You are on page 1of 9

International Journal of Fatigue 31 (2009) 13091317

Contents lists available at ScienceDirect

International Journal of Fatigue


journal homepage: www.elsevier.com/locate/ijfatigue

Fatigue crack growth prediction in concrete slabs


Cristian Gaedicke a, Jeffery Roesler a,*, Surendra Shah b
a
b

Department of Civil and Environmental Engineering, University of Illinois at Urbana Champaign, Newmark Laboratory, 205 N. Mathews Ave., Urbana, IL 61801-2352, USA
Department of Civil and Environmental Engineering, Northwestern University, A134 Technological Institute, 2145 Sheridan Road, Evanston, IL 60208-3109, USA

a r t i c l e

i n f o

Article history:
Received 15 November 2008
Received in revised form 3 February 2009
Accepted 24 February 2009
Available online 5 March 2009
Keywords:
Concrete fatigue
Crack growth
Fracture mechanics
Slabs on ground
Concrete pavement

a b s t r a c t
A new method to predict crack growth and fatigue life of concrete slabs is presented based on similar
models for 2-D specimens. Four large-scale concrete slabs on ground subjected to three different stress
ranges have been tested to fatigue failure. Geometric correction factors to calculate the effective crack
length and stress intensity factor from the experimental fatigue compliance are derived based on 3-D
nite element modeling. Crack length versus the number of cycles curves are constructed and tted to
a modied logistic function, which enables for the calculation of the crack propagation rates and critical
crack lengths from the rst and second derivative of the logistic function. Fatigue crack growth is separated into a decelerating and accelerating crack growth functions based on the initial crack length and the
applied stress intensity factor, respectively. The proposed method is able to predict the fatigue life of the
tested slabs and the fatigue resistance of several independent slabs cast with the same geometry and
material. The analysis of the tested slabs shows that load pulses with higher minimum loads generated
more fatigue damage as indicated by greater crack growth rates. The principles of this slab fatigue crack
growth procedure can now be extended to estimate the remaining life in uncracked or partially-cracked
concrete slabs on ground.
Published by Elsevier Ltd.

1. Introduction
The increasing trafc demands on concrete pavements and the
variation in available concrete material constituents pose unique
challenges in the structural design of concrete pavements and prediction of the slabs performance throughout its service life. Rigid
pavements structurally deteriorate, in the form of fatigue cracks,
over their service life, which can worsen with the introduction of
higher applied loads or changes in the boundary conditions, such
as loss of support or slab curling. Fatigue damage development is
especially critical in aireld pavements, with the introduction of
heavier aircraft and new generation gear congurations.
The traditional empirical method to predict the fatigue life of
concrete pavements requires calculating the critical tensile stress
in the slab and application of a fatigue equation. Fatigue equations,
also referred to as SN curves [112], relate the applied stress ratio
(tensile bending stress divided by concrete exural strength) to the
allowable repetitions until material failure. Miners Hypothesis has
been used for decades to linearly sum the fatigue damage of individual events that occur at different stress levels such that the
damage at failure is less than or equal to 1.0. Beam fatigue
equations used to predict fatigue life of a slab [2,3,13,14] inher-

* Corresponding author. Tel.: +1 217 265 0218; fax: +1 217 333 1924.
E-mail address: jroesler@illinois.edu (J. Roesler).
0142-1123/$ - see front matter Published by Elsevier Ltd.
doi:10.1016/j.ijfatigue.2009.02.040

ently assume that the fatigue and fracture of the concrete material
is independent of the specimen size and boundary conditions.
Full-scale fatigue tests on concrete slabs [1517] have conrmed that concrete slab fatigue cannot be accurately predicted
from beam fatigue curves. Slab fatigue curves always demonstrate
a greater resistance to cracking than that predicted by beam fatigue curves due to the change in specimen geometry and boundary
conditions [18]. The error in beam to slab fatigue prediction is also
dependent on the specic concrete materials utilized and, in particular linked to the concretes fracture properties. An additional
limitation in fatigue calculations is that Miners Hypothesis assumes that the fatigue damage caused in previous cycles does
not affect the stress state in the slab for future fatigue cycles. In
the reality, cracks initiate in the slab under repeated loading and
propagate progressively until failure occurs. In summary, current
empirical methods for fatigue damage analysis do not account
for the specimen geometry or boundary condition effects, the progressive crack growth, and the materials fracture resistance.
In the present paper, a fracture-based method to predict the fatigue crack growth of small-scale specimens is extended to predict
the crack propagation in concrete slabs supported by a soil foundation. This new slab fatigue model is then applied to four concrete
slabs tested in cyclic loading at the same load level but different
minimum loads. This methodology allows for the future prediction
of the remaining fatigue life of new and partially-cracked concrete
slabs for a variety of pavement applications.

