applied
sciences
Article
Simulation of Crack Propagation in Reinforced
Concrete Elements
Jonas Cramer, Sara Javidmehr and Martin Empelmann *






Citation: Cramer, J.; Javidmehr, S.;
Empelmann, M. Simulation of Crack
Propagation in Reinforced Concrete
Elements. Appl. Sci. 2021, 11, 785.
https://doi.org/10.3390/app11020785
Received: 15 December 2020
Accepted: 12 January 2021
Published: 15 January 2021
Publisher’s Note: MDPI stays neu-
tral with regard to jurisdictional clai-
ms in published maps and institutio-
nal affiliations.
Copyright: © 2021 by the authors. Li-
censee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and con-
ditions of the Creative Commons At-
tribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
Appl. Sci. 2021, 11, 785. https://doi.org/10.3390/app11020785
iBMB, Division of Concrete Construction of TU Braunschweig, Beethovenstraße 52,
38106 Braunschweig, Germany; j.cramer@ibmb.tu-bs.de (J.C.); s.javidmehr@ibmb.tu-bs.de (S.J.)
* Correspondence: m.empelmann@ibmb.tu-bs.de; Tel.: +49-531-391-5409
Abstract: The crack widths in reinforced concrete structures are affected by various influencing
parameters, such as the tensile strength of the concrete, which is a highly variable parameter. While
safe predictions of crack widths are possible with the existing models supplied in the code provisions
accompanied by suitable safety factors, there are large discrepancies between the experimentally
measured crack widths and the calculated values. Even within a uniaxial tensile test on reinforced
concrete components, several crack parameters (spacings and widths) are usually measured. To
further investigate the effects of scattering input parameters on the distribution of crack parameters,
a crack propagation model (CPM) is developed, which is a powerful tool that can be used to observe
sequential cracking of uniaxially loaded reinforced concrete components. In the present paper, the
background of the scattering of the concrete tensile strength will be explained and the procedure
for the simulation in the CPM, including the investigation of different bond–slip relationships, will
be introduced. Finally, the CPM will be used to calculate experimental crack parameters and the
distribution of the calculated crack parameters will be discussed.
Keywords: crack widths; crack spacing; rheological model; crack propagation; serviceability limit
state; concrete structures; concrete tensile strength; bond–slip behavior; statistical distribution
1. Introduction
In reinforced concrete structures, cracks are unproblematic, as long as the crack widths
are limited to a specific level depending on the environmental conditions. For this purpose,
crack width verification is carried out in the design of a structural component, which
ensures the requirements are met in terms of the serviceability and durability of the
structural component by means of a normatively defined calculated value of the crack
width. Different calculation models exist in the respective standards that can be used
to verify the calculated crack width, which are nowadays usually mechanically based.
Essential differences in the models result from the reference values used for the crack width.
In some models, the mean value of the crack widths is used as a reference value and the
characteristic crack width is determined using a corresponding increase factor. A further
group of models directly calculate the characteristic crack widths. For the comparison of
such characteristic crack widths with the measured crack widths in the tests, statistical
considerations are required. The 95th percentile of the distribution is associated with the
characteristic crack width value. Therefore, for both procedures, sufficient knowledge of the
stress-dependent crack formation process is required, as well as knowledge of the resulting
scattering and distribution density of the occurring crack widths and crack spacings.
The experimental determination of the stress-dependent crack width distribution is
associated with a very high experimental effort, because uniaxial tension tests on long
reinforced concrete specimens are required, in which a statistically relevant number of
cracks emerge. Such tests were carried out in the 1960s (e.g., [1,2]) and provided a good data
basis for structural elements with steel reinforcement and the common concrete properties
at that time. For new concrete mixtures, high reinforcement ratios, large-diameter steel
https://www.mdpi.com/journal/applsci
 diameter steel bars, or modern non-metallic reinforcement, the amount of available data
on existing tests for the statistical distribution of crack parameters (i.e., crack spacing and
crack widths) is very limited or does not even exist.
An alternative approach involves the use of numerical models of the crack formation
Appl. Sci. 2021, 11, 785
process involving a consistent mechanical background. The performance and prediction 2 of 19
accuracy of such models must be proven via validation with experimental tests. In [3,4],
efficient models of the simulation of random crack formation have already been
developed, which were each developed and used for specific tasks. For more
bars, or modern non-metallic
comprehensive considerations reinforcement, the amount
of crack formation of availabletensile
in reinforced data on existing tests
components or
for the statistical
flexural tension distribution
zones [5], a of crackpropagation
crack parameters model (i.e., crack spacing
(CPM) was and crack widths)
developed, with the is
very limited
ability or does not
to determine theeven exist.
distribution functions of crack widths and crack spacings in a
An alternative
general form. approach involves the use of numerical models of the crack formation
process In this paper, the model setup andbackground.
involving a consistent mechanical features ofThe theperformance and prediction
CPM are explained and a
accuracy
comparison with experimental tests is carried out for deformed bars. tests.
of such models must be proven via validation with experimental In [3,4],
Furthermore,
efficient
possiblemodels
extensions of the simulation
of the CPM forofinvestigation
random crack formation
of the have already
crack parameters been devel-
of recently used
oped, which were each developed and used for specific tasks.
concrete mixtures (e.g., steel-fiber-reinforced concrete, high-strength concrete, carbon For more comprehensive
considerations of crack formation in reinforced tensile components or flexural tension
concrete) are presented.
zones [5], a crack propagation model (CPM) was developed, with the ability to determine
the distribution
2. General Crack functions
Propagation of crack widths and crack spacings in a general form.
In this paper, the model setup and features of the CPM are explained and a comparison
If an uncracked
with experimental testsreinforced
is carried outconcrete specimen
for deformed bars.(uncracked
Furthermore,stage) is subjected
possible extensionsto a
steadily increasing centric tensile load, cracks will appear
of the CPM for investigation of the crack parameters of recently used concrete mixtures at the point of the lowest
concrete
(e.g., tensile strength (initial
steel-fiber-reinforced concrete,cracking stage). concrete,
high-strength In the crack, the concrete)
carbon concrete arestrain value is
presented.
zero and the steel strain value is maximal. Due to the bond stresses activated on both sides
2.ofGeneral
the crack, tensile
Crack stresses are introduced into the concrete and the steel tensile stresses
Propagation
are reduced until the
If an uncracked reinforced steel andconcrete
concretespecimen
strains are identical stage)
(uncracked again at the end oftothe
is subjected a
transmission length l t (Figure 1). With a further increase of the load, the next crack occurs
steadily increasing centric tensile load, cracks will appear at the point of the lowest concrete
in the strength
tensile remaining uncracked
(initial crackingcomponent zone
stage). In the at the
crack, point
the withstrain
concrete the lowest
valueconcrete
is zero andtensile
the
strength (crack formation stage; Figure 1, stage 2). This process
steel strain value is maximal. Due to the bond stresses activated on both sides of the crack, is repeated until at no
point are
tensile the steel
stresses are and concreteinto
introduced strains
the identical.
concrete and At this
the point the stabilized
steel tensile stressescracking stage
are reduced
is reached (Figure 1, stage 3), however it must be noted that with
until the steel and concrete strains are identical again at the end of the transmission length further load increase
ltbeyond
(Figurethis1). state,
With further
a further cracks can form.
increase of the load, the next crack occurs in the remaining
uncracked component zone at theformation
The crack width in the crack point with stage
the results
lowest from the integration
concrete of the(crack
tensile strength strain
differences between the reinforcement and surrounding concrete
formation stage; Figure 1, stage 2). This process is repeated until at no point are the over the transmission
length
steel and onconcrete
both sides of theidentical.
strains observedAt crack.
this In the stabilized
point the stabilizedcracking stage,stage
cracking the crack width
is reached
is determined
(Figure 1, stageby 3), integrating
however it the muststrain difference
be noted overfurther
that with half ofload
the crack spacing
increase beyond on this
both
sidesfurther
state, of the considered
cracks can crack.form.

