You are on page 1of 10

Engineering Structures 126 (2016) 618–627

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Crack width analysis of reinforced concrete under direct tension by finite


element method and crack queuing algorithm
A.K.H. Kwan ⇑, F.J. Ma
Department of Civil Engineering, The University of Hong Kong, Hong Kong, China

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

Article history: For serviceability and durability, the maximum crack width of a reinforced concrete structure has to be
Received 18 November 2015 controlled to within certain limits, depending on the environmental conditions. However, crack width
Revised 10 August 2016 analysis is not easy because the crack width is associated with the bond slip of reinforcing bars at indi-
Accepted 12 August 2016
vidual cracks, which has to be allowed for using discrete modelling of the cracks and reinforcing bars.
Available online 24 August 2016
Herein, a new finite element method for discrete crack analysis, which models the reinforcing bars as dis-
crete bar elements and uses interface elements to allow for the bond slip, is developed. It does not use any
Keywords:
discrete crack elements and therefore there is no need to generate discrete crack elements to follow the
Crack width
Crack queuing algorithm
crack propagation. Instead, it employs a crack queuing algorithm to simulate the stress redistribution
Finite element analysis during cracking. The method is applied to analyze tension specimens tested by others to verify its accu-
racy and applicability, and then used to study the effects of various parameters on the crack width.
Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction bars. There are a number of relatively simple one-dimensional ana-


lytical or numerical simulation methods for crack analysis [8–15],
In the design of reinforced concrete structures, both the ulti- but these are applicable only in the simplest cases and some are
mate limit state (ULS) and the serviceability limit state (SLS) have only for the evaluation of tension stiffening [11,12,14]. For the
to be considered. Most previous research focused on the structural analysis of crack formation, crack propagation and crack width
behavior at the ULS. However, whilst the ULS is important from the development in the general cases, the more rigorous and powerful
point of structural safety, the SLS is at least as important because finite element analysis method has to be employed.
the performance at the SLS affects not only the serviceability but Historically, for crack analysis, both the discrete crack model
also the durability. For instance, the formation of cracks in concrete and smeared crack model have been attempted. The discrete crack
could cause aesthetic, water leakage and steel bar corrosion prob- model requires either adaptive re-meshing to allow for crack prop-
lems. Somehow, there has been little research on the crack forma- agation during the iteration process of nonlinear analysis as pro-
tion and crack width of reinforced concrete structures. Up to now, posed by Ngo and Scordelis [16], or pre-determination of crack
although all design codes impose certain limits on the crack width, locations to simplify the re-meshing process as proposed by Kana-
there is no established analytical method for the evaluation of kubo et al. [17]. However, the computer coding is fairly compli-
crack width. What we have are only empirical formulas in the cated and the computer time needed is often much too long. In
design codes, such as BS 8110-2: 1985 [1], ACI 224R-01 [2] and contrast, the smeared crack model assumes that the whole con-
Eurocode 2: 2004 [3], which do not agree with each other. crete element would crack or, in other words, the crack formed
The relatively little research on crack analysis [4,5] is due could be smeared into many minor cracks throughout the concrete
mainly to the fact that the conventional smeared crack analysis element as proposed by Rashid [18]. No adaptive re-meshing or
would not give any crack widths directly, although there are some pre-determination of crack locations is needed and thus the
sophisticated methods of estimating crack widths by global track- smeared crack model generally requires less computer power
ing [6,7]. Hence, discrete crack analysis is preferred. Moreover, the and time. That is why most existing finite element analysis pro-
crack widths are dependent on the bond slip of steel bars, which grams are based on the smeared crack model. However, although
can only be accounted for by modelling the steel bars as discrete the tensile strain across the cracks in the cracked element can be
evaluated, since the tensile strain across a crack is actually depen-
dent on the gauge length [19], the crack width cannot be deter-
⇑ Corresponding author. mined from the tensile strain.
E-mail address: khkwan@hku.hk (A.K.H. Kwan).

http://dx.doi.org/10.1016/j.engstruct.2016.08.027
0141-0296/Ó 2016 Elsevier Ltd. All rights reserved.
A.K.H. Kwan, F.J. Ma / Engineering Structures 126 (2016) 618–627 619

