A Study of In Situ Pavement Material Properties
Determined from FWD Testing
December 2004
RSCH008-936
Vermont Agency of Transportation
Pavement Design Committee
“The information contained in this report was compiled for the use of the Vermont Agency of
Transportation. Conclusions and recommendations contained herein are based upon the research data
obtained and the expertise of the researchers, and are not necessarily to be construed as Agency policy.
This report does not constitute a standard, specification, or regulation. The Vermont Agency of
Transportation assumes no liability for its contents or the use thereof.”
1. Report No. 2. Government Accession No. 3. Recipient’s Catalog No. 2004-6
4. Title and Subtitle 5. Report Date December 2004
6. Performing Organization Code
A Study of In Situ Pavement Material Properties Determined from FWD Testing
7. Author(s) 8. Performing Organization Report No.
Michael Pologruto, P.E.
9. Performing Organization Name and Address 10. Work Unit No.
11. Contract or Grant No.
Vermont Agency of Transportation Materials and Research Section
National Life Building Montpelier, VT 05633-5001
12. Sponsoring Agency Name and Address 13. Type of Report and Period Covered
14. Sponsoring Agency Code
Federal Highway Administration
Division Office Federal Building
Montpelier, VT 05602
15. Supplementary Notes
16. Abstract The Vermont Agency of Transportation (Agency) developed pavement design procedures patterned after the American Association of State Highway and Transportation Officials (AASHTO) pavement design model described in the AASHTO 1993 Pavement Design Guide (Guide). While the Guide provides one of the most widely used empirical design models for flexible pavement design, a factor complicating its utility is the use of an abstract quality, the structural number (SN), to quantify the strength of the total pavement structure. A consequence of the SN is the need for structural layer coefficients (ai) to characterize the component materials of the pavement structure. The Agency found it difficult to quantify these design parameters because they are difficult to assess directly and consequently found it equally difficult to calibrate the AASHTO model to Vermont conditions. However, the Agency has developed and tested a method for determining layer coefficients using a falling weight deflectometer (FWD), and the resulting layer coefficients are representative of the in situ behavior of the pavement materials. This method is based on a model provided in the Guide for assessing the effective SN of a pavement structure. The Agency found layer coefficients determined for unbound subbases to be reasonable, while layer coefficients estimated for ACC materials were generally 25-35% higher than AASHTO’s implied maximum of 0.44. 17. Key Words 18. Distribution Statement Pavement Design, Layer Coefficient, Falling Weight Deflectometer, subbase
19. Security Classif. (of this report)
20. Security Classif. (of this page) 21. No. Pages
22. Price
17
1
Table of Contents
Executive Summary ………………………………………………………………………………………………………………………… 3
Objective ……………………………………………………………………………………………………………………………………….. 4
Background ……………………………………………………………………………………………………………………………………. 4
AASHTO Method …………………………………………………………………………………………………………………………… 5
Development of Experimental Model ………………………………………………………………………………………………… 5
Pilot Project to Test Experimental Model …………………………………………………………………………………………. 10
Data Analysis ……………………………………………………………………………………………………………………………….. 11
Fwd Results ……………………………………………………………………………………………………………………………… 11
Layer Coefficients …………………………………………………………………………………………………………………….. 11
Final Structure Simulation ………………………………………………………………………………………………………….. 12
Comparative Analysis of Layer Coefficients …………………………………………………………………………………. 13
Discussion of Results …………………………………………………………………………………………………………………….. 14
Conclusions ………………………………………………………………………………………………………………………………….. 17
Recommendations …………………………………………………………………………………………………………………………. 18
References ……………………………………………………………………………………………………………………………………. 20
2
Table of Figures
Figure 1 Seasonal Variation in SN ………………………………………………………………………………………………….. 6
Figure 2 Granular Subbase Behavior ………………………………………………………………………………………………. 8
Figure 3 Subbase Simulation of Layer Coefficient…………………………………………………………………………….. 9
Figure 4 FWD Testing Progression ……………………………………………………………………………………………….. 11
Figure 5 Determination of Layer Coefficients from FWD Testing …………………………………………………….. 12
Figure 6 ACC Layer Coefficient vs. Depth to Subgrade …………………………………………………………………… 16
List of Tables
Table 1 p-values from Paired t-Testing of FWD Computed and Simul 14
Table 2 14
Table 3 17
Table 4 Recommended Material Properties for Design Using the AASH 19
3
EXECUTIVE SUMMARY
The Vermont Agency of Transportation (Agency) developed pavement design procedures patterned after the
American Association of State Highway and Transportation Officials (AASHTO) pavement design model
described in the AASHTO 1993 Pavement Design Guide (Guide). While the Guide provides one of the most
widely used empirical design models for flexible pavement design, a factor complicating its utility is the use
of an abstract quality, the structural number (SN), to quantify the strength of the total pavement structure. A
consequence of the SN is the need for structural layer coefficients (ai) to characterize the component
materials of the pavement structure. The Agency found it difficult to quantify these design parameters
because they are difficult to assess directly, and consequently found it equally difficult to calibrate the
AASHTO model to Vermont conditions.
