International Journal of Fatigue 33 (2011) 11401150

Contents lists available at ScienceDirect

International Journal of Fatigue

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

Numerical studies on multistage strength degradation of notched

concrete beams under sequential loading
Zihai Shi , Yukari Nakamura, Masaaki Nakano
Research and Development Center, Nippon Koei Co., Ltd., 2304 Inarihara, Tsukuba-shi, Ibaraki 300-1259, Japan

a r t i c l e i n f o a b s t r a c t

Article history: Initial aws and defects exist in all engineering materials and thus in all structural members; only the
Received 21 July 2010 degree of imperfection varies. Under cyclic loading, the material weakening process in structural mem-
Received in revised form 19 November 2010 bers inevitably involves multiple cracking originating from some of these spatially-distributed initial
Accepted 22 November 2010
imperfections, and therefore diverse cracking behaviors are expected. As a threshold value of crack prop-
Available online 26 November 2010
agation is approached, new cracking behaviors can abruptly emerge and replace the previous ones, caus-
ing strength degradation. In the present study, by applying sequential loads at different locations of the
Keywords:
same FE model of a notched beam, it is shown that this unique strength degradation mechanism can
Multiple cracks
Fatigue crack growth
repeatedly occur amid a variety of cracking behaviors, leading to a multistage, periodic reduction of
Strength the load-carrying capacity.
Damage accumulation 2010 Elsevier Ltd. All rights reserved.
Stress singularity

1. Introduction
In testing and analysis of various structures under fatigue, the
time-varying nature of cyclic loads (induced by winds, waves, vehi-
cles, etc.) is often oversimplied by considering loads of varying
amplitudes but at xed positions. Experimental studies have
shown that during fatigue tests on reinforced concrete beams
and plates, the use of a moving cyclic load in a simulation of trafc
loads may reduce the maximum load with respect to the one ob-
tained under xed-point monotonic or cyclic loading conditions
[1,2]. Though these experimental observations have shed some
light on the effects of changing the loading positions during cyclic
loading on the load-carrying capacity of a structure, the exact
cause of the reduction is not well understood.
A lot of research has been done to clarify the mechanisms of fa-
tigue and, despite the extremely complicated nature of the prob-
lem and its interrelated factors, it has been revealed that fatigue
is closely related to multiple crack activities during cyclic loading
[310]. However, it remains to be claried how the degradation
of material strength in structural members takes place during cyc-
lic loading amid various cracking activities [11]. In a series of
numerical studies on notched concrete beams, in which the size
of a specic notch was enlarged incrementally to calculate the
load-carrying capacity, there was a critical size at which the crack-
ing behavior and fracture process changed abruptly and the maxi-
Corresponding author. Tel.: +81 29 871 2032; fax: +81 29 871 2022.
E-mail address: a4739@n-koei.co.jp (Z. Shi).
mum load of the new failure mode was signicantly reduced [12].
Fractal concepts have been used to explain size effects in the study
of fatigue crack growth [13,14].
To obtain evidence of the transition of cracking behavior and
the subsequent reduction of the load-carrying capacity, an experi-
mental study has recently been conducted [15]. As is shown in
Fig. 1, the tests have focused on the maximum loads which can
be sustained by notched concrete beams under four-point bending
and on the corresponding failure modes. A multistage strength
degradation relation with notch size has been obtained by enlarg-
ing the sizes of several randomly-chosen initial notches. Numerical
analyses have also been carried out to reproduce the fracture pro-
cesses and calculate the maximum loads, which both have shown a
good agreement with the test results. The mix proportions of the
concrete used for the fracture tests are summarized in Table 1,
and the material properties of the test specimens are given in
Table 2: elastic modulus E, Poissons ratio m, compressive strength
fc, tensile strength ft, and fracture energy GF.
The importance of this study is twofold. Firstly, it reconrms the
previous nding that, for a given load condition, increasing the size
of an initial notch may not affect the potential failure mode and the
maximum load, providing that it is lower than a critical value.
When the size is increased beyond that threshold value, however,
a fundamental change in the failure mode can take place, and the
load-carrying capacity can signicantly drop. Secondly, the study
shows that this strength degradation process can repeatedly occur,
i.e., the load-carrying capacity decreases in multiple stages, and at
each stage a unique cracking behavior governs the fracturing
processes, while the strength of the previously weakened material

