Naz Zal 2007

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Evaluating the Light Falling Weight

Deflectometer Device for In Situ


Measurement of Elastic Modulus
of Pavement Layers
Munir D. Nazzal, Murad Y. Abu-Farsakh, Khalid Alshibli,
and Louay Mohammad

Field and laboratory testing programs were conducted to evaluate the The limitations of some testing devices encouraged the develop-
potential use of the light falling weight deflectometer (LFWD) device for ment of new devices that can easily measure the in situ elastic modu-
measuring the in situ elastic modulus of pavement layers and subgrades. lus of highway materials and yet must be robust and accessible to
The field tests were conducted on several highway sections selected from different construction sites. One of the newly developed devices is the
different projects in Louisiana. In addition, nine test sections were con- light falling weight deflectometer (LFWD). It is a portable FWD that
structed and tested at the Pavement Research Facility site of Louisiana was developed in Germany as an alternative in situ device to the PLT.
Transportation Research Center. All sections were tested using the Prima Different types of LFWD are available on the market, including the
100 model-LFWD in conjunction with the falling weight deflectometer German dynamic plate (GDP), the Transport Research Laboratory
(FWD), plate load test (PLT), and dynamic cone penetrometer (DCP) tests (prototype) foundation tester (TFT), and the Prima 100 LFWD.
that were used as reference measurements. Linear regression analyses Although they exhibit many similarities in their mechanics of opera-
were carried out on the collected test data to develop models that could tion, they have differences in design and mode of operations, which
directly relate the LFWD stiffness modulus with moduli obtained from have led to variations in the measured results. Generally, the LFWD
FWD and PLT and the DCP penetration rate. In addition, multiple non- consists of a loading device that produces a defined load pulse, a load-
linear regression analyses were conducted to develop models that could ing plate, and at least one center geophone sensor to measure the
predict FWD and PLT moduli on the basis of the LFWD elastic moduli and deflection of the center of the plate surface (δc). The LFWD elastic
selected soil properties (moisture content and void ratio) of the tested mate- modulus (ELFWD) is then calculated from the applied load pulse and δc.
rials. The results showed that the FWD, PLT moduli, and DCP-penetra- Several studies had been conducted in the last 10 years to evaluate
tion rate could be predicted directly with LFWD at a significant confidence and assess the LFWD measurements. These studies used the PLT and
level. However, the prediction models were improved when the soil prop- FWD as benchmarks for comparison. Fleming compared the GDP
erties were included as variables. Laboratory tests also were conducted to with FWD (1). The results demonstrated that the correlation ratio
determine the influence depth of the LFWD, and the results of these tests between the GDP and the FWD moduli is approximately 0.5. How-
showed that the LFWD influence depth ranged from 270 to 280 mm. ever, Fleming (1, 2) reported that extensive field-stiffness measure-
ments on in situ construction sites showed a relatively consistent
correlation of 0.6 between the elastic moduli of the GDP and FWD.
Recent developments in pavement design make it essential to improve Livneh and Goldberg (3) suggested that the GDP elastic moduli are
the quality control–quality assurance procedures based on a stiffness– approximately 0.3 to 0.4 times the conventional FWD moduli.
strength criterion. The falling weight deflectometer (FWD) and plate Fleming et al. (4) conducted field tests to correlate the moduli of three
load test (PLT) have been used for many years to evaluate the stiff- main types of LFWD (TFT, GDP, Prima 100) with that of the FWD.
ness of pavement layers and subgrades. However, these devices have Their results showed that the FWD resilient modulus (MFWD) corre-
some limitations when they are used in the assessment of pavement lated well with moduli obtained from the Prima 100 with a correlation
layers during construction. The FWD faces accessibility problems for coefficient of 1.031 (i.e., MFWD = 1.031 ELFWD). However, the correla-
roads under construction. Conversely, the PLT is a relatively slow test tion coefficients with other LFWD types were as follows: MFWD = 1.05
that can take a few hours and usually needs a loaded truck or frame to to 2.22 EGDP, and MFWD = 0.76 to 1.32 ETFT.
jack against, which can be an obstacle for using it on roads. Kamiura et al. (5) studied the relationship between the LFWD
(Prima 100) and the PLT measurements for subgrade materials that
M. D. Nazzal, K. Alshibli, and L. Mohammad, Department of Civil and Environmen-
include volcanic soil, silty sand, and mechanically stabilized crushed
tal Engineering, and M. Y. Abu-Farsakh, Louisiana Transportation Research Center, stone. A linear relation was obtained between log(kLFWD /k30) and
Louisiana State University, 4101 Gourrier Avenue, Baton Rouge, LA 70808. log(kLFWD), where kLFWD is the ratio of the stress on the loading plate
Corresponding author: M. Y. Abu-Farsakh, mabuf@ltrc.lsu.edu. of the LFWD to the measured deflection at this stress, and k30 is the
ratio of stress on a 300-mm diameter PLT to the measured deflection
Transportation Research Record: Journal of the Transportation Research Board,
No. 2016, Transportation Research Board of the National Academies, Washington, at this stress. Kamiura et al. (5) also indicated that the ratio of kLFWD /k30
D.C., 2007, pp. 13–22. is affected by the grain size of the tested material, such that this ratio
DOI: 10.3141/2016-02 increases with increasing grain size.

