Review of Laboratory Asphalt Fatigue Prediction Models

Kiplagat Chelelgo1, Kaluli, J.W.2, Gariy, Z.A.3 and Mutuku, R.N. N.4

Kiplagat Chelelgo1,
PhD Student (Civil Engineering),
Pan African University Institute for
Basic Sciences, Technology and Innovation,
P.O. Box 62000-00200, Nairobi - Kenya,
Mobile Numbers: +254 716 217 856,
Email: kkchelelgo@yahoo.com

Kaluli James Wambua2,

Professor of Water and Environmental Engineering,
Department of Biomechanical and Environmental Engineering,
Jomo Kenyatta University of Agriculture and Technology (JKUAT),
P.O. Box 62000-00200, Nairobi - Kenya,
Mobile Numbers: +254 722 560 231/ +254 735 556 019,
Email: wambuak@gmail.com

Gariy Zachary C Abiero3,

Professor of Transportation Engineering,
Department of Civil, Construction and Environmental Engineering,
Jomo Kenyatta University of Agriculture and Technology (JKUAT),
P.O. Box 62000-00200, Nairobi - Kenya,
Mobile Numbers: +254 722 883 077,
Email: zagariy@yahoo.co.uk

Mutuku Raphael Ngumbau Ndisya4,

Professor of Structural Engineering,
Department of Civil, Construction and Environmental Engineering,
Jomo Kenyatta University of Agriculture and Technology (JKUAT),
P.O. Box 62000-00200, Nairobi - Kenya,
Mobile Numbers: +254 733 751 376/ +254 721 980 460
Email: rnmutuku@yahoo.com

Review of Laboratory Asphalt Fatigue Prediction Models

Abstract

Fatigue in asphalt pavements is the incremental damage that asphalt materials suffer as they receive multiple

variable loadings that do not necessarily exceed their yield strength. Accurate prediction of asphalt fatigue

comes in handy in the development of asphalt materials with high fatigue life as well as in the estimation of

fatigue life of asphaltic pavement materials. This review paper looks at the phenomenological models,

dissipated energy models and fracture mechanics models, which are the three models that have been used over

the years to model asphalt fatigue behaviour in the laboratory. The phenomenological models relate the

numbers of load cycles applied on an asphalt sample and the resultant asphalt tensile strains. The approach has

been found to be deficient in tracking crack initiation and progression. The dissipated energy models on the

other hand are based on the premise that work is done whenever a material is loaded. The models have been

successfully used to track crack initiation and propagation and to isolate dissipated energy that goes into

fatigue from that which performs mechanical work. Despite that success, the dissipated energy coefficients

have been noted to change with the mode and frequency of loading as well as the temperature. The third

approach, fracture mechanics models, is based on the premise that asphaltic materials have inherent

discontinuities that form the basis of crack initiation. Fracture mechanics models predict asphalt fatigue using

either linear fracture mechanics approach or non-linear fracture mechanics approach. The Linear fracture

mechanics method assumes that asphalt will always operate within the linear-elastic region while non-linear

fracture mechanics method assumes that it will at times be loaded beyond the linear elastic region. The non-

linear fracture mechanics approach comes across as the most promising method as it considers both the

inherent materials discontinuities and the loading regimes that are expected in the field. Both linear and non-

linear fracture mechanics approaches face challenges in finding specimen geometry that match laboratory

sample preparation procedures to field coring shapes. The three models discussed above are 2-dimensional

and do not consider seasonal and diurnal temperature dynamics in laboratory work. A 3-dimensional model

using non-linear elastic fracture mechanics approach is suggested as an alternative.

Key Words: Asphalt Fatigue, Dissipated Energy, Phenomenological Approach, Fracture Mechanics

2
1. Introduction

Asphalt concrete is a three-phase composite material composed of a coarse aggregates skeleton, a viscous-

elastic bituminous mortar and air in the voids between the aggregates. The bitumen mortar bind the course

aggregates together to produce a mix that can sufficiently resist rutting and fatigue failure, and stiff enough to

distribute applied surface loads to the sub-grade. Asphalt concrete is heterogeneous, in that, the constituent

aggregates are randomly sized, randomly shaped and randomly distributed within the bituminous matrix

(Arago et al, 2011). Existence of air voids and de-bonded interfaces between bitumen and aggregates

increases inhomogeneity and raise chances of appearance of discontinuities in the form of micro-cracks. These

discontinuities are the avenues through which various types of distresses in flexible pavements are formed.

The two major pavement distresses considered in analytical flexible pavement design are resistance to

permanent deformation (rutting) and resistance to fatigue failure (cracking). Besides fatigue cracks, asphalt

pavement experience other types of cracks such as reflective cracks, longitudinal cracks, block cracks,

slippage cracks, edge cracks, transverse cracks, thermal cracks and surface checking (Adlinge and Gupta,

2013). This paper looks at asphalt fatigue with a focus on models utilized in asphalt fatigue prediction.

1.1. Cracks in Asphalt Pavement

Cracks in asphalt pavements are caused by a combination of tensile and shearing stress concentrations in the

pavement layers. Once the tensile strength of asphalt is exceeded, cracks are initiated and propagated either

downwards or upwards depending on their position (Erkens and Moraal, 1996). These stresses could be due to

surface loads, thermal loads or moisture damage.