0142-1123/$ - see front matter 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijfatigue.2010.11.017
 Z. Shi et al. / International Journal of Fatigue 33 (2011) 11401150 1141

Fig. 1. Experimental and numerical results on multistage strength degradation due to change of failure mode in bending tests of notched concrete beams [15].

remains basically constant. Even though these observations are

based on studies performed on notched beams by altering the sizes
of initial notches, their implications for understanding the real
cracking behavior and the mechanism of strength degradation dur-
ing cyclic loading are obvious.
Initial aws and defects exist in all engineering materials and
thus in all structural members; only the degree of imperfection
varies. Under actual cyclic loading, not only the amplitude but also
the loading position may change. Eventually, the material weaken-
ing process in structural members caused by repeated loading
inevitably involves multiple cracking originating from some of
these spatially-distributed initial imperfections, and diverse crack-
ing behaviors are expected. Numerical simulations carried on by
applying sequential loads at different loading points have shown
that cracks can propagate from various locations of the notched
beam, so simulating a variety of cracking behaviors under cyclic
loading. Thus, the present study aims at showing that the multi-
stage strength degradation theory, (which in principle cannot be
veried directly on the same test specimen by experiments since
the load-carrying capacity can only be obtained by actually fractur-
ing the test specimen itself), can be corroborated on the basis of the
results of these numerical-simulations of crack propagation. To
achieve this, crack analyses under different loading conditions
need to be repeatedly carried out in correlation with each others.
The scheme of the study is explained in the following. Note that,
by applying a single load sequentially at different locations in the
same specimen, the stress tensors at the tip of a crack rotate.
2. Scheme of numerical study
As a further investigation on the multistage strength degrada-
tion phenomenon, the same notched beam problem of the previous
Table 2
Material properties of concrete used in fracture tests.
E (kN/mm2) m fc (N/mm2) ft (N/mm2) GF (N/mm)
29.2 0.18 34.3 3.17 0.142
study [15] discussed above has been selected (Fig. 2). The beam
contains ve initial notches equally spaced at an interval of
100 mm, except for the spacing between notches A and E, which
is deliberately set at a smaller value of 80 mm to embed an unsym-
metrical effect in the problem. The notch size is assumed to be
10 mm for notches A, B and C, and it is equal to 20 mm for notches
D and E. Further geometric details of the problem are shown in
Fig. 2.
In order to simulate crack propagation, sequential loads are
applied above the initial notches in four loading steps, following
a random order (D, E, C and A), as is shown in Fig. 2a. In each
loading step, a single load is applied to extend the crack from
the notch below to a certain length before moving to the next
loading point. Obviously, during each of the four loading
processes, multiple cracks may propagate from other notches,
too. At each step of crack growth the load-carrying capacity of
the beam under three-point bending is evaluated through crack
analysis, as is schematically illustrated in Fig. 2b, where the
cracks at the initial notches are treated as pre-existing cracks
for strength evaluation. Hence, crack analysis is carried out
alternately using two loading systems: sequential loading and
three-point bending.
The nite element (FE) model of the notched beam is shown in
Fig. 3, and the bilinear tension-softening relation for crack analysis

Table 1
Composition of concrete used in fracture tests.

