Prediction of Reflective Cracking in HMA Overlays - Wavelength Lytton 2010

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 27

Prediction of Reflection Cracking in Hot Mix Asphalt Overlays

Fang-Ling Tsai (Corresponding Author)


Graduate Research Assistant
Zachry Department of Civil Engineering
Texas A&M University
College Station, TX 77843
Tel: 979-845-9914
Fax: 979-845-0278
Email: lynne5280@tamu.edu

Robert L. Lytton, Ph.D., P.E.


Professor, Fred J. Benson Chair
Zachry Department of Civil Engineering
Texas Transportation Institute
Texas A&M University
College Station, TX 77843
Tel: 979-845-9964
Fax: 979-845-0278
Email: r-lytton@tamu.edu

Sang-Ick Lee
Graduate Research Assistant
Zachry Department of Civil Engineering
Texas A&M University
College Station, TX 77843
Tel: 979-845-7511
Fax: 979-845-0278
Email: s-lee@ttimail.tamu.edu

Word Count: 4974 Words + 4 Tables + 11 Figures = 8,724

Resubmitted Date: November 15, 2009

TRB 2010 Annual Meeting CD-ROM Paper revised from original submittal.
Tsai et al. 1

ABSTRACT
Reflection cracking is one of the main distresses in hot-mix asphalt (HMA) overlays. It has been
a serious concern since early in the 20th century. Since then, several models have been developed
to predict the extent and severity of reflection cracking in HMA overlays. However, only limited
research has been performed to evaluate and calibrate these models. In this paper, mechanistic-
based models are calibrated to field the data of over 400 overlay test sections to produce a design
process for predicting reflection cracks. Three cracking mechanisms: bending, shearing traffic
stresses, and thermal stress are taken into account to evaluate the rate of growth of the three
increasing levels of distress severity: low, medium, and high. The cumulative damage done by
all three cracking mechanisms is used to predict the number of days for the reflection crack to
reach the surface of the overlay. The result of this calculation is calibrated to the observed field
data (severity and extent) which has been fitted with an s-shaped curve.
In the mechanistic computations, material properties and fracture-related stress intensity
factors are generated using efficient Artificial Neural Network (ANN) algorithms. In the bending
and shearing traffic stress models, the traffic was represented by axle load spectra. In the thermal
stress model, a recently developed temperature model was used to predict the temperature at the
crack tips. This process was developed to analyze various overlay structures. HMA overlay over
either asphalt pavement or jointed concrete pavement in all four major climatic zones are
discussed in this paper. The results of this calculated mechanistic approach showed its ability to
efficiently reproduce field observations of the growth extent and severity of reflection cracking.
The most important contribution to crack growth was found to be thermal stress. The computer
running time for a twenty-year prediction of a typical overlay was between one and four minutes.

TRB 2010 Annual Meeting CD-ROM Paper revised from original submittal.
Tsai et al. 2

INTRODUCTION
Reflection cracking is one of the primary distresses in hot-mix asphalt (HMA) overlays for
flexible and rigid pavements. It permits the penetration of water and foreign objects into the
cracks and deteriorates the pavement structure and reduces the ride quality, thus shortening the
service life of the pavement. In order to improve the resistance of overlay to the reflection
cracking, it is necessary to analyze and predict the reflection cracking phenomenon in HMA
overlays. Several researchers have studied methods to predict reflection cracking (1, 2, 3). In
most cases, the researchers focused on only one of the three cracking mechanisms instead of
combining them as is done in this papers approach.
Reflection cracking occurs as existing cracks or joints in the overlaid pavement surface
layer grow through the overlay as illustrated in Figure 1. The crack growth is induced by bending
or shearing from passing traffic loads or by temperature changes. Every pass of a traffic load will
induce two peak shearing stresses and one bending peak stress in the HMA overlay (See Figure
1b). In addition, crack initiation and propagation is also influenced by the existing pavement
structure and conditions, reflection cracking countermeasures (e.g. reinforcing, interlayers),
HMA mixture properties, the degree of load transfer at joints and cracks, etc.. This paper
presents mechanistic-based models which were calibrated to field data to produce a design
process that is compatible with the Mechanistic-Empirical Design Program (4).

Bituminous
surfacing

Thermal crack Traffic crack


growth growth

Lean concrete Thermal expansion Traffic


roadbase and contraction movement

Sub-base

(a) Mechanisms of reflection cracking (after Nunn (5))

TRB 2010 Annual Meeting CD-ROM Paper revised from original submittal.
Tsai et al. 3

Tip of the crack

Overlay

Old surface course


Stresses at the tip of the

Bending
stress

B Void
crack

A
Shearing
stress

C
(b) Traffic mechanisms

FIGURE 1 Mechanisms of Reflection Cracking (5, 6).