1310

C. Gaedicke et al. / International Journal of Fatigue 31 (2009) 13091317

Nomenclature
a
a0
ac
c1, n1
c2, n2
DKI
dimax ; dimin

crack length
initial notch length, a0 = 0 for unnotched specimens
critical crack length
calibration parameters for fatigue deceleration stage
calibration parameters for fatigue acceleration stage
cycle variation of mode I stress intensity factor
maximum and minimum vertical actuator displacement

2. Background on fracture-based fatigue


Application of fracture mechanics offers a promising approach
to overcome several limitations associated with SN curves and
Miners Hypothesis. Initial attempts to use linear elastic fracture
mechanics [1922] and Paris Law [23,24] had limited success
due to their dependence on a signicant level of calibration of
the Paris Law with experimental data. Signicant calibration was
required since use of LEFM ignored the large fracture process zone
present in the concrete during the progressive failure and the
dependence of failure on the specimen size. Most researchers agree
that concrete shows a quasi-brittle failure response [2529], and
fracture is not dependent on only one parameter (i.e. the critical
stress intensity factor).
The efforts to model the quasi-brittle nature of concrete in fatigue have been focused on developing equations that consider the
fracture process zone or into combining nonlinear fracture
mechanics and Paris Law. Sain and Kishen [30,31] modied
Slowick et al. [32] LEFM-based fatigue equation to account for concretes tensile softening behavior on the fracture process zone.
Zhang et al. [33] proposed a semi-analytical model based on a cyclic crack bridging law [34] to predict the fatigue cracking behavior
of plain and ber reinforced (FRC) beams. Bazant and Xu [35] employed nonlinear elastic fracture mechanics (NLFM), i.e., the size
effect model [26], along with the Paris Law to take into account
the specimen size effect on the fatigue of concrete. Other researchers [36,37] have combined ParisErdogan Law and fractals to predict the fatigue life of plain and brous concrete beams. All of these
advanced approaches for concrete fatigue prediction employ multiple material parameters to better capture the effects of loading
and material variation on the specimens fatigue life.
Among the advanced methods which extend Paris Law into
NLFM, is Subramaniam et al. [3841] 2-D fracture-based fatigue
method. This effective crack method, based on the two-parameter
fracture model [42], calculates the crack propagation in concrete

Ci

Pimax ; P imin maximum and minimum vertical actuator load


compliance for cycle i
Ci
normalized compliance for cycle i
C nor
i
critical number of cycles
Nc
number of cycles on deceleration stage
Ndecc
number of cycles on acceleration stage
Nacc
number of cycles to slab failure
Nfailure

specimens based on experimental cyclic compliance (Fig. 1a),


which is dened as the ratio between displacement and applied
load per fatigue cycle. As is expected, the effective crack length increases during the fatigue life of a specimen as is schematically
shown in Fig. 1b. The instantaneous crack propagation rate (Da/
DN) can be calculated as:

Da ai  aj

i  j
DN

where ai and aj are the crack length for cycles i and j, respectively.
Fig. 1c shows that the crack propagation rate decreases until the
crack length reaches a minimum (ac) and then increases up to the
specimen failure. The crack propagation can be divided into the
two stages: deceleration stage and acceleration stage. Subramaniam
et al. [41] concluded that the deceleration stage could be described
by the following equation:

Da
C 1 a  a0 n1
DN

for a < ac

where a is the crack length, a0 is the initial notch, C1 and n1 are calibration constants for a given structural geometry, boundary conditions, and material. The critical crack length (ac), which is dened by
the inection point on the crack length versus number of fatigue cycles curve (Fig. 1b), was also found to be equal to about the crack
length at the peak monotonic load level [41]. The crack growth
characteristics in the acceleration stage are governed by the change
in stress intensity factor (DKI) during each load cycle, which is
quantitatively dened by Paris Law [23,24]:

Da
C 2 DK I n 2
DN

for a  ac

where DKI is the stress intensity factor range for each fatigue cycle;
C2 and n2 are calibration constants related to the structural geometry, boundary conditions, and material type. The main advantage of
this method over the LEFM and empirical fatigue approach is that it

Log(a/N)

a
=C 1( a a0 ) n1
N

afailure

Deceleration
Deceleration
Crack
propagation
rate

ac

a0

Acceleration
N

(a)

a
=C 2( KI ) n 2
N

Acceleration

(b)

a0

a\c

afailure

(c)

Fig. 1. Schematic of (a) compliance (Ci) measured from fatigue tests; (b) fatigue crack length (a) versus number of cycles (N); (c) crack growth rate versus crack length.

C. Gaedicke et al. / International Journal of Fatigue 31 (2009) 13091317

1311

uses an effective crack approach based on the measured compliance


to determine the crack growth in the slab versus number of applied
fatigue cycles. The derived model can then be used to predict the
number of cycles (N) needed to reach any given crack length (a),
by simply integrating Eqs. (2) and (3):

ac

1
da
C 1 a  a0 n1

a0

ac

1
da
C 2 DK I n2

In this paper, this 2-D fracture-based method proposed by Subramaniam et al. [3841] is extended to predict the fatigue life of
concrete slabs.
3. Slab fatigue life prediction method
The method herein proposed to predict fatigue life of concrete
slabs uses similar concepts to the method by Subramaniam et al.
[3841] with specic additions needed for slabs on ground. The
proposed method links the measured slab compliance from the
slab fatigue tests to the required fatigue parameters (C1, n1, C2,
n2, ac) shown in Eq. (4). First of all, crack length and stress intensity
factor for a given fatigue cycle are calculated from the experimental slab compliance. The theoretical relationships are derived from
3-D nite element models specically developed for this slab
geometry, loading conguration, and boundary condition. The next
step is to determine the crack propagation rate (Da/DN) for each
cycle and obtain the critical crack length. Finally, the deceleration
parameters (C1, n1) are calculated from Eq. (2) using the crack
propagation rate and crack length data, for a < ac. Likewise, the
acceleration parameters (C2, n2) are computed from Eq. (3) using
the crack propagation rate and crack length data, for a P ac.
4. Experimental program and results
4.1. Experimental program and test setup
An experimental program has been developed to verify the proposed fracture-based fatigue method on large-scale slabs subjected
to several loading conditions. Three different load pulses, which
simulate the applied load of a triple dual tandem aircraft gear have
been used, as is shown in Fig. 2. The pulse duration is one second
long with three distinct 90 kN peak loads. The maximum stress ratio is the same for all slabs, but the minimum applied load between