In addition, there is the issue of whether the cracks in concrete 2.1. Crack queuing algorithm
can rotate after formation. In this regard, there are two schools of
thought. Those who adopt the rotating crack model [18,20,21] In conventional finite element methods, at each iteration step,
assume that the cracks can rotate whereas those who adopt the every concrete element is judged one by one against the cracking
non-rotating crack model [22,23] assume that the cracks cannot criteria and all concrete elements satisfying the cracking criteria
rotate after formation. There is also the issue of how the steel rein- are allowed to crack at the same time. This would lead to the for-
forcing bars should be modelled. Two different approaches, mation of many closely spaced parallel cracks as if the crack is
namely: the discrete bar approach of modelling the steel bars by smeared throughout the adjacent concrete, but in reality, right
discrete bar elements [24–27] and the smeared bar approach of after the formation of the first crack at where the tensile stress/
smearing the steel bars within the concrete element [28,29], have strain is highest, the concrete stresses would be redistributed
been adopted. The discrete bar approach is capable of allowing for and the tensile stress perpendicular to the crack would be relieved
the bond slip of steel bars by inserting interface elements between to prevent the adjacent concrete from cracking. Hence, it is impor-
the discrete bar elements and the concrete elements but the tant to cater for the immediate stress redistribution caused by the
smeared bar approach is not capable of allowing for the bond slip formation of the first crack before considering the possible forma-
of steel bars. tion of other cracks.
In conventional finite element analysis of concrete cracking, To cater for the immediate stress redistribution due to crack
many closely spaced parallel cracks are often formed. This is not formation, the crack queuing algorithm [30] is employed. With this
reasonable because after the formation of one crack, the tensile algorithm, the external loading is applied in small increments
stress perpendicular to the crack should be immediately relieved starting from zero and at each loading step, an iteration procedure
due to stress redistribution and the adjacent concrete should not of allowing at most one concrete element to crack at one iteration
crack. To remedy this situation, the first author had developed a step is executed. At each iteration step, the concrete element hav-
crack queuing algorithm of allowing at most one concrete element ing the highest tensile stress or strain (depending on the cracking
(the one with the highest tensile stress or strain, depending on the criteria) is identified. If the cracking criteria have been met, then
cracking criteria) to crack at one time [30]. After allowing one con- this concrete element is allowed to crack. No other concrete ele-
crete element to crack, the whole structure is re-analyzed to cater ment is allowed to crack. After cracking, the stiffness matrix of
for the stress relief caused by the formation of the crack before the cracked concrete element is reformulated with the secant stiff-
considering the possible formation of any other crack. This would ness perpendicular to the crack reduced to a very small value.
avoid the formation of closely spaced parallel cracks, which should Then, the whole structure is reanalyzed so as to cater for the stress
not happen at all. redistribution due to cracking. This iteration procedure is repeated
Recently, the authors have developed a new discrete crack until no more concrete element satisfies the cracking criteria.
model for crack width analysis, which does not require any dis- When no more crack is formed, the nonlinear analysis proceeds
crete crack elements or adaptive re-meshing of finite elements to the next loading step by adding another loading increment to
near crack tips, and is therefore computationally more efficient the structure. The crack patterns so generated by such finite ele-
and easier to apply [31]. The non-rotating crack model and discrete ment method and crack queuing algorithm generally show clearly
bar approach are adopted. Moreover, the crack queuing algorithm defined discrete cracks well separated from each other, which
is employed to cater for the immediate stress relief during crack should be more realistic than the apparently smeared crack pat-
formation so as to generate a realistic crack pattern. The validity terns generated by the conventional finite element methods.
of the new discrete crack model has been verified by applying to
analyze beam specimens tested under flexure by others and com-
paring the theoretical results with the experimental results. In this 2.2. Crack width calculation
paper, this discrete crack model is extended to incorporate quadri-
lateral elements and deal with both deformed bars and smooth No special crack element is used to simulate the formation of
bars. The extended discrete crack model is first applied to analyze cracks in concrete. Instead, the crack is assumed to be formed
concrete specimens tested under direct tension by others to verify inside the cracked concrete element at the centroid, as shown in
its accuracy and applicability, and then used to conduct parametric
studies on the crack width of tension members with different
vK
material properties. The finite element analysis results are also vL
compared to the formulas given in the design codes to find out uK
how these formulas perform.
uL
Node L Node K

Concrete
2. Finite element method y
element

As the details of the finite element method have been given


before in a previous paper [31], only the additional features and α
x
the special numerical techniques are presented herein. For more
general applications, the finite element method is extended by Normal to crack
incorporating a 2-dimensional, 4-noded quadrilateral element for
the concrete and a bond-slip constitutive model applicable to both vJ
deformed and smooth bars for the steel bar-concrete bond. Other- Crack uJ
wise, as before, a 1-dimensional, 2-noded discrete bar element is Node J
used for the steel bars and a 1-dimensional, 4 noded interface ele- vI
ment is used for the steel bar-concrete interfaces. To capture the
Node I
post-crack and post-peak behavior of the reinforced concrete uI
structure, the same secant stiffness iteration method as before is
also adopted for the nonlinear analysis [32]. Fig. 1. Crack width calculation.
620 A.K.H. Kwan, F.J. Ma / Engineering Structures 126 (2016) 618–627