Maximum size of coarse Slump (cm) Air (%) W/C ratio (%) Sand-coarse Water Cement Fine Coarse Admixture
aggregate (mm) aggregate ratio (%) aggregate aggregate AE (g/m3)
W (kg/m3) C (kg/m3) S (kg/m3) G (kg/m3)
15 11.5 4.8 55 46 178 324 810 965 16.20
1142 Z. Shi et al. / International Journal of Fatigue 33 (2011) 11401150

CL

(b)

(a)
Fig. 2. Scheme of numerical studies on a notched beam: (a) geometry of a notched beam and sequence of loading; and (b) computation of load-carrying capacity of the
notched beam with pre-existing cracks.

is illustrated in Fig. 4. The material properties used for the present

study are listed in Table 3.
To facilitate the convergence of numerical solutions amid the
load point movements, the crack path from an initial notch is as-
sumed to be vertical. This assumption avoids the need of crack-
path remeshing and makes the task of setting the pre-existing
cracks at the initial notches under three-point bending much eas-
ier. However, this simplication may also cause some minor inac-
curacy in the crack analysis, which will be addressed when
discussing the results of the numerical analyses.
For crack analysis the extended ctitious crack model (EFCM) is
employed, which solves a multiple-crack problem discretely
through crack equations that are formulated based on relevant
Fig. 3. FE model of the notched beam used for crack analysis. cracking modes; for details of the EFCM and its numerical imple-
mentation refer to Appendix A. Also, the singularity issue encoun-
tered when formulating the EFCM is claried in Appendix B. The
open-source computer program, CAIC-M1.FOR [16], is used for
the crack analysis in the following numerical studies.
ft
3. Results and discussion
Tensile stress

Wc = 5GF /ft Figs. 58 present four scenes of crack propagation which have
been selected as examples to summarize the results of crack anal-
W1 = 0.75GF /ft ysis under sequential loading and three-point bending, respec-
tively. Note that these cases are connected to two threshold
regions characterized by the transition of failure mode and the de-
ft/4 crease of load-carrying capacity under three-point bending, as the
size of a propagating crack approaches a critical value. The ob-
tained relation between the load-carrying capacity and crack
W1 Wc length is shown in Fig. 9, which clearly exhibits multistage
Crack-opening displacement strength degradation.
In case 1 (Fig. 5a), with a single load applied above notch D, a
Fig. 4. Bilinear tension-softening relation of concrete. crack propagates from that notch for six steps, accompanied by a
one-step growth of a crack from notch C. No cracks emerge from
other notches. Based on the FE mesh shown in Fig. 3, it is known
Table 3 that one step of crack growth equals the spacing between two
Material properties of a notched beam used for numerical analysis. neighboring nodes, i.e., 10 mm. As is shown in Fig. 5b, the initial
E (kN/mm2) m fc (N/mm2) ft (N/mm2) GF (N/mm) conditions set for calculating the load-carrying capacity under
three-point bending are illustrated in the notched beam by two
26.6 0.2 37.5 2.86 0.1
pre-existing cracks, i.e., a crack of 10 mm at notch C (one step of
 Z. Shi et al. / International Journal of Fatigue 33 (2011) 11401150 1143

Case 1

(a)

Note: Beam failure at B; Jump at steps 3, 14, 15

(b)
Fig. 5. Results of crack analysis in case 1: (a) crack propagation due to sequential loading; and (b) computation of load-carrying capacity under three-point bending with pre-
existing cracks.

Case 2

(a)
Note: Beam failure at D; Jump at step 2
(b)
Fig. 6. Results of crack analysis in case 2: (a) crack propagation due to sequential loading; and (b) computation of load-carrying capacity under three-point bending with pre-
existing cracks.

propagation) and a crack of 60 mm at notch D (six steps of propa-
gation). To account for the cumulative damage effect by cyclic
loads on the crack surfaces, i.e., smoothing out the rough crack sur-
faces due to their repeated opening and closing, the cohesive forces
at the pre-existing cracks are deliberately ignored in calculating
the beam strength. The results of crack analysis show that the
potential failure mode is beam fracture at notch B, and there is
no strength degradation of the beam up to this stage of crack prop-
agation, as is shown in Fig. 9.
In case 2 (Fig. 6a), if crack D propagates for one step further than
in case 1 as the load at notch D is increased, the failure mode of the
notched beam will abruptly change and the beam will break up at
1144 Z. Shi et al. / International Journal of Fatigue 33 (2011) 11401150