120
T-01
T-04
T-07, T-07R

100

Pmax

Fig. 3. Fatigue test setup for slabs on ground.

peaks is different for each load pulse (rmin/rmax = 0.1, 0.4, and 0.7).
The slab specimens are identied based on the applied stress range
as T-01, T-04, T-07, respectively. For model verication, an additional slab identied as T-07R was also tested with a stress ratio
equal to rmin/rmax = 0.7.
Each concrete slab is 2000 mm  2000 mm  150 mm thick.
The slab is placed on the top of a 200 mm soil layer contained in
a test frame, as is shown in Fig. 3. The load is applied through a
445 kN hydraulic actuator at the mid-slab edge. The actuator operated under a precision servo-valve control in a closed-loop servo
hydraulic system powered by a 30 gpm hydraulic power supply
and regulated by a service manifold. The slab vertical displacements are measured through strategically-placed LVDTs. The soil
layer is a low-plasticity clay with a 5% California Bearing Ratio
(CBR) at 17% moisture content, and a correlated modulus of subgrade reaction equal to 17.7 MPa/m.
The concrete mixture design selected is based on mixtures commonly used at regional airports in Illinois. The coarse aggregate is a
crushed limestone with a 19 mm nominal maximum size, whereas
the ne aggregate is natural sand. The mixture proportions are
1085 kg/m3 of coarse aggregate and 707 kg/m3 of ne aggregate.
The mixture has a water to cementitious ratio equal to 0.40 using
290 kg/m3 of Type I cement, 77 kg/m3 of Type C y ash and 145 kg/
m3 of water. A water reducing admixture is used to obtain a target
slump equal to 100 mm. The average exural strength at the slab
testing age (120 days) is 5.4 MPa. The modulus of elasticity and
Poissons ratio are equal to 24 GPa and 0.15, respectively.

Load (kN)

80

4.2. Experimental fatigue test results

P min
60

40

20

P unloaded
0

125

250

375

500

625

750

875

1000

Time (ms)
Fig. 2. Shape of the load pulse for T-01: R-value = 0.1, T-04: R-value = 0.4, and T-07/
T-07R: R-value = 0.7.

The applied load and displacement history for each fatigue cycle
is collected during the fatigue tests, as is schematically depicted in
Fig. 4. The individual slabs compliance is calculated for every fatigue cycle, in order to determine the effective crack length and
crack length change per loading cycle. As is shown in Fig. 4 and
Eq. (5) below, the compliance is calculated for each loading cycle
(C1, . . ., Ci, . . ., Cfailure) as the ratio between the difference of the maximum (dimax) and minimum (dimin ) vertical displacement and the
difference between the maximum (P imax ) and minimum (Pimin ) applied load:

C i dimax  dimin =Pimax  Pimin

1312

C. Gaedicke et al. / International Journal of Fatigue 31 (2009) 13091317

5. Implementation of fracture-based fatigue method for


concrete slabs

P
Pimax

The proposed method has required development of equations,


by 3-D nite element analysis, to relate normalized compliance
to the effective crack length and applied stress intensity factor
for the given slab geometry, loading conguration, and boundary
conditions. The method also implements a logistic function to
translate the crack length into a continuous function in order to
simplify the fatigue crack growth analysis.

Cycle i

Ci =

P min
imin

i
i
max
min
i
i
Pmin
Pmax

imax

5.1. Crack length and stress intensity factor analysis

Fig. 4. Schematic of a fatigue cycle with its maximum and minimum load and
displacement for compliance calculation.

Due to inherent variations in the slab support and thickness between different tests, the calculated compliances are normalized
using the initial compliance (C1):

C nor
C i =C 1
i

The normalized compliance against the number of loading cycles is then used to theoretically evaluate the crack length at any
loading cycle.
Fig. 5 shows the relationship between the normalized compliance and the total number of loading cycles up to failure for each
tested slab. The fatigue life for slabs T-01, T-04, T-07 and T-07R
is equal to 61.84  103, 26.24  103, 4.384  103, and 5.44  103
cycles, respectively. As was expected, specimens that show higher
compliance growth rates (T-07 and T-07R) have a shorter fatigue
life. For each slab, the plots also show a rst stage with a decreasing compliance rate followed by a second stage with an increasing
compliance rate and then sudden failure. This compliance behavior
is similar to the results determined by Subramaniam et al. [41] for
concrete beam specimens. Slabs T-07 and T-07R, subjected to the
same load pulse with a stress range equal to 0.7, have demonstrated a shorter fatigue life compared to slab T-01 with a stress
range equal to 0.1, which was unexpected based on previous beam
fatigue results published by Tepfers [35].