Figure1.1.Stages
Figure Stagesof
ofcrack
crackpropagation
propagationin
inreinforced
reinforcedconcrete
concretespecimens
specimensininuniaxial
uniaxialtensile
tensiletests.
tests.

The crack width in the crack formation stage results from the integration of the strain
differences between the reinforcement and surrounding concrete over the transmission
length on both sides of the observed crack. In the stabilized cracking stage, the crack width
is determined by integrating the strain difference over half of the crack spacing on both
sides of the considered crack.
Hence, the concrete tensile strength values (specifically the lower concrete tensile
strength values) over the length are significant in terms of initiation of cracks, while the
bond–slip behaviour is the major parameter affecting the transmission of stresses between
the steel bar and the concrete. Therefore, to simulate the whole crack propagation process
with increasing load level (i.e., increasing steel stress σs ), the crack propagation model
(CPM) is configured with the following features:
 Hence, the concrete tensile strength values (specifically the lower concrete tensile
strength values) over the length are significant in terms of initiation of cracks, while the
bond–slip behaviour is the major parameter affecting the transmission of stresses between
Appl. Sci. 2021, 11, 785 3 of 19
the steel bar and the concrete. Therefore, to simulate the whole crack propagation process
with increasing load level (i.e., increasing steel stress 𝜎s ), the crack propagation model
(CPM) is configured with the following features:
• For For random
random cracking, concrete
concrete tensile
tensilestrength
strengthscattering
scatteringalong
alongthe
thelongitudinal
longitudinal axis
axis of
of
thethe specimen
specimen is required.
is required. Therefore,
Therefore, the the
CPMCPM
can can
mapmap different
different statistical
statistical distribu-
distributions;
 tions;
It must be possible to calculate steel and concrete strains separately over the entire
• Itlength
must of
bethe
possible
specimen. to calculate
Thus, both steelmaterials
and concrete strains
shall be separately
represented over the entire
separately;
 length of the specimen. Thus, both materials shall be represented separately;
A bond stress–slip relationship is required for the interaction of both materials. This
• A bond be
should stress–slip
definable relationship
in a general is required for the
mathematical interaction
form of both
to be able materials.
to take This
into account
should
differentbeloading
definable in aand
levels general
materialmathematical form to be able to take into account
configurations.
different loading levels and material configurations.
Using these features, the crack formation can be considered in a very variable and
Using
specific way these features,
(e.g., through theadaptation
crack formation
of inputcanparameters
be considered andincalculation
a very variable
steps and
and
specific way (e.g., through adaptation of input parameters and
isolated consideration of influence parameters with almost any number of calculation calculation steps and
isolated consideration
runs). Using a randomofdefinition
influence for parameters with tensile
the concrete almost strength
any numberover of calculation
the specimen
runs). Using a random definition for the concrete tensile
length, the random locations of the crack initiation and random crack spacingsstrength over the specimen
can be
length, the random
determined. locations
Furthermore, the of the crack
statistical initiation of
distribution andtherandom crack widths
output crack spacingsandcan be
crack
determined. Furthermore, the statistical distribution of the output
spacings can be investigated with several runs and numerous random parameters withincrack widths and crack
spacings can be investigated with several runs and numerous random parameters within a
a specific range.
specific range.
However, to obtain a mechanically reasoned model with very good prediction
However, to obtain a mechanically reasoned model with very good prediction ac-
accuracy, specific considerations concerning the tensile strength distribution are needed
curacy, specific considerations concerning the tensile strength distribution are needed to
to reach element-size-independent and reasonable results. Such aspects of the CPM model
reach element-size-independent and reasonable results. Such aspects of the CPM model
will be explained in the following.
will be explained in the following.
3. Crack
3. Crack Propagation
Propagation Model
Model (CPM)
(CPM)
3.1. Model
3.1. Model Setup
Setup
To implement
To implementthe thefeatures
featureslisted
listed
in in
thethe previous
previous section,
section, the the rheological
rheological modelmodel
de-
described
scribed in [6,7]
in [6,7] waswas extended.
extended. TheThe
basicbasic
modelmodel consists
consists of two
of two uniaxial
uniaxial bars bars (concrete
(concrete and
and reinforcing
reinforcing steel),steel),
whichwhich are connected
are connected with springs
with springs at a defined
at a defined distancedistance
(Figure(Figure 2).
2). Here,
Here, the springs represent the bond between the steel bar and the
the springs represent the bond between the steel bar and the concrete bar. The extended concrete bar. The
extended
crack crack propagation
propagation model (CPM) model (CPM)
allows allows
“free “free
crack crack formation”
formation” of the concrete
of the concrete tensile
tensile strength
strength scattering scattering
along thealong
bar orthe bar or component
component axis in
axis (cf. models (cf.[3,4]).
models Thein [3,4]). The
scattering of
the concreteoftensile
scattering strengthtensile
the concrete is randomly generated
strength according
is randomly to a given
generated distribution
according (e.g.,
to a given
normal distribution,
distribution logarithmic
(e.g., normal normallogarithmic
distribution, distribution).
normal distribution).

Figure 2.
Figure 2. Crack
Crack propagation
propagation model
model (CPM).
(CPM).

In contrast to the rheological model in [6,7], the specified model length l of the CPM
corresponds to the entire length of the simulated component. The crack formation is
recorded via stiffness reduction of the corresponding concrete element (EA)c,I = 0 (Figure 3).
ci. 2021, 11, x FOR PEER REVIEW 4 of 19

Appl. Sci. 2021, 11, 785 In contrast to the rheological model in [6,7], the specified model length l of the CPM 4 of 19
corresponds to the entire length of the simulated component. The crack formation is
recorded via stiffness reduction of the corresponding concrete element (EA)c,I = 0 (Figure 3).

Figure 3. Cracked element in Figure

the CPM.
3. Cracked element in the CPM.