However, the Agency has developed and tested a method for determining layer coefficients using a
falling weight deflectometer (FWD), and the resulting layer coefficients are representative of the in situ
behavior of the pavement materials. This method is based on a model provided in the Guide for assessing
the effective SN of a pavement structure. The Agency found layer coefficients determined for unbound
subbases to be reasonable, while layer coefficients estimated for ACC materials were generally 25-35%
support for the predictive qualities of FWD derived layer coefficients to approximate layer coefficients
simulated from the in situ conditions expected to prevail in the final pavement structure.
4
OBJECTIVE
Ever since the Vermont Agency of Transportation (Agency) adopted the American Association of State
Highway and Transportation Officials (AASHTO) pavement design method in 1993, one of the most vexing
problems facing Agency pavement designers has been the calibration of the AASHTO pavement design
procedure for Vermont conditions. Key to this calibration is the determination of the layer coefficients
necessary for characterizing Vermont pavement materials. It has been well established by others that there is
no direct method for quantifying the layer coefficient for a particular material. The AASHTO Pavement
Design Guide (Guide) does provide relationships for determining layer coefficients for several pavement
materials (1), however, these relationships were unique to the materials used to build the AASHO Road Test
in the 1950s. The Guide cautions against using the relationships provided to characterize local materials.
The Guide also suggests that each design organization determine relationships unique to the materials they
use to build pavement structures (1). Unfortunately, the Guide stops short of recommending any procedure
that may be used to determine layer coefficients, or how to develop models for predicting layer coefficients
tee (Committee) undertook a
serious investigation into determining layer coefficients for Vermont pavement materials. The Committee
evaluated as much research as was available on the topic before proposing the multi-year investigation
summarized in this report.
BACKGROUND
There have been several investigations reported using a falling weight deflectometer (FWD) to characterize
the structural properties of pavement materials. Zhou, et al. (2), Hossain, et al. (3), and Janoo (4) all used an
FWD in one way or another to determine material properties for the constituent pavement materials, some
conducting FWD testing on top of each material as the structure was being constructed. However, the
Committee did not consider the methods described for determining layer coefficients, utilizing the AASHTO
modulus/coefficient relationships provided in the Guide, desirable.
While the layer coefficient relationships provided in the Guide are convenient and tempting to use
once a resilient modulus has been established, their use is not necessarily appropriate. The Guide gives no
specific direction, but it does emphasize the importance for designers to calibrate various components of the
design model to local conditions and experience before implementing the AASHTO procedure. Layer
coefficients are certainly no exception to this caveat. Layer coefficients themselves are believed to be a
function of material thickness, underlying material support, and stress state. Further, the modulus/coefficient
relationships provided in the AASHTO Guide were developed for AASHO Road Test materials as they were
constructed at the Road Test site in 1958. The usage of these relationships for materials considerably
different from those used at the Road Test is unsubstantiated and can be misleading. Ideally, AASHTO
should have provided a procedure for designers to develop their own layer coefficient relationships for the
5
materials with which they commonly build pavement structures.
The AASHTO approach to flexible pavement performance quantifies the pavement structure as a
structural number (SN) and further divides the pavement structure into three constituent parts: surface, base,
and subbase. Although it is not very clear what conditions or stress states constitute or distinguish the
surface, base, or subbase from each other, the interplay among the three pavement components and how they
work in concert as a single structure is illustrated by Equation 1,
332211 DaDaDaSN (1)
where ai represents the layer coefficient and Di is the thickness of the material.
Accordingly, layer coefficients for a particular material can be thought to represent the SN-
contribution per unit thickness of that material to the total SN of the pavement structure.
Ideally, what is needed is a way to measure the SN provided by a particular material as a component
of a final pavement structure. This method should be relatively easy to perform so that a variety of
conditions may be surveyed.
AASHTO METHOD
It was not until the publication of the 1993 edition of the AASHTO Guide that a procedure was provided by
AASHTO for determining the in-place SN of a pavement structure using FWD deflection data. This
procedure is described in Appendix L of the 1993 Guide and provides a method for determining the
eff. However, Ioannides expressed concern about the
development of this method, particularly the introduction of mechanistic properties into the
statistical/empirical AASHTO model (5). Regardless, the Committee considered the possibility of deriving
layer coefficients from SNeff
efforts with this model have given this method tacit legitimacy. Specifically, if FWD testing were performed
on the top surface of each component material in a manner similar to that described by Zhou, et al., and
Janoo, the SNeff may be characterized for individual components of a pavement structure. It would follow
that layer coefficients should result from dividing the SNeff-contribution for each material by the thickness of
that material. The veracity of these resulting layer coefficients should then be supported by a comparison
with the layer coefficients that would be expected for the final pavement structure under in situ conditions.