1.3

Normalized Compliance

T-01

T-07R

1.25

T-07

1.2

1.15


4

3

2
a 52; 262 C nor  282; 952 C nor 561; 486 C nor
 nor 
14; 9661
 480; 420 C

1.1
T-04
1.05

1
0

10.000

20.000

5.1.1. Fatigue crack length


A set of nite element models have been developed in ABAQUS
to derive the relationship between the slabs compliance and the
respective crack length, and also to predict the resultant stress
intensity factor. Such relationships (crack length against compliance and crack length against stress intensity factor) have been
determined by varying the initial crack length in the slab between
0 and 1980 mm, to represent different crack propagation stages.
The models assume that crack propagation in the slab is idealized
by a full-depth crack starting from the loaded edge of the slab and
uniformly propagating across the slab. The assumption that the
through-crack grows along the symmetry line of the concrete slab
is generally consistent with the actual crack path observed in the
experiments. The proposed models use symmetry along the direction of crack propagation to reduce the overall size of the models,
as is shown in Fig. 6a. The example shown in this gure represents
the model for the concrete slab with a fatigue crack that has propagated 600 mm. The tested concrete modulus of elasticity and
Poissons ratio, and the soil modulus of subgrade reaction have
been used as material properties in the model.
The nite element mesh employs three-dimensional 20-node
quadratic reduced integration brick elements (C3D20R) with an
average element size of 25  25  6 mm. A 200 mm semi-circular
radial mesh is built around the crack tip to improve the stress
intensity factor computation, with 40 and 34 elements in the angular and radial direction, respectively, as is shown in Fig. 6b. A
renement factor (called Bias Ratio in ABAQUS) equal to 2.0 is used
in the radial direction.
The compliance in the FEM slab models is computed as the ratio
between the vertical displacement (d) and the total load. The vertical displacement on the models is calculated at the center of the
200  200 mm plate in the experiments, to ensure consistency between measured and calculated responses. The compliance is calculated for each crack length and then normalized with respect
to the unnotched slabs compliance, i.e., for a = 0. The normalized
compliance at a crack length equal to zero is then equal to 1.0.
Fig. 7a plots the theoretically-derived relationship between
crack length and the normalized compliance. A predictive equation
that relates the normalized compliance to the crack length can
then be developed so that the effective crack length of the experimental slabs can be calculated. The least squared approximation
between crack length and normalized compliance for this slab conguration is the following:

30.000

40.000

50.000

60.000

Number of Cycles
Fig. 5. Normalized compliance versus number of fatigue cycles for slabs.

where a is the crack length in millimeters and Cnor is the normalized


compliance. In the theoretical model, the Cnor is approximately 1.5
for a fully propagated crack (i.e., a = 2000 mm) and therefore for
measured Cnor > 1.5 the crack length is automatically set to
2000 mm. The effective crack length for each slab at any number

1313

C. Gaedicke et al. / International Journal of Fatigue 31 (2009) 13091317

Fig. 6. (a) Example of nite element model a slab with a crack length of a = 600 mm and (b) mesh around crack tip.

-3

2000

(b)
Normalized Stress Intensity Factor, K/P (mm-3/2)

(a)

1800

Crack Length (mm)

1600
1400
1200
1000
800
600
400
200
0
1

1.05 1.1

1.15 1.2

1.25 1.3

1.35 1.4

1.45 1.5

Normalized Compliance

x 10

5.5
5
4.5
4
3.5
3
2.5
2
1.5
1
0.5
400

600

800

1000

1200

1400

1600

1800

2000

Crack Length (mm)

Fig. 7. Finite element analysis relation between (a) normalized slab compliance and crack length and (b) crack length and normalized stress intensity factor (KI).

of cycles is calculated with Eq. (7) using the normalized slab compliances shown in Fig. 5.
The diagram of crack length against number of cycles is presented in Fig. 8 for the four slabs. Slabs T-07 and T-07R, which have
signicantly lower fatigue lives, are unloaded the least (i.e., to
70 kN) with respect to the other two slabs. In contrast, slab T-04
is subjected to a minimum load equal to 40 kN and slab T-01 to
a minimum load equal to 9 kN. Slabs T-01 and T-04 have a significant plateau section (with a small crack growth rate) compared to
slabs T-07 and T-07R.
The crack propagation rate for T-07 and T-07R specimens is
much higher, suggesting that the stress range affects the fatigue
life of the soil-supported concrete slabs differently from a simply

supported beam. A loading pulse that keeps higher minimum


stresses on the slab-soil system for a longer period of time, such
as in T-07 and T-07R, generates more plastic deformation in the
soil which subsequently increases the stress state in the concrete
slabs. Fig. 8 also shows that the effective crack length at which
unstable crack propagation occurs is different for each specimen,
i.e. 1450, 620, 920, and 1420 mm, for T-01, T-04, T-07, and
T-07R, respectively.
5.1.2. Stress intensity factor (DKI)
The acceleration stage of fatigue crack growth depends on the
variation of the stress intensity factor (DKI). A relationship between KI for different crack lengths (a) and applied loads (P) is

1314

C. Gaedicke et al. / International Journal of Fatigue 31 (2009) 13091317

is linear elastic and does not soften or yield under higher stress
states. The application of this equation is valid since Eq. (8) is only
used in the fatigue acceleration stage which starts for crack lengths
greater than 550 mm.
Eq. (8) is then modied to calculate the change in stress intensity factor, DKI, from the change in the applied load during each fatigue cycle and the effective crack length (a) (mm):

2000
1800
T-07R

Crack Length (mm)

1600
1400

T-01
1200

"

T-07

DK I wPmax  P min 1:27894  103

1000
800

4:98771  102

2000  a0:788467

where Pmax  Pmin is the difference between the maximum and


minimum load p
(N)
during each fatigue cycle, and w is a conversion
factor, i.e. w 10=100 to express DKI in MPa m1/2.