notch D, as is shown in Fig. 6b. Based on the strength-degradation
graph in Fig. 9, this failure mode transition will decrease the load-
carrying capacity by approximately 5% from the previous strength
level.
Case 3 (Fig. 7) and case 4 (Fig. 8) display cracking behaviors of
the beam in the second stage of the sequential loading: while
retaining the previous crack history, the sequential load is now
moved to notch E so to propagate cracks from that side of the
beam. As a result, two new cracks emerge at notches E and A, while
the older cracks at notches D and C are forced to close due to the
rotation of stress tensors at the tips of these cracks. As for cases
1 and 2, also cases 3 and 4 are two consecutive cases in terms of
crack growth: crack E propagates for ve steps in case 3 (Fig. 7a)
and six steps in case 4 (Fig. 8a). In both cases, crack A propagates
for only one step.
Since material damage due to fracture is irreversible, it is clear
that, in calculating the load-carrying capacity of the beam, the old-
er cracks from the previous loading history should also be treated

Case 3

Note: Beam failure at D; Jump at step 2

(a) (b)
Fig. 7. Results of crack analysis in case 3: (a) crack propagation due to sequential loading; and (b) computation of load-carrying capacity under three-point bending with pre-
existing cracks.

Case 4

Note: Beam failure at E

(a) (b)
Fig. 8. Results of crack analysis in case 4: (a) crack propagation due to sequential loading; and (b) computation of load-carrying capacity under three-point bending with pre-
existing cracks.
 Z. Shi et al. / International Journal of Fatigue 33 (2011) 11401150 1145

Fig. 9. Numerical results on multistage strength degradation due to change of cracking behavior and potential failure mode in a notched beam.

(a) (b)

(c) (d)

Fig. 10. Loaddisplacement relations of notched beams with pre-existing cracks under three-point bending: (a) pre-existing cracks due to sequential load 1; (b) pre-existing
cracks due to sequential load 2; (c) pre-existing cracks due to sequential load 3; and (d) pre-existing cracks due to sequential load 4.
1146 Z. Shi et al. / International Journal of Fatigue 33 (2011) 11401150
as pre-existing cracks (even though they are closed under the new
load system). Therefore, the notched beam under three-point
bending contains four pre-existing cracks in case 3 and case 4, as
is shown in Figs. 7b and 8b. The sizes of the pre-existing cracks
(in the following order: E, A, C, D) are equal to 50 mm, 10 mm,
10 mm and 70 mm for case 3, and 60 mm, 10 mm, 10 mm and
70 mm for case 4. As shown by the results of crack analysis under
three-point bending, the beam experiences a transition in failure
mode as the sequential load extends crack E from ve steps in case
3 to six steps in case 4, i.e., from notch D fracture to notch E failure.
As is shown in Fig. 9, the load-carrying capacity remains constant
up to the fth step of propagation of crack E, and then drops sud-
denly by approximately 5% at the cracks sixth step of extension.
As the sequential loading continues at notch C and then at notch
A so to propagate more cracks at different locations, the load-car-
rying capacity of the beam is shown to decrease in a similar fash-
ion. For details of the numerical analyses, refer to Fig. 9. Obviously,
the obtained multistage strength degradations can continue as
more loads are added to the sequential loading. In reality, however,
this process of strength degradation under repeated loading cannot
continue forever, because an abrupt structural failure will stop the
process when the excessively-reduced load-carrying capacity of
the beam can no longer sustain the design loads.
Concerning the failure modes in Fig. 9, as crack C propagates for
the rst three steps under the sequential load, the failure mode un-
der three-point bending changes, i.e., from E failure to B failure and
then to E failure again. This is believed to be the result of adopting
a xed-path strategy for crack analysis in the present study. If
crack-path remeshing is employed, the beam is expected to fail
only at notch E at this stage of crack propagation. However, this
is judged to be a minor inaccuracy, since the change of failure
mode from E to B has little inuence on the maximum load, as is
shown in Fig. 9. In other words, these two failure modes are com-
peting cracking modes, and with a slight change in the computa-
tional conditions, either of them can become the actual mode of
failure.
Fig. 11. An additional example on multistage strength degradation due to change of cracking behavior and potential failure mode in a notched beam.
Fig. 10ad presents the loaddisplacement relations obtained
under three-point bending for the notched beam with pre-existing
cracks. As expected, the greatest reduction in structural stiffness
occurs when the failure mode changes, as it is clearly shown in
each of the four gures that correspond to the four sequential
loads. To illustrate the existence of many possibilities for multi-
stage strength degradation, two more cases are studied under dif-
ferent conditions and their results are presented in Figs. 11 and 12,
respectively. The graph in Fig. 11 is obtained by reversing the order
of sequential loading of Fig. 2a and changing the loading condition
of Fig. 2b to four-point bending. The obtained relation demon-
strates once again the degradation of strength as the results of
changes in the cracking behaviors and the potential failure modes.
Fig. 12 shows another graph obtained by slightly enlarging notch B
in Fig. 2a to 20 mm and changing the sequential loads. Under these
conditions, it is found that all the potential failure modes under
three-point bending are the mid-span failure (fracture at B). Even
though there is no change in the potential failure mode here, the
periodically spontaneous propagation of crack B amid the multi-
ple-cracking activities in the beam does represent the sudden
occurrence of a new cracking behavior, which precipitates strength
degradation.
Concrete is a heterogeneous material consisting of aggregates
and cement paste bonded together at the interface, and the mate-
rial is inherently weak in tension due to the low bonding strength
and various pre-existing microcracks and aws formed during
hardening of the matrix. Consequently, fracture of concrete in-
volves complicated micro-failure mechanisms that include micro-
cracking, crack deection, crack branching, crack coalescence and
debonding of aggregates from the matrix. Obviously, each of these
micro-failure mechanisms can extensively develop under cyclic
loads, and their transitions (from one form to another in the hier-
archical order of crack formation) at certain threshold points of
crack formation and growth may signify the emergence of new
cracking behaviors and new failure modes. Based on the analysis
and results of the present study, the concrete weakening process
 Z. Shi et al. / International Journal of Fatigue 33 (2011) 11401150 1147

