Soil Dynamics and Earthquake Engineering 90 (2016) 476–479

Contents lists available at ScienceDirect

Soil Dynamics and Earthquake Engineering

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

Fatigue cracking failure criterion for flexible pavements under moving

vehicles
Niki D. Beskou a, Stephanos V. Tsinopoulos b, George D. Hatzigeorgiou c,n
a
Department of Civil Engineering, University of Patras, 26500 Patras, Greece
b
Department of Mechanical Engineering, Technological Educational Institute of Western Greece, 26334 Patras, Greece
c
School of Sciences and Technology, Hellenic Open University, 26335 Patras, Greece

art ic l e i nf o a b s t r a c t

Article history: An improved fatigue cracking criterion for flexible pavements is developed. It relates the allowable
Received 16 September 2016 number of load repetitions with the maximum tensile strain at the bottom of the asphalt concrete layer
Accepted 16 September 2016 that causes cracking there. The criterion is established by combining an empirical relationship providing
Available online 22 September 2016
the number of load repetitions as a function of the layer thicknesses and the axle loading as obtained by
the AASHO Road Test with that maximum tensile strain developed at the bottom of the asphaltic layer.
Because this maximum tensile strain is computed by a FEM that analyses pavements under moving
vehicles and assumes the asphaltic layer to behave as a linear viscoelastic material, the resulting failure
criterion takes into account speed and viscoelastic material behavior. Thus, this criterion is an improved
version of the existing ones based on static or dynamic loading and linear elastic material behavior.
& 2016 Elsevier Ltd. All rights reserved.

1. Introduction Various investigators, like Witczak [2], Rauhut and Kennedy [3]
and the Austin Research Engineers [4], have developed failure
The mechanistic-empirical methods of designing flexible pa- criteria of the type of Eq. (1) by combining numerical elastostatic
vements combine principles of mechanics for the determination of response results with experimental ones taken from the AASHO
the pavement response to vehicle loads with empirically obtained Road Test [5]. The Austin Research Engineers [6] constructed one
failure criteria (Huang [1]). Among these criteria, the most im- more criterion, such as that of Eq. (1), which takes into account
portant ones are those dealing with fatigue cracking, rutting and heavier loads and the effect of temperature. Other failure criteria
thermal cracking [1]. of the form of Eq. (1) have been developed by Verstraeten et al. [7]
This work deals with the first of these criteria. Fatigue cracking for the Belgian Road Research Centre, Powell et al. [8] for the
is caused by the maximum tensile strain at the bottom of the as- Transport and Road Research Laboratory (U.K.) and Thompson [9]
phaltic layer of a flexible pavement under vehicular load after a for the Illinois Department of Transportation (U.S.A).
large number of load repetitions. Thus, the failure cracking failure All the above fatigue cracking criteria, do not take into account
criterion is a relationship between the allowable number of load neither the vehicle speed effect nor the inelastic material behavior
repetitions with that maximum tensile strain. The general form of of the layers, simply because the response analyses used were
such a relationship is [1] elastostatic. The importance of the speed and inelasticity effects on
the pavement response has been stressed by many investigators,
Nf = f1( εt )−f2 (1) e.g., in [1,10–13]. The first to consider the effect of speed were
Sebaaly and Mamlouk [14], who developed a fatigue cracking
where Nf is the allowable number of load repetitions for crack
criterion by combining the empirical relation for Nf from the AA-
development due to fatigue, εt is the maximum tensile strain at
SHO Road Test [5] with numerical elastodynamic response results.
the bottom of the asphaltic layer and f1 and f2 constant coefficients Those results were obtained by their method of the system of
to be determined from experiments. The vehicular load associated horizontal layers in the frequency domain for analysing flexible
with the above relation is the standard single axle double tires pavements under moving loads with constant speed. Hysteretic
load of 18 kips ¼80 kN. damping was also taken into account.
In this Note, the idea of [14] is extended here to the case of a
n
Corresponding author. linear viscoelastic material, which is the appropriate model to si-
E-mail address: hatzigeorgiou@eap.gr (G.D. Hatzigeorgiou). mulate the dynamic behavior of asphaltic layer of a pavement [13].

http://dx.doi.org/10.1016/j.soildyn.2016.09.019
0267-7261/& 2016 Elsevier Ltd. All rights reserved.
 N.D. Beskou et al. / Soil Dynamics and Earthquake Engineering 90 (2016) 476–479 477