1.1.1.Reflective Cracks

As the name suggests, these cracks form in the underlying pavement layers and are reflected on the surface

above the position where a crack existed before rehabilitation. The underlying pavement layer could either be

a cracked flexible or rigid pavement that has received an asphalt overlay. When the cracked slabs move

relative to each other at the joints, stresses are induced at the bottom of the asphalt layer and cracks are

initiated. These cracks are then propagated upwards and appear at the surface as reflective cracks (FHWA,

2003). Some researchers have postulated that reflective cracks are initiated from the surface and propagated

downwards (Said et al, 2008) but laboratory tests results from the works of Scarpas et al (as cited by Said et

al, 2008) indicated that cracks initiated at the bottom travel faster than those initiated at the top. This

perspective is still a subject of research.

3
 Fig 1.1.1: Reflective Cracks (Baek, 2010)

1.1.2.Block Cracks

These appear on the pavement surface as a series of interconnected cracks that form roughly rectangular

patterns. They could be caused by cyclic pavement temperatures, weak sub-base support, sub-grade

settlement, insufficient surfacing thickness or reflective cracks. Appearance of block cracks is an indicator of

an aged asphalt pavement. Sometimes, lateral and longitudinal cracks intersect to form block cracks (FHWA,

2003).

Fig 1.1.2: Block Cracks (FHWA, 2003)

1.1.3.Longitudinal Cracks

These are cracks that run parallel to the center line of the road. These cracks occur in the direction of traffic

and usually at the edge of the wheel path. Discontinuous cracks along the wheel path are usually associated

4
with asphalt fatigue and point to the beginning of alligator cracking. These are associated with poor joint

construction, asphalt shrinkage or longitudinal segregation due paver operations.

Fig 1.1.3: Longitudinal Cracks (Adlinge and Gupta, 2013)

1.1.4.Transverse Cracks

These are also called Thermal Cracks. They are caused by thermal shrinkage of asphalt and they form at

regular intervals almost perpendicularly to the traffic direction. Like the longitudinal cracks, these cracks are

not load induced.

Fig 1.1.4: Transverse Cracks (Rababaah, 2005)

5
 1.1.5.Slippage Cracks

These are crescent-shaped cracks formed by turning, acceleration or breaking movements of traffic around

horizontal curves, junctions or hills. Horizontal forces generated by the traffic shove the mix in the forward

direction and form a series of horse-shoe shaped cracks. The cracks are related to poor quality surface asphalt

or de-bonding between the asphalt layer and the layers below.

Fig 1.1.5: Slippage Cracks (Adlinge and Gupta, 2013)

1.1.6.Edge Cracks

These are cracks that form close to the edges of asphalt pavements. They are common in roads with unpaved

shoulders and are an indicator of insufficient lateral edge support. It is possible that during construction, sub-

base or base materials at the point of failure didnt receive sufficient compaction. Infiltration through the

unsealed surface can weaken the sub-grade and thus compound the problem of settlement at the edges.

6
 Fig 1.1.6: Edge Cracks (FHWA, 2003)

1.1.7. Fatigue (Alligator Cracks)

Cracks encountered in asphalt pavements are predominately of this type. Just like block cracks, they are made

up of interconnecting longitudinal and lateral cracks that form some polygons. They form blocks that are

smaller and more irregular than those found in block cracks. Sometimes, they are called Alligator Cracks

due their resemblance to the skin of the alligator (Abojaradeh, 2013).Fatigue cracks are the ones considered in

analytical flexible pavement design due to their direct relationship with dynamic traffic loads. These loads are

believed to initiate tensile cracks in the wheel paths at the bottom of the asphalt layers. These cracks are then

propagated upwards to appear as alligator cracks at the pavement surface.

7
 Fig 1.1.7: Transverse Cracks (Abojaradeh, 2013)

1.1.8. Surface Cracks

These cracks are also called Roller Checks. Much emphasis has been placed by researchers on bottom-up

cracking but in the recent past, some researchers have shown that top-down cracking can have substantial

impact on pavement structural integrity. The most common surface cracks are those that develop during the

compaction process. They appear as short closely spaced shallow cracks that run perpendicular to the traffic

direction. Some scholars (Said et al, 2008) have researched extensively on this phenomenon and have blamed

the conventional steel drum roller on the appearance of the cracks. They have proposed a new type of roller,

Asphalt Multi-Integrated Roller (AMIR), to take the place of Pneumatic Tried Roller (PTR) which has

traditionally been used to clear surface cracks after compaction with a vibratory drum steel roller. AMIR was

subsequently fabricated in 1989 by Carleton University in conjunction with the National Research Council of

Canada (Said et al, 2008).