Since a crack canSince a crack

only emerge canbetween
only emerge between
two nodes, i.e.,two nodes,
in the middle i.e.,of
inathe middlebar
concrete of a concrete bar
element (Figure element (Figure
3), the crack 3), the
width is crack widthbyisusing
calculated calculated bydifference
the slip using the slip difference
of the left and of the left and
rightinto
right nodes, taking nodes, taking
account theinto account
steel the steel
bar strains. bar strains.
Considering thatConsidering that the
the strain values in strain values
the remaining “concrete stumps” (between concrete nodes KCi and KCi+1) equal izero andi+1 ) equal zero
in the remaining “concrete stumps” (between concrete nodes KC and KC
the steel strainand the steel
values strain
in the steel values in the steel
bar element bar element
i (between steeli (between
nodes KSsteeli andnodes
KSi+1)KSi and KSi+1 )
correspond to
correspond to εs,i = (uKS,i+1 − uKS,i
ε = (u
s,i ) ⁄ li,KS,i+1 − u
the calculated
KS,i ) ⁄ l , the calculated crack width
icrack width can be simplified as can bethe
simplified as the
difference between the translation u of the
difference between the translation u of the concrete nodes KCi+1 and KCi: i+1concrete nodes KC and KC i:

𝑤cal,i = 𝑠i − 𝑠i+1 + 𝜀s,i ∙ 𝑙i wcal,i = si − si+1 + ε s,i ·li

= (𝑢KS,i − 𝑢KC,i ) − (𝑢KS,i+1 − =
𝑢KC,i+1
(uKS,i −𝜀us,iKC,i
)+ ∙ 𝑙i) − (uKS,i+1 − uKC,i+1 ) + ε s,i ·li (1) (1)
= 𝑢KC,i+1 − 𝑢KC,i = uKC,i+1 − uKC,i
where si is thewhere
slip atsinode
is thei; slip
uKS,iat
is node
the translation of translation
i; uKS,i is the the steel barofatthenodesteeli; bar
uKC,iatisnode
the i; uKC,i is the
translation of the concrete bar at node i; ε is the strain of the steel bar i; and l is the length
translation of the concrete bar at node i; εs,i is the strain of the steel bar i; and li is the length
s,i i

of element i. of element i.
Using such simplified
Using such linear displacement
simplified methods, themethods,
linear displacement element the sizeelement
shouldsize be should be se-
selected carefully to carefully
lected obtain reasonable results. In results.
to obtain reasonable the CPM, error of
In the CPM, thetheerrordetermined
of the determined crack
crack widths iswidths
usuallyisless than less
usually 3%, than
with 3%,6 to with
8 elements
6 to 8 between
elementstwo cracks.
between twoTherefore,
cracks. Therefore, the
the element length l is generally selected with regard to the expected crack
element length le is generally selected with regard to the expected crack spacing
e spacing sr, sr , which
which is assumed over a reasonable
is assumed range forrange
over a reasonable the investigated specimen specimen
for the investigated (e.g., considering
(e.g., considering the
the reinforcement ratio, rebarratio,
reinforcement diameter,
rebar and bond–slip
diameter, relationship).
and bond–slip relationship).

3.2. Distribution3.2. Distribution

of Tensile of Tensile Strength
Strength
The experimental determination determination
The experimental of tensile
of tensile strength strength
has been has been
the subject of the subject of recent
recent
research (e.g., research (e.g.,several
[8,9]), and [8,9]), and several
indirect andindirect
direct and direct
testing testing exist.
methods methods exist. Uniaxial con-
Uniaxial
concrete tensile crete tensileisstrength
strength determinedis determined most effectively
most effectively using a direct using a direct (uniaxial)
(uniaxial) tensile tensile test
(Figure 4). To determine the scattering of the concrete tensile strength
test (Figure 4). To determine the scattering of the concrete tensile strength of a single batch, of a single batch, the
statistical distribution of several direct tensile tests is required.
the statistical distribution of several direct tensile tests is required. This scattering is This scattering is caused
by a wide variety of reasons, such as flaws due to insufficient
caused by a wide variety of reasons, such as flaws due to insufficient compaction, the compaction, the presence
of inhomogeneity and anisotropy caused by aggregates,
presence of inhomogeneity and anisotropy caused by aggregates, as well as internal stress as well as internal stress states
caused by shrinkage and hydration
states caused by shrinkage and hydration of cement. of cement.
To describe theTodistribution
describe the of distribution of tensile
tensile strength, strength,
a probability a probability
function function
is required. In is required.
[10], a logarithmic normal distribution is assumed. In contrast, [11] and some codes in some codes
In [10], a logarithmic normal distribution is assumed. In contrast, [11] and
[12,13] assumein a[12,13]
normal assume a normal
distribution, distribution,
while a summary while is aprovided
summaryinis [14]. provided
Both in [14]. Both
distributions are shown in Figure 5 (probability density function) for an expected tensile
distributions are shown in Figure 5 (probability density function) for an expected tensile
strength value of 2.0 N/mm2 and coefficients of variation of vk,1 = 10% and vk,2 = 20%.
strength value of 2.0 N/mm and coefficients of variation of vk,1 = 10% and vk,2 = 20%.
2
Appl. Sci. 2021, 11, 785 5 of 19
Appl. Sci. 2021, 11, x FOR PEER REVIEW 5 of 19
Appl. Sci. 2021, 11, x FOR PEER REVIEW 5 of 19

Figure 4. Direct tensile tests (cylinder Ø 150 × 300 mm, C30/37, dg = 16 mm).
Figure 4. Direct tensile tests (cylinder Ø150 × 300 mm, C30/37, dg = 16 mm).
Figure 4. Direct tensile tests (cylinder Ø 150 × 300 mm, C30/37, dg = 16 mm).