SCP(K)

Fig. 12. An example on multistage strength degradation due to occurrence of new cracking behavior without any change in the potential failure mode in a notched beam.

caused by cyclic loading appears to evolve in multiple stages of Fig. A1 illustrates two cracks of the mode-I type, crack A and
strength decrease. crack B, where crack propagation is set in the direction normal to
the tensile force at the tip of each ctitious crack. In formulating
4. Conclusions crack equations, subscripts a and b represent, respectively, crack
A and crack B, and l stands for the limit or critical value of a nodal
In the material weakening process of concrete under cyclic force (the tensile strength of concrete times the surface area re-
loading, strength decreases in successive stages amid multiple lated to a nodal point). Superscripts i, j and k denote the corre-
crack activities that involve changes in the cracking behaviors sponding nodes at designated cracks. For clarity, the cohesive
and with or without change in the potential failure mode. forces and the crack-opening displacements (CODs) of the inactive
Under cyclic loading, a threshold value of crack propagation ex- crack are marked by asterisks. Initially, crack A is assumed to be
ists, and the load-carrying capacity of a structural member can re- the sole propagating crack. Hence, the tensile force at its tip must
main constant until this value is approached. Beyond it, new reach the nodal force limit Qla, given by
cracking behaviors abruptly emerge and the load-carrying capacity
X
N X
M
is reduced. Q la CRa  Pa CIiaa F ia CIjab F j A1
b
This strength degradation can take place in multiple stages and, i1 j1
at each stage, a very different cracking behavior may dominate the
fracturing process while the maximum load remains basically un- where N and M are the number of nodes inside each ctitious crack,
changed. This process of strength degradation can continue until respectively. Note that the tensile forces at the tip of crack A, CRa,
the residual material strength of the structure can no longer sus- CIiaa and CIjab , are due to a unit external load, a pair of unit cohesive
tain the level of stress produced by the design load, leading to an forces at the ith node of crack A, and a pair of unit cohesive forces at
abrupt structural failure. the jth node of crack B, respectively. The external load Pa is the re-
quired load for propagating crack A, while crack B remains inactive.
Appendix A It should be noted that the tip force components due to the cohesive
forces of crack B in Eq. (A1) represent the crack interaction.
A.1. Formulation of the EFCM for mode I-type fracture The CODs along the two ctitious cracks are obtained as
follows:
In the discrete modeling of crack propagation, the fundamental
difference between single-crack and multiple-crack problems is X
N X
M
W ia BK ia  Pa AK ik k
aa F a AK ijab F j
b A2
whether the next-step of cracking behavior can be uniquely deter- k1 j1
mined or not. The mathematical formulation for a single-crack
problem is based on the mode of crack propagation, which is the X
N X
M
only valid mode of reaction for a single crack under tension. When W j j
b BK b  P a AK jiba F ia AK jk k
bb F b A3
multiple cracks are involved, however, the next-step of cracking i1 k1
behavior cannot be uniquely predicted because each crack pos-
sesses three potential modes of reaction to the external loading, where i = 1, . . . , N and j = 1, . . . , M. Here, the compliances BK ia at
i.e., crack propagation, crack arrest and crack closure. Combina- crack A and BK jb at crack B are due to the external load. The inu-
tions of these modes among multiple cracks generally lead to mul- ence coefcients AK ik ij
aa and AK ab are the CODs at the ith node of crack
tiple cracking modes, which must be considered when formulating A, due to a pair of unit cohesive forces at the kth node of crack A,
a multiple-crack problem. The following formulations are based on and a pair of unit cohesive forces at the jth node of crack B, respec-
the single-active-crack modes, in which only one crack is assumed tively. Similarly, the inuence coefcients AK jiba and AK jk bb represent
to be active while the rest remains inactive [17,18]. the CODs at the jth node of crack B due to a pair of unit cohesive
1148 Z. Shi et al. / International Journal of Fatigue 33 (2011) 11401150