13
14 Transportation Research Record 2016

The objective of this paper is to evaluate the potential use of the ing any test operation, the center deflection (δc) of the loading plate is
LFWD device to reliably measure the in situ elastic modulus of measured and used to calculate the LFWD elastic modulus (ELFWD).
different pavement layers and subgrades. This work is part of a The expression used to calculate ELFWD is similar to the one used
comprehensive study that was conducted at the Louisiana Trans- to calculate the surface modulus of a layered media assuming a uni-
portation Research Center (LTRC) to assess the use of different in form Poisson’s ratio (ν) and constant loading on an elastic half space
situ devices in the construction control of pavement layers and embank- (Boussineq elastic half space). This expression is described by
ments (6). For this purpose, field tests were conducted on several Equation 1:
highway sections selected from different projects in Louisiana. In
addition, six test sections and three trench sections were constructed 2 (1 − ν2 ) σ × R
and tested at the LTRC–Pavement Research Facility (PRF) site. The ELFWD = (1)
δc
field-testing program included conducting LFWD tests in compan-
ion with other standard tests such as the PLT, FWD, and DCP tests.
A comprehensive statistical analysis was conducted on the collected In Equation 1, σ is the applied stress, and R is the radius of the loading
plate. A Poisson’s ratio of 0.35 was used for soils tested in this study.
data. Laboratory experiments also were conducted to evaluate the
influence depth of the LFWD device.
Falling Weight Deflectometer
TEST DEVICES
The FWD is a nondestructive field test that has been widely used in the
Light Falling Weight Deflectometer evaluation of pavement layers. The FWD test implies applying an
impulse load by dropping a weight from a certain height to a 305-mm-
The Prima 100 model LFWD was used in this study (Figure 1). It was diameter circular loading plate that is in contact with the surface of the
developed and marketed by Carl Bro Pavement Consultants (previ- test section. Different load magnitudes can be generated by varying the
ously known as Phφnix) in Denmark. It weighs 26 kg and has a 10 kg mass of weight and drop height. A 40-kN load was used in this study.
falling weight that impacts a spring to produce a load pulse of 15 to A load cell is used to measure the applied load during the test. The sur-
20 ms. For safe operation, the drop weight is supported with a trans- face deflections are measured using seven geophone sensors located at
portation lock pin and guide rod with stabilizer. The Prima 100 has a various distances from the center of the loading plate. These geophones
load range of 1 to 15 kN (i.e., up to 450 kPa with its 200 mm diameter are positioned at 0, 305, 457, 610, 914, 1219, and 1524 mm away from
loading plate). It measures both the applied force and center deflection the center of the loading plate, as shown in Figure 2. These geophones
using a velocity transducer with a deflection range of 2.2 mm (4). Dur- enable the FWD to determine the response of pavement layers.

10 kg falling
weight

Geophone

FIGURE 1 Prima 100 LFWD.


Nazzal, Abu-Farsakh, Alshibli, and Mohammad 15

TOW VEHICLE
DROP ELECTRONIC
HEIGHT CONTROL
RAISE/LOWER
SENSORS BOX BATTERY BAR

DROP
WEIGHTS

DEFLECTION
SENSORS

DEFLECTION BASIN

FIGURE 2 FWD used in study.