Figure 5. Comparison between the normal distribution and log-normal distribution of tensile
strength,
Figure
Figure with coefficients
Comparison
5.5.Comparison between of variation
between thethe of distribution
normal
normal vk =distribution
10% andand
vk =and
20%.log-normal
log-normal distribution
distribution of tensile
of tensile
strength,
strength,with
withcoefficients
coefficients of
of variation
variation of
of vvkk==10%
10%and
andvvkk==20%.
20%.
It can be seen that both distributions are very similar for a coefficient of variation vk,1
ItItcan
= 10%, can be
with beanseen
seen thatboth
that
overlap both
of distributions
distributions
96.2%. For vk,2 =are are
20%, veryvery
the similar
similar
overlap for for a coefficient
a coefficient
is 92.4%, but theof of variation
variation
visible vk,1
deviation
=v10%,= with
10%,5an
in Figure
k,1 with an overlap
isoverlap of 96.2%.
significantly of For
larger96.2%. For
vk,2 =for
than 20%,vvk,1k,2
the =overlap
= 10%.20%, the overlap
is 92.4%,
Consequently, isthe
butthe 92.4%,
visiblebut
logarithmic the
deviationvis-
normal
ible
in deviation
Figure in vFigure
k < 20%5larger
5 is significantly
distribution for is significantly
can be than for vk,1larger
approximately = 10%. than for vk,1
Consequently,
represented by = 10%. Consequently,
the logarithmic
a normal distribution. normalthe
logarithmic
distribution normal
for
For ranges vk <of20% distribution forfor
can be approximately
variation vk < 20%
different can bevalues,
represented
expected approximately
by a anormal represented
must be by
distribution.
distinction madea
normal
between distribution.
For ranges of variation
a constant standard for deviation
different expected
and a constant values,coefficient
a distinction must be made
of variation. For the
betweenFor aranges
concrete tensileofstrength—in
constant variation
standard for different
deviation
contrast toexpected
and a constant
the values,
concrete a distinction
coefficient
compressive must be
of variation.
strength made
For the
(standard
between
concrete
deviation a constant
tensile standard
strength—in
constant [15,16])—the deviation
contrast and a constant
to theofconcrete
coefficient coefficient
variationcompressive of variation.
is consideredstrength
a constant For
(standard the
quantity.
concrete
deviation
With regardtensile
constant strength—in
to the[15,16])—the contrast
scattering coefficient to
of the tensilethe concrete
of variation
strength iscompressive
atconsidered strength (standard
a constantinquantity.
failure, evaluations de-
[11,17,18]
viation
With
show aconstant
regard to the
coefficient [15,16])—the
scattering
of variation coefficient
ofvthe tensile
k < 10%.
ofModel
In variation
strength Codeatisfailure,
considered
2010 anda Eurocode
evaluations
[12] constant quantity.
in [11,17,18]
2 [13], the
With
show regard
coefficient to
a coefficient the scattering
of variation
of variation of the
of the vtensile
k < 10%.
tensile
In Model
strength strength Code
of the at failure,
2010 [12]isand
component evaluations
vk =Eurocode in
18% (see 2[4]). [11,17,18]
[13], the
Similar
show
resultsa coefficient
coefficient of variation
are of variation
demonstrated in v[19],
of the tensile k <strength
10%. InofModel
by considering Code
the component 2010is [12]
the influences and
vk = 18% Eurocode
(see
of different 2 [13],
[4]). Similar
concrete
the coefficient
results
mixturesare [17]. ofThus,
variation
demonstrated for one of concrete
in the tensile
[19], by batch, strength
considering ofthe
theinfluences
a coefficient component
of variation vk k< =10%
ofisvdifferent 18% (see [4]).
concrete
seems to be
Similar
mixtures results
realistic.[17]. are
ForThus, demonstrated
several forconcrete
one concrete in
batches[19],
batch,by considering
in a acomponent
coefficient or the
of ininfluences
variation of different
vk < 10%
the absence seems
of tensile concrete
to bevk
tests,
mixturesFor
realistic. [17]. Thus,concrete
for one concrete inbatch, a coefficient inofthe
variation vof
k <tensile
10% seems
tests, vtok a
= 18% may severalbe reasonable. For batches a component
the probability function or of absencestrength,
tensile however,
be
= 18%realistic.
may For several
be reasonable. concrete
Forassumedbatches
the probability in a component
functioninofboth or in the
the tensile absence of tensile tests,
normal distribution can be approximately cases. strength, however, a
vk = 18% may be reasonable. For the probability function of the tensile strength, however,
normal distribution
The can be assumed
aforementioned distributionapproximately
function applies in both cases.
to the tensile strength values
a normal distribution can be assumed approximately in both cases.
The aforementioned
obtained from several uniaxial distribution
tensile function
tests and applies
cannot be todirectly
the tensile strength to
implemented values
define
obtained
a tensilefrom several
strength uniaxial tensile
distribution in the tests
CPMand cannot be
to simulate directlybar.
a tensile implemented
This is dueto todefine
the fact
a tensile strength distribution in the CPM to simulate a tensile bar. This is due to the fact
 Appl. Sci. 2021, 11, 785 6 of 19

Appl. Sci. 2021, 11, x FOR PEER REVIEW 6 of 19

The aforementioned distribution function applies to the tensile strength values ob-
tained from several uniaxial tensile tests and cannot be directly implemented to define a
tensile strength distribution in the CPM to simulate a tensile bar. This is due to the fact
that in uniaxial tensile tests
that in uniaxial tensile (oftests
unnotched
(of unnotched specimens), the tensile
specimens), the strength appliesapplies
tensile strength to the to the
weakestweakest
point of theofspecimen.
point the specimen. In the following,
In the following, thisthisvalue
valuewill
willbe
bedesignated
designated as as the
thefracture
fracturetensile
tensilestrength
strength f . Previous experimental investigations (e.g.,
f ct,f . Previous experimental investigations (e.g., [17,20]) show a difference
ct,f [17,20]) show a
difference between the distribution of the tensile strength within
between the distribution of the tensile strength within a specimen and the distribution of a specimen and the
distribution of the tensile
the fracture fracture tensilef ct,f
strength strength
. Regarding fct,f. Regarding
this point,this point, Empelmann
Empelmann [3] defined a[3]term for
definedthe
a term
specimenfor thetensile
specimen tensile
strength f ct ,strength
which isfctexplained
, which is below.
explained below.
Figure Figure
6a shows the fracture
6a shows processprocess
the fracture in a uniaxial tensile tensile
in a uniaxial specimen. The existing
specimen. The existing
micro-cracks accumulate
micro-cracks accumulate at the weakest
at the weakest point
pointofofthe specimenshortly
the specimen shortlybefore
beforethe the
maximum
maximum loadload is reached
is reached [21]. The [21]. The minimum
minimum tensile shown
tensile strength strength in shown
Figure 6b in corresponds
Figure 6b to the
corresponds
fracture to the fracture
tensile strength tensile strength
f ct,min = f ct,f .fct,min
Figure = fct,f6b
. Figure 6b schematically
schematically shows the shows
tensilethe
strength
tensile strength
values invalues in representative
representative elements (which elements (which
will will be later).
be discussed discussed
Basedlater). Based
on this point, it is
logicalitthat
on this point, the mean
is logical that the valuemeanof allvaluef ct,i of
values
all fct,iisvalues
higheristhan
higherthethan
mean thevalue several f ct,f
meanofvalue
values
of several obtained
fct,f values using using
obtained severalseveral
uniaxial tensiletensile
uniaxial tests. tests.

(a) (b)
Figure 6.Figure
(a) Fracture of concrete
6. (a) Fracture under uniaxial
of concrete under tension
uniaxial[21]. (b) Distribution
tension of specimen
[21]. (b) Distribution of tensile
specimen tensile
strengthstrength
[3]. [3].

