70 ■ Transportation Research Record 1806

Paper No. 02- 2039

Automated Dynamic Cone

Penetrometer for Subgrade
Resilient Modulus Characterization
Ashraf M. Rahim and K. P. George

Subgrade soil characterization, expressed as resilient modulus (MR), dures have been proposed to estimate subgrade moduli. Of particu-
is crucial for pavement design. For a new design, MR is generally ob- lar interest in this paper is an automated version of the dynamic cone
tained by conducting repeated triaxial tests on reconstituted and un- penetrometer (DCP) as a potential device for estimating MR through
disturbed cylindrical specimens. Because of the test’s complexities, in correlation.
situ tests are desirable, if reliable correlation is established. This study The DCP consists of a steel rod with a cone at one end that is
investigated the viability of using the automated dynamic cone pen- driven into the subgrade by means of a sliding hammer. The material’s
etrometer (ADCP, abbreviated as DCP) for subgrade characterization resistance to penetration is measured in terms of the DCP index
through correlation between DCP index (penetration per blow) and (DCPI) in millimeters/blow (3 ). Figure 1 shows the fully portable
MR. The sensitivity of DCP results to changes in subgrade stiffness, trailer-mounted automated DCP during operation in the field. The
expressed in modulus values, was also investigated. Twelve as-built DCP was originally designed and used to determine the strength
subgrade sections, reflecting a range of typical Mississippi subgrade profile of the subgrade because of its ability to provide a continuous
materials, were selected and tested with DCP. Undisturbed samples record of relative soil strength with depth. Numerous studies have been
were extracted with a Shelby tube and tested in a repeated triaxial conducted investigating the use of the DCP and factors affecting its
machine for MR. After other soil physical properties were determined, results (4– 8 ).
the soil was classified according to AASHTO procedure. DCP tests The direct use of the DCP in pavement design is yet to be estab-
were repeated atop the subgrade through drilled holes after construc- lished. However, it has been correlated to commonly used soil
tion of pavement layers. Falling weight deflectometer (FWD) testing parameters, for example, California bearing ratio (CBR) (5 ). Burn-
was conducted near the DCP test locations. Results suggest two rela- ham and Johnson (9 ) stated that the DCP could be used to provide
tionships, for fine-grain and for coarse-grain soils, in correlating the a reasonable estimate of unconfined compressive strength of soil-
DCP index to laboratory MR. Other physical properties helped improve lime mixtures. The shear strength of soils was correlated with the
the robustness of the regression models. For model verification, the DCP DCPI, for different confining pressures, in a laboratory study con-
and MR tests were repeated and produced good agreement between pre- ducted by Ayers et al. (7 ). Employing a DCPI–CBR relation fol-
dicted and actual MR values. The DCP index before and after pavement lowed by another one between the CBR and MR, the DCP results were
layer’s emplacement suggests subgrade stiffness enhancement, owing converted to the MR, which showed good agreement with both
to pavement overburden, which agrees with FWD–backcalculated laboratory-measured and falling weight deflectometer (FWD)–
moduli. backcalculated values (8 ).
Only a few studies attempted to directly correlate the MR to DCPI.
In the original AASHTO Guide for Design of Pavement Structures, Hassan (4 ) developed a simple regression model correlating the
published in 1961 and revised in 1972, the subgrade stiffness was MR with DCPI for fine-grain soils at optimum moisture content.
accounted for by assigning a soil support value (SSV) on a scale Chai and Roslie (10 ) used the results of DCP tests and CBR–DCP
ranging from 1 to 10. In 1986, the AASHTO guide was substantially relationships developed in Malaysia during the 1987 National Axle
revised to include replacement of SSV with subgrade resilient mod- Load Study to determine in situ subgrade elastic modulus. Jianzhou
ulus (MR ) (1 ). MR values may be estimated directly from labora- et al. (11 ) analyzed the FWD deflection data and DCP results of
tory testing, indirectly through correlation with other laboratory and six pavement projects in Kansas to develop a relationship be-
field tests, or backcalculated from deflection measurements. tween the DCPI and backcalculated subgrade moduli. Adopting the
For a new design, MR values are generally obtained by con- one-dimensional projectile penetration, originally developed by
ducting repeated triaxial tests on reconstituted and undisturbed Yankelevsky and Adin (12), Chua (13) related the DCP test results to
cylindrical specimens. The laboratory test is a tedious, costly, and elastic moduli of subgrade soil.
time-consuming procedure. Large numbers of samples need to be The objective of this paper is to explore the feasibility of using
collected and tested for reasonably accurate results. Even then, it is the DCP to characterize subgrade soil MR, via correlating the DCPI
difficult to reproduce the in situ sample conditions (2). Considering to laboratory resilient modulus. For two different soil groups,
the complexity of the repeated load triaxial test, field testing proce- namely, fine-grain and coarse-grain soils, two different regression
models are attempted correlating the MR (as dependent variable)
A. M. Rahim, California Department of Transportation, Central Region, Fresno, CA
with the DCPI and other soil physical state variables as independent
93726. K. P. George, Department of Civil Engineering, University of Mississippi, variables. Influence of pavement overburden on subgrade stiffness
Carrier Hall, Room 203, University, MS 38677. (moduli), and in turn on the DCPI, is also investigated.
Rahim and George Paper No. 02- 2039 71