600
400
T-04
200
0

5.2. Fatigue data ltering


0

10,000

20,000

30,000

40,000

50,000

60,000

A signicant challenge in the application of the fracture-based


fatigue model for slabs is to explain the inherent variability of
the experimental slab fatigue results. As can be observed in
Fig. 8, the crack length plots show cyclic variation, which is typical
in fatigue tests. The calculation of the crack propagation rates (Da/
DN) using Eq. (1) would require choosing certain crack lengths and
loading cycle numbers (i.e. ai, aj, i, j). As a result, the calculated
propagation rate would depend on how cycles i and j are chosen,
which means that Da/DN calculated using small cycle increments
would be different to Da/DN computed from larger cycle intervals.
Hence, the method would not be consistent which would limit the
applicability of the fracture-based fatigue method to concrete
slabs. In order to overcome this limitation, an intermediate step
is added to t the calculated crack lengths with a smooth continuous function. The following logistic function [43] relating the crack
length and loading cycles is chosen to improve the stability of the
crack growth rate calculations:

Number of Cycles
Fig. 8. Crack length versus loading cycles on slab specimens.

therefore needed. The applied stress intensity factor (KI) is calculated from the nite element model using contour integrals around
the crack tip to determine the J-Integral. Collapsed C3D20R brick
elements with quarter point nodes are used around the crack tip
to ensure the contour integral convergence. The stress intensity
factor is calculated for each crack length as the average of the
stress intensity factors for the 3rd, 4th and 5th contour. The variation between these three contours is on average less than 0.1
percent.
The relationship between the stress intensity factor and crack
length derived from the FEM models is shown in Fig. 7b. The equation that relates the normalized (KI/P) Mode I stress intensity factor
(mm3/2) to a given crack length (a, in mm) is the following:

a s ln

2

K I =P 1:27894  103

4:98771  10

0:788467

2000  a

(b)


10

2000

1800

1800

1600

1600

1400

1400

T-01
T-04
T-07
T-07R

T- 07R

Crack Length (mm)

Crack Length (mm)

2000

bN
Nm2  bm1

where a is the crack length, N is the loading cycle number, b, m1, m2


and s are experimental calibration constants. As is demonstrated in
Fig. 9a for slab T-01, the effective crack length versus number of
loading cycles can be accurately represented using the above function. The diagrams of crack length versus number of loading cycles
for all slabs are shown in Fig. 9b based on the predictive equation.
The continuous Eq. (10) for crack length versus fatigue cycles

This equation was specically derived based on the FEM model results in the range of 400 < a < 2000 mm, as it is meant to be used for
the acceleration stage which starts for crack lengths greater than
550 mm. The stress intensity factor decreases slightly with an increase in crack length for initial crack lengths less than 400 mm
due to the geometry of the slab and the assumption that the soil

(a)

1200
a
c
1000
800
600

Acceleration

Deceleration

400

T- 01

T- 07

1200
1000
800

T- 04

600
400

Experimental - T-01

200
0

Predicted by S-Curve

Nc
0

10,000

20,000

30,000

40,000

Number of Cycles

50,000

60,000

200
0

10,000

20,000

30,000

40,000

50,000

60,000

Number of Cycles

Fig. 9. (a) Raw and ltered crack length versus number of cycles curves for slab T-01; (b) idealized crack length versus loading cycles for all slab specimens.

1315

C. Gaedicke et al. / International Journal of Fatigue 31 (2009) 13091317

enables the calculation of the crack propagation rate as the rst


derivative of Eq. (10):

da
bm2  m1 s

dN b  Nbm1  Nm2

Table 1
Slab fatigue data, calibrated parameters, and fracture-based fatigue model
predictions.

11

The use of Eq. (11) instead of Eq. (1) standardizes the process and
eliminates user subjectivity in choosing discrete crack lengths and
cycle numbers.
Fig. 10 shows the relationship between the crack growth rate
(logarithmic scale) and the crack length. In this gure, the minimum crack growth rate (shown as a negative value) represents
the critical crack length (ac), which is the transition between decelerating and accelerating crack growth. As shown in Fig. 10, specimens T-01 and T-04 have lower crack propagation rates (given
by lower negative values) compared to T-07 and T-07R, which is
consistent with their longer fatigue life. The critical crack length
(ac) and critical number of loading cycles (Nc) are obtained by setting the second derivative of Eq. (10) to zero. The equation for the
critical crack length is:

ac s ln1=m2

12

T-04

T-07

T-07R

Measured fatigue values


1084
ac (mm)
31,682
Nc = Ndecc
30,158
Nacc
61,840
Nfailure
0.512
Nc/Nfailure

T-01

557
13,191
13,049
26,240
0.503

790
2341
2043
4384
0.534

1225
2795
2645
5440
0.514

Fatigue parametersa
c1
n1
c2
n2

1.28E+18
6.740
1.68E27
40.605

3.37E+14
6.110
1.57E101
173.147

2.29E+12
4.688
1.71E49
81.487

5.61E+14
5.150
3.29E12
17.122

Predicted fatigue values


Nc = Ndecc
Nacc
Nfailure
Nc/Nfailure

31,229
30,053
61,283
0.510

13,914
15,800
29,714
0.468

2325
2277
4601
0.505

2851
3081
5931
0.481

Prediction error (%)