DEVELOPMENT OF EXPERIMENTAL MODEL
The Committee decided to evaluate the SNeff method described in the Guide on several years of seasonal
FWD data initially collected to support the Strategic Highway Research Program (SHRP). Ultimately, over
five years of data, collected at eight different locations throughout the state and representing close to 30,000
deflection basins, provided a comprehensive assessment of the variation in SN for Vermont pavement
structures due to annual seasonal variability. It was observed after spring thaw, a somewhat elusive
6
phenomenon to capture, the SNeff remained fairly stable between days 100 and 300 and exhibited a
coefficient of variation under 10%. This stable time period corresponds very well with the typical April 15
to November 1 construction timeframe established for Agency construction projects. A summary of these
findings for five of the eight sites is illustrated in Figure 1, with SNeff values plotted against the Julian day of
the year (1-365).
SN Seasonal Variation
0
5
10
15
20
25
0 50 100 150 200 250 300 350
Day of year
SN
Berlin – Site 1 Berlin – Site 2 Charlotte New Haven South Hero
Figure 1 Seasonal Variation in SN
The foregoing findings led the Committee to form several assumptions:
if FWD testing were restricted to the May through October timeframe, fairly stable, essentially
unchanging, effective SNs may be expected at a given location,
barring any extreme fluctuations in temperature or moisture conditions, the SN contribution of any
component material, hence the layer coefficient, should also remain fairly stable during the May
through October timeframe, and
the SN contribution of any pavement structure component is independent of the stress states
produced from the range of loads (6,000 to 16,000 pounds) applied to the surface.
The first two assumptions seemed rather obvious from observation of the data presented in Figure 1.
The third assumption was a result of evaluating the daily results and recognizing that all seasonal locations
were tested using the SHRP FWD protocol, which targets four different loads: 6,000, 9,000, 12,000, and
16,000 pounds. Upon a detailed observation, the effective SNs derived from the SHRP protocol loading
range were surprisingly consistent for a given testing day and the coefficient of variation on the range of
7
effective SNs characteristic for any given day was typically about 1%. Put another way, 95% of the effective
SNs for a particular location and developed during a given day of testing were within less than 1% of the
average SN for that day.
The consistency in the SN was unexpected and truly remarkable. Considering the impulse load
more than doubles during the FWD test, the stress-dependency of the modulus for the unbound materials,
and the visco-elasticity of the asphalt stabilized materials, it seemed highly unlikely that the interplay among
the various material stiffnesses would exactly compensate to provide a constant SN to such a precise degree.
It seemed more plausible that each constituent SN associated with the surface, base, and subbase, should
remain relatively constant on its own.
If the foregoing is true, this last assumption supports the notion that the SNeff established for a
particular material may remain reasonably stable from its placement to its service in the final structure if:
1. All construction and FWD testing activities take place during May through October,
2. No extreme temperature or moisture fluctuations occur prior to FWD testing, and
3. FWD target loads for the base and subbase materials are within the magnitude of stresses likely for
the final structure under normal loadings and do not induce shear failure in the unbound materials.
While strongly implied from the analysis of the seasonal data, the Committee nonetheless attempted
to analytically corroborate the second assumption of a stable SN contribution from any component material.
Unfortunately, this analysis of the SNeff method described in the Guide proved beyond a simple algebraic
manipulation of the SNeff model. A more practicable solution considered was to perform a simulation of the
expected behavior of typical Vermont subbase materials using an elastic layer simulation (ELS).
Two conditions were simulated with the ELS to evaluate the behavior of a pavement structure
subjected to an FWD test. Of particular interest in this simulation is the behavior of the granular subbase
material. Two different stages of the pavement construction were examined. The first condition simulated
the FWD test on the stress-dependent granular subbase resting on a stress-dependent fine-grained subgrade.
The second condition simulated the FWD test of a constant-modulus surface material on stress-dependent
granular base, subbase, and fine-grained subgrade materials. The material properties and performance of the
subbase were compared as illustrated in Figure 2.
8
Figure 2 Granular Subbase Behavior
Resilient moduli for these stress-dependent materials in both simulations were determined using a
simple K-theta model as illustrated in Equation 2,
2
1
k
R kM (2)
where: MR is the resilient modulus,
k1 and k2 are material-specific regression constants, and
Under the initial conditions, FWD tests were simulated on the surface of each component material.
This was a straightforward analysis from which deflections, loading plate pressures, and subgrade properties
were readily available. However, when simulating the final condition, the loading plate pressures for the
soils engineers may agree on a Boussinesq stress-distribution for a point load, a typical pavement structure
does not behave the same as an equivalent, relatively homogeneous, soil mass. A different approach is
necessary to model the stress-distribution occurring beneath a circular load on a relatively stiff upper layer
into a less stiff (by an order of magnitude) unbound aggregate. Noureldin and Al Dhalaan (6) proposed a
stress-
the loading plate to a circular area with a radius corresponding to the depth from the surface within a depth
BaseYields an SNeff, but does this equal the SNeff contributed by the subbase
in the final structure?