Fig. 1. Numerically, the tensile strain across the crack can be calcu- Moreover, the Poisson’s ratio of the concrete is assumed to be 0.2
lated from the nodal displacements of the cracked concrete ele- and the fracture toughness is obtained from the research of the
ment. However, since the tensile strain across a crack is first author [37] for all the specimens. For the steel bar-concrete
dependent on the gauge length and therefore is not well defined, bond-slip properties, the bond-slip properties for deformed bars
the crack width should not be evaluated from the tensile strain and smooth bars given in CEB-FIP Model Code 1990 [38] are
across the crack. To overcome this problem, the crack width is adopted. From Table 3, it can be seen that for deformed bars, the
evaluated directly from the nodal displacements. Firstly, at each peak bond stress and residual bond stress range from 9.3 to
side of the crack, the node furthest from the crack is identified. 12.5 MPa and from 1.4 to 1.9 MPa, respectively, whereas for
Let the node at the left hand side furthest from the crack be the smooth bars, the peak bond stress and residual bond stress are
node L, and the node at the right hand side furthest from the crack both equal to 1.5 MPa. Besides, the bond slips at peak bond stress
be the node J, as in Fig. 1. The displacement of node L away from for deformed bars and smooth bars are 0.6 mm and 0.1 mm,
the crack dL and the displacement of node J away from the crack respectively. Since the initial bond stiffness (the gradient of the ini-
dJ are given respectively by: tial portion of the bond stress-slip curve) given by CEB-FIP Model
Code 1990 is infinitely large, a finite value equal to the secant stiff-
dL ¼ uL cos a  v L sin a ð1aÞ ness at one-fifth of the peak bond stress is taken as the initial bond
stiffness to avoid numerical difficulties.
dJ ¼ þuJ cos a þ v J sin a ð1bÞ The test set-up with a tension force T applied to each end of the
in which uL and v L are the displacements of node L in the x- and y- specimen is shown in Fig. 2(a) whilst the mesh used for the finite
directions; uJ and v J are the displacements of node J in the x- and y- element analysis is shown in Fig. 2(b). In the analysis of the first set
directions; and a is the crack angle (angle between the normal to of specimens, the concrete member is meshed into 10  350 num-
the crack and the x-axis). Then, the crack width w can be calculated bers of 2-dimensional concrete elements. On the other hand, the
by summing up the displacements of the two nodes away from the steel reinforcing bar is meshed into 350 numbers of 1-
crack, as in the following equation. dimensional steel bar elements. Besides, there are 350 numbers
of steel bar-concrete interface elements connecting the steel bar
w ¼ dL þ dJ ð2Þ to the concrete. In the analysis of the second set of specimen, the
concrete, steel bar and steel bar-concrete interface are meshed into
When the crack width w is negative, the crack would be under com-
10  230 concrete elements, 230 steel bar elements and 230 inter-
pression and may be regarded as closed.
face elements. In the analysis of the third set of specimens, the con-
crete, steel bar and steel bar-concrete interface are meshed into
3. Crack width analysis of tension specimens 10  220 concrete elements, 220 steel bar elements and 220 inter-
face elements. Finer meshes have been tried but almost identical
In order to verify the accuracy and applicability of the finite ele- numerical results have been obtained indicating that the meshes
ment method, three sets of tension specimens tested by others are used are fine enough to achieve convergent results.
analyzed. The first set are those made with deformed or smooth
bars tested by Radnić and Markota [33]. The second set are those 3.2. Stress redistribution during cracking
made with deformed bars tested by Farra and Jaccoud [34], and
the third set are those made with deformed bars tested by Wu Before studying the crack results, the stress redistribution dur-
and Gilbert [35]. ing cracking are presented by plotting the distributions of concrete
tensile stress, steel stress and bond slip of a typical specimen, D12-
3.1. Tension specimens analyzed RA, in Figs. 3–5. In each of these figures, the stress distributions at
three stages (before formation of first crack when T = 10.4 kN; after
In the first set, there are totally 6 tension specimens. All the formation of first crack when T = 11.0 kN; and after formation of
specimens have a length of 700 mm and a cross section of three cracks when T = 13.6 kN) are plotted to see the changes dur-
70 mm  70 mm. In each specimen, only one steel reinforcing bar ing cracking.
is embedded at the centre of the concrete section. Three specimens From Fig. 3, it can be seen that before cracking, the concrete ten-
(D8-RA, D10-RA and D12-RA) are made with deformed bars sile stress increases from zero at one end to a maximum value at
whereas the other three specimens (D8-GA, D10-GA and D12- the middle and then decreases to zero at the other end. After initi-
GA) are made with smooth bars. The steel ratios (steel area to con- ation of the first crack, the tensile stress at the crack drops to zero
crete area ratios) are 1.03%, 1.6% and 2.31% for the three specimens and the tensile stress elsewhere is redistributed such that the ten-
made with either deformed or smooth bars. sile stress variation from one end to one crack is symmetrical with
In the second set, there is only one tension specimen (S20-14). zero stress at one end or crack and maximum stress at mid-way.
It has a length of 1150 mm, a cross section of 100 mm  100 mm Upon further loading, more cracks are formed and the tensile stress
and one steel reinforcing bar embedded at the centre of the con- redistributes again such that at a crack, the tensile stress is zero
crete section. The steel reinforcing bar is a deformed bar and the and at mid-way between cracks, the tensile stress is maximum.
steel ratio is 1.54%. From Fig. 4, it is evident that before cracking, the steel stress is
In the third set, there are two tension specimens (STN-12 and maximum at the two ends and minimum at the middle. After a
STN-16). Both specimens have a length of 1100 mm, a cross section crack is formed, the steel stress at the crack increases sharply to
of 100 mm  100 mm and one steel reinforcing bar embedded in a certain peak value whereas the steel stress elsewhere is redis-
each specimen. The steel bars embedded are deformed bars and tributed such that the steel stress is maximum at one end or crack
the steel ratios are 1.14% and 2.05%. and minimum at mid-way between one end and the crack. When
Tables 1–3 list the dimensions and steel properties, the concrete more cracks are formed, the steel stress redistributes again with
properties, and the bond properties, respectively, of the tension a peak occurring at each crack.
specimens analyzed. The concrete compressive and tensile From Fig. 5, it is found that before cracking, the maximum bond
strengths are as given in the references [33–35]. Where the elastic slip occurs at the two ends. As the first crack is formed, the bond
modulus of the concrete is not given, it is calculated from the con- slip near the newly formed crack changes abruptly such that the
crete compressive strength as per the formula in ACI 318R-02 [36]. concrete at the left hand side slips to the left away from the crack
A.K.H. Kwan, F.J. Ma / Engineering Structures 126 (2016) 618–627 621

Table 1
Dimensions and steel properties of tension specimens.

Specimen Steel bar diameter (mm) Steel bar type Cross section Steel ratio (%) Steel yield strength Steel ultimate strength
number (mm  mm) (MPa) (MPa)
D8-RA 8 Deformed 70  70 1.03 400 500
D10-RA 10 bar 70  70 1.60 400 500
D12-RA 12 70  70 2.31 400 500
D8-GA 8 Smooth bar 70  70 1.03 500 560
D10-GA 10 70  70 1.60 500 560
D12-GA 12 70  70 2.31 500 560
S20-14 14 Deformed 100  100 1.54 540 600
STN-12 12 bar 100  100 1.14 540 600
STN-16 16 100  100 2.05 540 600

Notes: (1) In each specimen, there is only one steel bar. (2) The initial elastic modulus, tensile strain at start of strain hardening and ultimate tensile strain are taken as
200 GPa, 1.0% and 10.0%, respectively.

Table 2
2.4
Concrete properties of tension specimens.
T = 10.4 kN, before formation of any crack
Specimen Compressive Tensile Elastic Poisson’s Fracture T = 11.0 kN, after formation of 1 crack
number strength strength modulus ratio toughness 2.0 T = 13.6 kN, after formation of 3 cracks
(MPa) (MPa) (GPa) (MN m2/3)

Concrete stress (MPa)


D8-RA 24.1 1.8 23,244 0.2 1.3
D10-RA 24.1 1.8 23,244 0.2 1.3 1.6
D12-RA 24.1 1.8 23,244 0.2 1.3
D8-GA 24.1 1.8 23,244 0.2 1.3
D10-GA 24.1 1.8 23,244 0.2 1.3 1.2
D12-GA 24.1 1.8 23,244 0.2 1.3
S20-14 38.8 2.8 29,300 0.2 1.3
STN-12 21.6 2.0 22,400 0.2 1.3 0.8
STN-16 21.6 2.0 22,400 0.2 1.3