Consequently, an adaption
Consequently, an adaption is required to derive
is required to derive thethedistribution
distribution ofofthe
thespecimen
specimen tensile
tensile strength f ctct (all (all values in aa component)
component) based based on onthe
thegiven
givendistribution
distributionofofthe thefracture
fracturetensile
tensilestrength
strengthvalues
values f ct,ffct,f(weakest
(weakest values
values in in a component
a component in uniaxial
in uniaxial tensile
tensile specimens
specimens without without a notch).
a notch).
According to [3], such
According adaption
to [3], such adaption can be built
can be bybuilt
assuming a normal
by assuming distribution
a normal for
distribution for
fct,f and fctct,f
andandsetting
f ct andthesetting
mean the
value mean of the specimen
value of the tensile
specimen strength
tensilefct,m equal tof ct,m
strength the equal
95th to the
percentile95th percentile
value of the value
fractureof tensile
the fracture tensile
strength strength
fct,f,95% . f ct,f,95% .
For theFor meanthevalue
mean ofvalue
the offracture the fracture concrete
concrete tensiletensile
strength fct,f,m =f ct,f,m
strength 2, the 2 , the
= 2.0 N/mm
2.0 N/mm
following following
percentilepercentile values
values are obtained are obtained
by settingby thesetting the assumed
assumed coefficient coefficient of variation
of variation (of (of
the fracture tensile strength)
the fracture tensile strength) as equal to vk,f = 10%: as equal to v k,f = 10%:

𝑓ct,f,5% = 𝑓ct,f,m ⋅ (1f − 1.645 ⋅ 𝑣k,f ) = 2.0 N⁄mm2 ⋅ (1 − 1.645 ⋅ 0.1) =21.67 N⁄mm2 (2) 2
ct,f,5% = f ct,f,m · (1 − 1.645 · vk,f ) = 2.0 N/mm · (1 − 1.645 · 0.1) = 1.67 N/mm (2)

2 2
𝑓ct,f,95% = 𝑓ct,f,m ⋅ (1f ct,f,95%
+ 1.645 k,f ) =· 2.0
=⋅ 𝑣f ct,f,m ⁄mm2· ⋅v(1
(1 +N1.645 k,f )+= 2.0 N/mm
1.645 · (1 +
⋅ 0.1) = 2.33 mm2 · 0.1) = 2.33 N/mm
N⁄1.645 (3) (3)
Thevalue
The mean meanofvalue of the specimen
the specimen tensile strength
tensile strength corresponds
corresponds to: to:

𝑓ct,m f= 𝑓ct,f,95%
ct,m = 2.33=
= f ct,f,95% mm2N/mm2
N⁄2.33 (4) (4)

Assuming equal coefficient variations vk = vk,f = 10%:

𝑓ct,m,5% = 𝑓ct,m ⋅ (1 − 1.645 ⋅ 𝑣k ) = 2.33 N⁄mm2 ⋅ (1 − 1.645 ⋅ 0,1) = 1.95 N⁄mm2 (5)
Appl. Sci. 2021, 11, 785 7 of 19

Appl. Sci. 2021, 11, x FOR PEER REVIEW Assuming equal coefficient variations vk = vk,f = 10%: 7 of 19

f ct,m,5% = f ct,m · (1 − 1.645 · vk ) = 2.33 N/mm2 · (1 − 1.645 · 0.1) = 1.95 N/mm2 (5)

𝑓ct,m,95% = 𝑓 ct,m ⋅ (1 +
f ct,m,95% f ct,m ⋅· 𝑣(1k )+=1.645
=1.645 ) =2 ⋅2.33
2.33 ·Nv⁄kmm 1.645 ⋅2 0.1)
(1 −N/mm · (1 −
= 1.645
2.71 N ⁄mm
· 0.1 2 2.71 N/mm2 (6)
) = (6)
where ffct,f,m
where ct,f,misisthe
themean
meanvalue
value ofof the
the fracture
fracture tensile strength; vvk,f
tensile strength; is the
k,f is the coefficient
coefficient of
of
variation of the fracture tensile strength;
variation of the fracture tensile strength; fct,m f is the mean value of the specimen
ct,mis the mean value of the specimen tensile tensile
strength; v k is the coefficient of variation
strength; vk is the coefficient of variation of the of the specimen
specimen tensile
tensile strength;
strength; 5% 5% is
is the
the 5th
5th
percentile value; and 95% is the 95th percentile
percentile value; and 95% is the 95th percentile value. value.
This calculation
This calculation method
method is is depicted
depicted inin Figure
Figure 7.7.

Figure
Figure 7. Fracture tensilestrength
Fracture tensile strength distribution
distribution andand the corresponding
the corresponding specimen
specimen tensile strength
tensile strength distribution.
distribution.
3.3. Element Size and Autocorrelation
3.3. Element Size andbar
The concrete Autocorrelation
of the CPM consists of a finite elements series, with each element
being assigned
The concrete a random
bar of the tensile consists fof
CPMstrength ct,iaaccording
finite elements to theseries,
applied withdistribution.
each element As
being assigned a random tensile strength fct,i according to the applied distribution. Asa
shown in Section 3.2 (Figure 6), each tensile strength value is assigned according to
mechanically
shown reasoned
in Section representative
3.2 (Figure 6), eachelement
tensile size ∆x, which
strength valueisisoften defined
assigned based ontothe
according a
maximum aggregate size d g . Here, ∆x is assumed to
mechanically reasoned representative element size ∆x, which is often definedbe in the range of 2 · dg to 3 ·dgbased
[3,22,23].
on
A maximum
the maximumaggregate
aggregate size
size of d16 mm results in a representative element size ∆x of about
g. Here, ∆x is assumed to be in the range of 2∙dg to 3∙dg

[3,22,23]. A maximum aggregate sizedirectly

32–48 mm. The element size affects of 16 mm theresults
prediction accuracy of theelement
in a representative model and sizethe∆x
accuracy of the crack spacing values, as a crack can only emerge
of about 32–48 mm. The element size affects directly the prediction accuracy of the model in the mid-length area of
each element. Therefore, to enable the simulation of fine cracking
and the accuracy of the crack spacing values, as a crack can only emerge in the mid-length patterns, the element
length
area of leeach
should also beTherefore,
element. very small. to To consider
enable these aspects
the simulation ofand
finesimultaneously
cracking patterns, achievethe
a nearly mesh-independent model, a smaller element length le is selected and the concrete
element length le should also be very small. To consider these aspects and simultaneously
tensile strengths of the adjacent elements within a correlation length Lx are correlated to
achieve a nearly mesh-independent model, a smaller element length le is selected and the
avoid mechanically inconsistent oscillation of the tensile strengths in the smaller elements.
concrete tensile strengths of the adjacent elements within a correlation length Lx are
This approach is known as spatial autocorrelation [24].
correlated to avoid mechanically inconsistent oscillation of the tensile strengths in the
The expression of the linear dependence or correlation between two points x1 and x2 is
smaller elements. This approach is known as spatial autocorrelation [24].
described by the correlation function ρ(x1 ,x2 ). A common form is the quadratic exponential
The expression of the linear dependence or correlation between two points x1 and x2
(Gaussian) correlation function [4,25,26].
is described by the correlation function ρ(x1,x2). A common form is the quadratic
exponential (Gaussian) correlation function [4,25,26]. | x1 − x2 | 2
ρ( x1 , x2 ) = e−( Lx ) 2 (7)
|𝑥 −𝑥 |
−( 1 2 ) (7)
𝜌(𝑥1function
, 𝑥2 ) = 𝑒 according𝐿x
Figure 8 illustrates the correlation to Equation (7). For ρ(x1 ,x2 ) = 0,
thereFigure
is no linear correlation
8 illustrates between xfunction
the correlation 1 and x2 ,according
whereas ρ(x 1 ,x2 ) = 1 shows
to Equation (7). Fora ρ(x
full1,x
depen-
2) = 0,
dencyisof no
there bothlinear
points.
correlation between x1 and x2, whereas ρ(x1,x2) = 1 shows a full
dependency of both points.
 Appl.Sci.
Appl. Sci.2021, 11,x785
2021,11, FOR PEER REVIEW 88ofof19
19

1.0

0.8

0.6

ρ [-]
0.4

0.2

0.0
0.00 0.50 1.00 1.50 2.00 2.50
(x1-x2)/Lx [-]

Figure
Figure8.8.Quadratic
Quadraticexponential
exponentialcorrelation
correlationfunction.
function.