Subgrade
Subbase
Surface
FWD
FWD
9
ral numbers for base and subbase materials in the
final structure simulation.
Subbase layer coefficients determined from the simulation results of the initial condition described
above were generally within 5% of the layer coefficients determined for the subbase performing in the final
condition and are illustrated in Figure 3. The Committee interpreted the results of this pavement simulation
to validate the assumption that the SN for any component of a pavement structure may remain stable enough
for the design of flexible pavement structures. Without finding any research to contradict the findings of the
simulation, the Committee decided to sponsor a pilot study to determine real world layer coefficients from
FWD testing.
0.950
1.000
1.050
1.100
1.150
14,000 16,000 18,000 20,000 22,000 24,000 26,000 28,000 30,000 32,000
Subbase Modulus – K1 (psi)
Subbase Simulation of Layer Coefficient
Initial to Final Condition
R at
io o
f a 3
K2 = 0.6
K2 = 0.4
K2 = 0.5
Figure 3 Subbase Simulation of Layer Coefficient
In summary, the layer coefficient determination model consisted of the following steps:
1) Assume the SN for any material is a fixed property and remains constant throughout the
construction operation, after it has reached its design condition,
2) Collect FWD deflection data on the top surface of each pavement material, during the
construction season of April 15 through November 1,
3) Use backcalculation software to determine the subgrade MR at the centerline of the load for each
10
FWD test,
4) Correct any deflections taken directly on the pavement, or asphalt cement concrete (ACC), to 68°
F,
5) Determine the SNeff appropriate for each successive build-up of pavement material, and
6) Determine each layer coefficient for each material by taking the difference in the SNeff
determined directly on top and directly below the material layer, and dividing by the material thickness.
Note: The SNeff on top of the subgrade is defined as zero.
PILOT PROJECT TO TEST EXPERIMENTAL MODEL
The next step was to identify a pilot project and collect real data representative of materials used for the
construction of pavement structures in Vermont.
Since analysis of the seasonal data would seem to indicate the drop weight used has little effect on
the SNeff finally determined for any given pavement structure, this study focused on the deflection basins
generated by a single target weight for each material. The target drop weights applied on the surface of each
material were consistent with the effects that would be expected from a 100-psi tire pressure applied at the
plate pressures below 10 psi difficult. This only presented a concern with the sand subbase, which should
have been tested using a pressure in the range of two to three psi. But, testing the sand subbase at 10 psi
yielded no evidence of shear failure due to overstressing and backcalculation results exhibited root-mean-
square (RMS) variations from the FWD-measured deflection basins of less than 25%. The Committee
considered this compromise to be satisfactory for a sand subbase.
The layer coefficients for the pilot project were 0.074, 0.163, and 0.639 for the sand borrow
subbase, dense-graded crushed stone (DGCS), and ACC, respectively. These findings were encouraging
since the layer coefficients established for the unbound materials were within the ranges established by
AASHTO for these materials.
The layer coefficient for the ACC was not discounted outright. Although 0.639 is almost 50%
higher than the 0.44 upper limit established by AASHTO for ACC surface course, two other indicators of
layer coefficients for ACC, a Marshall stability of 2,730 lbf. and a resilient modulus of 580,000 psi, were
also beyond the upper AASHTO limits of 2,100 lbf. and 450,000 psi respectively.
The findings from the data analysis of the pilot project were encouraging. Consequently, the
Committee considered the experimental model developed thus far to be a success. The Committee endorsed
further collection of FWD data, using the experimental model developed with the pilot project, at several
more projects to determine if the method developed was capable of providing satisfactory estimates of
material properties and that these properties are representative of in-service performance. In all, nearly 50
test sites were evaluated for this next phase of the research.
11
DATA ANALYSIS
FWD Results
Backcalculations were performed on all deflection basins to determine the resilient modulus of the subgrade,
a necessary input for the SNeff calculations. Two independent applications were used: ELMOD 4.0 and
EVERCALC 5.0. These two applications perform similar functions, using different algorithms. Both
attempt to achieve convergence between the FWD measured deflection basin and a calculated deflection
basin based on the backcalculated layer moduli.
lent thickness
developed by Odemark and described by Ullidtz (7), was used to spot check a random sample of ELMOD
and EVERCALC output, to ensure reliability of the backcalculation results.
In order to control the quality of the backcalculation findings, goodness-of-fit thresholds were
established for deflection basins taken on the sand, DGCS, and ACC surfaces of 25, 10, and 2% RMS,
respectively. That is, if a backcalculation for a sand deflection basin could not produce a solution with an
RMS less than 25%, that site was removed from further consideration in this study. Similarly, if either the
DGCS or ACC backcalculation failed to meet the appropriate RMS threshold, the entire site was considered
compromised and removed from the study. Figure 4 illustrates how the SNeff progresses as FWD testing is
conducted on each successive pavement material.