lists the locations, soil classifications, and other properties of the soils
in each section. A schematic of the test activities and evaluation
strategies can be seen in Figure 2.

Field Testing and Sampling

Dynamic Cone Penetrometer Tests

Before the experimental program in the field, side-by-side tests were

conducted with a manual DCP and an automated DCP, establishing
that they provide statistically similar results. For the purposes of this
discussion, however, the penetration test results will be referred to
as DCP results, though most of the results were gathered employing
an automated DCP. The grading operation was completed in early
1999, and the subgrade of each section was tested in the summer of
1999. The scheme for the DCP investigation consisted of testing at
30 m (100 ft) intervals to a depth of 1 m (3 ft) in the subgrade. At the
same locations, the DCP tests were repeated in the spring and sum-
mer of 2000, after completion of pavement construction. After graph-
ing the results of penetration versus number of blows, as shown in
Figure 3, possible subgrade layering was sought by investigating
slope changes in the trend line. Through a visual inspection of the
plots, the points of slope change were identified, thus determining
different layers, each with a finite slope. The slope of a segment is
designated by the DCPI in units of millimeter/ blow. As shown in
Figure 3, most of the 60 sample locations exhibited three layers. The
calculated DCPI for each layer was used for correlation with the cor-
responding laboratory-measured MR, as will be discussed in detail
in a later section.

Falling Weight Deflectometer Tests

FIGURE 1 Automated DCP in operation.
Side-by-side FWD tests were also conducted in all 12 test sections,
before and after emplacement of pavement layers. The stiffness change
TESTING PLAN
observed in the FWD test series will be compared with the corre-
Test Sections sponding change in the results of DCP test series. How sensitive the
two tests are to the overburden will be discussed in a later section.
Selected were 12 test sections of 244 m (800 ft) in completed grad- Detailed results of FWD–backcalculated moduli before and after
ing projects, reflecting a range of soil types in Mississippi. Table 1 pavement construction can be found in Rahim and George (14 ).

TABLE 1 Soil: Physical Properties of Test Sections

Proctor Test Shelby Tube Samples

a 3
Section Roadway/ AASHTO Max. Dry Density , kN/m (pcf) Dry Density b, kN/m3
No. County Station Classification / Optimum Moisture, % (pcf) / Moisture b, %
1 SR25/Rankin 1303-1311 A-6 17.4 (110.8) / 14.0 18.0 (114.6) / 16.8
2 SR25/Rankin 1347-1355 A-6 18.2 (115.9) / 12.0 19.0 121.0) / 12.5
3 SR25/Rankin 1590-1598 A-6 17.1 (108.9) / 14.3 17.0 (108.3) / 16.9
4 SR25/Rankin 1696-1704 A-6 18.00 (114.6) / 13.0 18.4 (117.2) / 12.8
5 SR25/Leake 522-530 A-6 18.4 (117.2) / 14.0 19.2 (122.3) / 11.7
6 US45/Monroe 88-96 A-2-4 16.7 (106.4) / 15.0 17.5 (111.5) / 14.2
7 US45/Monroe 108-116 A-2-4 16.3 (103.8) / 16.0 17.4 (110.8) / 16.5
8 US45/Monroe 170-178 A-2-4 17.1 (108.9) / 14.5 18.4 (117.2) / 12.4
9 US45/Monroe 260-266 A-2-4 15.7 (100.0) / 17.5 16.8 (107.0) / 16.3
10 US45/Monroe 461-469 A-6 17.4 (110.8) / 14.5 18.1 (115.3) / 16.6
11 US45/Monroe 490-498 A-6 17.1 (108.9) / 15.5 18.2 (115.9) / 13.7
12 US45/Monroe 668-676 A-3 15.7 (100.0) / 15.5 16.5 (105.1) / 17.2
a Standard Proctor.
b Average dry density/moisture determined from Shelby tube samples.
72 Paper No. 02- 2039 Transportation Research Record 1806