The first stage of predicting reflection cracking is the identification and separation of the
three cracking mechanisms present in the overlay: bending, shearing and thermal stresses. Paris
fracture law (7) was used in this study to evaluate the length increase of the reflection cracks per
day when the overlays were subjected to the three mechanisms. Each mechanism was modeled
as a two-phase process. The first phase is when the crack grows from the bottom of the overlay
to Position 1, as illustrated in Figure 2, and all three mechanisms contribute to crack growth. In
the second phase, the crack grows from Position 1 to the overlay surface and bending produces
no additional crack growth due to a negative bending stress. Because the bending stress is
negative, the only causes of further crack growth are shearing and thermal stresses.

NfS2 NfT1

Overlay
NfB1 NfS1 NfT1
Position I

Bending Stress Shearing Stress Thermal Stress


FIGURE 2 Definitions of the Numbers of Days and Two-phase Crack Growth.

The flow charts shown in Figure 3a and 3b present the calculation processes for thermal
stress and traffic stresses due to bending and shearing. Figure 3a illustrates an algorithm used to
compute crack growth due to the thermal stress. The first step in the process consists in

TRB 2010 Annual Meeting CD-ROM Paper revised from original submittal.
Tsai et al. 4

calculating the pavement temperature at the current tip of the growing reflection crack. The
hourly pavement temperature is used to evaluate the properties of the binder and mixture and the
viscoelastic thermal stress. The model used to calculate the pavement temperature is presented
later in this paper. The daily increments of crack length are computed and summed each day to
check the total crack length against the thickness of the overlay and the numbers of days for the
crack to reach the overlay surface are reported. Similar processes were used for calculating
bending and shearing crack growth and are described in Figure 3b. As is done in the MEPDG
software, traffic load spectra were used to characterize the daily traffic loading. The method used
to present axle and tire loading spectra is described later in this paper. After completing the
calculation of the number of days for each mechanism to reach Position 1 and then to reach the
surface of the overlay, the set of five numbers of days (i.e., three numbers of days reach the
Position 1 due to bending, shearing, and thermal stresses, and two numbers of days to reach
the overlay surface due to shearing and thermal stresses) were used in calibrating all overlays of
the same structural type within the same climatic zone to the observed field distress data.

TRB 2010 Annual Meeting CD-ROM Paper revised from original submittal.
Tsai et al. 5

Weather Data

Hourly Solar Daily Wind Daily Air Emissivity Coefficient


Radiation Speed Temperature Absorption Coefficient
Albedo
a
d
Hourly Wind Hourly Air
Speed Temperature

Pavement Temperature (T) Binder Properties

1999 Model?
2006 Model?

1999 Model 2006 Model

Gradation Gradation
Volumetric Composition Volumetric Composition
Frequency (fc) Phase Angle (b)
Viscosity () Shear Modulus of Asphalt (G*)

No
Collocation E Relaxation Modulus at Crack Tip
Viscoelastic Thermal Stress (T)
inverse (Artificial Neural Network Model)

Thermal Stress Intensity Factor (SIF)


Fracture Properties A,n
(Artificial Neural Network Model)

Crack Growth
C=A[J]n N

Is C Overlay No. of Days


Yes
Thickness NfT1,NfT2

(a) Crack growth computation due to thermal

TRB 2010 Annual Meeting CD-ROM Paper revised from original submittal.
Tsai et al. 6

Weather Data

Hourly Solar Daily Wind Daily Air Emissivity Coefficient


Radiation Speed Temperature Absorption Coefficient
Albedo
a
d
Hourly Wind Hourly Air
Speed Temperature

Daily Traffic Vehicle


Pavement Temperature (T) Binder Properties
Class Distribution

1999 Model?
2006 Model?

1999 Model 2006 Model

Gradation Gradation
Each Vehicle Class Volumetric Composition Volumetric Composition
(Axle and Tire Loads) Frequency (fc) Phase Angle (b)
Viscosity () Shear Modulus of Asphalt (G*)
No

Bending (Shearing) Stress Intensity


Layer Relaxation Modulus
Factor (SIF)
(Artificial Neural Network Model)
(Artificial Neural Network Model)

Crack Growth
Fracture Properties A,n
C=A[J]n N

Is C Overlay No. of Days


Yes
Thickness? NfB/NfS1,NfS2

(b) Crack growth computations due to Bending (shearing)

FIGURE 3 Computation Processes of the Flow Charts for Reflection Cracking.