forces at the ith node of crack A, and a pair of unit cohesive forces at simple models shown in Fig. A1ac. With the number of equations
the kth node of crack B, respectively. According to the reciprocity (2N + 2M + 1) matching the number of unknowns (2N + 2M + 1), the
theorem: AK ik ki jk kj ij ji
aa AK aa ; AK bb AK bb , and AK ab AK ba . problem can be solved uniquely because these equations are line-
Finally, imposing the tension-softening law of concrete along arly independent.
each ctitious crack, we get: When crack B is assumed to be the sole propagating crack, the
crack equations can be obtained readily from Eqs. (A1)(A5) by
F ia f W ia A4 interchanging the subscripts a and b, the superscripts i and j, the
number of nodes N and M, and by reassigning asterisks to the cohe-
F j j
b f W b A5 sive forces and the CODs of crack A. Upon solving the two sets of
crack equations, the true cracking mode is identied based on
where i = 1, . . . , N and j = 1, . . . , M. Eqs. (A1)(A5) form the so-called the minimum load criterion, which predicts the onset of crack
crack equations, that is, the conditions for crack A to propagate. propagation at the minimum load, that is:
Note that the various coefcients employed in these equations
can be obtained from linear elastic FE computations based on the P minPa ; Pb A6

Unit external loads (P = 1.0)

Tip tensile
Tip tensile force CRb
force CRa
j
Compliance BK b

Compliance BK ai
Crack B
Crack A
(a)

Tip tensile Tip tensile

Tip tensile k
force CI ba Tip tensile k
force CIbb
k
force CIaa k
force CIab F bk = 1.0
Influence
Fak = 1.0
jk Influence
coefficient AK ba
Fak = 1.0 Fbk = 1.0 jk
coefficient AK bb
Influence ik Influence ik
coefficient AKaa Crack B coefficient AK ab Crack B
Crack A Crack A
(b) (c)

External loads (Pa) External loads (Pb)

Limit nodal
Limit nodal force Qlb
force Qla
*j
Cohesive force F b Cohesive force F bj

Cohesive Cohesive
force F i Displacement W *j force F *i
a Displacement W bj
a b
i Crack B (inactive) *i Crack B (active)
Displacement W a Displacement W a
Crack A (active) Crack A (inactive)