Study Plan

Laboratory Tests Field Testing

Shelby Tube FWD DCP

Before and after pavement Before and after pavement
layers construction layers construction

Undisturbed
cylindrical samples
Comparison of DCP and FWD
results before and after pavement
layers construction

Atterberg limits, sieve analysis, Laboratory

moisture content, dry density resilient modulus

LL, PI, wc, cu, γd, % passing #200 MR DCP index

1- Fine-grain soil correlation; MR = f (DCPI, LL, wc,)

2- Coarse-grain soil correlation; MR = f (DCPI, cu, ….)

FIGURE 2 Schematic chart showing field and laboratory tests and data analysis.

Soil Sampling and Testing Laboratory Tests

Composite bulk samples were collected from every section for rou- Resilient Modulus Test
tine laboratory testing, including the Standard Proctor compaction
(T99-90), with those results presented in Table 1. For MR testing, Laboratory MR tests, following the AASHTO TP46 Protocol, were
Shelby tube samples were obtained from five locations at 61-m conducted employing the Mississippi Department of Transportation
(200-ft) intervals to a depth of 1.5 m (5 ft). Retrieved from each foot repeated triaxial machine furnished by Industrial Process Control.
was one test cylinder of 71 mm (2.8 in.) in diameter by 142 mm The deformation in the samples was monitored employing two lin-
(5.6 in.) in height, with the top three samples tested for the MR in the ear variable differential transformers mounted outside of the testing
laboratory, accumulating 15 MR values from each test section. At chamber. The average MR values for the last five loading cycles of a
completion of the MR test, the density and moisture contents of each 100-cycle sequence yielded the MR. Figure 4 presents a typical plot
sample were determined. The average values per section are listed of MR versus deviatoric stress for a fine-grain soil sample. In total,
in Table 1. As can be verified, the actual densities of the Shelby tube 180 plots were prepared, from which the MR of each sample was
samples are somewhat higher than the maximum Standard Proctor interpolated for the stress state representative of field conditions.
densities (by 1.5% to 8.8%). It could be the case that those samples As a general rule, the modulus values of the first-foot samples of
have undergone recompaction and densification while the Shelby tube fine-grain soil sections were always higher than the corresponding
was being pushed in extracting a sample. This problem is aggravated coarse-grain soil values. This result could be attributed to desicca-
in the top layer because of desiccation and shrinkage. tion and shrinkage and consequent stiffening of the soil. Increase in
Rahim and George Paper No. 02- 2039 73

No. of Blows
0 10 20 30 40 50 60
0

Cumulative Penetration (mm)

200

400

600

800

1000

1200

1400

FIGURE 3 Typical ADCP test results (Monroe County, Station 8900).

soil suction in fine-grain soil, as a result of desiccation, in turn, gives etration versus blows plot (typical plot in Figure 3), three DCPI val-
rise to larger modulus values. ues were extracted corresponding to the same sample location. Since
the MR is a function of stress state, the question arises as to selecting
an MR from an MR–stress state relation. Relying on the results of
Laboratory Routine Tests
Thompson and Robnett (16), Elliot (17 ) suggested using a zero con-
Routine tests for soil classification (M145-87) (15 ) were conducted fining pressure and a 41.6-kPa (6-psi) deviatoric stress when selecting
on composite samples collected from each section (see Table 1). an MR value from laboratory test data. With the recognition that
The same routine tests determining soil state variables were per- in-place subgrade must sustain the overburden of pavement layers,
formed on soil from each test cylinder at completion of the MR test. in addition to the standard 18-kip axle load, in situ stress due to a
Those properties comprised some of the explanatory variables in the typical pavement is combined with stress due to a 20-kN (4,500-lb)
regression analysis. wheel load at a tire pressure of 690 kPa (100 psi). Stresses calcu-
lated by KENLAYER (18 ) yielded a stress state of 37 kPa (5.4 psi)
deviatoric stress and 14 kPa (2 psi) lateral stress. Making use of
RESILIENT MODULUS RELATED TO DCPI the foregoing stress combination, one MR value for each sample was
interpolated from plots such as those in Figure 4.
Resilient Modulus Determination