Figure 4 FWD Testing Progression
Layer Coefficients
The estimation of layer coefficients (ai) uses the SNeff contributed by each pavement material. Figure 5
Subgrade
Sand
DGCS
ACC
SNeff=10.71
FWD
FWD
FWD
SNeff=5.37
SNeff=1.52
12
illustrates as the SNeff is established for each material interface, the change in SNeff for any two adjacent
material interfaces represents the SN contribution for the material bounded by these adjacent interfaces. The
resulting layer coefficient is the SN contribution for any particular material divided by the thickness of that
material layer. But, if the thickness has not been accurately assessed, this will have a corresponding adverse
effect on the layer coefficient.
Subgrade
Sand
DGCS
ACC
SNeff=10.71
SNeff=5.37
SNeff= 0
t = 7.54 in.
t = 24.36 in.
t = 19.80 in.
SNeff=1.52
SNACC = 10.71-5.37 = 5.34
and
aACC = 5.34÷7.54 = 0.708
SNDGCS = 5.37-1.52 = 3.85
and
aDGCS = 3.85÷24.36 = 0.158
SNSand = 1.52, and
aSand = 1.52÷19.80 = 0.077
Figure 5 Determination of Layer Coefficients from FWD Testing
The development of layer coefficients using the procedure just outlined is relatively easy and
applicable to the materials in question. The issue of whether layer coefficients developed in this manner are
characteristic of material performance of the final (in-place) structure and appropriate for design must be
supported.
Final Structure Simulation
When evaluating the suitability of layer coefficients for use as design parameters, the only pertinent standard
should be their prediction of layer coefficient behavior in the final structure. Ideally, a fully instrumented
pavement structure, with a full array of stress and strain sensors to monitor the behavior of each material
interface, would provide the necessary data to make this comparison. However, based on past experience,
the Committee considered subsurface instrumentation too unreliable.
Instead, an ELS was conducted to simulate the response of the final structure. The simulation was
carried out on a model of the final structure using the actual layer thickness and backcalculated resilient
modulus for each material.
13
calculated below the
surface as proposed by Noureldin and Al Dhalaan, to simulate the behavior of the final structure under an
FWD load and to estimate the SNeff at each material interface. Once the SNeff was determined at the surface
of each material, the layer coefficients were calculated as illustrated previously in Figure 5.
Comparative Analysis of Layer coefficients
In all, six experimental projects amounting to almost 50 test sites were evaluated to establish the significance
between calculating layer coefficients from FWD testing and how well they represent the in situ conditions
estimated by simulation of the final structure. Although cursory observation revealed satisfactory agreement
between centerline deflections measured with the FWD and deflections predicted by the ELS, a more
detailed statistical analysis of the layer coefficients developed from FWD test data and the ELS was done to
provide a more objective means of establishing that no significant difference existed between the results of
the two methods. If no statistically significant difference is found, then any distinction observed may be
attributable to normal variation in the material properties i.e., the materials do not exhibit linear elastic,
isotropic, and homogeneous properties and normal error in data acquisition. Also, if both methods yield
similar results, it would further substantiate the assumption that the SN contributed by a pavement material is
a fixed property and, more importantly, layer coefficients determined via FWD tests are suitable for design
of the final structure.
The statistical analysis was carried out using a paired t Test, which assumes the difference between
pairs of data to average zero. Ordinarily, a low p-value (a statistical metric that quantifies the rarity of an
occurrence) resulting from a paired t Test indicates little relationship between the two data sets being
compared. For this research, a high p-value (>0.05) suggests a statistically significant correlation exists
between the paired data sets. Thus, the p-values determined by this analysis indicate the layer coefficients
determined via FWD tests are suitable for design of the final structure as indicated in Table 1.
-value (that is, for those results determined at each project
-value (representing the results of an analysis carried out as the
results of each successive project are added to the cumulative database). These results indicate a significant
level of agreement, or correlation between, the two data sets suggesting no statistically significant difference
exists between the two methods: FWD- and ELS-derived layer coefficient determinations. Thus, it may be
concluded, with a high degree of certainty, that FWD-derived layer coefficients are sufficiently accurate to
predict in situ behavior to be useful pavement design parameters.
14
Table 1 p-values from Paired t-Testing of FWD Computed and Simulated Layer Coefficients
p-value (at 95% level of confidence)
Project specific Research cumulative
Vergennes-Ferrisburgh 0.29 0.29
Montpelier State Highway 0.48 0.34
Bolton-South Burlington 0.10 0.28
Burlington 0.09 0.13
Colchester 0.09 0.33
Addison 0.09 0.32
DISCUSSION OF RESULTS
As the results of the statistical analysis supported
determined from FWD testing are sufficiently representative of in situ conditions (exhibited via simulation of
the final structure) an evaluation of the results determined up to this point was warranted.