0.9

13.2

5.0

9.0

Fatigue parameters are presented in for a in mm and DKI in MPa m0.5.

and critical number of loading cycles is given by:

Nc bm1 m2 =2m2

13

The critical crack length values are shown in Table 1. The critical
number of loading cycles (Nc) ranges between 49 and 52 percent of
the number (Nfailure) of loading cycles up to failure for each slab.
The critical crack length is different for each slab and does not
show a direct trend with fatigue life of the slabs, which could be
a result of the general assumption on how the crack is idealized
to propagate across the slab and the fact the support condition cannot be guaranteed to be identical for all slab tests.
Subramaniam et al. [41] found that three-point-bending beam
specimens made with the same geometry and the same concrete
material had approximately the same critical crack length. They
discovered that the critical crack length (ac) in fatigue could be
approximated by the equivalent crack length at the peak load in
a monotonic test. The present paper shows that the critical crack
length varies signicantly which is likely due to the boundary condition change (i.e. specimen soil support) and the assumption of
the crack geometry (uniform propagation of a through-crack). In
the slab-on-ground fatigue tests, the minimum load level is not
constant for all specimens, and this minimum load level affects
the support conditions. This difference affects the overall fatigue

The deceleration stage is dened by a reduction in the crack


propagation rate until the critical crack length (ac) is reached. Eq.
(2) can be easily calibrated using a logarithmic transformation
and then applying linear regression (see Fig. 11a). Fatigue coefcients C1 and n1 are obtained from the following linear regression:

log

da
dN

logC 1 n1 loga  a0

14

and coefcients C1 and n1 are presented in Table 1.


The acceleration stage is dened by an increase in the crack
propagation rate for crack lengths larger than the critical crack
length, and is dened by Eq. (3). As is shown in Fig. 11b, the relationship between log (da/dN) and log (DKI) is linear, and therefore,
Eq. (3) can be expressed as:


log

da
dN


logC 2 n2 logDK I

15

5.4. Slab fatigue model verication

-0.5

Log(da/dn)

5.3. Slab fatigue model calibration

where DKI is the variation of the stress intensity factor


K PI max  K PI min for each fatigue cycle, calculated using Eq. (9). The
calibrated coefcients C2 and n2 are shown in Table 1. Several of
such C2 coefcients are extremely small, which physically represent
the crack length variation with load cycles as the stress intensity
factor variation approaches zero.

0.5

T- 07
T- 07R

-1

-1.5
T-01
T-04
T-07
T-07R

T- 04
T- 01

-2

-2.5

behavior, as specimens that are subjected to higher minimum load


levels show shorter fatigue life.

200

400

600

800

1000

1200

1400

1600

1800

2000

Crack Length (mm)


Fig. 10. Crack propagation rate (log) versus crack length for all slab specimens.

The accuracy of the previously calibrated parameters can be


checked by calculating the total fatigue life of the specimens using
Eq. (4) and comparing it with the experimental results. The total
fatigue life (Nf) of the slab specimen is calculated by integrating
the deceleration Eq. (2) over the crack length between a0 and ac,
and the acceleration Eq. (9) over the crack length starting from ac
until its failure crack length af (af = 2000 mm). A small a0 value
greater than zero (i.e., a0 = 0.01) is used in Eq. (4) for slabs with
no notch to avoid a singularity in the deceleration integral. DKI
must be expressed in terms of the crack length (a) and the maximum and minimum fatigue loads (Pmax, Pmin) using Eq. (9) in the
acceleration integral. The integrals to calculate the fatigue life of
the slabs are:

1316

C. Gaedicke et al. / International Journal of Fatigue 31 (2009) 13091317

(b)
T-01
T-04
T-07
T-07R

0.5

Crack Propagation Rate, Log(da/dn)

Crack Propagation Rate, Log(da/dn)

(a) 1

T- 07

-0.5

T- 07R
-1

-1.5

T- 01
T- 04

-2

-2.5
2.6

2.7

2.8

2.9

3.1

1
T-01
T-04
T-07
T-07R

0.5

T- 07

T- 07R

-0.5

-1

-1.5

T- 04

T- 01

-2

-2.5
0.55

0.575

0.6

0.625

0.65

0.675

Log (KI), (log(MPa m ))

Log Crack Length, (log(mm))

0.5

Fig. 11. (a) Calibration of the Fatigue deceleration coefcients by linear regression; (b) calibration of the fatigue acceleration coefcients by linear regression.

ac

1
da
C 1 a  a0 n1
1
n
h
ion2 da
4:98771102
C 2 wPmax  Pmin 1:27894  103 2000a
0:788467