TRB 2010 Annual Meeting CD-ROM Paper revised from original submittal.
Tsai et al. 7

FRACTURE MECHANICS OF CRACK GROWTH IN HMA OVERLAYS


Paris law (7) is the key concept in fracture mechanics for modeling crack propagation.
Expressed in Equation 1, Paris law has been successfully applied to hot-mix asphalt by many
researchers for the analysis of experimental test data and the prediction of reflection- and low
temperature-cracking (1, 8).
dc
= A ( K ) ak (1)
n

dN
where:
c = the crack length;
N = the number of loading cycles;
A and n = fracture properties of the asphalt mixture (as shown in Equations 2 and 3);
K = the Stress Intensity Factor (SIF) amplitude.
The magnitude of which depends on the stress level, geometry of the pavement structure,
fracture mode (tensile, shearing or tearing), and current crack length. The healing factor takes
into account the reflection crack propagation for different traffic cases. It prolongs the number of
days required for the crack to reach the surface of the overlay based on the average time between
passing vehicles, t, as shown in Equation 4. Fracture property ak in Equation 5 is the
viscoelastic stress pulse effect. It is a function of the normalized load wave shape w(t) and is the
principal difference between elastic and viscoelastic fracture. The load wave shape is then raised
to the power n (Pariss law coefficient) and is integrated between the time limits of zero and t,
the time required for the axle group to pass a given point on a pavement. Since the load wave
shapes were different for different axle combinations, ak was calculated for single, tandem,
tridem, and quadrem axles in both the bending and shearing cases as shown in Figure 4.
However, ak=1 for the thermal case.
g
=n g0 + 1 (2)
mmix
g g
log A = g 2 + 3 log D1 + 4 log t (3)
mmix mmix
1 + g5 ( t )
SFhealing = (4)
g6

t
ak = w(t ) n dt (5)
0
where:
mmix = the slope of the graph of the relaxation modulus (Ei) vs. loading time (ti);
D1 = the coefficient in the master creep compliance power law equation;
t = undamaged tensile strength;
gi = fatigue calibration coefficients (i=1~6).
The fatigue calibration coefficients gi were developed in the SHRP A-003A project and
reported in the SHRP Report A-357 for four different climate zones (8).

TRB 2010 Annual Meeting CD-ROM Paper revised from original submittal.
Tsai et al. 8

4.0 ft

Overlay
Lj Lj

Old Surface Crack or Joint

2.0 ft 2.0 ft

5.0 ft Lj Lj 5.0 ft

W( t ) 0.82 0.82
Load 0.72 0.72
Wave
0.92 0.095 0.92
Shape
0.0 0.0

(14 + Lj) ft.

[W (t)]n (0.92)n (0.92)n

(0.82)n (0.82)n
(0.72)n (0.72)n
(0.095)n
0.0 0.0

(a) Load wave shape for bending crack propagation

TRB 2010 Annual Meeting CD-ROM Paper revised from original submittal.
Tsai et al. 9

Overlay
Lj
Old Surface Crack or Joint

Lj

5.0 ft Lj 5.0 ft

1.0

W( t )
Load
Wave
Shape

1.0

(10 + Lj) ft.

1.0

[W (t)]n

1.0

(b) Load wave shape for shearing crack propagation

FIGURE 4 Load Wave Shape for Tandem Axle in Bending and Shearing Crack
Propagation.

TRB 2010 Annual Meeting CD-ROM Paper revised from original submittal.
Tsai et al. 10

ESTIMATION OF THE STRESS INTENSITY FACTOR AND MATERIAL


PROPERTIES
In order to use Paris law for calculating crack propagation, the determination of the method to
compute the Stress Intensity Factor (SIF) became an important issue for the design program. The
speed and accuracy of computing SIF values became a very critical aspect of the crack
propagation analysis. Nearly 100,000 runs of a finite element program were made to calculate
the stress intensity factors at the tip of the reflection crack for a wide variety of crack lengths and
pavement structures. This data base of stress intensity factors was modeled with 18 separate
Artificial Neural Network algorithms: 6 for thermal, 6 for bending, and 6 for shearing stresses.
For the thermal load SIF calculation, a 2-D finite element program specifically for pavement
thermal SIF analysis was employed (9). For the traffic load SIF calculation, the finite element
model used a Fourier Series to represent the effects of loads acting at some lateral distance from
the 2-D plane where the calculations were made (10).
The method chosen to model the computed results from the finite element simulations
was the Artificial Neural Network (ANN) algorithm (11). The coefficient of determination (R2)
of all these models except one was above 0.99. Two examples of fitting an ANN model to the
computed SIF are given in Figure 5. The R2 for these cases are 0.9982 for an HMA overlay of
an existing asphalt concrete pavement (as Figure 5a) and 0.9998 for an HMA overlay of an
existing jointed concrete pavement (as Figure 5b). The results from both of these examples show
the ANN algorithm adequately modeling the results from finite element simulations.

(a) Case for hot mix asphalt overlays over cracked asphalt surface layer

TRB 2010 Annual Meeting CD-ROM Paper revised from original submittal.
Tsai et al. 11

(b) Case for hot mix asphalt overlays over jointed concrete surface layer

FIGURE 5 Comparison of Stress Intensity Factors (SIF) with Artificial Neural Network
Model Predictions.