The plan called for correlating the MR to DCPI ( possibly in con- MR –DCPI Correlation
junction with other soil physical state properties). At each Shelby
tube location, samples from the first, second, and third foot yielded By necessity, the 180 test cylinders from 12 test sections were clas-
MR–stress state curves, such as are shown in Figure 4. From the pen- sified into two groups: fine-grain and coarse-grain soil, in accordance

120
Conf. pressure = 41 kPa

100 Conf. pressure = 28 kPa

Resilient Modulus (MPa)

Conf. pressure = 14 kPa

80

60

40

20
0 10 20 30 40 50 60 70

Deviatoric Stress (kPa)

FIGURE 4 Typical plot of laboratory M R test results (Monroe County, Station

670+ 00, Sample 1).
74 Paper No. 02- 2039 Transportation Research Record 1806

with AASHTO M145-87 (15). For each group, one regression model to determine the regression coefficients. After an exhaustive search,
relating the MR to the corresponding DCPI was attempted. examining many different forms and interaction terms, the following
Because the DCP test is destructive in nature, it is not realistic to model forms were selected: for fine-grain soil,
expect a one-to-one relation between the MR and DCPI. Therefore,
  LL  
a3
it appeared prudent to include basic soil state properties as indepen-
MR = ao ( DCPI )a1  γ a2
dr +  
dent variables in the regression models. After an exhaustive search   wc  
for variables that may have some bearing on the MR, the following
were selected in the regression analysis: dry density (γd ), moisture R2 = 0.71 RMSE = 31.6 (1)
content (wc ), liquid limit (LL), plasticity index (PI ), percentage pass-
ing the #200 sieve, and uniformity coefficient (cu ). Table 2 lists the and for coarse-grain soil,
range of dependent and independent variables for the two soil groups.
b1
The Statistical Package for the Social Sciences (SPSS) (19) was  DCPI 
employed to perform the regression analysis in this study. For identi-
MR = bo  
 log cu 
( γ b2dr cr )
+ w b3
fying individual variables that exhibit high correlation with the MR, a
simple correlation analysis was conducted. Explanatory variables with R2 = 0.72 RMSE = 12.1 (2)
poor correlation with the MR were identified and in turn excluded from
where R2 is the coefficient of determination, and RMSE is the root-
further consideration. The PI for fine-grain soils and the percentage
mean-square error.
passing the #200 sieve for coarse-grain soils are examples. For a reli-
Listed in Table 4 are the values of regression coefficients and
able regression model, there should not be a strong correlation among
summary statistics of the derived models. The F-test for multiple
the explanatory variables. Explanatory variables, if they are highly cor-
related, would weaken the predictive power of the model, a problem regression relation was conducted to validate the significance of
referred to as multicollinearity. Multicollinearity is always associated the relationship between the MR and all of the explanatory variables
with unstable estimated regression coefficients and can seriously included in the models (20). With F * values for fine-grain and coarse-
limit the use of regression analysis for inference and forecasting (20, grain soil much higher compared with F(0.95, p − 1, n − p), the evi-
21). If such multicollinearity does exist, its effect can be eliminated dence is sufficient to show that a relationship does exist between the
by coining or transforming variables. Simple correlations were con- MR and other explanatory variables. Note that p is the number of
ducted exploring multicollinearity, among the transformed variables. explanatory variables, and n is the number of observations. The sig-
For illustrative purposes, the correlation matrix for coarse-grain soil is nificance of individual coefficients was tested employing the t-test.
presented in Table 3. It is clear from the table that the correlation coef- At a confidence level of 95%, all of the coefficients are significant,
ficients between each pair of transformed variables are relatively low, as t* > 1.96 (see Table 4).
less than 0.40 for all combinations. In addition, the coefficient of cor- To check the robustness of the developed models, residuals versus
relation of the MR with each transformed variable has improved com- the predicted MR values for the two soil groups were plotted. Such plots
pared with that before transformation. More detailed results and are used to examine multicollinearity among the explanatory variables
analysis can be found in Rahim (22). The listed transformed variables after developing the model. The residual-MR plots for fine-grain and
were, therefore, employed for developing the regression model. coarse-grain soil are presented in Figure 5a and 5b, respectively.
In developing regression models, the curve estimation option in No distinct pattern is observed, ruling out multicollinearity among
SPSS was employed. The best forms of relation between the MR and explanatory variables. Therefore, the models are well specified.
each of the explanatory variables were investigated, employing the The laboratory MR values were plotted against the predicted MR
coefficient of determination (R 2 ) as the best-fit criterion. The explana- for fine-grain and coarse-grain soil, as shown in Figure 6a and 6b,
tory variables were then combined, and different model forms were respectively. The plotted points clustering along the line of equality
examined. The nonlinear regression option in SPSS was employed are an indication of the robustness of the presented models.