The analysis performed on any FWD data is aimed at determining The load is usually transmitted to the plate by a hydraulic jack acting
the resilient modulus of each layer in a pavement section and the depth against heavy mobile equipment. In this study, the PLT test was per-
of the underlying bedrock layer. The FWD resilient modulus, MFWD, formed according to ASTM D1196 standard. The resulting load-
is determined through a backcalculation process that is performed deflection curve from PLT can be used to determine the elastic
using computer programs such as MODULUS, ELMOD, and EVER- modulus of the tested layer using the following equation:
CALC. In this study, the ELMOD 4.0 program was used to determine
the MFWD. The Dynatest model 8000 FWD system was used in this σ 2 × P (1 − ν )
2

study as a reference test to evaluate the LFWD measurements. EPLT = = (2)


 πRδ

where
Plate Load Test
EPLT = elastic modulus,
The PLT (Figure 3a) has been used as a useful site investigation tool  = strain,
for many years. Furthermore, it has been used for proof testing of P = applied load,
pavement layers in many European countries. Currently, it is used for δ = deflection of the plate.
both rigid and flexible pavements. The test consists of loading a cir-
cular plate, usually 305 mm in diameter, that is in contact with the sur- Different elastic moduli can be obtained from the PLT. In this study,
face and measuring deflections at the different loading increments. the initial elastic modulus (EPLT(i)) and the reloading modulus (EPLT(R ))
2

ε1 ε2
Stress

R2)
i)
EPLT(

LT(
EP

(a) Strain
(b)

FIGURE 3 PLT: (a) device and (b) definition of E PLT(i) and E PLT ( R 2 ) .
16 Transportation Research Record 2016

were determined for all PLTs. The initial modulus (EPLT(i)) was deter- TEST RESULTS
mined from the initial slope of the load–displacement curve. The
reloading modulus (EPLT(R )) was determined using the deflection
2 The variation of the ELFWD with the number of passes measured dur-
under the second loading cycle of the plate and the applied load by ing the construction of some test sections at the PRF site is shown in
the end of that cycle. Figure 3b describes the deflections and Figure 5. It is clear that for these test sections, the ELFWD increased with
stresses used for determining the EPLT(i) and EPLT(R ) from the load
2 the increase in the number of passes and thus with the increase of the
displacement curve. dry density. Figure 6 presents the variation of ELFWD with time for the
cement-treated, lime-treated, and blended calcium sulfate (BCS) base
sections that gain strength with time. It can be seen that for the cement
Dynamic Cone Penetrometer soil section (1), the ELFWD values increased with time until 33 days; after
that, they stabilized. However, for the other sections, the LFWD in-
The DCP was initially developed in South Africa for in situ evalua- creased with the time until reaching a maximum value between 3 to
tion of pavements (7). Since then, it has been used in South Africa, 7 days; after that, they either stabilized or decreased slightly. The
the United Kingdom, Australia, New Zealand, and the United States.
strength of these soils was expected to continue increasing with time
In this study, the DCP test was used as a standard test for compari-
because of the chemical reactions that usually occur in the cement- and
son with the LFWD measurements. The DCP test is a simple, fast,
lime-treated soils, mainly the pozzolanic reactions and dehydrations.
and economical test that can provide continuous measurements of the
However, the two cement-treated sections were covered. Thus, their
in situ strength and stiffness of pavement layers and subgrades. The
DCP test is conducted by dropping a weight (8 kg) from 575 mm curing was retarded, which affected the strength gain with time for these
height and recording the number of blows versus depth. The penetra- sections. In addition, the lack of curing resulted in minor shrinkage
tion rate (PR) in mm/blow is then calculated. The DCP-PR has been cracks, which might have affected the results of the LFWD tests. Con-
correlated to many engineering properties such as the California versely, the lime-treated soil and BCS base sections were not covered
Bearing Ratio, shear strength of granular materials, the subgrade and were influenced by the rain that occurred during the testing period.
resilient modulus (MR), elastic modulus (Es), and soil classification. For the lime-treated section, the LFWD measurements taken during the
rainy period (days 12 to 25) were reduced. The results of LFWD on
BCS base section showed that this material has a high elastic modulus
FIELD-TESTING PROGRAM and can be a strong supportive base layer for pavements. However, this
material is sensitive to moisture content. Consequently, the LFWD
The field-testing program included conducting a series of tests on
measurements on BCS also were influenced by the rain that occurred
selected highway sections from different construction projects
during the testing period. It should be noted that the purpose of this
throughout Louisiana. In addition, six base and subgrade sections and
three trench sections were constructed and tested at the LTRC-PRF study was to evaluate the LFWD, not to evaluate the materials, and all
site (Figure 4a). Table 1 presents a summary of the tested sections and tests were conducted under the same conditions. The error bars of ELFWD
the properties of materials used in these sections. In each section, at in Figure 6 suggest that the variability of the test measurements did not
least five LFWD measurements were first acquired. One PLT, one have much influence on the trend of elastic modulus with time. The
FWD, and two DCP tests were then conducted in each test section and variations of LFWD modulus with dry unit weight for the three trench
used as reference measurements in the analysis. The dry unit weight sections constructed at the PRF site are presented in Figure 7, which
and moisture content also were measured with the nuclear density shows that both the dry unit weight and LFWD modulus increased with
gauge device. Figure 4b describes the layout of the field tests. For the increase of the compaction effort. These results suggest that there is
further information on field testing, refer to Abu-Farsakh et al. (6). a general trend between the elastic moduli and the dry unit weight of