Two other ANN programs were used to compute the complex modulus of the asphalt
mixture material in the overlay (11). The material properties of the overlay were calculated with
an ANN algorithm which replicates Witczaks 1999 and 2006 models of asphalt mixture moduli
(12, 13). The input requirement includes the binder properties, gradations of the aggregate and
volumetric composition of the mix. Figure 6 shows the comparisons with Witczaks model and
the ANN algorithm. The R2 for Witczaks 1999 regression model is 0.68 and for the ANN is
0.98 as shown in Figure 6a. Figure 6b shows the R2 for Witczaks 2006 model is 0.77 and for the
ANN is 0.96. These comparisons of ANN and Witczaks models show that the ANN algorithm
provided higher accuracy in calculating the complex modulus of the asphalt mixture.

TRB 2010 Annual Meeting CD-ROM Paper revised from original submittal.
Tsai et al. 12

(a) Witczaks 1999 model

(b) Witczaks 2006 model

FIGURE 6 Comparison of Witczaks Models with Artificial Neural Network


Algorithm (11).

TRB 2010 Annual Meeting CD-ROM Paper revised from original submittal.
Tsai et al. 13

ACCURATE METHOD OF ESTIMATING TRAFFIC EFFECTS


The number of axle loads can be determined using the traffic loads categorized based on
the Federal Highway Administration (FHWA) vehicle classification (classes 1 to 13)
(14), axle types (single, tandem, tridem, and quadrem), and the number of tires. The
estimated number of axle loads will be use to estimate their influence on reflection
cracking. The matrix of vehicle class and axle types can be categorized according to the
number of tires because all axles of vehicles in classes 4 and 5 and all single axles in
classes 6 and 7 have single tires while the others have dual tires. When the steering and
non-steering axles are combined together in a single axle type, the matrix can be
characterized into eight categories based on the vehicle class, axle types, and the number
of tires. The total number of axle loads for each category is used to further determine the
axle load distribution factor for the analysis of traffic load effects on reflection cracking.
Table 1 shows the categories of traffic loads as related to the vehicle class. The odd-
numbered categories have single tires and the even-numbered categories have dual tires.
The axle load distribution factor is defined as the percentage of the total axle
applications in each load interval by an axle type for a specific vehicle class (classes 4 to
13). The load ranges and intervals for each axle type are represented in Table 2. The
determination of the axle load distribution requires Weight-In-Motion (WIM) (15) data,
which is the number of axles measured within each axle load range by axle types of each
vehicle class. The normalized axle load distribution factors add up to 100 percent for
each axle type within each vehicle class.

TABLE 1 Vehicle Class Related to Axle and Tire Categories

Vehicle Class Single Axle Tandem Axle Tridem Axle Quad Axle

4
No. 3 Single Tire No. 5 No. 7
5
No. 1
6
7
8 Dual Tires
9
10
No. 2 No. 4 No. 6 No. 8
11
12
13
14

TRB 2010 Annual Meeting CD-ROM Paper revised from original submittal.
Tsai et al. 14

TABLE 2 Load Intervals for Each Axle Type

Axle Type Axle Load Interval


Single Axles 3,000 ~ 40,000 lb. at 1,000 lb. intervals
Tandem Axles 6,000 ~ 80,000 lb. at 2,000 lb. intervals
Tridem Axles
12,000 ~ 102,000 lb. at 3,000 lb. intervals
Quad Axles

ACCURATE METHOD TO ESTIMATE THERMAL EFFORTS


Pavement temperature is also an important element for predicting the reflection cracking
of an overlay. The one dimensional model developed by Han, Jin, and Glover (16) is
based on radiation and conduction energy balance fundamentals. This model considers
multiple sources of heat transfer at the pavement surface: solar radiation and reflection of
the solar radiation at the surface by a fraction ~ , the albedo, absorption of atmospheric
down-welling long-wave radiation on the pavement surface, emissivity by long-wave
radiation to the atmosphere, and convective heat transfer between the pavement surface
and the air close to the surface, which is enhanced by the wind. Below the surface, heat is
transferred by conduction. The heat transfer process is depicted in the schematic shown in
Figure 7. The inputs for this model include air temperature, solar radiation, and wind
speed which can mostly be collected from the LTPP data base (17), as well as albedo,
emissivity, and absorption coefficients. For accuracy, hourly air temperature and hourly
solar radiation were used in this model.

Outgoing longwave
Solar radiation radiation
Atmospheric downwelling
Heat convection
longwave radiaiton
by wind

Heat conduction
Pavement

FIGURE 7 Schematic Representation of Heat Transfer Model of Pavement.