TABLE 2 Ranges of Dependent and Independent Variables for Fine-Grain and

Coarse-Grain Soils

Variable
Soil Type Variable notation Description Range
Dependent MR Laboratory resilient 31(4,436) – 269 (38,986)
modulus a, MPa (psi)
DCPI Penetration Index, mm (in.) 3.7 (0.14) – 66.7 (2.63)
Fine-grain γd Dry density, kN/m3 (pcf) 15.1 (96.0) – 20.6 (131)
Independent wc Moisture content, % 10.6 – 31.1
LL Liquid limit 20 - 57
PI Plasticity index 2 - 31
Dependent MR Laboratory resilient 28 (4,058) – 158 (22,899)
modulus a, MPa (psi)
Coarse- DCPI Penetration index, mm (in.) 5.6 (0.22) – 40.0 (1.6)
grain γd Dry density, kN/m3 (pcf) 15.7 (99.7) – 19.1 (121.6)
Independent wc Moisture content, % 12.4 – 22.0
cu Uniformity coefficient 2.8 - 925
% #200 Percent passing # 200 sieve 7 - 33
a MR values calculated at 37 kPa deviatoric stress and 14 kPa confining pressure.
Rahim and George Paper No. 02- 2039 75

TABLE 3 Correlation Matrix of Transformed Explanatory Variables for

Coarse -Grain Soil

MR DCPI/ Log cu γdr wcr

MR 1 -0.45 0.35 -0.42
DCPI / Log cu -0.45 1 -0.20 0.03
γdr 0.35 -0.20 1 -0.40
wcr -0.42 0.03 -0.40 1
γdr = Actual density / standard Proctor maximum density.
wcr = Actual moisture content / optimum moisture content.