300 mm

LFWD
DCP

PLT&
FWD
D
LFWD LFWD LFWD

DCP D: Nuclear density

LFWD

300 mm
(a) (b)

FIGURE 4 Field-testing program: (a) testing of sections at the PRF site and (b) layout for field testing.
Nazzal, Abu-Farsakh, Alshibli, and Mohammad 17

TABLE 1 Properties of Materials Tested at Different Sections

Soil Classification
Tested
Section Material Stations AASHTO USCS Gs γd (kN/m3) m.c. (%)

Highway US-190 Crushed limestone base 12 + 530 A1 GW-GC 2.56 18.9 7.8
12 + 650 2.56 18.9 8.9
15 + 800 2.56 — —
Highway LA-182 Lime-treated subgrade 47 + 10 A6 CL 2.71 15.9 21.2
Cement-treated base 173 + 68 — — 2.71 17.8 10.7
Cement-treated subbase 319 + 00 A4 CL-ML 2.73 16.4 12.9
319 + 10 2.73 16.3 12.8
Lime-treated subbase 503 + 90 A4 CL 2.71 17.0 9.5
504 + 00 2.71 17.5 10.2
504 + 10 2.71 — —
Highway US-61 Clayey subbase Pass 1 A6 CL-ML 2.71 — —
Pass 4 2.71 16.0 15.6
PRF base and Clayey silt — A4 CL-ML 2.71 16.5 16.6
subgrade test Cement soil–1 — A4 CL-ML 2.73 15.7 16.2
sections Cement soil–2 — A4 CL-ML 2.73 15.2 13.5
Lime-treated soil — A4 CL-ML 2.71 15.2 16.2
Crushed limestone — A-1-a GM-GC 2.56 19.2 7.5
Blended calcium sulfate — — — 2.96 14.9 30.6
PRF trench Crushed limestone Light A1-a GP-GM 2.56 18.9 4.9
test sections Moderate 2.56 19.1 5.2
Heavy 2.56 21.1 5.6
RAP Light A-1-b SP 2.56 15.8 11.9
Moderate 2.56 16.9 11.4
Heavy 2.56 18.0 11.6
Sand Light A-1-a GP 2.65 16.1 3.3
Moderate 2.65 17.2 2.9
Heavy 2.65 17.3 2.7

RAP = recycled asphalt pavement; Gs = specific gravity; and m.c. = moisture content.

the material, at the same moisture content. However, this relation de- was evaluated using the coefficient of variation, Cv, of the five mea-
pends primarily on the type of the tested material and its behavior. surements taken at each test section. Figure 8 presents the variation of
the Cv with its corresponding average LFWD elastic moduli. It can be
observed that the Cv of the LFWD measurements ranged from 2.1%
Repeatability of LFWD Measurements to 28.1%. The general trend for the points in this figure indicates that
the Cv values decreased as the LFWD elastic moduli increased. This
The reliability of the LFWD measurement is significantly influenced was also seen during field testing, where it was difficult to conduct
by its repeatability. In this study, the repeatability of the LFWD device LFWD tests on weak subgrades. The reason for this is that very weak

110
Crushed limestone (second layer)
100 Cement soil (2)
400
Cement soil (1)
Clayey silt (third layer)
90 350 Cement soil (2)

80 300 Lime-treated soil


ELFWD (MPa)

70 BCS base
ELFWD (MPa)

250
60
200
50
150
40
100
30

20 50

10 0
1 2 3 4 5 6 7 0 3 6 9 12 15 18 21 24 27 30 33 36 39
Number of passes Days after construction

FIGURE 5 E LFWD with number of passes. FIGURE 6 E LFWD with time.