TRB 2010 Annual Meeting CD-ROM Paper revised from original submittal.
Tsai et al. 15

UNIFORM METHOD OF CHARACTERIZING OBSERVED


REFLECTION CRACKIND DISTRESS
Jayawickrama and Lytton (1) proposed an S-shaped empirical model of reflection
cracking that described the severity of reflection cracking on the asphalt overlay to the
elapsed time since construction:



= 100 e
RFS DTotal
(6)
where:
RFS = reflection cracking severity, ranging from 0 to 100%;
DTotal = total number of days since the overlay was constructed;
and = calibration parameters.
The parameter is the scale factor for the reflection cracking severity and is
always equal to the number of days required to reach 36.8% (= 1/e) of the total amount of
expected reflection cracking. The parameter is a shape factor which describes the slope
of the s-curve as shown in Figure 8.
However, because the database used for calibration consisted of the length of
observed transverse cracks before and after overlay at the date when the survey was
performed, the variables of the original reflection cracking model were modified as
follows:



D ( N i )( % ) = e Ni
(7)
where:
D(Ni) = percent of the ratio of the current reflection crack length to the maximum crack
length;
i = ith crack observation;
Ni = number of days after overlay construction.
The percent of the total reflection crack length, D(Ni), at each observation was
calculated by dividing the observed length of the transverse crack after overlay
construction by the total length of the transverse cracks on an existing pavement surface
just before the construction of the overlay. Based on the field data obtained from test
sections, the parameters and were calibrated for three different severity levels: H and
H for high severity level, MH and MH for medium and high severity levels, and LMH
and LMH for low, medium, and high severity levels as shown in Figure 8.
A System Identification (SID) process was used to develop the and of the s-
shaped curve models. The observed reflection cracking data and the mathematical model
are identified when the error between them is minimized or satisfies the error criterion;
otherwise, the parameters of the mathematical model are adjusted until the error is
reduced sufficiently (18). An example of the field observed data and its relation to the
SID generated results for the three severity levels is shown in Figure 8. Table 3 shows the
generated values of and of the three severity levels by the SID model and coefficients
of determination (R2).

TRB 2010 Annual Meeting CD-ROM Paper revised from original submittal.
Tsai et al. 16

75

50
% Crack Length

100/e

25

LMH H
0 MH
0 2000 4000 6000 8000
No. of Days
LMH (Field) MH (Field) H (Field)
LMH (Pred.) MH (Pred.) H (Pred.)

FIGURE 8 Parameters in Reflection Cracking Severity Model.

TABLE 3 Calibrated Parameters from SID Process


Severity , days , days R2
Low-Medium-High 0.308 2130.56 0.868802
Medium-High 0.581 4583.75 0.841267
High 1.510 7799.83 0.991139

CALIBRATION METHOD AND RESULTS


Each computation of the mechanistic model (bending, shearing, and thermal) produced
five numbers of days, which are the numbers of days for a crack to reach position 1 with
bending, shearing, and thermal stresses, and the number of days for the needed for the
crack to travel from position 1 to the surface of the overlay with only shearing and
thermal stresses. These five numbers of days can be combined in several ways to model
the observed values of and (i.e., the scale and shape parameters of the model).
Equations 8, 9, and 10 are the models used in this study. Similar linear models were used
to model the shape parameters .
N N N
LMH = N fB1 0 + 1 fB1 + 2 fB1 + N fT 2 3 + 4 fT 2 (8)
N N N
fT 1 fS 1 fS 2
N N N
MH = N fB1 5 + 6 fB1 + 7 fB1 + N fT 2 8 + 9 fT 2 (9)
N N N
fT 1 fS 1 fS 2

TRB 2010 Annual Meeting CD-ROM Paper revised from original submittal.
Tsai et al. 17

N fB1 N fB1 N fT 2
H = N fB1 10 + 11 + 12 + N fT 2 13 + 14 (10)
N fT 1 N fS 1 N
fS 2
where:
, = scale and shape parameters of the model;
NfBi = number of days crack growth due to bending (i=1 at Position 1);
NfSi = numbers of days crack growth due to shearing (i=1 at Position 1, i=2 at the
surface of the overlay);
NfTi = numbers of days crack growth due to thermal (i=1 at Position 1, i=2 at the
surface of the overlay);
i = calibration coefficients.
Once these five numbers of days were identified from each mechanistic
computation, and the sets of and for low, medium, and high severity were obtained
from the System Identification Method, the calibration coefficients, i and i in equations
8, 9, and 10 were determined using a linear regression method. These 11 sets of
calibration coefficients were found based on the four different climate zones and
pavement structures. Table 4 shows a summary of the calibrated results for different
climate zones and pavement structures. Only the HMA overlay over existing asphalt
concrete (AC) model in a wet-no freeze climate zone has no observed high severity
reflection cracking data. The results of using the linear regression method are shown in
Figure 9 and Figure 10.