8
 Fig 1.1.8: Surface Cracks (Baek, 2010)

2. Materials and Methods

2.1. Phenomenological Methods

Phenomenological and energy method approaches employ bending beam fatigue tests in the laboratory to

predict asphalt fatigue behaviour. Bending beam fatigue tests can be conducted using four-point bending

(4PB), three-point bending (3PB), two-point bending (2PB) or trapezoidal fatigue loading arrangements in

accordance with ASTM D7460 and AASHTO T321 procedures (Rowe et al 2012). For a 4PB test, for

instance, asphalt sample of dimensions 300mmx50mmx50mm cut from 300mmx300mm slabs are used. The

set-up involves loading of the test sample with a sinusoidal or haversine load through a servo-hydraulic

actuator connected to a fatigue frame through a steel shaft. Linear Variable Differential Transducers (LVDTs)

are used to automatically measure vertical deflection of the beam as it is loaded. Vertical deflection at the

center of the beam is used to calculate both stresses and strains induced in the specimen. Logarithmic plots of

stiffness versus fatigue are automatically done by a desktop computer connected to the test set-up. Stress-

strain hysteresis loops are also automatically plotted by the desktop computer. Traditionally, fatigues failure

has been assumed to take place once stiffness of the sample drops by 50%

9
 Fig 2.1 (a): Four-Point Bending Fatigue Test Apparatuses IPC (Rowe et al, 2012)

Fig 2.1 (b): Geometry of the Four Point Bending Test (Erkens and Moraal, 1996)

2.2. Fracture Mechanics Methods

There are various tests that can be used in the laboratory to predict and determine fracture energy and critical

j-integral of asphaltic materials. Notched specimens are used to predict crack initiation while concepts of

fracture mechanics are employed in tracking crack propagation. The most common among laboratory

procedures include the Single-Edge notched Beam SE (B), Disk-shaped Compaction Tension test DC (T) and

10
Semi-Circular Bending test (SCB). They generally apply the same concept of initiating a crack and tracking

its progression.

2.2.1.Semi-Circular Bending test (SCB)

In this test method, 150mm diameter and 135mm thick gyratory specimens are prepared in the laboratory and

sliced into 25mm thick cylindrical samples. These samples are then cut into semi-circles that are notched to

25mm depth from the center of the flat face of the half-cylinder (Wang, & Yang, 2011). The geometry of

samples in this set-up makes it easier for both laboratory and field specimens to be tested. Bending is

conducted at three notch depths 25.4mm, 31.8mm and 38.0mm and the specimen is loaded at an average

deformation rate of 0.5 mm/min till failure occurs. Critical value of j-integral
Jc is determined from

the load-deformation curves plotted. The higher the value of

J c , the higher is the fracture toughness of the

sample (Mohammad et al, 2012). There are reports in literature that during the tests, the crack starts curving as

it approaches the top on the sample (Mir et al, 2012)

Fig 2.2.1: The Semi-Circular Bend [SC (B)] test (Mir et al, 2012)

11
 2.2.2.Disk-shaped Compaction Tension test DC (T)

This is the standard test adopted by ASTM under ASTMD 7313-07. The test uses specimens of 150mm

diameter. Due to its geometry, the sample can be easily fabricated in the laboratory or cored from the field.

The specimen is loaded in tension at a rate of 1 mm/min through loading holes in the specimen as shown

in Fig: 2.2.2 below. The challenge with this test is that it only works best at temperature below 10 . It

has also been reported that the fabrication process weakens the loading surfaces (Mir et al, 2012)

Fig 2.2.2: DC (T) Geometry and Experimental test set-ups (Kim, & Buttlar, 2009)

Fracture energy of a concrete sample in determined by plotting the applied force against Crack Mouth

Opening Displacement (CMOD). CMOD is preferred over Crack Tip Opening Displacement (CTOD) due to

the ease of measuring but in a study they conducted, (Kim and Buttlar, 2009) used measurement of

25 , as shown in the figure above, to determine CTOD. CMOD overestimates fracture

displacement at

energy since it measures both displacements due to fracture and machine compliance.

2.2.3.Single-Edge notched Beam SE (B) Test

This test was the first to be used when the concept of fracture mechanics in asphalt pavement was first

studied. In this set-up, a pre-notched beam specimen is simply supported on two rollers and loaded directly

through the line of the notch. Displacement of the notch (CMOD) is directly measured using a Linear Variable

Differential Transducer (LVDT) (Wang and Yang, 2011). The challenge with SE (B) is that samples of the

required geometry can neither be fabricated in the laboratory nor cored from the field pavement due to the

12
configuration of the existing fabrication and coring moulds. The specialized fabrication equipment required to

perform the test in the laboratory substantially increases the cost of this test (Wang and Yang, 2011).

Fig: 2.2.3: Single-Edge Notch Beam [SE (B)] (Wang, & Yang, 2011)

3. Results and Discussions

3.1. Fatigue Prediction Models in Asphalts

Fatigue can be defined as failure or accumulation of damage in a material subjected to repeated fluctuating

load of stress levels that do not necessarily exceed its yield strength. Fatigue resistance of an asphalt mix can

therefore be defined as its ability to resist a given number of repeated loads without significant accumulation

of damage (Pais et al, 2002). Fatigue in asphalt is believed to initiate with cracking of plastically damage

regions at the bottom of the specimen followed by propagation of a crack to the top of the specimen. It has

however been found that for thick pavement layers, stresses at the top generate surface cracks, unlike in

thinner pavements, where cracks are generated at the bottom (Pais et al, 2002).

13
 Fig 3.1: Stresses in Asphalt Pavement (Walubita, 2000)

Accurate and reliable prediction of asphalt fatigue is instrumental in the analytical pavement design process.

