xxx

to investigate this effect in the future study. Finally, it is suggested to use this methodology for pavement quality assessment using the GPR and FWD test data. Acknowledgements The financial support by the New Mexico Department of Transportation (NMDOT) is highly acknowledged. The authors would like to thank the initiative and effort by Jeff Mann and Virgil Valdez, NMDOT personnel, to make the arrangements for the field tests on the selected pavement sections. Special thanks are extended to the members of the NMDOT field exploration team. References [1] J. Corley-Lay, C.S. Morrison, Layer Thickness Variability for Flexible Pavements in North Carolina, Transportation Research Record: Journal of Transportation Research Board, 1778, Washington, D.C., 2001, pp. 107–112. [2] Y. Cao, B.B. Guzina, J.F. Labuz, Pavement Evaluation Using Ground Penetrating Radar Final Report, Department of Transportation, Minnesota, 2008. [3] A.S. Noureldin, K. Zhu, S. Li, D. Harris, Network pavement evaluation using falling weight deflectometer and ground penetrating radar, in: Proceedings, TRB 82nd Annual Meeting, Washington, D.C., 2003. [4] Z. Ahmed, K. Helali, A.A. Jumikis, R.A. Khan, Enhancing the pavement management systems database through incorporation of GPR and core data, in: Proceedings, TRB 83rd Annual Meeting, Washington, D.C., 2004. [5] K.R. Maser, Non-Destructive Measurement of Pavement Layer Thickness. Report FHWA/CA/OR-2003/03 prepared for the California Department of Transportation, 2003. [6] D.A. Willet, B. Rister, Ground Penetrating Radar Pavement Layer Thickness Evaluation. Research Report KTC-02-29/FR101-00-1F, Kentucky Transportation Center, 2002. [7] R.A. Tarefder, M.U. Ahmed, Consistency and accuracy of selected FWD backcalculation software for computing layer modulus of airport pavements, Int. J. Geotech. Eng. 7 (1) (2011) 21–35. [8] M.D. Nazzal, L.N. Mohammad, Estimation of Resilient Modulus of Subgrade Soils Using Falling Weight Deflectometer, Transportation Research Record: Journal of Transportation Research Board, No. 2186, Transportation Research Board of the National Academies, Washington, D.C., 2010, pp. 1–10. [9] M.U. Ahmed, R.A. Tarefder, A.K. Maji, Variation of FWD modulus due to incorporation of GPR predicted layer thicknesses, in: Proceedings of the 15th International Conference on Ground Penetrating Radar, IEEE, Brussels, Belgium, June 30–July 4, 2014.

11

[10] D.H. Chen, J. Xie, T. Scullion, Using GPR and FWD to assist in selecting the optimal pavement rehabilitation strategy, Pavement. Mater. ASCE (2011) 63– 70. [11] R. Haas, W.R. Hudson, J. Zaniewski, Modern Pavement Management, Krieger Publishing Company, Florida, 1994. [12] R.A. Tarefder A.S.M. Rahman Enhanced Statewide and Independent Assurance Test for Dynamic Modulus of NMDOT Superpave Mixes for the Implementation of Mechanistic Empirical Pavement Design Guide (MEPDG), Final Report, Research Bureau, New Mexico Department of Transportation, 2016, pp. 1–188. [13] H.H. Titi, M.B. Elias, S. Helwany, Determination of Typical Resilient Modulus Values for Selected Soils in Wisconsin Grant No. WHRP 0092-03-11, Wisconsin Department of Transportation, Wisconsin, 2006. [14] K. Montoya, Improvements on manual pavement distress data collection to conform to state and federal requirements MSc Thesis, University of New Mexico, New Mexico, 2010. [15] K.C.P. Wang, O. Smadi, Automated Imaging Technologies for Pavement Distress Surveys Transportation Research Circular, E-C156, Pavement Monitoring and Evaluation Committee, 2011. [16] C. Plati, A. Loizos, Using ground-penetrating radar for assessing the structural needs of asphalt pavements, Nondestr. Test. Eval., Taylor & Francis 27 (3) (2012) 273–284. [17] J. Hu, P.K.R. Pavana, D.J. White, I. Beresnev, Pavement thickness and stabilised foundation layer assessment using ground-coupled GPR, Nondestr. Test. Eval., Taylor & Francis 31 (3) (2016) 267–287. [18] K.D. Smith, J.E. Burinsma, M.J. Wade, K. Chatti, J.M. Vandenbossche, H.T. Yu, Using Falling Weight Deflectometer Data with Mechanistic-Empirical Design and Analysis. Volume 1: Final Report, Federal Highway Administration, Washington, DC, 2017. [19] AASTOWare-ME, AASHTOWareÒ Pavement ME DesignTM v.2.0. Software Manual by AASHTO, Help Version 2.0.1, 2014. [20] A. John, R. A. Fisher and the making of maximum likelihood 1912–1922, Stat. Sci. 12 (3) (1997) 162–176. [21] ASTM D 4694-96, Standard Test Method for Deflections with a Falling WeightType Impulse Load Device, Road and Paving Materials: Paving Management Technology, Annual Book of ASTM Standards, vol. 4, No. 3, 1996. [22] R.A. Tarefder, M.M. Hasan, Advanced Statewide Calibration of MEPDG for NMDOT Final Report, New Mexico Department of Transportation, Albuquerque, New Mexico, 2017. [23] M.U. Ahmed, R.A. Tarefder, Optimal Use of Falling Weight Deflectometer and Ground Penetrating Radar Report No. NM12SP-01, New Mexico Department of Transportation, 2016. [24] M.U. Ahmed, M.M. Hasan, R.A. Tarefder, Investigating stress dependency of unbound layers using falling weight deflectometer and resilient modulus tests, ASTM Geotech. Test. J. 39 (6) (2016).

Please cite this article in press as: M.U. Ahmed, R.A. Tarefder, Incorporation of GPR and FWD into pavement Mechanistic-Empirical design, Constr. Build. Mater. (2017), http://dx.doi.org/10.1016/j.conbuildmat.2017.06.105