A summary of the findings for the first six projects studied in this investigation are presented in
Table 2, listing the layer coefficients and resilient moduli so far determined and the number of test locations
for which all quality control criteria were met.
Table 2 Summary of Material Properties for First Six Projects
Sand DGCS ACC I ACC II ACC III ACC IS ACC IIS ACC IIIS
ai 0.073 0.152 0.386 0.687 0.855 0.839 0.588 0.495
Mr (psi) 18,900 41,900 397,000 343,000 360,000 321,000 153,000 346,000
N 47 47 30 30 30 15 17 17
Of particular interest are the layer coefficients determined for the unbound materials. The sand
the fact that this material is much deeper than the unbound subbase materials were at the Road Test. Since
the sand is placed so deeply in Vermont pavement structures, where it would experience lower stress states
than Road Test unbound subbases, it may be performing like a fine-grained material and may explain why
the resilient modulus of 18,900 psi falls on the high side of the AASHTO scale, in relation to the layer
coefficient. The value of 0.152 for the DGCS falls on the higher end of the range established by AASHTO
for an unbound base material. This higher layer coefficient is consistent with the higher resilient modulus of
41,900 psi determined for DGCS and also conforms to the behavior one would expect of a stress-stiffening
coarse-graded granular material. By comparison, laboratory testing of these materials has established
15
estimates of the resilient modulus to be 25-35% of the backcalculated resilient modulus for sand (8) and 30-
45% of the backcalculated resilient modulus for DGCS (9).
Probably the most conspicuous eccentricity with the results established thus far in this effort is the
unusually high layer coefficients established for the ACC materials. Although there appears to be nothing
fundamentally wrong with the layer coefficients determined for the ACC materials i.e., the same procedure
was used to derive reasonable unbound layer coefficients and the elastic layer simulation would seem to
indicate an accurate prediction of in-place behavior their use with the AASHTO design model presented
some concerns. Most obviously, any layer coefficients over 0.50 represent a range of conditions as of yet
unsubstantiated for the empirically derived AASTHO model. Also, the ACC layer coefficients developed
under this investigation were established for materials that were designed using much lower layer coefficients
(0.32-0.39) with the AASHTO model. And finally, if the layer coefficients presented here (>0.50) are used
for an AASHTO design under typical Vermont traffic loading, almost no base material (DGCS) is called for
because all of the strength (SN) is provided by a few inches of ACC. The Committee considered several
mechanisms likely to generate layer coefficients outside the traditional range established by AASHTO.
First, environmental conditions in Vermont necessitate thick pavement structures to mitigate the
effects of frost penetration. These substantial structures are likely far beyond anything studied at the Road
Test.
ikely to be different from,
if not an improvement upon, those materials from which the AASHTO relationships have been derived.
Vermont is fortunate to have readily available, high-quality, and affordable aggregates. The Agency has also
traditionally used stiff asphalt cements and high compactive efforts in an attempt to minimize distresses
Third, the ELSs were conducted using the elastic moduli determined from backcalculations of the
FWD deflection basins taken on the surface of the finished pavement structure. Even though many of the
ACC moduli were consistently in excess of the 450,000-psi upper limit published by AASHTO, the layer
coefficients determined via ELS still corroborated the layer coefficients determined from the FWD deflection
data.
And finally, the FWD measures in situ behavior. It is not unreasonable to contend that laboratory-
supported AASHTO modulus/coefficient relationships may not accurately predict in situ behavior for any
material, whether unbound or asphalt stabilized. Indeed, Figure 6 illustrates how the ACC layer coefficients
Interestingly enough, when analyzed using the top of the unbound portion of the structure as the subgrade,
the ACC layer coefficients thus determined cluster within the more traditional range of 0.20-0.44 established
for ACC materials used in the AASHTO model. This interplay between ACC layer coefficients and its
support structure may be analogous to the synergism of a concrete bridge deck supported by steel girders.
Neither is adequate to the task in isolation, but when acting in unison, they achieve an effect of which each is
16
individually incapable.
ACC Layer Coefficients vs. Depth to Subgrade
0
0.2
0.4
0.6
0.8
1
1.2
0 10 20 30 40 50 60 70
Depth to Subgrade (in)
A C
C L
a y
er C
o ef
fi ci
en t
assumed
on top of
stone
assumed
on top of
sand
ordinarily
defined
Figure 6 ACC Layer Coefficient vs. Depth to Subgrade
Valid resilient moduli for the various types of ACC (I, II, III, etc.) materials used by Agency
designers may have to be determined indirectly, since backcalculation limitations cannot distinguish such
subtleties within the FWD loading plate radius of the testing surface. Marshall stabilities were considered
useful for estimating the resilient moduli of the ACC materials, assuming there exists a correlation between
Marshall stabilities and resilient moduli (a notion implied by AASHTO). The Marshall stabilities may give
an indication of the relative proportions of the individual resilient moduli compared to the resilient modulus
backcalculated for the total ACC thickness. Another possibility may be an indirect tension test (like ASTM
D4123), which establishes the resilient modulus for ACC samples. For this investigation, Marshall stabilities
were used, when available, as a proxy to isolate the resilient moduli for different ACC types.