Bymultiplying
By multiplyingthe thecorrelation
correlationfunction
functionwith
withthe
thestandard
standarddeviations
deviations σ(x
σ(x11))and
andσ(x
σ(x2)2 )
(here ) = ) = the value of covariance is obtained and the positive symmetric
(here σ(x1) = σ(x2) = σ), the value of covariance is obtained and the positive symmetric
σ(x 1 σ(x2 σ),
covariance matrixCCcan
covariancematrix canbebecomposed:
composed:
2
σ𝜎2 ρ𝜌12 𝜎22 ·⋯· · 𝜌ρ1n 𝜎 22
 
12σ 1n σ
 ρ21 σ2 σ2 · · · ρ2n σ2 
C = 

𝜌21. 𝜎 2 𝜎. 2 .⋯ 𝜌2n.𝜎 2  (8)
.. .. .. .. 
𝑪=  (8)
ρm1⋮σ2 ρm2⋮σ · ⋱2 ·· σ⋮ 2

[𝜌m1 𝜎 2 𝜌correlated
The generation of normally distributed m2 𝜎
2
⋯ random 𝜎 2 ] numbers can be performed
by means of the Cholesky decomposition. When using the Cholesky decomposition,
The generation ofuncorrelated
normally distributed correlated
(a1 , arandom numbers can be performed
normally distributed random values 2 , . . . am ) with a mean value of zero
by means of the Cholesky decomposition. When using
and a standard deviation of one must first be generated. These the Cholesky decomposition,
represent the size of the
normally distributed
scattering area. uncorrelated random values (a1, a2, … am) with a mean value of zero

and a The
standard deviation
Cholesky of one must
decomposition first
of the be generated.
correlation matrix These
leadsrepresent
to: the size of the
scattering area.
= L · LT matrix leads to:
The Cholesky decomposition of theCcorrelation (9)

Subsequently, the normally distributed 𝑳 ⋅ 𝑳T values (a , a , . . . a ) are multiplied

𝑪 = random (9)
1 2 m
withSubsequently,
the triangularthe
matrix L:
normally distributed random values (a1, a2, … am) are multiplied
with the triangular matrix
 ∗ L:
···
   
a1 L11 0 0 a1
∗
 a∗  𝑎1  L𝐿2111 L22  𝑎
0 ·⋯
·· 00 1 a2 
 2 
 ..  =∗  .. · (10)
   
.. .⋯ . ..
 .  𝑎2  𝐿. 21 𝐿. 22 .. 0..  𝑎

2 . 
∗ = ⋅ (10)
am ⋮ Lm1⋮ Lm2 ⋮ · ·⋱· Lmn
⋮ ⋮ am
∗ 𝐿mn ] [𝑎m ] which must be added to
[𝑎m
This results in a normally ] [𝐿m1 correlated
distributed 𝐿m2 ⋯ distribution,
the mean value of
This results in the desired distribution
a normally f ct,m ) [25,27]:
(here this isdistribution,
distributed correlated which must be added
to the mean value of the desired distribution (here this∗is fct,m) [25,27]:
f ct,i = f ct,m · (1 + ai ) (11)
𝑓ct,i = 𝑓ct,m ⋅ (1 + 𝑎i∗ ) (11)
As laid out before, Lx should be defined in order to give a meaningful correlation by
As laid out
considering the before, Lx should
representative be defined
element in order
size and so as to give a meaningful
generate correlationwith
a good dependency by
considering thecorrelation
regard to the representative element
function at the size andtime.
same so as According
to generateto a good dependency
the correlation with
function
regard to the
in Figure correlation
8, there function
is almost at the same
no correlation time.
in the According
case of le /Lx =to2.the correlation
This means thatfunction
almost
inuncorrelated
Figure 8, there is almost
random noare
values correlation
generatedinfor theLxcase of leBased
= le /2. /Lx = 2.onThis means
Section 3.2,that
thisalmost
should
uncorrelated random values are generated for Lx = le/2. Based on Section 3.2, this should
be valid for a representative element size ∆x of 2∙dg to 3∙dg. Therefore, the correlation length
can subsequently be assumed according to:
Appl. Sci. 2021, 11, 785 9 of 19

Appl. Sci. 2021, 11, x FOR PEER REVIEW 9 of 19

be valid for a representative element size ∆x of 2·dg to 3·dg . Therefore, the correlation
length can subsequently be assumed according to:

1.01.0
· dg⋅ ≤
𝑑g L≤ 𝐿≤x 1.5
x
≤ 1.5
· dg⋅ 𝑑g (12)
(12)
which means
which means aa correlation
correlation length
length LLxxofofbetween
between1616and and 2424mm mm forfor
an anaggregate sizesize
aggregate dg =
16 mm.
dg = 16 mm.
Figure 99illustrates
Figure illustratesthe
the influence
influence of of
thethe correlation
correlation lengthlength for uncorrelated
for uncorrelated and
and cor-
related distributions of the tensile strength along the model axis for element sizes ofof55
correlated distributions of the tensile strength along the model axis for element sizes
and 20
and 20 mm.
mm. All
All distributions
distributions areare generated
generatedrandomly
randomlyaccording
accordingtotothe thevalues
valuesspecified
specified in
Figure
in Figure9. For uncorrelated
9. For uncorrelatedvalues, the distribution
values, of le =of5 lmm
the distribution has too many local minima
e = 5 mm has too many local
and maxima
minima compared
and maxima to le = 20 to
compared mm.le =The20 distributions along the axis
mm. The distributions alongarethenotaxis
comparable
are not
(although the
comparable probability
(although density functions
the probability are almost
density functions arethe same),
almost leadingleading
the same), to a mesh
to a
mesh dependency. In the case of correlated values, there is almost no mesh dependency, the
dependency. In the case of correlated values, there is almost no mesh dependency, as as
numbers
the numbersof local minima
of local minimaandandmaxima
maxima areare
similar forfor
similar le =le20=mm20 mmandand
le = l5e mm.
= 5 mm.

(a) (b)

Figure9.9. Element
Figure length llee vs.
Element length vs. correlation lengthLLxxfor
correlationlength for(a)
(a)uncorrelated
uncorrelatedand
and(b)
(b)correlated
correlateddistributions
distributionsalong
along the
the model
model axis.
axis.