18 Transportation Research Record 2016

160 compacted with a dynamic compactor. The cylindrical mold was


Sand
then removed, which resulted in a cylindrical cavity of a 300-mm
140 Crushed limestone diameter surrounded by the soil to be tested. This procedure was
RAP done to simulate similar soil boundary condition effects. To evaluate
120 the influence depth, the tested material was placed in the cylindrical
mold and compacted in 75-mm-thick layers up to a 300-mm thick-
100
ELFWD (MPa)

ness. Each layer was well compacted using a standard hammer. On


completion of the compaction of each layer, three LFWD measure-
80 ments were taken at the center of the compacted soil. To clearly
define the influence zone for the LFWD, stiff soil was constructed
60
on top of soft soil and vise versa. For example, BCS was placed on
top of the softer clay layer, whereas clay soil was placed separately
40
on top of the stiff BCS layer.
20 Figures 9a and 9b present the average ELFWD values plotted against
the layer thickness. In Figure 9a, clay soil on top of BCS layer, the
0 ELFWD values decreased with increasing thickness until reaching an
15 16 17 18 19 20 21 22 asymptote at approximately 280 mm thickness. However, for BCS
Dry unit weight (kN/m3) on top of clay soil in Figure 9b, the ELFWD values increased with
increasing thickness and gradually stabilized after a thickness of
FIGURE 7 E LFWD and dry unit weight for different 270 mm.
trenches.
These results suggest that the influence depth for the LFWD device
ranges between 270 and 280 mm, which is approximately 1.5 times
the diameter of the loading plate. This depth is less than the influ-
subgrade usually undergoes plastic permanent deformation during the
ence depth for the PLT, which is approximately two times the diam-
LFWD test, which causes extra soil densification. Conversely, the
eter of the plate. This finding agrees with previous results reported
LFWD performance was enhanced for well-compacted and stiffer
by Fleming (2), which showed that quicker load application reduces
pavement layers. Similar findings were reported by Fleming (2). The
the influence depth of the stress.
results of his study suggested that field tests conducted with LFWD
Because in some cases the influence depth of the PLT and LFWD
and FWD had a greater variation on weak subgrade materials compared
exceeded the thickness of the tested pavement layer reaching to the
with those conducted on stiffer subbase and base course materials.
underlying layers and subgrade, the modulus obtained directly from
these tests represented the composite modulus of the layers within the
Influence Depth of LFWD influence depth rather than the true modulus of the tested layer. To
overcome this situation, a multilayered system solution had to be used
The influence depth of the LFWD device was investigated by conduct- to backcalculate the modulus of the tested layer. The Odemark (8)
ing laboratory tests inside two test boxes (1,824 mm long, 912 mm method, known as the method of equivalent thickness, was used in
wide, and 912 mm deep) located at the geotechnical research labora- this study to backcalculate the layer’s modulus on multilayer systems.
tory of the LTRC. To prepare for testing, a cylindrical mold 305 mm In this method, layers of different stiffness were the first transformed
in diameter and 305 mm tall was placed at the center of each box. to an equivalent layer of the same stiffness, such that Boussinesq’s
Then a 300-mm-thick layer of soil was placed around this mold and equations for homogeneous elastic half-space media could be used to
predict stresses and deflections. For sections with unknown subgrade
moduli (E1), the subbase or base modulus (E2) was backcalculated
32 assuming that the E1/E2 ratio for these test sections was the same as
the E1/E2 ratio for the FWD.
28

24 Regression Analysis

20 Comprehensive statistical analyses were carried out on the collected


field measurements to develop regression models between the LFWD
CV%

16 elastic modulus (ELFWD) and the measurements obtained by the stan-


dard tests (FWD, PLT, and DCP). The ELFWD values were correlated
12 with the FWD backcalculated resilient modulus (MFWD), initial and
reloading PLT moduli (EPLT(i) and EPLT(R )), and DCP-PR. SAS was
2

8
used to perform the analyses.
The variation in the moisture content and dry unit weight of the
4
tested materials in the different sections required the inclusion of
0 variables that can describe these properties in the developed regres-
0 50 100 150 200 250 300 350 400 sion models. Therefore, two sets of the model were investigated for
ELFWD (MPa) each test device: a direct LFWD model and an LFWD soil property
model. To develop the LFWD direct model, linear regression analy-
FIGURE 8 C v variation with LFWD modulus. ses were conducted to directly relate the ELFWD modulus to those
Nazzal, Abu-Farsakh, Alshibli, and Mohammad 19

110

100 H BCS layer


100
Clay layer
90
90 H Clay layer
BCS layer
80
ELFWD (MPa)

ELFWD (MPa)
80

70 70

60 60

50 50

40 40

60 90 120 150 180 210 240 270 300 330 40 80 120 160 200 240 280 320 360
Thickness (H) (mm) Thickness (H) (mm)
(a) (b)

FIGURE 9 E LFWD versus thickness for (a) clay and (b) BCS.