This material behavior effect is taken into account during the with dual tires of 18 kips ¼80 kN and tire pressure of
numerical computation of the maximum strain εt of Eq. (1) by 75 psi¼ 517 kPa moving with speeds of 0, 30 and 60 mph (0, 48.27
using the time domain FEM, as explained in detail in [13]. and 94.54 km/h) was determined.
It was assumed that the top asphalt layer in all sections was
linear viscoelastic, while all the other layers were linear elastic. The
shear modulus of the viscoelastic material versus time is shown in
2. Response of flexible viscoelastic pavements to moving loads
Fig. 1. The curve with the broken line corresponds to the asphalt
The dynamic response of a flexible viscoelastic pavement to a material 900901 in Berthelot et al. [15] scaled up from G(0)¼
vehicle load moving with constant speed is numerically obtained 817 MPa to G(0)¼E/2(1þv)¼3171.59/2(1þ0.35)¼ 1174.66 MPa to
by the time domain finite element method (FEM) of the ANSYS be compatible with the elastic properties of the top layer as de-
computer program, as described in [13]. This method has been scribed in Table 1. Furthermore, the resulting curve is compressed
to the left (by dividing the horizontal scale by 3) in order to exhibit
validated by means of other numerical methods and experimental
reduction of G(t) at early times and thus to lead to a more pro-
results and applied to various inelastic pavement models in [13]. It
nounced effect of viscoelasticity on the response. The broken line in
was found in [13] that use of the so called viscoelastic model, i.e., a
Fig. 1 can be expressed in a Prony series form with three terms
pavement model with its top asphalt layer being linear viscoelastic
reading
and the other layers elastic, can successfully simulate pavement
response at a reasonable computational cost. 3 − t
G( t) = G∞ + τiG
Following the idea of Sebaaly and Mamlouk [14] for the con- ∑ Gie
i=1 (2)
struction of a fatigue cracking criterion, 27 pavement sections of
the AASHO Road Test [5] consisting of four layers with material where τ1G ¼2.0, τ2G ¼ 0.2
and τ3G ¼0.02
s. A nonlinear regression
properties those of Table 1 and variable thicknesses, as shown in analysis can provide the necessary values for Gi and G1 of Eq. (2) for
the first few columns of Table 2, were analysed by the aforemen- the best fit in the form
tioned time domain FEM and their response to a single axle load
G1 = 123.40, G2 = 435.36, G3 = 204.37, G∞ = 411.53 MPa (3)
Table 1 Eq. (2) with Gi and G1 those in (3) is represented in Fig. 1 by
Material properties of pavement sections of AASHO Road Test [5].
the solid line and is used in the computations.
Layer Material E, ksi (MPa) ν ρ, pcf (kg/m3) In this work one is interested in the maximum tensile strain εt
at the bottom of the asphalt layer, which is obtained from the
Surface Asphalt concrete 460 (3171.59) 0.35 145 (2322.90) relation
Base Crushed stones 40 (275.79) 0.40 140 (2242.80)
Subbase Sandy gravel 20 (137.89) 0.40 135 (2162.70) εt = [(εx +ε y )/2] + [[(εx −ε y )/2]2 + (εxy )2]0.5 (4)
Subgrade Category A-6(6) 5 (34.47) 0.45 115 (1842.30)
where εx, εy and εxy are the maximum strain components of the
response with the x axis being the axis of the vehicle's path. The
Table 2 values of εt computed for all the 27 sections of the AASHO Road
Number of load repetitions and maximum tensile strains at the bottom of the as- Test [5] are shown in Table 2 for three values of speed, i.e., 0, 30
phalt layer for various AASHO Road Test pavement sections. and 60 mph or 0, 48.27 and 94.54 km/h.
AASHO Layer thicknesses (in) Nf From εt εt εt
Section Eq. (6) 10  6 in/ 10  6 in/ 10  6 in/
in in in 3. Allowable number of load repetitions for fatigue cracking
0 mph 30 mph 60 mph
Surface Base Subbase
AASHO Road Test [5] provides an empirical expression from
710 2 3 4 11,967 293 278 270 experimental measurements for the allowable number Nf of
717 1 3 4 3837 402 382 371 moving load repetitions for fatigue cracking in flexible pavements
727 1 0 4 1138 582 555 545
755 1 6 0 3560 416 391 385
758 2 6 0 11,224 274 260 252
111 2 6 8 68,857 255 237 224
140 4 6 8 306,054 198 181 173
145 4 3 0 32,896 281 256 248
161 4 6 0 72,387 238 218 206
575 4 3 12 318,786 225 197 189
583 4 0 4 34,767 313 288 280
619 4 0 8 79,918 255 240 231
625 4 6 13 664,098 205 190 181
427 5 9 12 1,710,544 145 127 114
439 5 3 4 164,943 224 208 198
445 5 6 12 1,036,698 156 143 132
473 4 6 4 154,246 226 213 204
477 4 9 12 984,141 171 151 142
261 5 3 12 605,461 172 151 141
297 6 3 8 616,817 160 145 135
319 5 3 8 325,321 192 172 160
333 6 9 16 4,555,401 126 104 97
336 6 3 12 1,091,658 155 140 129
719 1 6 4 10,867 403 378 368
156 3 6 8 150,816 223 198 189
325 6 6 8 1,054,746 161 146 140
260 5 6 8 583,288 172 152 143
Fig. 1. Shear modulus versus time for viscoelastic material.
478 N.D. Beskou et al. / Soil Dynamics and Earthquake Engineering 90 (2016) 476–479