TRB 2010 Annual Meeting CD-ROM Paper revised from original submittal.
Tsai et al. 18

TABLE 4 Summary of Modeling Coefficients for Different Pavement Structures


and Climate Zones
Pavement Structures
0 1 2 3 4
(Old surface/Overlay)
AC/AC -27.45 -806.75 -1.39 1.70 0.12
WF
JRC(JPC)/AC 1534.41 -310.11 -310.11 13.77 4.58
LMH WNF AC/AC 2.71E+10 1.36E+10 1.36E+10 2.42 23.65
DF AC/AC -3.4E+10 -1.7E+10 -1.7E+10 -43.10 52.65
DNF AC/AC 1283.50 -59.20 -59.20 6.39 19.36
0 1 2 3 4
AC/AC 0.24 0.0046 -0.40334 0.00112 3.4E-05
WF
JRC(JPC)/AC 4.94 1.83 1.83 0.035 0.0731
LMH WNF AC/AC 4.32 2.45 2.45 0.00997 0.005
DF AC/AC 3568.9 -2.30 3570.20 0.0168 0.048
DNF AC/AC 1.03E+08 51317546 51317556 -0.0119 0.0096

0 1 2 3 4
AC/AC 461.98 1802.5 -5090.2 0.994 0.293
WF
JRC(JPC)/AC 1939.734 -715.43 -715.43 99.52726 302.123
MH WNF AC/AC 1.63E+10 1.63E+10 2276.096 45.702 12.44
DF AC/AC -1.5E+07 11038523 -2.6E+07 39.37872 -4.95955
DNF AC/AC 948.12 429.195 -814.488 0.1297 -0.0138
0 1 2 3 4
AC/AC 0.534 0.10467 0.0075 0.0071 0.00219
WF
JRC(JPC)/AC 3.1775 1.3415 1.3415 0.0316 -0.00028
MH WNF AC/AC 0.2981 3.6754 3.6754 0.06189 0.0241
DF AC/AC -241.469 -151.783 -92.808 -0.0201 0.04339
DNF AC/AC 14277102 7138252 7138848 -0.0007 -6.9E-06

0 1 2 3 4
AC/AC 9.35E+08 2.89E+09 -2E+09 3.720311 1.43812
WF
JRC(JPC)/AC -7.6E+09 -7.6E+09 -5707.83 -76.3259 -1127.66
H WNF AC/AC N/A N/A N/A N/A N/A
DF AC/AC 82.1351 3.49921 -2659.23 69.3191 66.7197
DNF AC/AC 722.887 -1.419 213.211 1.7771 0.0152
0 1 2 3 4
AC/AC -2.8114 -9.17575 4.75347 0.000672 4.76E-05
WF
JRC(JPC)/AC 1.11E+09 1.11E+09 3.6543 0.092 1.2978
H WNF AC/AC N/A N/A N/A N/A N/A
DF AC/AC 0.3211 -0.1337 -3.879 0.12684 0.12484
DNF AC/AC -0.7942 -0.4896 -0.03659 0.008545 -3.5E-05

TRB 2010 Annual Meeting CD-ROM Paper revised from original submittal.
Tsai et al. 19

Figure 9 and Figure 10 show the comparison between the and values
predicted using the mechanistically-computed numbers of days and the and values
that were fitted to the observed and normalized distress field data, respectively. The
first example in Figure 9a shows the observed data versus the predicted values of for an
HMA overlay over asphalt concrete pavement in a wet-freeze climate zone. Figure 9b
shows the observed versus the predicted value of for an HMA overlay over asphalt
concrete pavement in a wet-no freeze climate zone. Figure 9c shows Low, Medium, and
High severity level regression result for the coefficient for an HMA overlay over an
existing asphalt concrete pavement in a dry-freeze climate zone. The last example shows
the observed versus predicted value of for Medium and High severity levels of a HMA
overlay over an existing asphalt concrete pavement in the dry-no freeze climate zone and
is displayed in Figure 9d. The corresponding regression results for shape parameter ()
are shown in Figure 10. The results show that the majority of the predicted data correlates
well to the observed data. Some of the data shows a higher degree of scatter that may be
caused by an error during field data collection or from the field data model used in the
System Identification (SID) generation process.

10000

8000
_MH_Predicted

6000

4000

2000

0
0 2000 4000 6000 8000 10000
_MH_Observed
(a) Wet-freeze climate zone
10000

8000
_MH_Predicted

6000

4000

2000

0
0 2000 4000 6000 8000 10000
_MH_Observed
(b) Wet-no freeze climate zone

TRB 2010 Annual Meeting CD-ROM Paper revised from original submittal.
Tsai et al. 20

4000

_LMH_Predicted 3000

2000

1000

0
0 1000 2000 3000 4000
_LMH_Observed
(c) Dry-freeze climate zone
10000
8000
_MH_Predicted

6000
4000
2000
0
0 2000 4000 6000 8000 10000
_MH_Observed
(d) Dry-no freeze climate zone
FIGURE 9 Regression Results of Scale parameter for AC over AC Pavement in
Different Climate Zone.