0.4

Table 3
Bond properties of tension specimens. 0.0
0 100 200 300 400 500 600 700
Specimen Peak s1 s2 s3 Residual Initial bond Distance along specimen (mm)
number bond (mm) (mm) (mm) bond stiffness (N/
stress stress mm3) Fig. 3. Distribution of concrete tensile stress in specimen D12-RA.
(MPa) (MPa)
D8-RA 9.8 0.6 0.6 1.0 1.5 183.0
D10-RA 9.8 0.6 0.6 1.0 1.5 183.0 and the concrete at the right hand side slips to the right away from
D12-RA 9.8 0.6 0.6 1.0 1.5 183.0
the crack. On the other hand, the steel bar always slips relative to
D8-GA 1.5 0.1 0.1 0.1 1.5 73.7
D10-GA 1.5 0.1 0.1 0.1 1.5 73.7 the concrete towards the crack. As more cracks are formed, the
D12-GA 1.5 0.1 0.1 0.1 1.5 73.7 bond slip redistributes again with the nearby concrete slipping
S20-14 12.5 0.6 0.6 1.0 1.9 232.2 away from the newly formed cracks. Hence, at each crack, the bond
STN-12 9.3 0.6 0.6 1.0 1.4 173.3 slip at one side of the crack and the bond slip at the other side of
STN-16 9.3 0.6 0.6 1.0 1.4 173.3
the crack are always opposite in direction, thus causing discontinu-
Note: s1, s2 and s3 are the slip at start of peak bond stress, slip at end of peak bond ity of the bond slip there.
stress and slip at start of residual bond stress, respectively. The above stress distributions could not be obtained if the crack
queuing algorithm is not adopted in the finite element analysis.

Steel reinforcing bar Concrete

T T

(a) Test set-up

T T

(b) Finite element mesh


Fig. 2. Test set-up and finite element mesh.
622 A.K.H. Kwan, F.J. Ma / Engineering Structures 126 (2016) 618–627

240 Among the three sets of specimens, only the specimen S20-14
T = 10.4 kN, before formation of any crack tested by Farra and Jaccoud [34] and the specimens STN-12 and
T = 11.0 kN, after formation of 1 crack STN-16 tested by Wu and Gilbert [35] have been provided with
200 T = 13.6 kN, after formation of 3 cracks measured crack number and spacing results. The crack number
and spacing results at the final loading stage of these specimen
obtained analytically by finite element analysis are compared to
Steel stress (MPa)

160
the measured results in Table 4. From the comparison, it can be
seen that at the final loading stage, the finite element analysis
120 tends to slightly overestimate the crack number and underesti-
mate the crack spacing. Actually, the crack number and spacing
vary with the loading or steel stress. For more detailed comparison,
80 the measured and analytical crack spacing results for STN-12 and
STN-16 are plotted against the steel stress in Fig. 7. From the
40 curves plotted, it is found that the finite element analysis tends
to overestimate the steel stress at first crack (or predict a later for-
mation of the first crack) but later on tends to overestimate the
0 crack number and underestimate the crack spacing (or produce
0 100 200 300 400 500 600 700
more cracks at closer spacing) at higher steel stresses.
Distance along specimen (mm)

Fig. 4. Distribution of steel stress in specimen D12-RA. 3.4. Crack width results

For the first set of specimens tested by Radnić and Markota [33],
the analytical and measured maximum crack widths are compared
0.06 T = 10.4 kN, before formation of any crack in Figs. 8 and 9, where the crack widths predicted by the design
T = 11.0 kN, after formation of 1 crack formulas in ACI 224R-01 [2] and Eurocode 2: 2004 [3] are also plot-
T = 13.6 kN, after formation of 3 cracks ted to see if they agree with the analytical or measured results.
0.04 Fig. 8 presents the crack width results of D8-RA, D10-RA and
D12-RA, which are made with deformed bars. For the specimen
Bond slip (mm)

0.02
D8-RA, the analytical crack width is somewhat larger than the
measured result when the steel stress is equal to 240 MPa, but
the discrepancy gradually diminishes to almost zero when the
0.00 steel stress increases to 400 MPa. For the specimen D10-RA, the
analytical and measured crack widths agree with each other at
all steel stress levels to within 7% error. For the specimen D12-
-0.02 RA, the analytical crack width is slightly larger than the measured
result when the steel stress is lower than 250 MPa but slightly
smaller than the measured result when the steel stress is higher
-0.04
than 250 MPa. Overall, the analytical crack widths of the speci-
mens made with deformed bars agree quite well with the mea-
0 100 200 300 400 500 600 700
sured results. Lastly, comparing the design formulas with the
Distance along specimen (mm)
analytical and measured crack widths, it can be seen that the
Fig. 5. Distribution of bond slip in specimen D12-RA. design formula in ACI 224R-01 fits better with the analytical and
measured results than the design formula in Eurocode 2: 2004,
which tends to overestimate the crack width.
Fig. 9 presents the crack width results of D8-GA, D10-GA and
3.3. Crack pattern and crack spacing results D12-GA, which are made with smooth bars. For the specimens
D8-GA and D10-GA, the analytical crack width is generally larger
The crack patterns at different loading stages of the typical than the measured result at all steel stress levels. However, the dis-
specimen D12-RA obtained analytically by finite element analysis crepancy between the analytical and measured crack widths is not
are compared to the measured results by Radnić and Markota entirely caused by the finite element analysis. The crack width
[33] in Fig. 6. When T = 11 kN, 4 cracks could be seen in the mea- should be dependent on the steel bar size but the measured crack
sured crack pattern but only 1 crack appears in the analytical crack widths of D8-GA (bar size = 8 mm) and D10-GA (bar size = 10 mm)
pattern. At T = 20 kN, the measured crack number increased to 5 are very much the same, indicating that there are some random
but the analytical crack number is only 3. At T = 30 kN, the mea- variations or errors in the experiment. For the specimen D12-GA,
sured crack number was 7 whereas the analytical crack number the analytical crack width agrees quite well with the measured
is 6. Then, at T = 40 kN, the measured crack number increased to result at all steel stress levels, although it tends to be slightly lar-
12 and the analytical crack number becomes 15. Overall, the mea- ger. On the whole, the analytical crack widths of the specimens
sured crack pattern and spacing show some randomness but the made with smooth bars are significantly larger than the measured
analytical crack pattern and spacing are fairly regular. This is results. One possible reason is the overly conservative peak bond
because in reality, the intrinsic random variations in material prop- stress in the bond-slip model adopted in the finite element analy-
erties could significantly affect the loading and locations at which sis. The bond-slip model adopted is the one given in CEB-FIP Model
the cracks are formed but in the finite element analysis, the mate- Code 1990 [38]. As can be seen from Table 3, for smooth bars, the
rial properties are assumed to be perfectly uniform. Nevertheless, peak bond stress and residual bond stress are both taken as
the finite element method augmented with a crack queuing algo- 1.5 MPa. In theory, the bond strength (the peak bond stress) should
rithm has successfully produced crack patterns showing clearly be higher than the residual bond stress, but due to its sensitivity to
defined discrete cracks. surface condition and unreliability, the bond strength of smooth
A.K.H. Kwan, F.J. Ma / Engineering Structures 126 (2016) 618–627 623