in the form

( )
log Nf = 5.484 + 7.275 log(0.33 D1 + 0.10 D2 + 0.08 D3 + 1)+

+ 2.947 log(L 2)−3.136 log(L 1 + L 2) (5)

where D1, D2 and D3 are the thicknesses in inches of the surface,

base and subbase layers, respectively, L1 the nominal axle load in
kips and L2 ¼1 for the case of a single axle and 2 for the case of a
tandem-axle. Thus, for the case of a single axle load of
18 kips ¼80 kN considered here, Eq. (5) reduces to the simpler
form

( )
log Nf = 1.474 + 7.275 log( 0.33 D1 + 0.10 D2 + 0.08 D3 + 1) (6)

The above equation is used to determine Nf for all 27 sections

considered here and the resulting values are recorded in Table 2.
Application of a nonlinear regression analysis on the data for Nf
and εt in Table 2 for every speed, enables one to determine the
factors f1 and f2 of Eq. (1) and express it in the form
−5.63
( )
Nf = 2.5⋅1018 εt ⋅106 , V = 0 mph (0 km/h)
−5.25
Nf = 2.0⋅1017 ( ε ⋅10 )
t
6
, V = 30 mph( 48.27 km/h)
−5.04
Nf = 5.0⋅1016 ( ε ⋅10 )
t
6
, V = 60 mph( 94.54 km/h) (7a)

or (see Eq. (1)),

−f2
(
Nf = f1 εt ⋅106 )
(
f1 = f1(V) = 1015 2500 − 112.5V + 1.194V 2 )
f2 = f2 (V) = 5.63 − 0.015V + 0.000086V 2 (7b)
2
with a coefficient of correlation R ¼0.97, 0.98 and 0.97, respec-
tively, and V is expressed in mph. The above expressions for the
fatigue failure criterion take into account the effect of speed as
well as the effect of linear viscoelasticity. It has been shown in [13],
that viscoelastic material behavior in the asphalt concrete layer
leads to a response that (i) is higher than that for elastic material
behavior, (ii) is closer than the elastic one to the real response
obtained from tests and (iii) decreases, in general, with speed.
At this point one should observe that Eqs. (7a) and (7b) should
be used with caution because they were produced by combining
strains, which are functions of speed, with Nf, which, generally, are Fig. 2. Allowable number of repetitions Nf as obtained by different fatigue cracking
not functions of speed (Eq. (5)). However, it has been found – both criteria.