10

8
_MH_Predicted

0
0 2 4 6 8 10
_MH_Observed

(a) Wet-freeze climate zone

TRB 2010 Annual Meeting CD-ROM Paper revised from original submittal.
Tsai et al. 21

15
12
_MH_Predicted 9
6
3
0
0 5 10 15
_MH_Observed
(b) Wet-no freeze climate zone

12
10
_LMH_Predicted

8
6
4
2
0
0 4 8 12
_LMH_Observed

(c) Dry-freeze climate zone


15

12
_MH_Predicted

0
0 5 10 15
_MH_Observed
(d) Dry-no freeze climate zone
FIGURE 10 Regression Results of Shape Parameter for AC over AC Pavement
in Different Climate Zone.

TRB 2010 Annual Meeting CD-ROM Paper revised from original submittal.
Tsai et al. 22

MODELING RESULTS COMPUTED WITH FIELD


OBSERVATIONS
The predicted reflection cracking results presented in this paper are based on the
calibration models and coefficients that were generated from field observed data. Figure
11a illustrates three different levels of severity of reflection cracking for asphalt concrete
overlays over a jointed reinforced concrete pavement in a wet-freeze climate zone
(Beaver, Pennsylvania). Pittsylvania, Virginia was taken as an example of an HMA
overlay over asphalt concrete pavement in a wet-no freeze climate zone in Figure 11b.
Low, medium, and high distresses appeared early, around 1000 days after overlay
construction, and medium and high severity levels began to appear around five years with
a sharply increasing rate. No high level of severity distress was observed. Figure 11c
shows the predicted extent and severity of transverse reflection of an HMA overlay over
asphalt concrete pavement in a dry-freeze climate zone (Deaf Smith County, Texas). In
this case, all three levels of severity appeared very early. The high level of severity
appeared within 80 days after overlay construction. Figure 11d illustrates the reflection
cracking extent and severity of an HMA overlay over a cracked asphalt concrete
pavement in a dry-no freeze climate zone (San Bernardino, California). All three
severities appeared after 5 years in this overlay. In addition, the medium and high
severity levels show a sharp increase after 4 years.

90
80
% Total Length of Cracks

70
60
50 L+M+H
40 M+H
30 H
20 100/e
10
0
0 1000 2000 3000 4000 5000 6000
No. of Days
(a) AC over JRC pavement at Beaver, Pennsylvania (WF)

TRB 2010 Annual Meeting CD-ROM Paper revised from original submittal.
Tsai et al. 23

120
% Total Length of Cracks 100

80
L+M+H
60
M+H
40 H (Not calibrated)
20 100/e

0
0 1000 2000 3000
No. of Days
(b) AC over AC pavement at Pittsylvania, Virginia (WNF)

120

100
% Total Length of Cracks

80

60 L+M+H
M+H
40 H
20 100/e

0
0 100 200 300 400 500 600
No. of Days

(c) AC over AC pavement at Deaf Smith County, Texas (DF)

TRB 2010 Annual Meeting CD-ROM Paper revised from original submittal.
Tsai et al. 24

100

% Total Length of Cracks 80

60
L+M+H
40 M+H
H
20 100/e

0
0 1000 2000 3000 4000 5000 6000
No. of Days
(d) AC over AC pavement in San Bernardino, California (DNF)

FIGURE 11 Total Percentage of Crack Length for Different Severity Levels.

CONCLUSIONS
This paper presents the development and computed results of a reflection cracking design
method to predict the reflection cracking of HMA overlays over asphalt concrete or
jointed concrete pavements. The results were calibrated to observed distress in over 400
pavement test sections in most of the States of the United States. This program was
developed based on a mechanistic model that predicts the reflection cracking lives of a
specified hot mix asphalt overlay due to bending and shearing traffic stresses and thermal
stresses. The relationship between the mechanistically computed service lives of an
asphalt overlay and the appearance, extent, and severity of observed reflection cracking
in the field depends upon the characteristics of the pavement structure, overlay structure,
and of the traffic and climatic conditions at the project location.
The results of using the mechanistic-based model display its ability reproduce
field observations of the increase in the extent and severity of reflection cracking quickly
and efficiently. In general, the program completes a twenty-year reflection cracking
simulation in one to four minutes.
In this paper, HMA overlays of two different pavement structures were presented:
HMA overlays over asphalt and jointed concrete pavements. Several other sets of
calibrated mechanistic model coefficients were developed, but are not presented here
because of space limitations. The other models included asphalt overlays with strain-
absorbing membrane and reinforcing interlayers. Modeling the reflection cracking
performance of these types of overlays presents unique technical challenges and merits a
separate presentation. The methods to develop the sets of calibration coefficients for

TRB 2010 Annual Meeting CD-ROM Paper revised from original submittal.
Tsai et al. 25

overlays in this study can be used efficiently by interested agencies to develop their own
sets of calibration coefficients using their own overlay data.