Development of laboratory test regimes and reliable transfer functions to correlate field and laboratory data is

vital. To achieve this, several approaches have been developed and applied over the years. These approaches

fall within three categories: phenomenological based, energy methods based and those that borrow from

fracture mechanics (Maggiore et al, 2012). To obtain useful information from the above methods, some

laboratory or field tests will need to be set-up and calibrated. This could involve laboratory simulation,

construction of actual test-tracks or study of a road in operation. Models can be obtained from processing

data with the aid of tools such as regression analysis, numerical methods, artificial neural networks, discrete

elements method and finite elements methods (Suh et al, 2010).

3.1.1. Phenomenological Approach

This method makes use of fatigue curves in studying fatigue properties of bitumen. It focuses on the

relationship between the number of load cycles applied on a sample and the resultant asphalt tensile strain.

Laboratory samples and conditions are selected on the basis of their ability to simulate actual on-site

conditions and behaviour. The samples should reflect the average properties of the material that will be used

in the field. The failure criterion is developed by subjecting the samples to load cycles up to failure and

plotting the results of tensile trains versus load cycles on a log-log scale Fig: 3.1.1, (Pais et al, 2002). The
14
data is fitted using a linear regression model that assumes logarithmic linearity between the strain and the

number of load applications. The assumption of logarithmic linearity in the fatigue curve is based on the

concept of Miners law of cumulative pavement damage (Chiangmai, 2010). Laboratory tests are usually

conducted with few but large strain levels to minimize the number of samples required.

Fig 3.1.1: General Fatigue Curve (Chiangmai, 2010)

Fatigue properties of an asphalt mix can be characterized by statistically analysis the relationship between the

number of load cycles applied and the tensile strain developed. Equation 3.1.1 is a general equation that can

be used to determine the fatigue life of a particular asphalt mix, whether in controlled-stress or controlled-

strain (Chiangmai, 2010).

k2
1
N f =k 1
( ) Equation 3.1.1
t ; t

Where,
N f = Fatigue life (cycles to failure)

t = Tensile strain

t = Tensile stress

15
 k 1= Intercept of the fatigue line

k 2= Slope of the fatigue line

k1 k2
To obtain the constants and , at least two fatigue tests conducted at two strain levels are used.

k 1 is largely dependent on the loading mode and the type of

Some researchers have noted that the value of

test adopted. This discrepancy, together with the fact that they cant track fatigue damage accumulation have

made phenomenological methods unpopular in modern researches (Molenaar, 2001). Fatigue life of the

sample at any point is represented by the number of load cycles corresponding to the strain value at that point.

3.1.2. Energy Methods

a) Cumulative Dissipated Energy

Energy methods are based on the premise that work is done whenever a material is loaded. A linear-elastic

material regains its original shape and size when the stressing load is withdrawn. On the contrary, a visco

-elastic material will exhibit some degree of deformation once the load is withdrawn. A plot of stress versus

strain for an elastic material will trace the same path but that of a visco-elastic material will exhibit a phase

difference between the stress and strain (Maggiore et al, 2012). As asphalt traces the new path upon

withdrawal of load, some energy is dissipated as heat, mechanical work and fatigue damage. The loading and

unloading paths enclose a hysteresis loop whose area is equivalent to the amount of work done in the process.

Energy methods equate fatigue damage to energy depletion from one cycle to the next (Chiangmai, 2010).

16
Fig: 3.1.2 (a): Visco-Elastic Behaviour (Chiangmai, 2010)

17
 Fig 3.1.2 (b): Stress-Strain Hysteresis Loop by Carpenter (Maggiore et al, 2012).)

If an asphalt sample is loaded with a sinusoidal dynamic load in flexure and we assume that it is linear-visco-

elastic, the area under the hysteresis loop represents the dissipated energy per cycle and the equation below

can be used to calculate its value.

W i= i i sin i 3.1 .2(a)

W i= Dissipated energy at load cycle i ,

Where;

i= Stress amplitude at load cycle i ,

i = Strain amplitude at load cycle i ,

and
i = phase angle between the stress and strain wave signals.

18
Fig 3.1.2 (c): Stress & Strain of Typical Viscoelastic Materials under Sinusoidal Loading (FHWA, 2015)

As loading of the sample continues, the sample starts developing micro-cracks and experiences some

reduction in stiffness. This reduction in stiffness leads to variation in the dissipated energy in every loading

cycle. For a strain-controlled test (haversine load), the dissipated energy will progressively decrease with

increasing load repetitions, while for stress-controlled test (sinusoidal load), the dissipated energy will

progressively increase with increase in load repetitions. Initial change in the size shape of progressive

hysteresis loops is taken to signify crack initiation.

Fig 3.1.2 (d): Dissipated Energy vs Loading (Maggiore et al, 2014)

19
The cumulative dissipated energy approach assumes that all the dissipated energy goes into causing fatigue

damage in asphalt. The method thus sums up energy dissipated in all cycles leading to failure. That basically

means summing up the areas of hysteresis loops of all the load cycles. The method was proposed and applied

by Van Dijk, who in 1972 determined a formula that relates dissipated energy to the number of cycles to

failure (Yuan et al, 2013).