Model Verification through pavement layers. Results from these tests made it possible
to investigate the sensitivity of the DCP test to changes in subgrade
To verify the predictability of the developed models, the DCP test was stiffness brought about by overburden pressure.
conducted at four different locations in a construction site in Oxford, A comparison between backcalculated moduli from FWD deflec-
Mississippi. Field density and moisture content were measured for the tion data conducted directly on the subgrade and after pavement layer
four locations. Bulk soil samples were collected from each of the test construction revealed that modulus values were enhanced primar-
locations, and cylindrical specimens were reconstituted for labora- ily due to overburden imposed by pavement layers. On average, the
tory MR testing, duplicating the moisture and density in the field. modulus was enhanced by 40% and 100% for fine-grain and coarse-
Atterberg limits and grain size distribution were also determined, with grain soils, respectively. More details of the comparison, results, and
the soils classified as A- 4, putting them in the fine-grain soil group. analysis can be found in Rahim and George (14 ). The DCP results
The laboratory MR values were determined, for the three samples before and after pavement layer construction (DCPI1 and DCPI2 )
from each location, at stress combination of 37 kPa deviatoric stress were compared to investigate whether the stiffness change, captured
and 14 kPa confining pressure. The average of the three MR values by backcalculated moduli, could be in line with the DCPI change.
is listed in Column 7 in Table 5. With employment of the fine-grain Figure 7a and 7b compare DCPI1 and DCPI2 for fine-grain and
soil model (Equation 1), the MR values were predicted and compared coarse-grain soil, respectively.
with average laboratory values (Columns 8 and 7 of Table 5). As expected, the ratios of DCPI2 / DCPI1 are different for the two
To compare the predicted and actual MR values, the test of dif- soil types. With stiffening of the soil, the DCPI2 decreased in rela-
ference of paired samples was conducted (23). Twelve modulus val- tion to the DCPI1, providing percentage decreases of 20% and 34%
ues (three per location) form the sample size. On the basis of the for fine-grain and coarse-grain soil, respectively. A scrutiny of the
test results (t* = − 0.52 compared with  t0.025, 11 = 2.201), the null test conditions reveals that the subgrade is only partially confined
hypothesis, H0—no significant difference exists between predicted when tested through the core hole. A recent study by Chen et al. (8 )
and actual values—is accepted. Viewed differently, no evidence showed that the DCPI2, determined through a core hole, was 11%
of significant difference exists between actual and predicted mod- smaller than the DCPI1 with no overburden. The authors gave no
uli, which validates the predictability of one of the models devel- clue as to the type of soil tested, however. In the second part of their
oped in this study. experiment, the DCP was driven through the asphalt layer deter-
mining DCPIasp for fully confined conditions, observing 30%
reduction from DCPI2 to DCPasp. If this 30% reduction were to be
SENSITIVITY OF DCP TO CHANGES IN applied to our results, it could be shown that from a no-overburden
SUBGRADE STIFFNESS to full-overburden state, the DCPI could decrease by 44% and 60%
for fine-grain and coarse-grain soil, respectively. These changes
The MR –DCP correlation study was part of a larger study to char- (decreases) in the DCPI are somewhat in line with the changes
acterize Mississippi subgrade soils, in terms of MR. Because of the in backcalculated moduli of 40% for fine-grain and 100% for
need for MR values in both pavement design and rehabilitation, other coarse-grain soils.
tests were conducted, including FWD tests, before and after pave- A basic difference in the test procedure may be cited for the poor
ment layer construction, and DCP atop subgrade following coring or inferior correlation between the DCPI change and backcalculated

TABLE 4 Summary Statistics for Two Soil Group Models

Soil Type Coefficient Value t* F* RMSE R2

ao 27.86 4.33
a1 -0.114 2.05
Fine-grain a2 7.82 4.60 46.5a 31.45 0.71
a3 1.925 10.81
bo 90.68 9.99
b1 -0.305 10.48
Coarse-grain b2 -0.935 1.98 31.82b 12.12 0.72
b3 0.674 2.17
a Critical F = 2.50
b Critical F = 2.55
76 Paper No. 02- 2039 Transportation Research Record 1806

100 300
80
250
60

Predicted M R , MPa
40
Residuals, MPa

200
20
0 150
-20
100
-40
-60 50
-80
-100 0
0 50 100 150 200 250 300 0 50 100 150 200 250 300

Predicted MR , MPa Actual MR , MPa

(a) (a)

40 180
160
30
140

Predicted M R , MPa
20
Residuals, MPa

120
10
100
0 80
-10 60
-20 40
20
-30
0
-40
0 20 40 60 80 100 120 140 160 180
0 50 100 150
Predicted MR, MPa Laboratory MR , MPa

(b) (b)

FIGURE 5 Residuals versus predicted M R values for FIGURE 6 Laboratory versus predicted M R values for
(a ) fine -grain soil and ( b) coarse-grain soil. (a ) fine-grain soil and (b) coarse-grain soil.