Measured crack pattern when T = 10 kN

Analytical crack pattern when T = 11 kN

Measured crack pattern when T = 20 kN

Analytical crack pattern when T = 20 kN

Measured crack pattern when T = 30 kN

Analytical crack pattern when T = 30 kN

Measured crack pattern when T = 40 kN

Analytical crack pattern when T = 40 kN

Fig. 6. Crack patterns of D12-RA.

Table 4 of the specimens STN-12 and STN-16, two sets of measured results
Crack number and crack spacing of S20-14, STN-12 and STN-16. are given, namely: measured maximum crack width and measured
Specimen Measured Analytical Measured Analytical average crack width. This was probably because the measured
number crack crack crack spacing crack spacing crack width varied from one crack to another and thus both the
number number (mm) (mm) maximum and average values are given. However, the finite ele-
S20-14 6 7 192 165 ment analysis, which assumes that the material properties are per-
STN-12 5 7 195 138 fectly uniform, yields uniform crack widths and thus more or less
STN-16 5 9 165 104
the same analytical maximum crack width and analytical average
crack width. Comparing, it can be seen that for the specimen
STN-12, the analytical maximum and average crack widths are lar-
bars has been taken very conservatively to be the same as the ger than the measured average crack width but smaller than the
residual bond stress. It is the authors’ belief that the actual bond measured maximum crack width. For the specimen STN-16, the
strength should be higher than the residual bond stress and that analytical maximum and average crack widths agree very well
if the actual measured bond strength is used in the finite element with the measured average crack width but are substantially smal-
analysis, smaller and more accurate crack width results could be ler than the measured maximum crack width. Generally, the agree-
obtained. Lastly, it is found that the design formula in ACI 224R- ment between the analytical crack width and the measured
01 fits quite well with the measured results but the design formula average crack width is reasonably good.
in Eurocode 2: 2004 does not fit well with the measured results. To evaluate the overall accuracy of the crack width analysis, the
For the second and third sets of specimens tested by Farra and analytical crack widths at various steel stress levels of the nine
Jaccoud [34] and Wu and Gilbert [35], the analytical and measured specimens analyzed are plotted against the corresponding mea-
crack widths are compared in Fig. 10. All these specimens are made sured crack widths in Fig. 11. For the specimens STN-12 and
with deformed bars. STN-16, the measured average crack widths, rather than the mea-
For the specimen S20-14, the analytical crack width is generally sured maximum crack widths, are taken as the measured crack
smaller than the measured result at all steel stress levels. For each widths. In the figure, the data points for specimens made with
624 A.K.H. Kwan, F.J. Ma / Engineering Structures 126 (2016) 618–627

600 0.5

500 STN-12 Measured Measured


0.4

Crack width (mm)


Analytical
Crack spacing (mm)

Analytical
ACI 224R
400 0.3 Eurocode 2
300
0.2
200 D8-RA
0.1
100
0.0
0 50 100 150 200 250 300 350 400 450
0
0 50 100 150 200 250 300 350 400 450 500 550 Steel stress (MPa)
Steel stress (MPa)
0.5
600
Measured
0.4

Crack width (mm)


Measured Analytical
500
STN-16 ACI 224R
Crack spacing (mm)

Analytical
0.3 Eurocode 2
400
0.2
300
0.1 D10-RA
200

100 0.0
0 50 100 150 200 250 300 350 400 450

0 Steel stress (MPa)


0 50 100 150 200 250 300 350 400 450 500 550
Steel stress (MPa) 0.5

Measured
Fig. 7. Crack spacings of STN-12 and STN-16. 0.4
Crack width (mm)

Analytical
ACI 224R
0.3 Eurocode 2
deformed bars are marked by square symbols whereas the data
points for specimens made with smooth bars are marked by circu- 0.2
lar symbols. Statistical analysis of the square data points yields a D12-RA
mean error of 3.7% and a mean absolute error of 13.5%. Likewise, 0.1
statistical analysis of the circular data points yields a mean error of
0.0
36.6% and a mean absolute error of 37.5%. Hence, the finite element 0 50 100 150 200 250 300 350 400 450
method augmented with the crack queuing algorithm is suffi- Steel stress (MPa)
ciently accurate for predicting the crack width of tension members
made with deformed bars. For more accurate prediction of the Fig. 8. Crack widths of D8-RA, D10-RA and D12-RA.
crack width of tension members made with smooth bars, a better
bond-slip model is needed.
increasing steel stress. However, the crack width decreases with
increasing concrete compressive strength. The effect of concrete
4. Parametric studies compressive strength is not really large, but with the concrete
compressive strength increased from 20 MPa to 60 MPa, the crack
The proposed finite element method, which has been verified width could be decreased by about 10%. This is because as the con-
by comparing with experimental results, is now employed to study crete compressive strength increases, the bond strength also
the effects of various structural parameters on the crack width of increases, thus providing better bond slip restraints against widen-
tension members. Two parametric studies have been carried out, ing of the cracks.
one on the effect of concrete compressive strength and the other
on the effects of steel ratio and bar size.
4.2. Effects of steel ratio and bar size