N Ndecc Nacc

a0

ac

16
These integrals can be evaluated numerically using Mathematica, and the predicted fatigue cycles for slabs T-01, T-04, T-07 and
T-07R, are compared to the actual fatigue life of the specimens. The
predicted fatigue obtained using calibrated parameters for each
specimen (c1, n1, c2, n2, ac) is very similar to the actual fatigue life
for each specimen. The low prediction errors shown in Table 1 conrm the validity of the proposed method along with the accuracy
of the experimental calibration.
The proposed model is also used to explain the slab fatigue test
results (similar to T-01) conducted by Roesler et al. [16,17] by
employing specimens with similar geometry and concrete materials that never failed during cyclic loading. The rst slab, identied
as T5 in Ref. [17], was subjected to the same pulse shape as T-01,
but with maximum and minimum applied loads equal to 78.73
and 6.67 kN, respectively. This slab specimen did not fail in fatigue
even after 594.7  103 cycles, which is consistent with the model
that predicted 1.39  107 to failure under those load conditions.
For a second slab subjected to maximum and minimum applied
loads equal to 44.9 and 4.4 kN (see Ref. [16]), the model predicted
3.07  1010 loading cycles up to failure. This prediction was consistent with the fatigue test on this slab, which resisted 610.6  103
loading cycles without fatigue failure. These independent conrmations show that the model can predict the conditions under
which concrete slabs show high fatigue endurance limit. Fatigue
tests with different slab geometries, soil properties, initial notch
lengths, or applied load location would require different geometric
factors to calculate the crack length versus compliance and stress
intensity factor given a similar calibration process was employed.
Extrapolation of this particular slab fatigue model to other slab
geometries and material properties can result in predictive errors.
6. Conclusion
A new method to predict fatigue life and crack propagation in
concrete slabs is presented by extending an existing 2-D fracture-based fatigue model. The new method offers advantages over

traditional SN curve approaches for slab fatigue predictions such


as calculation of the progressive cracking in the slab, use of fracture
mechanics to explain crack growth, and the ability to analyze the
effects of partial depth cracks. The new method is able to predict
the fatigue life of large-scale concrete slabs on ground subjected
to different fatigue load pulses. The validity of the slab fatigue
model has been assessed by examining two independent slab fatigue tests with the same geometry, concrete material, and load
pulse shape.
Several unique features of the proposed slab fatigue model are
the calculation of the effective elastic crack length against the
number of fatigue cycles from the measured slab compliance data
and the application of a calibrated logistic function to calculate the
crack propagation rate and critical crack length resulting in a more
accurate determination of the deceleration and acceleration phases
of fatigue crack growth. The analysis shows that load pulses that
keep a higher minimum load between peak loads tend to generate
larger crack growth rates in slabs. Higher sustained loads for a
longer period of time increase the stress state in the soil (stress
softening behavior), prevent recovery of the soil, and nally induce
higher stresses in the concrete slab. This new model has the potential to be applied to determine the remaining life of uncracked or
partially-cracked concrete pavement slabs.
Acknowledgements
This paper was prepared from a study conducted in the Center
of Excellence for Airport Technology (CEAT), funded in part by the
Federal Aviation Administration (FAA). The CEAT is maintained at
the University of Illinois at Urbana-Champaign with partners at
Northwestern University. The contents of this paper reect the
views of the authors who are responsible for the facts and accuracy
of the data presented herein. The contents do not necessarily reect the ofcial views and policies of the FAA. This paper does
not constitute a standard, specication, or regulation.
References
[1] Parker F, Barker WR, Gunkel RC, Odom EC. Development of a structural design
procedure for rigid airport pavements. Contract no. DOT FA73WAI-377. US
Army Engineer WES, Vicksburg, MS; 1979. 300p.
[2] Darter MI. Design of zero-maintenance plain jointed concrete pavement, vol. 1:
development of design procedures. Federal Highway Administration report no.
FHWA-RD-77-III; 1977.