(d) (e)
Fig. A1. Crack-tip-controlled modeling of multiple cracks: (a) under unit external loads; (b) by a pair of unit cohesive forces at crack A; (c) by a pair of unit cohesive forces at
crack B; (d) load condition for the growth of crack A; and (e) load condition for the growth of crack B.
 Z. Shi et al. / International Journal of Fatigue 33 (2011) 11401150
After setting the true crack path for the next-step of crack prop-
agation, the stress and displacement elds are calculated under the
condition of the obtained load and cohesive forces, as is shown in
Fig. A1d and e. This process can be repeated until structural failure
is reached. Obviously, the solution procedure previously described
can be readily extended to problems with an arbitrary number of
cracks.
It should be noted that the numerical results obtained need to
be checked carefully to eliminate invalid solutions that will be
encountered when an assumed cracking mode is irrelevant for
the problem. In a situation like this, the crack tip is readjusted by
releasing or closing the tip nodes, and the solution of the problem
is recalculated. By relocating the crack-tip position of an inactive
crack, other cracking modes with geometrically admissible strain
elds may emerge, including simultaneous propagations of several
cracks, and crack growth accompanied by crack closure. For further
details on the solution procedures and how to extend them to solve
a mixed-mode fracture problem, refer to Shi [15,19].
Appendix B
B.1. Clarication of the stress singularity issue in the EFCM
The issue of stress singularity in evaluating the tip tensile force
using the simple models in Fig. A1ac, where traction-free cti-
tious cracks are subjected to unit external loads and unit cohesive
forces, needs to be addressed [15,19]. Linear-elastic stress analysis
of sharp cracks predicts innite stresses at the crack tip. In an FE
computation this problem is rarely encountered because the ten-
sile stress at the crack tip is often obtained as the averaged stress
from the surrounding continuum elements. However, is the value
thus obtained valid? To clarify the problem, it is important to note
that the physical crack model as shown in Fig. A1d and e, where the
cohesive forces are applied in the fracture process zone (FPZ), does
not contain any stress singularity at the crack tip. The issue arises
only when this physical model is decomposed into sub-problems
with simpler loading conditions, based on the principle of superpo-
sition in elasticity theory, as is shown in Fig. A1ac.
Based on the theory of linear elastic fracture mechanics, a plas-
tic zone forms at the crack tip, and therefore the tip stresses in the
plastic zone remain nite. If the size of plastic zone is extremely
small and well within the singularity-controlled region, the devia-
tion of the stress eld obtained by linear-elastic stress analysis in
( yy) =0
ys
(= ft) KI
2 [8] Shields EB. Fracture prediction of hole patterns with multiple cracks using the
Fictitious crack
O Plastic zone
O r
rp
Fig. B1. Imaginary inelastic crack-tip behavior in concrete and estimate of plastic
 