moduli change. The difference is that the DCP test is destructive in models for fine-grain and coarse-grain soil. A feature of the models
nature, whereas the FWD test is not. Because the soil is plasticized is that in addition to the DCPI, other soil state variables are found to
while the cone is driven into the soil, the effect of confinement be significant in MR prediction. The models were verified by repeat-
hardly arises in the DCP test; accordingly, the DCPI change from no ing the tests in another site and comparing measured and predicted
overburden to full overburden is not as significant as is the modulus moduli. Following is a summary of the significant conclusions:
change in the FWD test. Nonetheless, that the DCP change closely
follows the modulus change is an indication of the viability of the 1. The automated DCP is a simple and expedient device for field
DCP for monitoring modulus changes of in-place subgrade. testing of soils and particulate material.
2. Shelby tube samples extracted for MR testing had suffered sig-
nificant sample disturbance, especially those from the desiccated
SUMMARY AND CONCLUSIONS top layer.
3. As mandated by data, two models were developed—one for
The focus of this study was to investigate the use of the automated fine-grain and another for coarse-grain soil. The model predictabil-
DCP for subgrade soil characterization. In 12 sections of prepared ity is substantially increased by including soil state properties as
subgrade, side-by-side DCP and FWD tests were conducted, and additional explanatory variables.
undisturbed samples for MR tests were collected. MR values were 4. The predictability of the fine-grain soil model was substantiated
regressed against the DCPI other soil state properties, obtaining by testing an independent site with excellent comparison between

TABLE 5 Physical and Mechanical Properties of Samples Tested for Model Verification

Actual Actual Predicted

Location moisture Dry densitya, Moisture Density Liquid MRa, MRa,
# contenta, % kN/m3 ratioa ratioa limita MPa MPa
1 12.6 17.1 0.76 1.05 39.0 189 216
2 12.6 17.8 0.76 1.08 37.0 197 193
3 15.3 16.8 0.93 1.03 39.0 141 146
4 13.0 16.7 0.79 1.02 28.0 113 103
a Average of three samples.
Rahim and George Paper No. 02- 2039 77

80 3. Hammons, M. I., F. Parker, A. M. Malpartida, and J. M. Armaghani.

Development and Testing of an Automated Dynamic Cone Penetrom-
70
eter for Florida Department of Transportation. Draft Report, Contract
60
(DCPI)2, mm/blow

FLDOT-ADCP-WPI #0510751. Auburn University, Ala., 1997.

50 4. Hassan, A. The Effect of Material Parameters on Dynamic Cone
40
Penetrometer Results for Fine-Grained Soils and Granular Materials.
Ph.D. dissertation. Oklahoma State University, Stillwater, 1996.
30 5. Livneh, M., I. Ishai, and N. A. Livneh. Effect of Vertical Confinement
20 on Dynamic Cone Penetrometer Strength Values in Pavement and Sub-
10 grade Evaluation. In Transportation Research Record 1473, TRB,
National Research Council, Washington, D.C., 1995, pp. 1–8.
0
6. Livneh, M. Friction Correction Equation for the Dynamic Cone Pen-
0 20 40 60 80
etrometer in Subsoil Strength Testing. In Transportation Research
(DCPI)1, mm/blow Record: Journal of the Transportation Research Board, No. 1714, TRB,
(a) National Research Council, Washington, D.C., 2000, pp. 89 –97.
7. Ayers, M. E., M. R. Thompson, and D. R. Uzaraki. Rapid Shear
Strength Evaluation of In Situ Granular Materials. In Transportation
Research Record 1227, TRB, National Research Council, Washington,
70 D.C., 1989, pp. 134 –146.
60
8. Chen, D.-H., J.-N. Wang, and J. Bilyeu, Application of Dynamic Cone
Penetrometer in the Evaluation of Base and Subgrade Layers. In Trans-
(DCPI)2, mm/blow