numerically or based on in-situ observations – that the allowable

number of repetitions Nf is dependent on the speed of vehicles of repetitions Nf as obtained for the cases of Eqs. (7), (8) and (9). It
(see Eq. 7b). Furthermore, because Eq. (5) is also the result of re- is observed that, (i) Nf decreases with speed, (ii) Nf is higher for the
gression analysis with R2 ¼ 0.79 [5], the true values of correlation viscoelastic case than that of the elastic case for zero speed and
coefficient for Eq. (7a) are R2 ¼ 0.97*0.79 ¼0.77, 0.98*0.79 ¼0.77 (iii) the proposed method can satisfactorily evaluate the allowable
and 0.97*0.79 ¼ 0.77. number of repetitions Nf.
A comparison of the failure criterion (7b) against two other
failure criteria of the general form of Eq. (1) but obtained for zero
speed and elastic material behavior [9] or nonzero speed and
elastic material behavior with damping [14] has been done and 4. Conclusions
the results are shown in Fig. 2. The failure criterion of [9] has the
form On the basis of the preceding discussion, one can state the
following conclusions:
Νf = 5*10−6(εt )−3 (8)
1) A more realistic fatigue cracking failure criterion for flexible
while that of [14] for V ¼30 mph the form pavements relating the allowable number of load repetitions to
Νf = 3.63*10−31(εt )−9.40 the maximum tensile strain at the bottom of the asphalt con-
(9)
crete layer has been developed
Fig. 2 has to do with the relation of the maximum tensile strain 2) The above failure criterion takes into account the effect of
εt at the bottom of the asphalt layer versus the allowable number speed and viscoelastic material behavior.
 N.D. Beskou et al. / Soil Dynamics and Earthquake Engineering 90 (2016) 476–479 479

Acknowledgement of Transportation, 1986.

[7] Verstraeten J, Veverka V, Francken L. Rational and practical designs of asphalt
pavements to avoid cracking and rutting. In: Proceedings of 5th international
The first author (Niki D. Beskou) acknowledges with thanks the conference on the structural design of asphalt pavements, Vol.1. Delft, The
support provided to her by the ‘IKY fellowships of excellence for Netherlands: Delft University of Technology; 1982. p. 45–58.
postgraduate studies in Greece-Siemens program’ (Grant number: [8] Powell WD, et al. The structural design of bituminous pavements. TRRL La-
SPhD/11171/13β). All the authors acknowledge with thanks helpful
boratory Report 1132. London, U.K.: Transportation and Road Research
Laboratory.
discussions with Professor D.D. Theodorakopoulos. [9] Thompson MR. ILLI-PAVE based full-depth asphalt concrete pavement design
procedure. In: Proceedings of 6th international conference on the structural
design of asphalt pavements, Vol.1., Ann Arbor, MI, U.S.A.: University of Mi-
chigan; 1987. p. 13–22.
References [10] Beskou ND, Theodorakopoulos DD. Dynamic effects of moving loads on road
pavements: a review. Soil Dyn Earthq Eng 2011;31:547–67.
[1] Huang YH. Pavement Analysis and Design.New Jersey, USA: Pearson Prentice- [11] Monismith CL. Flexible pavement analysis and design-A half century of
Hall, Upper Saddle River; 2004. achievement. In: Rollins K, Zekkos D, editors. Geotechnical Engineering, State
[2] Witczak MW. Design of full-depth asphalt airfield pavements. In: Proceedings of the Art and Practice, GSP 226, Reston VA, U.S.A.: ASCE; 2012.
of 3rd international conference on structural design of asphalt pavements. [12] Beskou ND, Tsinopoulos SV, Theodorakopoulos DD. Dynamic elastic analysis of
Ann Arbor MI, U.S.A.: University of Michigan; 1972. p. 550–567. 3-D flexible pavements under moving vehicles: a unified FEM treatment. Soil
[3] Rauhut JB, Kennedy TW. Characterizing fatigue life for asphalt concrete pa- Dyn Earthq Eng 2016;82:63–72.
vements. Transp Res Rec 1982;888:47–56. [13] Beskou ND, Hatzigeorgiou GD, Theodorakopoulos DD. Dynamic inelastic
[4] Austin Research Engineers Inc. Asphalt Concrete Overlays of Flexible Pave- analysis of 3-D flexible pavements under moving vehicles: a unified FEM
ments. Vol.1: Development of New Design Criteria, Report FHWA-RD-75-75. treatment. Soil Dyn Earthq Eng 2016.
1975. [14] Sebaaly PE, Mamlouk MS. Development of dynamic fatigue failure criterion. J
[5] AASHO Road Test, Special Reports 61A-61G, Highway Research Board, Wa- Transp Eng ASCE 1988;114:450–64.
shington D.C., U.S.A, 1962. [15] Berthelot CF, Allen DH, Searcy CR. Method for performing accelerated char-
[6] Austin Research Engineers Inc. Evaluation of Increased Pavement Loading, acterization of viscoelastic constitutive behavior of asphaltic concrete. J Mater
Vol.1: Research Results and Finding, Report No 84-59/1 to Arizona Department Civ Eng ASCE 2003;15:496–505.