zone size r p .
1149
the vicinity of the crack tip from the actual stress eld can be ig-
nored. This is considered to be the case in the crack analysis of con-
crete because the ctitious crack model assumes that the FPZ is
long and innitesimally narrow [20]. Consequently, the size of
the plastic zone at the ctitious crack tip, where imaginary yielding
is assumed to take place in correspondence of the tensile strength
of concrete, must be extremely small. An evaluation is given below.
On the crack plane (h = 0) in Fig. B1, the rst-order estimate for
the size of plastic zone, r p , and the COD at O, /, are obtained for
mode-I fracture and plane stress conditions [15], as follows:
 2
1 KI
rp B1
2p rys
4K 2I
/ B2
pErys
By eliminating the stress intensity factor KI from the above
equations and replacing the yield strength rys with the tensile
strength of concrete, ft, the size of plastic zone is obtained as
follows:
1E
rp / B3
8 ft
With E/ft being a material constant, rp is proportional to /. As
the COD is innitesimally narrow near the tip of the ctitious crack
(in numerical simulations the obtained near-the-tip COD is gener-
ally lower than Wc/100), an extremely small rp can be expected. For
example, assuming Wc = 0.1 mm and / = Wc/100 = 0.001 mm, and
E/ft = 6000 for ordinary concrete, Eq. (B3) leads to rp = 0.75 mm. If
the stress redistribution in the shaded area in Fig. B1 is taken into
account, the second-order estimate [15], i.e.:
 2
1 KI
rp 2rp B4
p rys
becomes rp = 2r p = 1.5 mm. Compared with the typical dimensions
of concrete structural members and sizes of cracks, the value of sev-
eral millimeters for the size of the plastic zone is indeed negligible.
Based on this analysis, the stress singularity in the numerical solu-
tion of the EFCM seems to be too weak to seriously affect the accu-
racy of the elastic solution and, therefore, it can be ignored.
References
[1] Okada K, Okamura H, Sonoda K, Shimada I. Cracking and fatigue behavior of
bridge deck RC slabs. Proc Jpn Soc Civil Eng 1982;321(5):4961.
[2] Kawaguchi M, Kawaguchi T, Harada K, Takahashi M. Fatigue tests of reinforced
concrete slab models of a highway bridge and an attempt to diagnose their
residual lives. Proc Jpn Soc Civil Eng 1987;380(I7):28392.
[3] Murakami Y. Metal fatigue - effects of small defects and nonmetallic
inclusions. Tokyo, Japan: Yokendo; 1993.
[4] To S, Lambert SB, Burns DJ. A multiple crack model for fatigue in welded joints.
Int J Fatigue 1993;15:33340.
[5] Martin-Meizoso A, Martinez-Esnaola JM, Fuentes-Perez M. Interactive effect of
multiple crack growth on fatigue. Theor Appl Fract Mech 1995;23:21933.
[6] Pitt S, Jones R. Multiple-site and widespread fatigue damage in aging aircraft.
Eng Fail Anal 1997;4:23757.
[7] Wu Z, Sun X. Multiple fatigue crack initiation, coalescence and growth in blunt
notched specimens. Eng Fract Mech 1998;59:3539.
nite element method. Int J Fatigue 2001;23:1320.
[9] Wessel C, Cisilino A, Santi O, Otegui J, Chapetti M. Numerical and experimental
determination of three-dimensional multiple crack growth in fatigue. Theor
Appl Fract Mech 2001;35:4758.
[10] Carstensen JV, Magnin T. Characterisation and quantication of multiple crack
growth during low cycle fatigue. Int J Fatigue 2001;23:195200.
[11] Carpinteri A. Energy dissipation in R.C. beams under cyclic loadings. Eng Fract
Mech 1991;39:17784.
[12] Shi Z, Suzuki M. Numerical studies on load-carrying capacities of notched
concrete beams subjected to various concentrated loads. Constr Build Mater
2004;18:17380.
[13] Carpinteri A, Spagnoli A. A fractal analysis of size effect on fatigue crack
growth. Int J Fatigue 2004;26:12533.
1150 Z. Shi et al. / International Journal of Fatigue 33 (2011) 11401150

[14] Carpinteri A, Spagnoli A, Vantadori S. A multi fractal analysis of fatigue crack
growth and its application to concrete. Eng Fract Mech 2010;77:97484.
[15] Shi Z. Crack analysis in structural concrete theory and
applications. Burlington, MA, USA: Elsevier; 2009.
[16] Shi Z. CAIC-M1. The companion website of crack analysis in structural concrete;
2009. <http://elsevierdirect.com/companion.jsp?ISBN=9780750684460>.
[17] Shi Z, Ohtsu M, Suzuki M, Hibino Y. Numerical analysis of multiple cracks in
concrete using the discrete approach. J Struct Eng 2001;127:108591.
[18] Shi Z, Suzuki M, Nakano M. Numerical analysis of multiple discrete cracks in
concrete dams using extended ctitious crack model. J Struct Eng
2003;129:32436.
[19] Shi Z. Numerical analysis of mixed-mode fracture in concrete using extended
ctitious crack model. J Struct Eng 2004;130:173847.
[20] Gerstle WH, Xie M. FEM modeling of ctitious crack propagation in concrete. J
Eng Mech 1992;118:41634.