50 portation Research Record: Journal of the Transportation Research

40 Board, No. 1764, TRB, National Research Council, Washington, D.C.,
2001, pp. 1–10.
30
9. Burnham, T., and D. Johnson. In-Situ Foundation Characterization
20 Using the Dynamic Cone Penetrometer. Report MN-93/05. Minnesota
10
Department of Transportation, Maplewood, 1993.
10. Chai, G., and N. Roslie. The Structural Response and Behavior Predic-
0 tion of Subgrade Soils Using Falling Weight Deflectometer in Pavement
0 10 20 30 40 50 60 70 Construction. Proc., 3rd International Conference on Road and Airfield
(DCPI)1, mm/blow Pavement Technology, Beijing, China, April 1998, pp. 1481–1489.
(b) 11. Chen, J., M. Hossain, and T. M. LaTorella. Use of Falling Weight
Deflectometer and Dynamic Cone Penetrometer in Pavement Evalua-
FIGURE 7 Comparison of DCPI 1 and DCPI 2 (before and after tion. In Transportation Research Record: Journal of the Transporta-
pavement construction) for (a ) fine-grain soil and tion Research Board, No. 1655, TRB, National Research Council,
(b) coarse -grain soil. Washington, D.C., 1999, pp. 145–151.
12. Yankelevsky, D. Z., and M. A. Adin. A Simplified Analytical Method
for Soil Penetration Analysis. International Journal for Numerical and
Analytical Methods in Geomechanics, Vol. 4, 1980, pp. 233–254.
13. Chua, K. M. Determination of CBR and Elastic Modulus of Soils Using
a Portable Pavement Dynamic Cone Penetrometer. Proc., 1st Inter-
the measured and predicted MR values. The coarse-grain soil model, national Symposium on Penetration Testing, Orlando, Fla., March 1988,
however, awaits verification. pp. 407– 414.
5. The sensitivity of the DCP test procedure is substantiated by 14. Rahim, A., and K. P. George. Falling Weight Deflectometer for Estimat-
comparing the change in DCPI (before and after emplacement of pave- ing Subgrade Elastic Moduli. Journal of Transportation Engineering,
ment) with the corresponding change in FWD backcalculated moduli. ASCE, 2002 (in press).
15. AASHTO Standard Specifications for Transportation Materials and
Methods of Sampling and Testing, 15th ed. Washington, D.C., 1990.
By way of general conclusions and recommendation, it is advanced 16. Thompson M. R., and Q. L. Robnett. Resilient Properties of Sub-
that for the range of soils tested, the developed MR – DCPI models grade Soils, Final Report—Data Summary. Transportation Engineering
provide useful predictions of MR. Series 14, Illinois Cooperative Highway Research and Transportation
Program Series 160. University of Illinois at Urbana-Champaign, 1976.
17. Elliot, R. P. Selection of Subgrade Modulus for AASHTO Flexible
Pavement Design. In Transportation Research Record 1354, TRB,
ACKNOWLEDGMENTS National Research Council, Washington, D.C., 1992, pp. 39– 44.
18. Huang, Y. H. Pavement Analysis and Design. Prentice Hall, Engle-
This paper is a part of the study entitled Subgrade Characterization wood Cliffs, N.J., 1993.
for Highway Pavement Design, conducted by the Department of 19. Norusis, M. J. SPSS for Windows, Professional Statistics, release 6.0.
Civil Engineering, University of Mississippi, in cooperation with SPSS Inc., Chicago, Ill., 1993.
20. Neter, J., W. Wasserman, and G. A. Whitmore. Applied Statistics. Allyn
the Mississippi Department of Transportation (MDOT) and & Bacon, Inc., Boston, Mass., 1988.
FHWA. Technical help received from Alfred Crawley, Joy 21. Chatterjee, S., and B. Price. Regression Analysis by Example. John Wiley
Portera, Bill Barstis, Johnny Hart, and Alan Hatch of MDOT, and and Sons, Inc., New York, 1977.
Jianrong Yu, of the university, is acknowledged. 22. Rahim, A. In-Situ Tests for Subgrade Resilient Modulus Characterization.
Ph.D. dissertation. University of Mississippi, Oxford, 2001.
23. Cooleman, H. W., and W. G. Steele. Experimentation and Uncertainty
Analysis for Engineers. John Wiley and Sons, Inc., New York, 1989.
REFERENCES
The opinion, findings, and conclusions expressed in this paper are those of the
1. AASHTO Guide for Design of Pavement Structures, AASHTO, authors and not necessarily those of the Mississippi Department of Transportation
Washington, D.C., 1986. or FHWA. This paper does not constitute a standard, specification, or regulation.
2. Houston, W. N., M. S. Mamlouk, and R. W. S. Perera. Laboratory Versus
Nondestructive Testing for Pavement Design. Journal of Transportation Publication of this paper sponsored by Committee on Strength and Deformation
Engineering, ASCE, Vol. 118, No. 2, March/April 1992, pp. 207–222. Characteristics of Pavement Sections.