3.4. Bond–Slip Relationship

3.4. Bond–Slip Relationship
The bond–slip behaviour between the concrete bar and reinforcement bar is simulated
The bond–slip behaviour between the concrete bar and reinforcement bar is simulated
in the CPM using
in the CPM using springs
springs with
with aa stiffness
stiffness equal
equal to:
to:
c = (𝜏b (𝑠)/𝑠) ⋅ 𝑢s ⋅ 𝑙e (13)
c = (τb (s)/s) · us · le (13)
where τb(s) is the local bond stress; s is the local slip (relative displacement between
where τ b (s) is the local bond stress; s is the local slip (relative displacement between concrete
concrete and rebar); us is the perimeter of the reinforcement bar; and le is the element
and rebar); us is the perimeter of the reinforcement bar; and le is the element length.
length.
A bond–slip relationship according to Equation (14) is chosen in the first instance,
A bond–slip relationship according to Equation (14) is chosen in the first instance,
with which a large variety of bond laws can be integrated:
with which a large variety of bond laws can be integrated:
 α
s 𝑠 α
τb (s)𝜏b=(𝑠)τb,max · ⋅( )
= 𝜏b,max (14)
(14)
s1 𝑠1
This correlation
This correlation corresponds to the the ascending
ascendingbranch
branchofofthethebond
bondlaw lawaccording
according to
Model
to ModelCode
Code2010 [12].[12].
2010 For calculations in the serviceability
For calculations limit state,
in the serviceability limitusing theusing
state, ascending
the
branch ofbranch
ascending the bond–slip relationship
of the bond–slip is sufficient
relationship according
is sufficient to [28].
according ThisThis
to [28]. considers the
considers
expected
the maximum
expected maximum crack width
crack of 0.4ofmm,
width which
0.4 mm, corresponds
which to a slip
corresponds ofslip
to a 0.2 mm at mm
of 0.2 both
sides
at bothofsides
the of
cracks, whichwhich
the cracks, is clearly within
is clearly the ascending
within branch
the ascending of the
branch bond–slip
of the bond–
relationship.
slip relationship.
For the general form of the bond–slip relationship according to Equation (14), some
reference values based on the technical literature are listed in Table 1 and are illustrated
in Figure 10a for concrete strength class C30/37.
Appl. Sci. 2021, 11, x FOR PEER REVIEW 10 of 19

Table 1. Parameters for the ascending branch of the bond–slip relationship for normal-strength
Appl. Sci. 2021, 11, 785 concrete. 10 of 19

Model Bond Conditions τb,max s1 α

MC2010 [12] Good 2.5 ⋅ √𝑓cm 1.0 0.40
For the general form of the bond–slip relationship according to Equation (14), some
MC2010 [12] All other cases 1.25 ⋅ √𝑓cm 1.8 0.40
reference values based on the technical literature are listed in Table 0,95 1 and are illustrated in
Farra [29] Mean 0.22 ⋅ 𝑓cm 1.0 0.30
Figure 10a for concrete strength class C30/37.
König and Tue [30] 0.31 ⋅ 𝑓cm 1.0 0.30
2/3
Noakowski [31] Good 1.0 0.12
0.95 ⋅ 𝑓cmfor normal-strength concrete.
Table 1. Parameters for the ascending branch of the bond–slip relationship
Huang [32] Good 0.45 ⋅ 𝑓cm 1.0 0.40
Model Bond Conditions τ b,max s1 α
Even if the[12]
real bond behaviour differs significantly
2.5 · from the bond–slip relationship
p
MC2010 Good pf cm 1.0 0.40
in Equation (14), user-defined bond–slip relationships
MC2010 [12] All other cases · be
1.25can f cm defined in the CPM
1.8 0.40using
Farra [29] Mean · 0.95
0.22
slip values and the corresponding bond stress values. Between f cm the points, 0.30
1.0 a linear
König and Tue [30]
interpolation is used to determine the missing values. The2/30.31 · f cm 1.0
only prerequisite 0.30is the
here
Noakowski [31] Good · f cm
0.95exemplarily 1.0 0.12
starting point at the coordinate origin (τb(s = 0) = 0). An user-defined bond–
Huang [32] Good 0.45 · f cm 1.0 0.40
slip relationship is shown in Figure 10b for a splitting failure.

(a) (b)
Figure 10.
Figure Bond laws:
10. Bond laws: (a)
(a) according
according to
to Equation
Equation (14)
(14) and
and Table
Table 1;
1; (b)
(b) user-defined.
user-defined.

Even
A if the real
reduction of bond behaviour
the local differs
bond stress insignificantly
the vicinity from the bond–slip
of a separation relationship
crack, e.g., due in
to
internal cracking, can be considered by using a linear reduction factor accordingusing
Equation (14), user-defined bond–slip relationships can be defined in the CPM slip
to [12]:
values and the corresponding bond stress values. Between the points, a linear interpolation
is used to determine the missing values. 𝜏b,red (s,
The𝑥) only
= λ(𝑥) ⋅ 𝜏b (𝑠)
prerequisite here is the starting point(15)
at
the coordinate origin (τ b (s = 0) = 0). An exemplarily user-defined bond–slip relationship is
where:
shown in λ(x) is the
Figure factor
10b for afor a reduction
splitting failure.of the local bond stress at distance x from the
separation
A reduction of the local bond stressthe
crack; x is the distance from crack;
in the and τof
vicinity b(s) is the local bond stress (e.g.,
a separation crack, e.g., due to
according to Equation (14)).
internal cracking, can be considered by using a linear reduction factor according to [12]:
In Model Code 2010 [12], λ(x) is defined by:
τb,red (s, x ) = λ( x ) · τb (s) (15)
λ(𝑥) = 0.5 ⋅ 𝑥/𝜙 ≤ 1.0 (16)
where: λ(x)
As an is the factor
alternative forlinear
to the a reduction
reductionoffactor
the local bond to
according stress at distance
Equations x from
(15) and the
(16), it is
separation crack; x is the distance from the crack; and τ b (s) is the local bond stress
also possible to take into account the zone of the disturbed bond using a regionally separated (e.g.,
according
bond to Equation
stress–slip (14)). according to [33–35]. It must be noted that the numerical
relationship
In Model Code 2010 [12], λ(x) is time
calculation effort and the computation defined by: significantly with the used model.
increase
λ( x ) = 0.5 · x/φ ≤ 1.0 (16)

As an alternative to the linear reduction factor according to Equations (15) and (16),
it is also possible to take into account the zone of the disturbed bond using a regionally
separated bond stress–slip relationship according to [33–35]. It must be noted that the
numerical calculation effort and the computation time increase significantly with the
used model.
Appl. Sci. 2021, 11, 785 11 of 19
Appl. Sci. 2021, 11, x FOR PEER REVIEW 11 of 19