obtained using the standard test devices, whereas multiple non- and the FWD backcalculated resilient modulus (MFWD), the following
linear regression analyses were conducted to develop the LFWD soil regression model was obtained:
properties models.
The development of the LFWD soil properties models required the MFWD = 0.964 ELFWD (5)
selection of parameters that could best describe the effects of soil’s
dry unit weight and moisture content. On the basis of error optimiza- The coefficient of determination (R2) of this model was 0.94,
tion procedure of several model forms, it was concluded that the best whereas the standard error [the square root of the mean square errors
LFWD soil properties models should follow the general form shown (RMSE)] was 23.18. This result suggests that the LFWD and FWD
in Equation 3. In this analysis, the void ratio was selected to describe yield close modulus values. This model is similar to the one proposed
the dry unit weight, in which its determination requires the use of by Fleming (4) based on the results of several field tests conducted only
soil’s specific gravity (Equation 4). Thus, the LFWD soil properties on different subgrade soils:
models will account not only for the dry unit weight and moisture
content but also indirectly for the soil type as well. MFWD = 1.031 ELFW (6)
β1
Y = α × ELFWD [ wc ] eβ3 β2
(3) The results of the FWD–LFWD correlation are depicted in Fig-
ure 10a. Figure 10a also presents a comparison between the proposed
where regression model and the one suggested by Fleming (4). It can be
Y = elastic modulus obtained from the standard tests, seen that the model proposed by Fleming (4) lies within the 95%
wc = soil water content measured in the field, prediction interval of the model proposed in this study. Thus, the
α = regression constant, and proposed model is compatible with the Fleming (4) model. Equation 5
e = void ratio determined using the following soil volumetric indicates that good correlation can be attained between the moduli
relationship: obtained from both FWD and LFWD. It is often assumed that test
parameters determined from different test backgrounds with differ-
γ wGs ent applied stresses and strains may not provide reasonable correla-
e= −1 (4)
γd tions with each another. However, this is not always the case. For
example, earlier studies (1–3) showed acceptable correlations between
where the FWD and the LFWD test results. The soil elastic modulus is less
influenced by stress history. In fact, this parameter is controlled
γw = unit weight of water (9.81 kN/m3); by the same soil variables, which may have led to the development
γd = dry unit weight of soil measured in the field; and of better correlations between them. The strain history influence
Gs = soil specific gravity. also is expected to be insignificant in nondestructive repeated load
devices. Furthermore, the FWD and the LFWD tests were conducted
FWD–LFWD Direct Model on soil under identical conditions. In conclusion, the statistical
correlation presented in this paper (Equation 5) is empirical and
Based on the linear regression analysis conducted on the collected based on testing a wide range of soils that has a good coefficient of
field test data to relate the LFWD elastic modulus obtained (ELFWD) determination.
20 Transportation Research Record 2016

240 240
MFWD= 0.964*ELFWD
220 220 MFWD= 6.398*(ELFWD)0.732(e)-0.086(wc)-0.334
95% Prediction Interval
R2 = 0.985
200 Fleming et al. (4) 200
180 180

Predicted EPLT (R2) (MPa)


160 160
MFWD (MPa)

140 140
120 120
100 100
80 80
60 60
40 40
20 20
0 0
0 20 40 60 80 100 120 140 160 180 200 0 20 40 60 80 100 120 140 160 180 200
ELFWD (MPa) Measured MFWD (MPa)
(a) (b)

FIGURE 10 M FWD -E LFWD : (a) direct model and (b) soil properties model.

FWD–LFWD Soil Properties Model PLT–LFWD Soil Properties Model

The nonlinear multiple regression analyses that was conducted on Nonlinear multiple regression analyses were conducted on the field
the collected field test measurements to determine the parameters test data to find the best PLT–LFWD soil properties models. The
of the model described in Equation 4, yielded the following model: results yielded the models shown in the Equations 10 and 11:

M FWD = 6.398 × ( ELFWD ) (e )–0.086 ( wc )–0.334 EPLT( i ) = 1.03 × ( ELFWD ) (e )–0.524 ( wc )–0.0887
0.732 1.031
(7) (10)