C. Gaedicke et al. / International Journal of Fatigue 31 (2009) 13091317


[3] Tepfers R, Kutti T. Fatigue strength of plain, ordinary, and lightweight concrete.
J Am Concr Inst 1979;76:63552.
[4] Tepfers R. Tensile fatigue strength of plain concrete. J Am Concr Inst
1979;76:91933.
[5] Tepfers R. Fatigue of plain concrete subjected to stress reversals. ACI
Publication SP-75; 1982. p. 195217.
[6] Rollings R. Corps of engineers design procedure for rigid aireld pavements. In:
Proceedings of the second international conference on concrete pavement
design. West Lafayette, IN: Purdue University; 1981. p. 18598.
[7] Rollings RS, Witczak M. Structural deterioration model for rigid aireld
pavements. J Transport Eng 1990;116(4):47991.
[8] Rollings RS. Evolution of the US militarys concrete aireld pavement design
procedures. In: Proceedings, fourth international workshop on design theories
and their verication of concrete slabs for pavements and railroads, Bucaco,
Portugal; 1998. p. 26376.
[9] Kesler CE. Effect of speed of testing on exural fatigue strength of plain
concrete. In: Proceedings, highway research board, vol. 32; 1953. p. 2518.
[10] Murdock JW, Kesler CE. Effect of range of stress on fatigue strength of plain
concrete beams. J Am Concr Inst 1958;30(2):22131.
[11] Nieto AJ, Chicharro JM, Pintado P. An approximated methodology for fatigue
tests and fatigue monitoring of concrete specimens. Int J Fatigue
2006;28(8):83542.
[12] Hui Li H, Zhang M, Ou J. Flexural fatigue performance of concrete containing
nano-particles for pavement. Int J Fatigue 2007;29(7):1292301.
[13] Packard RG. Fatigue concepts for concrete airport pavement design. J Transport
Eng ASCE 1974;100(3).
[14] Packard RG, Tayabji SD. New PCA thickness design procedure for concrete
highway and street pavements. In: Proceedings, 3rd international conference
on concrete pavement design. West Lafayette, Ind.: Purdue University; 1985.
[15] Roesler JR. Fatigue of concrete beams and slabs. Ph.D. Thesis, University of
Illinois, Urbana, Illinois; 1998. 483p.
[16] Roesler JR, Littleton PC, Hiller JE, Long GE. Effect of stress state on concrete slab
fatigue resistance, Draft nal report for Federal Aviation Administration.
Urbana, IL: University of Illinois; 2004. 227p.
[17] Roesler JR, Hiller JE, Littleton PC. Large-scale aireld concrete slab fatigue tests.
In: Proceedings of the 8th international conference on concrete pavements,
vol. 3. Colorado Springs, CO; 2005. p. 124768.
[18] Roesler JR. Fatigue resistance of concrete pavements. In: 6th international DUT
workshop on fundamental modelling of design and performance of concrete
pavements, Belgium; 2006.
[19] Lytton RL, Shanmugham U. Analysis and design of pavements to resist thermal
cracking using fracture mechanics. In: Proceedings of the 5th international
conference on the structural design of asphalt pavements, vol. 1. Delft,
Netherlands; 1982. p. 81830.
[20] Ramsamooj DV. Analysis and design of the exibility of pavements. Ph.D.
Thesis, Ohio State University, Columbus, OH; 1970. 147p.
[21] Ramsamooj DV. Fatigue cracking of asphalt pavements. Transport Res Record
1980:438.
[22] Ramsamooj DV. Fracture of highway and airport pavements. Eng Fract Mech
1993;44(4):60926.

1317

[23] Paris PC. The growth of fatigue cracks due to variation in load. Ph.D. Thesis,
Lehigh University, Bethlehem, PA; 1962.
[24] Paris P, Erdogan F. A critical analysis of crack propagation laws. J Basic Eng,
Trans Am Soc Mech Eng 1963(December):52834.
[25] Kaplan MM. Crack propagation and fracture of concrete. J Am Concr Inst
1961;58(5):591609.
[26] Bazant ZP. Size effect in blunt fracture: concrete, rock, metal. J Eng Mech
1984;110(4):51835.
[27] Bazant ZP. Size effect on structural strength: a review. Arch Appl Mech
1999:70325.
[28] Ioannides AM. Fracture mechanics applications in pavement engineering: a
literature review. Contract no. DACA39-94-C-0121. US Army Engineer
Waterways Experiment Station, Vicksburg, MS; 1995.
[29] Ioannides AM. Fracture mechanics in pavement engineering. transportation
research record no. 1568. Transportation Research Board, National Research
Council, Washington, DC; 1997. p. 106.
[30] Sain T, Chandra Kishen JM. Residual fatigue strength assessment of
concrete considering tension softening behavior. Int J Fatigue 2007;29(12):
213848.
[31] Sain T, Chandra Kishen JM. Probabilistic assessment of fatigue crack growth in
concrete. Int J Fatigue 2008;30(12):215664.
[32] Slowik V, Plizzari G, Saouma V. Fracture of concrete under variable amplitude
loading. ACI Mater J 1996;93(3):27283.
[33] Zhang J, Stang H, Li VC. Fatigue life prediction of ber reinforced concrete
under exural load. Int J Fatigue 1999;21(10):103349.
[34] Zhang J, Stang H. Application of stress crack width relationship in predicting
the exural behavior of ber reinforced concrete. Cem Concr Res
1998;28(3):43952.
[35] Bazant ZP, Xu K. Size effect in fatigue fracture of concrete. ACI Mater J
1991;88(4):3909.
[36] Carpinteri A, Spagnoli A. A fractal analysis of size effect on fatigue crack
growth. Int J Fatigue 2004;26(2):12533.
[37] Carpinteri A, Spagnoli A, Vantadori S. An elasticplastic crack bridging model
for brittle-matrix brous composite beams under cyclic loading. Int J Solids
Struct 2006;43(16):491736.
[38] Subramaniam KV, Popovics JS, Shah SP. Monitoring fatigue damage in
concrete. In: Proceedings MRS fall meeting, vol. 503; 1997. p. 1517.
[39] Subramaniam KV, Popovics JS, Shah SP. Testing concrete in torsion: instability
analysis and experiments. J Eng Mech, Am Soc Civil Eng (ASCE)
1998;124(11):125868.
[40] Subramaniam KV, Popovics JS, Shah SP. Fatigue behavior of concrete subjected
to biaxial stresses in the CT Region. Mater J Am Concr Inst (ACI)
1999;96(6):6639.
[41] Subramaniam KV, ONeil E, Popovics JS, Shah SP. Flexural fatigue of concrete:
experiments and theoretical model. J Eng Mech, Am Soc Civil Eng (ASCE)
2000;126(9):8918.
[42] Jenq YS, Shah SP. Two parameter fracture model for concrete. J Eng Mech
1985;10:122741.
[43] Kingsland SE. Modeling nature. Chicago: University Of Chicago Press; 1995.

You might also like