W i=W f = A ( N f ) z Equation3.1 .2(b)

i=1

Where;

W f = The cumulative dissipated energy to failure

N f = Number of load cycles to failure

A , z = Coefficients dependent on mixture properties

The challenge with the dissipated energy approach is in its inability to capture the onset of crack initiation and

to distinguish energy dissipation due to damage and viscoelasticity. Values of the parameters computed using

Equation 3.1.2 (b) have been noted to change with the mode of loading, temperature and frequency (Maggiore

et al, 2014).

b) Dissipated Energy Ratio (DER)

Since the introduction of Dissipated Energy approaches by Van Dijk in the 1970s there have been positive

steps in refining them to address identified inefficiencies. Among the first proposals was the Dissipated

Energy Ratio which was later upgraded to Ratio of Dissipated Energy Change (RDEC). This was proposed by

Hopman & Pronk (Abojaradeh, 2013) to try to explain or locate crack initiation. Dissipated Energy Ratio is

given by the equation below.

nW o
DER= Equation3.1 .2(c )
Wi

W o = Energy dissipated in the first cycle

W i= Energy dissipated in the i th cycle

n= Load cycle

20
The peak value for the plots of energy ratio for different cycles versus the load cycles under controlled-stress

gave a peak value that researchers believed represented the number of load cycles at which cracks initiated.

This point of curvature can easily be determined through differentiating the curve equation and equating the

value to zero. For controlled-strain, crack initiation was considered to be the point at which the plots deviated

from the straight (Abojaradeh, 2013).

c) Ratio of Dissipated Energy Change (RDEC)

Rate of Dissipated Energy Ratio Change was proposed by Carpenter and Jansen (Cited by Abojaradeh, 2013).

This method was chosen over previous methods for the simple reason that it allowed the researches to get a

parameter that isolate dissipated energy that goes into fatigue from that which performs mechanical work or

that which goes into heat generation (Abojaradeh, 2013). It is based on the premise that relative change in

dissipation energy has direct bearing of damage accumulation. This change is measured from one cycle to the

next. Carpenter and Jansen developed the following relationship to define Ration of Dissipated Energy

Change

DE n+ 1DE n
RDEC= Equation3.1 .2(d)
DE n

Where;

RDEC = Average ratio of the dissipated energy change per load cycle,

DE n n
= Dissipated energy in cycle

DE n+1 = Dissipated energy in cycle n+1

In the study by Carpenter and Jansen, the change in dissipated energy was calculated from laboratory samples

every 100 cycles and the calculated value of RDEC plotted against the number of load cycles. Initially, a

downward trend was observed in the plots followed a long plateau and finally an upward trend. They

explained that the initial rapid drop in the value of RDEC was due to the specimens resisting deformation

being induced by the loads. The second phase was fairly stable indicating that fatigue damage was progressing

more uniformly. The value of RDEC at the level is called the (Plateau Vale-PV). Ghuzlan and Carpenter (Yuan

et al, 2013) estimated this value to correspond to 50% reduction in asphalt fatigue,
N f 50 . The final rapid

rise in the value of RDEC represents fatigue damage to the sample.

21
 Fig 3.1.2 (e).The Curve of the RDEC against the Number of Load Repetition (Fakhria et al, 2013)

The researchers went further to find a good correlation between the plateau values of several samples and the

corresponding load cycles to failure for each sample. When plotted on a log-log scale, a linear graph similar to

the traditional fatigue curve was obtained. The plots for both constant-stress and constant-strain bore striking

similarities.

22
 Fig 3.1.2 (f): Plateau Value of DER vs. Number of Load Cycles (Fakhria et al, 2013)

A new failure criterion for both controlled-strain and controlled-stress was developed and represented as the

equation below.

b
N f =C ( PV ) Equation 3.1.2( e)

Where;

Nf = Number of load cycles to failure

PV = The average value of RDEC in the second plateau region

C and b = Fitting parameters (dependent on the composition and properties of the AC mixture)

3.1.3.Fracture Mechanics Approach

Fracture mechanics concepts have traditionally been applied in fatigue design of steel structures but since the

mid-1970s, they have found useful application in asphalt pavement design (Wang, 1991). Its uniqueness from

previous methods lies in the consideration of discontinuities that exist in materials. The methods presented in

the earlier sections assume continuum theory holds for all materials but in actual fact, materials such as
23
asphalt have discontinuities called stress raisers in the form of micro-cracks and air voids. These naturally

occurring cracks in asphalt pavements will grow under cyclic fatigue loading to develop a fractured surface.

Fracture mechanics looks at energy expended in development and propagation of such cracks. Typically,

cracks develop and grow in three stages namely: initiation, propagation and unstable crack (Chiangmai,

2010).

To model fracture characteristics of a material, a sample of particular dimensions and properties is pre-cracked

and subjected to cyclic stress. The body of the sample will have materials at three states: macro crack state

(which in this case is the pre-cracking), process zone and zone of elastic behavior. The zone of elastic

behaviour represents that part of the sample far from the crack while the process zone is the plastic zone

around the tip of the crack.