ACKNOWLEDGEMENTS
The authors want to acknowledge the funding support provided by NCHRP Project 1-41.
The contributions by the Senior Program Officer for the NCHRP, Dr. Amir Hanna, were
instrumental in guiding this project. The research was conducted by the Texas
Transportation Institute of Texas A&M University. Dr. Robert L. Lytton, Professor of
Civil Engineering was the principal investigator and the co-principal investigator was
Fujie Zhou, Research Engineer. The Finite Element program which generated the stress
intensity factors was written by Sheng Hu and his research group. The Artificial Neural
Network models were provided by Dr. Halil Ceylan of Iowa State University. Dr. Charles
J. Glover and his research group in Texas A&M University developed the program to
predict pavement temperature. The viscoelastic thermal stress prediction program was
provided by Dr. Reynaldo Roque of the University of Florida in Gainesville. The authors
deeply appreciate the collaboration provided by these research groups.

REFERENCES
1. Jayawickrama, P.W., and R.L., Lytton. Methodology for Predicting Asphalt
Concrete Overlay Life Against Reflection Cracking, Proceedings 6th International
Conference on Structural Design of Asphalt Pavements, Volume I, 1987, pp. 912-
924.
2. Mallela, J. and H. L. V. Quintus. Investigation of Reflective Crack Control
Measures for Ac Overlays of PCC Pavements. F. Report. New York, Department of
Design and Construction, 2004.
3. Button, J. W. and A. Chowdhury. Field Tests Using Geosynthetics in Flexible and
Rigid Pavements to Reduce Reflection Cracking. FHWA/TX-06/0-1777-2, Texas
Transportation Institute, College Station. 2006.
4. NCHRP 1-37A Mechanistic Empirical Design of New and Rehabilitated Pavement
Structures. http://www.trb.org/mepdg/guide.htm, 2007.
5. Nunn, M. E. An investigation of reflection cracking in composite pavements in the
United Kingdom, Proceedings of 1st international RILEM conference on Reflective
Cracking in Pavements,1989, pp. 146-153.
6. Lytton, R.L. Use of Geotextiles for Reinforcement and Strain Relief in Asphaltic
Concrete. Geotextiles and Geomembranes, Vol. 8, 1989, pp. 217-237.
7. Paris, P. C., and E. Erdogan. A Critical Analysis of Crack Propagation Laws,
Journal of Basic Engineering, Transaction of the American Society of Mechanical
Engineering, Series D., Vol. 85, 1963, pp. 528-534.
8. Lytton, R. L., J. Uzan, E. G. Fernando, R. Roque, D. Hiltunen, and S. M. Stoffels.
Development and Validation of Performance Prediction Models and Specifications
for Asphalt Binders and Paving Mixes. SHRP A-357, National Research Council,
TRB, Washington, D.C., 1993.

TRB 2010 Annual Meeting CD-ROM Paper revised from original submittal.
Tsai et al. 26

9. Hu, S., F. Zhou, and T. Scullion. Thermal Reflective Crack Propagation,


International Conference on Advanced Characterization of Pavement and Soil
Engineering Materials, Athens, Greece, Vol 2, June 2007, pp.1173-1182.
10. Hu, S., X. Hu, F. Zhou, and L. F. Walubita. SA-CrackPro: New Finite Element
Analysis Tool for Pavement Crack Propagation, Transportation Research Record
2068, TRB, Washington, D.C., 2008, pp. 10-19.
11. Ceylan H., K. Gopalakrishnan, and S. Kim. Looking to the future: the next-
generation hot mix asphalt dynamic modulus prediction models, International
Journal of Pavement Engineering, 2009, pp. 1-12.
12. Andrei, D., M. Witczak, and M. W. Mirza. Development of a revised predictive
model for the dynamic (Complex) modulus of asphalt mixtures. NCHRP 1-37A Inter
Team Report, University of Maryland, 1999.
13. Bari, J. and M. Witczak. Development of a new revised version of the Witczak E*
predictive model for hot mix asphalt mixtures. Journal of the Association of Asphalt
Paving Technologists, 75, 2006, pp. 381-423.
14. FHWA. Traffic Monitoring Guide. FHWA-PL-01-021, US Dept. of Transportation,
Federal Highway Administration, Office of Highway Information Management,
2001.
15. NCHRP. Guide for Mechanistic-Empirical Design of New and Rehabilitated
Pavement Structures; Part 2. Design Inputs, Chapter 4. Traffic, National
Cooperative Highway Research Program, 2004.
16. Han, R., X. Jin, and C. J. Glover. Modeling of Pavement Temperature History for
Pavement Performance Prediction, submitted for presentation and publication to
the 88th Annual Meeting of the Transportation board, Washington, D.C., 2009.
17. Long Term Pavement Performance, LTPP Data base, http://www.ltpp-
products.com/index.asp.
18. Natke, H. G., Identification of Vibrating Structures, Springer-Verlag, New
York, 1982.

TRB 2010 Annual Meeting CD-ROM Paper revised from original submittal.

You might also like