4.1. Effect of concrete compressive strength In recent years, it has become more and more popular to use
high-strength steel with yield strength substantially higher than
In this parametric study, tension specimens similar to those before. This can reduce the amount of steel needed but the working
tested by Radnić and Markota [33] are analyzed using different val- stress of the steel reinforcement (usually around 60% of the yield
ues of concrete compressive strength varying from 20 to 60 MPa in strength) would tend to be higher, leading to potentially larger
steps of 10 MPa. The section size and steel bar size are set constant crack width of the reinforced concrete structure. The potentially
as 70 mm  70 mm and 12 mm, respectively. Only deformed bars larger crack width may cause both serviceability and durability
are used. The elastic modulus of concrete is calculated from the problems but it is still unclear how these problems could be dealt
concrete strength using the formula in ACI 318R-02 [36] and the with. Most of the design codes are limiting the use of high-strength
bond-slip properties are taken as those given in CEB-FIP Model steel. For instance, ACI 318R-02 [36] limits the use of steel with
Code 1990 [38]. For the finite element analysis, the same finite ele- yield strength exceeding 552 MPa and AASHTO [39] only permits
ment mesh as before is employed. The analytical crack width so the use of steel reinforcing bars with yield strength no greater than
obtained is plotted against the steel stress in Fig. 12. As expected, 517 MPa. Detailed study of how the steel stress and other param-
for all the specimens analyzed, the crack width increases with eters related to the steel reinforcement would affect the crack
A.K.H. Kwan, F.J. Ma / Engineering Structures 126 (2016) 618–627 625

0.6
0.8
Measured D8-GA Measured
0.5
Crack width (mm)

Analytical

Crack width (mm)


Analytical
0.6 ACI 224R
0.4
Eurocode 2
0.4 0.3

0.2
0.2 S20-14
0.1
0.0 0.0
0 50 100 150 200 250 300 350 400 450 500 0 50 100 150 200 250 300 350 400 450 500 550
Steel stress (MPa) Steel stress (MPa)

0.6
0.8 Measured maximum crack width
Measured D10-GA 0.5 Measured average crack width
Analytical
Crack width (mm)

Crack width (mm)


ACI 224R Analytical maximum crack width
0.6 0.4
Eurocode 2 Analytical average crack width
0.3
0.4
0.2
0.2 STN-12
0.1

0.0 0.0
0 50 100 150 200 250 300 350 400 450 500 0 50 100 150 200 250 300 350 400 450 500 550
Steel stress (MPa) Steel stress (MPa)

0.6
0.8 Measured maximum crack width
Measured
D12-GA 0.5 Measured average crack width
Analytical
Crack width (mm)
Crack width (mm)

ACI 224R Analytical maximum crack width


0.6 0.4
Eurocode 2 Analytical average crack width
0.3
0.4
0.2
0.2
0.1 STN-16

0.0 0.0
0 50 100 150 200 250 300 350 400 450 500 0 50 100 150 200 250 300 350 400 450 500 550
Steel stress (MPa) Steel stress (MPa)

Fig. 9. Crack widths of D8-GA, D10-GA and D12-GA. Fig. 10. Crack widths of S20-14, STN-12 and STN-16.

width is needed to improve our understanding of the cracking


behavior of concrete structures reinforced with high-strength 0.6
steel. For this purpose, Soltani et al. [40] have carried out a theoret-
ical study using their own numerical method.
0.5
Analytical crack width (mm)

Herein, a parametric study on the effects of steel stress, steel


ratio and bar size is carried out. The specimens analyzed are the
same as those of Soltani et al. and the material properties are as 0.4
presented in Table 5. Deformed steel bars of four sizes (12.7 mm,
19.0 mm, 25.4 mm and 32.0 mm) are used. They all have yield
0.3
strengths higher than 800 MPa. For each bar size, three different
steel ratios, 0.75%, 1.00% and 1.50%, are considered in the analysis.
The specimens analyzed are numbered in the form of D‘‘A”-‘‘B”, 0.2
where D means diameter of steel bar, A is the imperial bar size
and B is the steel ratio (in %), as depicted in the first column of
Table 5. 0.1 Tension specimen with deformed bar
The calculated crack widths of the specimens with different bar Tension specimen with smooth bar
sizes and steel ratios are plotted against the steel stress in Figs. 13– 0.0
15, where the respective results obtained by Soltani et al. are also 0.0 0.1 0.2 0.3 0.4 0.5 0.6
plotted for comparison. Both the crack width results by the present Measured crack width (mm)
analysis and by Soltani et al.’s analysis show the similar phenom-
Fig. 11. Analytical crack width versus measured crack width.
ena that (1) the crack width generally increases with the steel
stress; (2) at the same steel ratio, the crack width is larger when
the bar size is larger; and (3) at the same bar size, the crack width
to check carefully that the resulting crack width would not exceed
is smaller when the steel ratio is higher. Hence, when using high-
the permissible limit. One possible way of minimizing the crack
strength steel with higher working stress under SLS, it is advisable
width is to use a larger number of smaller diameter bars and keep
626 A.K.H. Kwan, F.J. Ma / Engineering Structures 126 (2016) 618–627

0.18 2.0

Concrete strength = 20 MPa D4-1.00 by present analysis D4-1.00 by Soltani et al.


0.16 1.8
Concrete strength = 30 MPa D6-1.00 by present analysis D6-1.00 by Soltani et al.
Concrete strength = 40 MPa D8-1.00 by present analysis D8-1.00 by Soltani et al.
0.14 1.6
Concrete strength = 50 MPa D10-1.00 by present analysis D10-1.00 by Soltani et al.
Concrete strength = 60 MPa 1.4
0.12
Crack width (mm)

Crack width (mm)


1.2
0.10
1.0
0.08
0.8
0.06
0.6
0.04
0.4
0.02
0.2

0.00 0.0
40 60 80 100 120 140 160 180 200 220 240 260 300 350 400 450 500 550 600 650 700 750 800
Steel stress (MPa) Steel stress (MPa)
Fig. 12. Crack widths at various concrete strengths. Fig. 14. Crack widths at steel ratio = 1.00%.