Fig 3.1.3 (a): Cracking in Semi-Brittle Materials (Erkens & Moraal, 1996)

a) Linear Elastic Fracture Mechanics Approach

Fracture mechanics approach tracks material behaviour through the three stages of crack propagation. The

traditional approach borrows heavily from behaviour of homogenous materials, such as steel, that have

relatively smaller grains. The first step in crack development begins with a tiny process zone, the size of

material grain size. Crack propagation at this stage is largely dependent on the grain properties of the material.

In the second stage, the process zone is larger than the grain size but still smaller than the crack size, therefore,

Linear Elastic Fracture Mechanics approach can be used to track crack progression. The fracture zone in the

unstable crack stage is way bigger than the crack and stresses involved way larger. Here, the material begins

24
to undergo plastic failure and crack behaviour can only be analysed using Non-Linear Elastic Fracture

Mechanics (Wang, 1991). The figure below shows this progression phenomenon in cracks.

Fig 3.1.3 (b): Three Stages in Cracking (Wei Wang 1991)

To simplify the analysis of fracture behaviour of asphalt, it assumed to remain within stage II while in service.

In this stage, the area around the crack tip undergoes plastic deformation but the rest of the body of the sample

remains linear elastic. Linear elastic fracture behaviour can be described by Paris law, which relates rate of

crack growth to the number of load cycles and a stress intensity factor (Chiangmai, 2010).

dc
=A ( K )n Equation3.1.3 (a)
dN

Where;

c= Crack length

N= Number of load repetitions

K = Stress intensity factor during a loading cycle (Driving Force)

25
 A , n= Coefficients dependent on material properties

The driving force, ( K =K max K min ) , is a value that depends on the geometry of the sample, applied

Kc
force and the flaw size. The resistance force on the other hand is a material property determined by

subjecting the sample to some laboratory tests. When the driving force exceeds the resistance force, unstable

crack growth fracture occurs (John and Stanley, 1999). The differential equation 3.1.3 (a) can be solved to

obtain a value for N and the critical flaw length that the material can withstand at particular load cycles,

thus:

c1
c1 h
dc n 1 1 c
=A ( K ) N = dc=h d Equation3.1 .3(b)
dN c A ( K )
o
n
c A ( K )o
n
h
h

Where;

c o= Crack length at the beginning of the test

c 1= Crack length at the end of the test

h= Height of the beam

A , n= Coefficients dependent on material properties

b) Non Linear Elastic Fracture Mechanics Approach

The assumption of elastic linearity of asphalt material over all loading regimes in Linear Elastic Fracture

Mechanics has been criticized as being inaccurate. It is appreciated that asphalt, within some temperature and

speed ranges, behaves in a visco-elasto-plastic manner. The aggregates in the skeleton tend to shift as the

number of loads increase and while doing so, they interlock leaving the material with some residual fatigue

strength. This therefore means that large plastic zone, represented by stage III in Fig: 3.1.3 is not completely

unable to resist some loads. Elastic-Plastic Fracture Mechanics or Non Linear Fracture Mechanics approach

therefore attempts to bridge this gap. Elastic-Plastic Fracture mechanics introduces two parameters, j-contour

26
 Gf ). J-integral characterize energy release rate for non-linear
integral (J-Integral) and fracture energy (

elastic materials and is determined as the plastic strain field of elasto-plastic materials. For linear materials, J-

Gf
integral is numerically equal to (Chiangmai, 2010). J-contour integral is a path independent line or

surface integral that encloses the cracked surface and characterizes energy release rate for a non-linear

material. It represents change in mechanical energy per unit area of the newly cracked surface. It can be

determined in the laboratory as the area under a load displacement curve for an elastic-plastic material. This

integral is represented by the basic equation below (Abojaradeh, 2013).

[
J = U d ( , ) dT

u
x ]
dx Equation3.1.3 (c)

Where;

= The curve surrounding the crack tip

T = The component of traction vector normal to the contour

U= The strain energy density

u= The displacement vector

v = Direction of the crack

y= Direction normal to the crack line

dx= The differential element representing incremental length along curve

= Stress (stress is monotonically increasing with the strain)

= Strain

In general, j-integral can be represented as rate of change of energy. For visco-elastic materials such as asphalt

concrete, this rate of change represents the variation in the development of j-integral. The equation below

gives the relationship between j-integral and rate of change of energy release

27
 1
J= ( )

B a A
Equation3.1 .3( d)

Where;

=( UF) Potential energy of the cracked body

U= Stored strain energy

F= Work done by the external force

B= Width or thickness of the sample

a= Crack length

A= Cracked area

J c , is achieved. Critical j-integral is a material

A crack starts propagating once a critical value of j-integral,

fracture property that can be found in the laboratory by subjecting a sample to some loads and measuring the

resultant displacements. The rate of crack growth is related to the rate of change of j-integral; therefore, crack

propagation life can be determined using the relationship below (Chiangmai, 2010). The other parameter in

Gf . This refers to the amount of work

elastic-plastic fracture mechanics is fracture energy, denoted as

expended in creating a new fracture surface of unit length of area. In the laboratory, this parameter is

determined by carrying out Crack Mouth Opening Displacement (CMOD) test and computing the area under

the load-displacement plot as fracture energy. The equation below is used in this computation.