At this time, it is uncertain why there exist such marked disparities between the layer coefficients
determined for Marshall and Superpave materials. The Committee debated this issue extensively and finally
conceded that Marshall and Superpave mixes are two different materials and layer coefficients may simply
be one more manifestation of these differences. The Committee endorsed further study to bolster or refute
some of these concerns with the ACC properties.
Ten additional projects were identified for further study to allow for additional data collection and
to improve the predictive capabilities of the subsequent estimates. Another benefit to further study was the
potential for investigation into additional materials. Two of the additional projects used gravel for subbase
17
instead of the DGCS usually required on the State system. One Interstate project provided for an
rvice for nearly
40 years. Also novel to the Interstate project was an experimental material to provide for better drainage: an
asphalt-treated permeable base (ATPB).
Table 3 summarizes the properties established from all 16 projects investigated.
Table 3 Summary of Material Properties
Material
Layer Coefficient Resilient Modulus (psi)
N Standard
deviation Average 95% Pre. N
Standard
deviation Average. 95% Pre.
Sand 139 0.013 0.078 2.9% 139 10,200 19,100 9.0%
Gravel 21 0.033 0.134 11.1% 21 12,500 29,600 19.2%
Old stone 21 0.021 0.102 9.2% 19 12,100 26,200 22.2%
DGCS 164 0.032 0.137 3.6% 164 16,800 29,700 8.7%
ATPB 21 0.067 0.398 7.7% 21 64,700 110,500 26.6%
ACC I 75 0.190 0.483 9.1% 76 169,800 357,600 10.8%
ACC II 62 0.284 0.630 11.5% 62 188,600 347,500 13.8%
ACC III 76 0.517 0.844 14.0% 76 200,500 304,500 15.0%
ACC IS 83 0.256 0.536 10.4% 21 85,300 191,200 20.2%
ACC IIS 102 0.184 0.504 7.2% 40 44,100 140,600 10.0%
ACC IIIS 93 0.170 0.533 6.6% 65 213,100 322,500 16.4%
ACC IVS 35 0.223 0.570 13.4% 35 49,700 92,400 18.5%
In addition to the number of data points (N), the standard deviation, and the average, Table 3
includes the level of precision on the average at the 95% level of confidence. Put another way, the level of
precision ensures that if one were to use the average value for design, it would be reasonable to assume that
the value provided under conditions of actual performance would be within the precision indicated 95% of
the time.
CONCLUSIONS
The AASHTO guide describes a procedure for determining the effective SN provided by a pavement
structure from FWD deflection data. While Ioannides presented compelling justification for questioning the
theoretical purity of this concept, the success of its practical application as investigated by this research is
difficult to ignore.
When FWD testing is conducted during the April 15 through November 1 construction season, and
no drastic temperature and moisture fluctuations occur, the SNeff and resulting layer coefficient associated
with a particular component of a pavement structure appear to remain reasonably stable, even after additional
material is placed.
The stress distribution described by Noureldin and Al Dhalaan appears to provide a reasonably
accurate portrayal of the effective plate radius that develops below the surface of a pavement structure for an
applied circular load, without which the simulated layer coefficients would have been difficult to determine.
18
It is paramount to accurately and precisely determine the thickness of each material being evaluated.
Depending upon the material, any error in the thickness assessment can have a corresponding error in the
layer coefficient determination, e.g., a 25% thickness error may lead to a 25% error in the layer coefficient
determination. While this magnitude of error is not desirable in any of the materials, it can certainly have
alarming consequences with the stiffer and thinner ACC materials.
The layer coefficients determined for the unbound materials appear reasonable, while the ACC layer
coefficients are outside the range typical for the AASHTO procedure. However, there does appear to be
substantiation for these higher ACC layer coefficients from other material properties, namely the Marshall
stabilities and backcalculated resilient moduli. Further, all the layer coefficients determined by the method
developed under this investigation are reasonably accurate estimates of the in situ behavior simulated by
elastic layer theory. Indeed, such high correlation between these two different procedures would be highly
unlikely, considering the variables that lead to their development.
Whether by serendipity or by design, the development of the AASHTO effective SN procedure
provides designers with a very powerful tool for the determination of layer coefficients.