Table 5
2.0
Specimens analyzed in parametric study on effects of steel ratio and bar size.
D4-1.50 by present analysis D4-1.50 by Soltani et al.
Specimen Steel bar diameter Steel ratio Cross section 1.8
D6-1.50 by present analysis D6-1.50 by Soltani et al.
number (mm) (%) (mm  mm)
1.6 D8-1.50 by present analysis D8-1.50 by Soltani et al.
D4-0.75 12.7 0.75 130.0  130.0 D10-1.50 by present analysis D10-1.50 by Soltani et al.
D6-0.75 19.0 0.75 194.4  194.4 1.4
D8-0.75 25.4 0.75 259.9  259.9
Crack width (mm)

D10-0.75 32.0 0.75 327.5  327.5 1.2


D4-1.00 12.7 1.00 112.6  112.6
D6-1.00 19.0 1.00 168.4  168.4 1.0
D8-1.00 25.4 1.00 225.1  225.1
D10-1.00 32.0 1.00 283.6  283.6 0.8
D4-1.50 12.7 1.50 91.9  91.9
D6-1.50 19.0 1.50 137.5  137.5 0.6
D8-1.50 25.4 1.50 183.8  183.8
D10-1.50 32.0 1.50 231.6  231.6 0.4
Notes: (1) In each specimen, there is only one steel bar. (2) For the concrete, the
compressive strength, tensile strength and elastic modulus are 34.5 MPa, 3.3 MPa 0.2
and 27.8 GPa, respectively. (3) For the steel-concrete bond, the peak bond stress is
11.7 MPa and residual bond stress is 1.8 MPa. 0.0
300 350 400 450 500 550 600 650 700 750 800
Steel stress (MPa)
the steel ratio sufficiently high. In this regard, more specific design Fig. 15. Crack widths at steel ratio = 1.50%.
guidelines to be developed by further research are needed.

2.0 On the other hand, comparing the crack width results by the
D4-0.75 by present analysis D4-0.75 by Soltani et al.
present analysis with those obtained by Soltani et al. it is seen that
1.8 although both sets of results reveal the same overall trends, due to
D6-0.75 by present analysis D6-0.75 by Soltani et al.
D8-0.75 by present analysis D8-0.75 by Soltani et al. differences in method of analysis, the results by the present anal-
1.6
D10-0.75 by present analysis D10-0.75 by Soltani et al. ysis sometime agree fairly well with the results by Soltani et al.
1.4 and sometimes are quite different from the results by Soltani
et al. Further research to develop a more generally accepted
Crack width (mm)

1.2 method of analysis is needed.


1.0

0.8 5. Conclusions

0.6 A new finite element method for discrete crack analysis without
the use of discrete crack elements or the need of adaptive re-
0.4
meshing of finite elements near crack tips has been developed.
0.2 To cater for the stress redistribution during crack formation, a
crack queuing algorithm of allowing at most one concrete element
0.0 to crack at each iteration step and re-analyzing the concrete struc-
300 350 400 450 500 550 600 650 700 750 800
Steel stress (MPa) ture before considering the formation of other cracks is employed.
This finite element method augmented with the crack queuing
Fig. 13. Crack widths at steel ratio = 0.75%. algorithm has been applied previously to concrete members under
A.K.H. Kwan, F.J. Ma / Engineering Structures 126 (2016) 618–627 627