with the R2 value of 0.986 and RMSE value of 5.89. The LFWD soil EPLT( R2) = 1.22 × ( ELFWD )1.064 ( wc ) – 0.195 ( e ) – 0.021 (11)
properties model has a higher R2 and much smaller RMSE values
compared with the direct LFWD model, which demonstrates that it
Equations 10 and 11 have coefficient of determination R2 values of
has a better prediction of the MFWD values. Figure 10b compares the
0.985 and 0.986 and RMSE values of 5.89 and 9.1, respectively. The
measured MFWD to those predicted by the model according to Equa-
PLT–LFWD soil properties models have slightly higher R2 and much
tion 7. It is clear that the data are less scattered about the model line
smaller RMSE values compared with the PLT–LFWD direct models.
than in Figure 10a.
This suggests that these models better fit the field data than the direct
models. Figure 12a and 12b compare the measured PLT elastic moduli
values to those predicted by the models described in Equations 10 and
PLT–LFWD Direct Model
11. Again, one can notice that the data points are less scattered about
Statistical regression analyses were conducted to develop direct cor- the model line when compared with those in Figures 11a and 11b.
relations between the ELFWD and both the PLT initial and reloading
moduli (EPLT(i) and EPLT(R )). The following linear regression model
2

was obtained between ELFWD and EPLT(i): Correlation Between LFWD and DCP

EPLT( i ) = 1.041 × ELFWD (8) A regression model was also developed between the LFWD elastic
modulus (ELFWD) and the DCP penetration rate (DCP–PR in mm/blow)
as follows:
with R2 and RMSE values equal 0.92 and 13.82, respectively. The
following regression model was obtained between ELFWD and EPLT(R2): 5301.54
ELFWD = (12)
8.31 + PR1.44
EPLT R2 = 0.875 ( ELFWD ) (9)
( )
The model has R2 and RMSE values of 0.87 and 20, respectively. The
In Equation 9, R2 equals 0.97 and RMSE equals 13.4. The results of result of this regression analysis is shown in Figure 13.
regression models are illustrated in Figures 11a and 11b. The figures Several correlations also were proposed by some researchers to
show that the regression models fit the data well. evaluate the elastic modulus of the tested material from the DCP–PR
Nazzal, Abu-Farsakh, Alshibli, and Mohammad 21

220 220
EPLT(i)= 1.04*(ELFWD) R2 = 0.97 EPLT(R2)= 0.875*(ELFWD) R2 = 0.97
200 95% Prediction Interval 200 95% Prediction Interval

180 180

160 160

140 140
EPLT(i) (MPa)

EPLT(i) (MPa)
120 120

100 100

80 80

60 60

40 40

20 20

0 0
0 20 40 60 80 100 120 140 160 180 200 220 0 20 40 60 80 100 120 140 160 180 200 220
ELFWD (MPa) ELFWD (MPa)
(a) (b)

FIGURE 11 Result of regression models: (a) E PLT(i) -E LFWD direct model and (b) E PLT(R2) versus E LFWD .

(e.g., 9, 10). For example, on the basis of a regression analysis on field log ( Es ) = 3.652 − 1.17log ( PR ) (14)
test measurements, Chen and Bilyeu (9) obtained the following rela-
tionship between the FWD backcalculated resilient modulus (MFWD)
Figure 13 presents a comparison between the different ELFWD–PR
and the DCP–PR (in mm/blow):
relations and the measured ELFWD values. It can be seen that the
proposed ELFWD–PR model (Equation 12) can predict the measured
MFWD = 338 ( PR )
–0.39
(13) data better than the other models. It is worth mentioning that the
models in Equations 13 and 14 were proposed only for subgrade
A different relationship was proposed by Pen (10) between the material; however, different pavement materials were investi-
subgrade’s elastic modulus (Es) (in MPa) and PR (in mm/blow) as gated in this study, which include the base, subbase, and subgrade
defined in Equation 14. materials.

200 200
EPLT(i)= 1.03*(ELFWD)1.031(e)-0.524 (wc)-0.0887
180 EPLT(i)= 1.22*(ELFWD)1.064(wc)-0.195(e)-0.207
R2 = 0.985 180
R2 = 0.986
160 160
Predicted EPLT(i) (MPa)

Predicted EPLT(R 2) (MPa)

140 140

120 120

100 100

80 80

60 60

40 40

20 20

0 0
0 20 40 60 80 100 120 140 160 180 200 0 20 40 60 80 100 120 140 160 180
Measured EPLT(i) (MPa) Measured EPLT(i) (MPa)
(a) (b)

FIGURE 12 Soil properties models: (a) E PLT(i) -E LFWD and (b) E PLT ( R 2 ) -E LFWD .
22 Transportation Research Record 2016

200 EPLT(R2) prediction models were improved by including the soil’s void
1.44 ratio and moisture content.
180 3939.9/[PR + 8.31]
• Good correlation also was obtained between ELFWD and the
Chen et al. (1999)
160 Pen (1990)
DCP–PR with R2 = 0.87.
• The LFWD is easier, faster, and less expensive than the other ref-
140 erence tests. In addition, the results of this study demonstrated that the
LFWD can be used to measure the in situ elastic moduli of the differ-
ELFWD (MPa)

120
ent pavement layers. However, more geophone sensors should be
100 added to make it more reliable and applicable to multilayer pavement
systems.
80

60
ACKNOWLEDGMENTS
40
This research was funded by the Louisiana Transportation Research
20 Center and Louisiana Department of Transportation and Develop-
0 ment. The authors express their thanks to Mark Morvant, Zhongjie
0 10 20 30 40 50 60 70 80 Zhang, and Gavin Gautreau at LTRC for providing valuable help and
PR (mm/blow) support in this study.