RECOMMENDATIONS
Considering the emphasis that will be placed upon mechanistic design in the next version of the AASHTO
to
calibrate the AASHTO pavement design model to Vermont materials and the conclusion to that effort as
conjunction with the current AASHTO pavement design model. The Committee considered the 85 th
–
percentile for ACC layer coefficients to ensure reasonableness of designs provided by the model.
Any follow up research should focus on supplementing the database for the mechanistic properties
thus far established. Work should continue on the resilient modulus for all unbound materials and the
pavement design guide for ACC materials.
19
Table 4 Recommended Material Properties for Design Using the AASHTO Model
Material
Layer Coefficient Resilient Modulus (psi)
N Standard
deviation Average Rec. N
Standard
deviation Average Rec.
Sand 139 0.013 0.078 0.078 139 10,200 19,100 19,100
Gravel 21 0.033 0.134 0.134 21 12,500 29,600 29,600
Old stone 21 0.021 0.102 0.102 19 12,100 26,200 26,200
DGCS 164 0.032 0.137 0.137 164 16,800 29,700 29,700
ATPB 21 0.067 0.398 0.331 21 64,700 110,500 110,500
ACC I 75 0.190 0.483 0.293* 76 169,800 357,600 357,600
ACC II 62 0.284 0.630 0.346 62 188,600 347,500 347,500
ACC III 76 0.517 0.844 0.327 76 200,500 304,500 304,500
ACC IS 83 0.256 0.536 0.280* 21 85,300 191,200 191,200
ACC IIS 102 0.184 0.504 0.320 40 44,100 140,600 140,600
ACC IIIS 93 0.170 0.533 0.363 65 213,100 322,500 322,500
ACC IVS 35 0.223 0.570 0.347 35 49,700 92,400 **
* If an ATPB is used, the layer coefficient for the base course (either ACC I or ACC IS) should be
increased to at least the 0.331 used for the ATPB.
** At this time, there is no recommendation for the ACC IVS resilient modulus.
ACKNOWLEDGEMENTS
This research would not have been possible without the persistent hard work of Duane Stevens and Jim
Pavement Design Committee, particularly Chris Benda, Jim Bush, Mike Hedges, Alec Portalupi, and Roger
Lyon-Surrey, for advice and review of the findings.
20
REFERENCES
(1) AASHTO Guide for Design of Pavement Structures. American Association of State Highway and
Transportation Officials, Washington, D.C., 1993.
(2) Zhou, H., G.R. Rada, and G.E. Elkins. Investigation of Backcalculated Moduli Using Deflections
Obtained at Various Locations in a Pavement Structure. In Transportation Research Record 1570, TRB,
National Research Council, Washington, D.C., 1997, pp. 96-107.
(3) Hossain, M., A. Habib, and T.M. LaTorella. Structural Layer Coefficients of Crumb Rubber-Modified
Asphalt Concrete Mixtures. In Transportation Research Record 1583, TRB, National Research Council,
Washington, D.C., 1997, pp. 62-70.
(4) Janoo, V.C. Layer Coefficients for NH DOT Pavement Materials. Special Report 94-30, U.S. Army
Corps of Engineers, Cold Regions Research & Engineering Laboratory, September, 1994.
(5) Ioannides, Anastasios M. Theoretical Implications of the AASHTO 1986 Nondestructive Testing
Method 2 for Pavement Evaluation. In Transportation Research Record 1307, TRB, National Research
Council, Washington, D.C., 1991, pp. 211-220.
(6) Noureldin, A.S. and M.A. Al Dhalaan. Establishment of Some Structural Parameters to Pavement
Evaluation Using the Falling Weight Deflectometer. A presentation given at the 70 th
TRB Annual Meeting,
Washington, D.C., January 1991.
(7) Ullidtz, P. Modeling Flexible Pavement Response and Performance. Polyteknisk Forlag, Denmark,
1998.
(8) Chitty, Daniel E., Blouin, Scott E., Quenneville, Steven R., and Beckwith, Daniel B. Laboratory Tests
and Analysis: Resilient Modulus and Low Strain Rate Modulus Testing of Sands. ARA Report Number
4835-2, Applied Research Associates, Inc., South Royalton, Vermont, June, 2001.
(9) Janoo, Vincent C. and Bayer II, John J. The Effect of Aggregate Angularity on Base Course
Performance. Technical Report ERDC/CRREL TR-01-14, U.S. Army Corps of Engineers, Cold Regions
Research & Engineering Laboratory, September, 2001.
What Students Are Saying About Us
.......... Customer ID: 12*** | Rating: ⭐⭐⭐⭐⭐"Honestly, I was afraid to send my paper to you, but splendidwritings.com proved they are a trustworthy service. My essay was done in less than a day, and I received a brilliant piece. I didn’t even believe it was my essay at first 🙂 Great job, thank you!"
.......... Customer ID: 14***| Rating: ⭐⭐⭐⭐⭐
"The company has some nice prices and good content. I ordered a term paper here and got a very good one. I'll keep ordering from this website."