flexure and in this research extended to incorporate quadrilateral [9] Vecchio FJ. Nonlinear finite element analysis of reinforced concrete
membranes. ACI Struct J 1989;86(1):26–35.
elements and bond-slip models for both deformed and smooth
[10] Gupta A, Maestrini SR. Post-cracking behavior of membrane reinforced
bars. concrete elements including tension-stiffening. J Struct Eng ASCE 1989;115
The finite element method augmented with the crack queuing (4):957–76.
algorithm has been applied to analyze three sets of concrete mem- [11] Chan HC, Cheung YK, Huang YP. Crack analysis of reinforced concrete tension
members. J Struct Eng ASCE 1992;118(8):2118–32.
bers under direct tension tested by others. The finite element anal- [12] Kwak H, Kim S. Bond-slip behavior under monotonic uniaxial load. Eng Struct
ysis results reveal that rather drastic redistributions of concrete 2001;23(3):298–309.
stress, steel stress and bond slip would occur during the formation [13] Nejadi S, Gilbert RI. Shrinkage cracking and crack control in restrained
reinforced concrete members. ACI Struct J 2004;101(6):840–5.
of cracks. Particularly, at a newly formed crack, the concrete stress [14] Yankelevsky DZ, Jabareen M, Abutbul AD. One-dimensional analysis of tension
would drop to zero, the steel stress would increase sharply to a cer- stiffening in reinforced concrete with discrete cracks. Eng Struct 2008;30
tain peak value and the bond slip at the two sides of the crack (1):206–17.
[15] Häubler-Combe U, Hartig J. Evaluation of concrete cracking due to restrained
would become opposite. Unlike the apparently smeared crack pat- thermal loading and shrinkage. ACI Struct J 2012;109(1):41–5.
terns generated by the conventional finite element methods, the [16] Ngo D, Scordelis AC. Finite element analysis of reinforced concrete beams. ACI J
crack patterns generated by this finite element method all contain Proc 1967;64(3):152–63.
[17] Kanakubo T, Sato Y, Uchida Y, Watanabe K, Shima H. Japan Concrete Institute
clearly defined discrete cracks well separated from each other. TC activities on bond behavior and constitutive laws in RC (Part 3 Application
Comparing the analytical crack number and spacing results with of constitutive laws for FEA). Proc Bond Concr 2012;2012:105–12.
the measured results, it is evident that the finite element analysis [18] Rashid YR. Ultimate strength analysis of prestressed concrete pressure vessels.
Nucl Eng Des 1968;7(4):334–44.
tends to overestimate the steel stress at first crack, overestimate
[19] Bažant ZP, Byung HO. Crack band theory for fracture of concrete. Mater Struct
the crack number and underestimate the crack spacing. Comparing 1983;16(3):155–77.
the crack width results with the measured results, it is seen that for [20] Suidan M, Schnobrich WC. Finite element analysis of reinforced concrete. J
members with deformed bars, the finite element analysis underes- Struct Div ASCE 1973;99(10):2109–22.
[21] Červenka V. Constitutive model for cracked reinforced concrete. ACI J Proc
timates the crack width by only 3.7% but for members with smooth 1985;82(6):877–82.
bars, the finite element analysis overestimates the crack width by [22] De Borst R, Nauta P. Non-orthogonal cracks in a smeared finite element model.
36.6%. Hence, for crack width analysis of concrete members with Eng Comput 1985;2(1):35–46.
[23] Willam K, Pramono E. Fundamental issues of smeared crack models. In:
deformed bars, the finite element method is sufficiently accurate. Proceedings of SEM/RILEM international conference of fracture of concrete and
For crack width analysis of concrete members with smooth bars, rock. Houston (Texas): Springer; 1987. p. 142–57.
a better bond-slip model with less conservative peak bond stress [24] Soltani M, An X, Maekawa K. Computational model for post cracking analysis
of RC membrane elements based on local stress-strain characteristics. Eng
adopted is needed. Struct 2003;25(1):993–1007.
After verifying the accuracy and applicability of the finite ele- [25] Jendele L, Cervenka J. Finite element modelling of reinforcement with bond.
ment method, two parametric studies have been carried out to Comput Struct 2006;84(28):1780–91.
[26] Kwak H, Kim S. Simplified monotonic moment-curvature relation considering
evaluate the effect of concrete strength and the combined effects fixed-end rotation and axial force effect. Eng Struct 2010;32(1):69–73.
of steel stress, steel ratio and bar size. It is found that the effect [27] Ng PL, Lam JYK, Kwan AKH. Tension stiffening in concrete beams. Part I: FE
of concrete strength is not large but with the concrete strength analysis. ICE Proc Struct Build 2010;163(1):19–28.
[28] Cope RJ, Rao PV, Clark LA, Norris P. Modelling of reinforced concrete behavior
increased from 20 MPa to 60 MPa, the crack width could be
for finite element analysis of bridge slabs. Numer Methods Non-linear
decreased by about 10%. It is also found that since the crack width Prob. Swansea: Pineridge Press; 1980. p. 457–70.
would increase with the steel stress, the use of high-strength steel [29] Gupta AK, Akbar H. Cracking in reinforced concrete analysis. J Struct Eng ASCE
could lead to an unacceptably large crack width. Nevertheless, the 1984;110(8):1735–46.
[30] Kwan AKH, Wang ZM, Chan HC. Mesoscopic study of concrete II: nonlinear
crack width could be minimized by using a larger number of smal- finite element analysis. Comput Struct 1999;70(5):545–56.
ler diameter bars and keeping the steel ratio sufficiently high. [31] Ma FJ, Kwan AKH. Crack width analysis of reinforced concrete members under
flexure by finite element method and crack queuing algorithm. Eng Struct
2015;105:209–19.
References [32] He XG, Kwan AKH. Modeling dowel action of reinforcement bars for finite
element analysis of concrete structures. Comput Struct 2001;79(6):595–604.
[1] British Standards Institution. BS 8110-2: 1985: structural use of concrete – [33] Radnić J, Markota L. Experimental verification of engineering procedures for
Part 2: Code of practice for special circumstances. London (UK): British calculation of crack width in concrete elements. Eng Model 2003;16(1):63–9.
Standards Institution; 1985. [34] Farra B, Jaccoud JP. Influence of concrete and reinforcement on cracking of
[2] ACI Committee 224. ACI 224R-01: control of cracking in concrete concrete structures. Test report of short-terms imposed strains on
structures. Farmington Hills (Michigan, USA): American Concrete Institute; ties. Lausanne (Switzerland): IBAP; 1994.
2001. [35] Wu HQ, Gilbert RI. Modeling short-term tension stiffening in reinforced
[3] Comité Européen de Normalisation. Eurocode 2: design of concrete structures: concrete prisms using a continuum-based finite element model. Eng Struct
Part 1-1: General rules and rules for buildings. London (UK): British Standards 2009;31(10):2380–91.
Institution; 2004. [36] ACI Committee 318. ACI 318R-02: building code requirements for structural
[4] Kwak H, Song J. Cracking analysis of RC members using polynomial strain concrete. Farmington Hills (Michigan, USA): American Concrete Institute;
distribution function. Eng Struct 2004;24(4):455–68. 2001.
[5] Gribniak V, Perez-Caldentey A, Kaklauskas G, Rimkus A, Sokolov A. Effect of [37] Chen HH, Su RKL, Kwan AKH. Fracture toughness of plain concrete made of
arrangement of tensile reinforcement on flexural stiffness and cracking. Eng crushed granite aggregate. Hong Kong Inst Eng Trans 2011;118(2):6–12.
Struct 2016;124(1):418–28. [38] Comité Euro-International du Beton. CEB-FIP model code 1990: model code for
[6] Dufour F, Pijaudier-Cabot G, Choinska M, Huerta A. Extraction of a crack concrete structures. London: Thomas Telford; 1993.
opening from a continuous approach using regularized damage models. [39] AASHTO. AASHTO LRFD bridge design specification. 5th ed. Washington (DC,
Comput Concr 2008;5(4):375–88. USA): American Association of State Highway and Transportation Officials;
[7] Dufour F, Legrain G, Pijaudier-Cabot G, Huerta A. Estimation of crack opening 2010.
from a two-dimensional continuum-based finite element computation. Int J [40] Soltani A, Harries KA, Shahrooz BM. Crack opening behavior of concrete
Numer Anal Meth Geomech 2012;36(16):1813–30. reinforced with high strength reinforcing steel. Int J Concr Struct Mater 2013;7
[8] Gilbert RI, Warner RF. Tension stiffening in reinforced concrete slabs. J Struct (4):253–64.
Eng ASCE 1978;104(12):1885–900.

You might also like