Af
Gf = Equation 3.1 .3(e)
BL

Where;

Gf = Fracture energy (J/m2),

Af = Area under Load-CMOD curve (kNmm),

B = Thickness of specimen (mm),

28
 L = Ligament or crack length (mm)

4. Conclusion and Recommendation

This review set out to critically look at the methods that have been used by engineers to predict fatigue

cracking in asphalt over the years. The methods evolved from the phenomenological to energy methods and

finally, fracture mechanics approach. The methods adopted by various researchers and agencies are

differentiated either by the laboratory procedures adopted or the method of data analysis used. From published

studies of a number of researches, it was found out that:

1. Phenomenological approaches underestimate the fatigue resistance of asphalt and cant effectively

track fatigue damage accumulation.

2. Among the energy methods, Ratio of Dissipated Energy Change serves as a better capture of fracture

damage since it separates energy consumed in crack formation from that dissipated through heat and

noise
3. Among the fracture mechanics approaches, Non-Linear Elastic Fracture Mechanics approach captures

the behaviour of asphalt better than Linear Elastic Fracture approach. Theres need though to

approach the analysis from a 3-dimensional perspective.

The research recommends that further research on the effect of variable amplitude laboratory load on fatigue

behaviour be considered. Theres also need to consider the likely effects asphalt healing in the rest periods

during laboratory testing.

Acknowledgements

I would like to acknowledge and appreciate the invaluable input, advice and guidance given by my course

supervisors Prof. Zachary Gariy, Prof. James Wambua and Prof. Raphael Mutuku. I would also like to

appreciate all the knowledge we freely shared with my Civil Engineering PhD classmates at Pan African

University of Science, Technology and Innovation. I am highly indebted to Pan African University for

awarding me a scholarship that made this work possible. The academic and no-academic staff of Pan African

University accorded me the most conducive environment for conducting the research that led to this paper.

List of Reference

Abojaradeh, M. (2013). Development of Fatigue Failure Criterion for Hot-Mix Asphalt Based on Dissipated

Energy and Stiffness Ratio, Jordan Journal of Civil Engineering, 7(1), Issue 1, pp. 54-69.

Adlinge, S.S., and Gupta, A.K. (2013). Pavement Deterioration and its Causes. Journal of Mechanical & Civil

Engineering, 6, pp. 9-15.

29
Ahmad, A. A., and Jung, S. J. (2013). Evaluation of Prediction Models of Fatigue for Asphalt Mixes.

International Journal of Pavement Research and Technology, 6(6), pp. 696-703.

Arago, F.T., Kim, Lee, J., Allen, D.H. (2011). Micromechanical Model for Heterogeneous Asphalt

Concrete Mixtures Subjected to Fracture Failure. Journal of Materials in Civil Engineering (ASCE), 23, pp.

30-38

Asphalt- The Magazine of the Asphalt Institute, (2016). Understanding Asphalt Pavement Distresses.

http://asphaltmagazine.com/understanding- asphalt-pavement-distresses-five-distresses-explained/

[Accessed on 26 June, 2016]

Chiangmai, C.N. (2010). Fatigue-Fracture Relation on Asphalt Concrete Mixtures, Doctoral Thesis,

University of Illinois at Urbana-Champaign.

Baek, J. (2010). Modeling Reflective Cracking Development in Hot-Mix Asphalt Overlays and

Quantification of Control Techniques, Doctoral Thesis, and University of Illinois at Urbana-Champaign.

Erkens, S.M.J.G., and Moraal J. (1996). Cracking in Asphalt, Road Research Laboratory, Faculty of

Civil Engineering, Delft University of Technology, HERON, 41(1), pp. 53-70

Fakhria, M., Hassania, K., and Ali Reza Ghanizadeha, A.R. (2013). Impact of Loading Frequency on the
Fatigue Behavior of SBS Modified Asphalt Mixtures Mansour, Procedia - Social and Behavioral Sciences,
104, pp. 69 78

Federal Highway Administration (FHWA), (2003). Pavement Distress Identification Manual for the NPS

Road Inventory Program Cycle 4, 2006-2009, Publication No. FHWA-RD 03- 031

Federal Highway Administration (FHWA), (2015). Premature Asphalt Pavement Cracking, Final Report SPR

734

John, M. B., and Stanley T. R. (1999). Fracture and Fatigue Control in Structures: Applications of Fracture

Mechanics. 3rd Edition. Philadelphia: PA.

Kim, H., and Buttlar, W.G. (2009). Discrete Fracture Modeling of Asphalt Concrete, International

Journal of Solids and Structures, 46, pp. 25932604.

30
Maggiore, C., Airey, G., and Marsac, P. (2014). A Dissipated Energy Comparison to Evaluate Fatigue

Resistance Using 2-Point Bending, Journal of Traffic and Transportation Engineering (English Ed.), 1(1), pp.

49-54

Maggiore, C., Grenfell, J., Airey, G., and Collop, A.C. (2012). Evaluation of Fatigue Life Using Dissipated

Energy Methods. 7th RILEM International Conference on Cracking in Pavements, Mechanisms, Modeling,

Testing, Detection, Prevention and Case Histories, Delft, Netherlands, pp. 643-652.