FIGURE 13 E LFWD versus DCP–PR.


REFERENCES

1. Fleming, P. R. Recycled Bituminous Planings as Unbound Granular


CONCLUSIONS Materials for Road Foundations in the UK. Proc., 5th International Con-
ference on Bearing Capacity of Roads and Airfields, Vol. 3, Trondheim,
Norway, 1998.
This paper presents the results of field tests conducted to assess the 2. Fleming, P. R. Field Measurement of Stiffness Modulus for Pavement
potential use of LFWD to measure the in situ elastic modulus of Foundations. Presented at 80th Annual Meeting of the Transportation
pavement layers and subgrades for future use in pavement character- Research Board, Washington, D.C., 2001.
ization. The LFWD elastic modulus (ELFWD) was compared with 3. Livneh, M., and Y. Goldberg. Quality Assessment During Road Forma-
tion and Foundation Construction: Use of Falling-Weight Deflectometer
moduli obtained from other standard tests (FWD, PLT) and with the and Light Drop Weight. In Transportation Research Record 1755, TRB,
DCP–PR. National Research Council, Washington, D.C., 2001, pp. 69–77.
Statistical analyses were conducted to determine the best correla- 4. Fleming, P. R., M. W. Frost, and C. D. F. Rogers. A Comparison of
tions. Laboratory tests also were conducted to estimate the influence Devices for Measuring Stiffness In-Situ. Proc., 5th International Confer-
ence on Unbound Aggregate in Roads, Nottingham, United Kingdom,
depth of the LFWD test device. Based on the results of this study, the 2000.
following conclusions can be made: 5. Kamiura, M., E. Sekine, N. Abe, and T. Meruyama. Stiffness Evaluation
of the Subgrade and Granular Aggregate Using the Portable FWD. Proc.,
• The results of the laboratory tests showed that the influence 5th International Conference on Unbound Aggregate in Roads, Notting-
ham, United Kingdom, 2000.
depth of the LFWD ranged from 270 to 280 mm (10.5 to 11 in.), 6. Abu-Farsakh, M. Y., K. Alshibli, M. D. Nazzal, and E. Seyman. Assess-
depending on the stiffness of the tested materials. ment of In Situ Test Technology for Construction Control of Base
• Field test results showed that the repeatability of the LFWD Courses and Embankments. Report No. FHWA/LA.04/389. Louisiana
depends on the stiffness of the tested material. Poor repeatability was Transportation Research Center, Baton Rouge, 2003.
7. Kleyn, E. G. The Use of the Dynamic Cone Penetrometer (DCP). Report
observed for weak subgrade layers, which was significantly improved 2/74. Transvaal Roads Department, Pretoria, South Africa, 1975.
for stiff and well-compacted layers. 8. Odemark, N. Investigations as to the Elastic Properties of Soils and
• Good agreement was obtained between the ELFWD and the FWD Design of Pavements According to the Theory of Elasticity. Statens
backcalculated resilient modulus (MFWD) with R2 = 0.94, which sug- Väginstitut, Mitteilung No. 77, Stockholm, Sweden, 1949.
9. Chen, D.-H., and J. Bilyeu. Comparison of Resilient Moduli Between
gests that the LFWD and FWD tests yield close modulus values. Field and Laboratory Testing: Case Study. Presented at 78th Annual
However, the MFWD prediction models were improved by including Meeting of the Transportation Research Board Washington, D.C., 1999.
the void ratio and moisture content as variables in the LFWD soil 10. Pen, C. K. An Assessment of the Available Methods of Analysis for Esti-
properties model. mating the Elastic Moduli of Road Pavements. Proc., 3rd International
Conference on Bearing Capacity of Roads and Airfields, Trondheim,
• The ELFWD correlated well with initial and reloading moduli
Norway, 1990.
obtained using the PLT. The R2 values were 0.97 and 0.98 for EPLT(i)
and EPLT(R2), respectively. Similar to the MFWD models, the EPLT(i) and The Soil and Rock Properties Committee sponsored publication of this paper.

You might also like