Minnesota Asphalt Pavement Association, (2014). Slippage Cracks, http://www.asphaltisbest.com/wp-

content/uploads/2014/02/Slippage-Cracks.pdf [Accessed on 26 June, 2016]

Mir, R., Martnez, A. , Prez-Jimnez F., Botella, R., and Valds G. (2012) Assessment of Cracking

Resistance of Bituminous Mixtures by Means of Fenix Test, 7th RILEM International Conference on

Cracking in Pavements, Mechanisms, Modeling, Testing, Detection, Prevention and Case Histories, Delft,

Netherlands, pp. 61-69.

Mohammad, L.N., Kim, M., and Elseifi, M.A. (2012). Characterization of Asphalt Mixtures Fracture

Resistance Using the Semi- Circular Bending (SCB) Test, 7th RILEM International Conference on

Cracking in Pavements, Mechanisms, Modeling, Testing, Detection, Prevention and Case Histories, Delft,

Netherlands, pp. 1-10.

Molenaar, A.A.A. (2001). Prediction of Fatigue Cracking in Asphalts, Do we follow the Right Approach?

Journal of the Transportation Research Board, DOI: 10.3141/2001-17

Pais, J.C., Pereira, P.A.A., and Picado, L. (2002). Variability of Laboratory Fatigue Life of Asphalt

Mixes Using Four Point Bending Test Results, International Journal of Pavements, 1(2,) pp. 12.

Pais, J.C., Pereira, P.A.A., Minhoto, M.J.C., Fontes, L., Kumar, D.S.N.V.A., and Silva, B.T.A. (2010).

The Prediction of Fatigue Life of Asphalt Mixtures Using Four-Point Bending Tests, Proceedings of 24th

ARRB Conference Building On 50 Years of Road and Transport Research, (pp. 1-11). Melbourne,

Australia. ARRB Group Ltd., pp. 1-10.

Pramesti, F.P., Molenaar, A.A.A., and Van De Ven, M.F.C. (2013). The Prediction of Fatigue Life Based

on Four Point Bending Test, Procedia Engineering, 5, pp. 851 862

31
Rababaah, H. (2005). Asphalt Pavement Crack Classification: A Comparative Study of Three AI

Approaches: Multilayer Perceptron, Masters Thesis, Indiana University South Bend.

Rowe, G.M., Blankenship, P., Sharrock, M.J., and Bennert, T. (2012), The Fatigue Performance of

Asphalt Mixtures in The Four Point Bending Beam Fatigue Test in Accordance With AASHTO and ASTM

Analysis Methods, 5th Eurasphalt & Eurobitume Congress, Istanbul, Turkey, pp 1-12

Said, D., Abd El Halim, A.O., and Pais, J.C. (2008). Study of The Causes and Remedial of Premature

Surface Cracking of Asphalt Pavements, EPAM3 3rd European Pavement and Asset Management

Conference, Coimbra, Portugal, pp. 1- 14.

Siavashani, A.Z, (2014). Development of Indirect Ring Tension Test for Fracture Characterization of Asphalt

Mixtures, Doctoral Thesis, University of Kentucky.

Suh, Y., Mun, S., and Insoo, Y.I. (2010). Fatigue Life Prediction of Asphalt Concrete Pavement Using a

Harmony Search Algorithm, Korean Society of Civil Engineers Journal of Civil Engineering, DOI:

10.1007/s12205-010-0906-x

Tang, S. (2014). Evaluate The Fracture And Fatigue Resistances Of Hot Mix Asphalt Containing High

Percentage Reclaimed Asphalt Pavement (RAP) Materials At Low And Intermediate Temperatures, Doctoral

Thesis, Iowa State University.

Walubita L.F. (2000). Effects of Traffic Loading on Asphalt Surfacing Pavements and Model APT Testing with

the Third-Scale MMLS Mk3 Machine, Masters Thesis, University of Stellenbosch, South Africa

Wang, H., and Yang, L. (2011). The Fracture Characteristics Analysis of the Asphalt Mixtures by Virtual

Loading Test, Procedia Environmental Sciences, 10, pp. 601 606

Wang, W. (1991), Plastic Energy Damage Accumulation Theory and its Application in Fatigue Problem.

Proceedings of The International Symposium of Fatigue and Fracture in Steel and Concrete Structures,

Madras, India, pp.153-174.

Xiao, F.P., Amirkhanian, S.N., and Juang, H.J. (2007). Rutting Resistance of the Mixture Containing

Rubberized Concrete and Reclaimed Asphalt Pavement, Journal of Materials in Civil Engineering, 19,

pp. 475-483

32
Xiulin, Z., (2001). On Some Basic Problems of Fatigue Research in Engineering, International Journal of

Fatigue, 23, pp. 751766

Yina, A., Yanga, X., Yang, Z. (2013). 2D And 3D Fracture Modeling of Asphalt Mixture with Randomly

Distributed Aggregates and Embedded Cohesive Cracks, Procedia IUTAM 6, pp. 114 122

Yuan, M.M., Zhang, X.N., Chen, W.Q., and Zhang, S.X. (2013). Ratio of Dissipated Energy Change-

Based Failure Criteria of Asphalt Mixtures, Research Journal of Applied Sciences, Engineering and

Technology, 6(14), pp. 2514-2519.

33