3.5. Model Features and Advantages

3.5. Model Features and Advantages
The model is characterized by its fast and stable calculation of cracking. In addi-
The model is characterized by its fast and stable calculation of cracking. In addition,
tion, the CPM is easily extendable, e.g., to simulate cracking of fiber-reinforced concrete
the CPM is easily extendable, e.g., to simulate cracking of fiber-reinforced concrete (FRC)
(FRC) by considering the fiber effects as a crack-width-dependent additional element load.
by considering the fiber effects as a crack-width-dependent additional element load.
Furthermore, simulation of high-performance concrete (HPC)and ultra-high performance
Furthermore, simulation of high-performance concrete (HPC)and ultra-high performance
concrete (UHPC) is possible by setting the properties of the concrete bar and adjusting the
concrete
bond–slip (UHPC) is possible
relationship. by setting an
Additionally, theextension
propertiestoofa the concrete
three-bar bar and
model (seeadjusting
Figure 11)the
is
bond–slip relationship. Additionally, an extension to a three-bar model
possible, which enhances the simulation of elements with different reinforcing elements (see Figure 11) is
possible, which with
(e.g., elements enhances the simulation
prestressed of elements
conventional steelwith different
rebars). reinforcingreinforcement
Non-metallic elements (e.g.,
elements with prestressed conventional steel rebars). Non-metallic
(such as carbon concrete, basalt, or glass fiber reinforcements) can be accommodated reinforcement (such by
as
carbon concrete, basalt, or glass fiber reinforcements) can be accommodated
adjusting the material parameters of the “reinforcement bar” and the bond–slip relation- by adjusting the
material
ship. Byparameters
adjusting theof the “reinforcement
support conditions,bar” and the bond–slip
constrained stresses relationship. By adjusting
can also be investigated.
the support conditions, constrained stresses can also be investigated. A transfer
A transfer to components subjected by flexure is possible if the tension and compression to components
subjected by flexure
zone are treated is possible
separately [5].if the tension and compression zone are treated separately [5].

Figure
Figure 11.
11. Extension
Extension of
of the
the CPM
CPM to
to aa three-bar
three-bar model.
model.

In addition
In addition toto the
the listed
listed expansion
expansion possibilities
possibilities for
for different
different materials,
materials, the
the CPM
CPM can
be expanded
be expanded withwith regard
regard toto the scattering
scattering of the material parameters. In In particular,
particular, the
scattering of
scattering of the
the bond
bond stress
stress should
should bebe mentioned
mentioned here.
here. Other
Other scattering
scattering parameters
parameters (e.g.,
(e.g.,
the modulus
the modulus of of elasticity
elasticity ofof the
the rebar
rebar and
and concrete)
concrete) are
are expected
expected to to be
be less significant
significant or
can be
can be represented by by already
alreadyexisting
existingscattering
scatteringparameters
parameters(e.g.,
(e.g.,scattering
scatteringdimensions
dimensionsof
thethe
of concrete
concretebar
barbybyscattering
scatteringofofthethetensile
tensilestrength,
strength,scattering
scattering diameter
diameter of the rebar by by
scattering of
scattering of bond
bond stress).
stress).

4. Simulation
4. Simulation andand Calculation
Calculation Procedure
Procedure
With the preceding explanations,
With the preceding explanations, the the CPM
CPM can
can be
be used
used to
to simulate
simulate the
the cracking
cracking of
of aa
reinforced concrete tensile specimen. In the following, the programme sequence will be
reinforced concrete tensile specimen. In the following, the programme sequence will be
briefly explained. A representation can be taken from Figure 12.
briefly explained. A representation can be taken from Figure 12.
In the first calculation step, the entire model or component is uncracked. A load
In the first calculation step, the entire model or component is uncracked. A load
increase follows until the concrete tensile strength is exceeded at the “weakest” point.
increase follows until the concrete tensile strength is exceeded at the “weakest” point.
Then, the stiffness of the cracked concrete element is reduced to zero ((EA) = 0) (also see
Then, the stiffness of the cracked concrete element is reduced to zero ((EA)c,ic,i = 0) (also see
Figure 3). The model is recalculated with the adjusted stiffness value (cracked concrete
Figure 3). The model is recalculated with the adjusted stiffness value (cracked concrete
element). The load is then increased further until the concrete tensile strength is exceeded
element). The load is then increased further until the concrete tensile strength is exceeded
at another point. The process is repeated until the desired load level Ns,end is reached.
at another point. The process is repeated until the desired load level Ns,end is reached.
The iterative calculation procedure in the CPM will be illustrated below using an
The iterative calculation procedure in the CPM will be illustrated below using an
example. The calculations are based on the parameters listed in Table 2.
example. The calculations are based on the parameters listed in Table 2.
Appl. Sci. 2021, 11, 785 12 of 19
Appl. Sci. 2021, 11, x FOR PEER REVIEW 12 of 19

Figure 12. 12.

Figure Iterative calculation
Iterative procedure
calculation for the CPM.
procedure for the CPM.

Table
Table 2. CPM2. CPM parameters
parameters forfor
thethe caseexample.
case example.

Geometry
Geometry
Reinforcement ratio ρ 2%
Diameter of reinforcing bar Reinforcement ratio ∅ ρ16 mm 2%
Length of model Diameter of reinforcing bar l ∅
2000 mm 16 mm
Element length Length of model le l20 mm 2000 mm
Element
Material
length le 20 mm
Concrete: C20/25 Material
Mean value of compressive strength
Concrete: C20/25of concrete fcm 28 N/mm2
Mean value of tensile strength of concrete
Mean value of compressive strength of concrete fct,m
f cmN/mm
2.2 2
28 N/mm2
Mean value of modulus of elasticity of concrete
Mean value of tensile strength of concrete E cm 30,000
f ct,m N/mm 2
2.2 N/mm2
Reinforcement: B500 Mean value of modulus of elasticity of concrete Ecm 30,000 N/mm2
Modulus of elasticity ofReinforcement:
reinforcing steelB500 Es 200,000 N/mm2
Bond–slipModulus
relationshipof elasticity of reinforcing steel Es 200,000 N/mm2
Model Code 2010 (good bond conditions)
Bond–slip relationship
Bond stress 𝜏b (𝑠) = 13.23 ⋅ (𝑠⁄1.0)0.4
Reduction factor Model Code 2010 (good bond conditions) λ(𝑥) = 0.5 ⋅ 𝑥/16 ≤ 1.0
Bond stress τb (s) = 13.23 · (s/1.0)0.4
Reduction factor λ( x ) = 0.5 · x/16 ≤ 1.0
Scattering
Tensile strength of concrete
Coefficient of variation vk 10%
Correlation length Lx 20 mm
Appl. Sci. 2021, 11, 785 13 of 19

Figure 13 shows the strains determined with the CPM for different load levels (ex-
pressed by the maximum steel stress σs,max ). At the cracks, the concrete stress is zero.
The first crack occurs with a steel stress of σs,max = 100 N/mm2 at the point of lowest
concrete tensile strength, depending on the distribution of the specimen tensile strength
(for comparison, without scattering of the concrete tensile strength vk = 0, cracking would
not occur until σs,max = 125 N/mm2 ). Outside the transmission length, the steel and
concrete stresses between the cracks are identical until the completed crack pattern with
σs,max = 114 N/mm2 is reached. Up to this loading level five cracks are formed, mainly
at the low points of the strength distribution. Due to the non-linear bond law, a further
increase in the load level leads to an increase in the bond forces, and thus also to a greater
introduction of force into the concrete. This leads to further cracking occurring in the
stabilized cracking stage (progressive stabilized cracking). These cracks predominantly
occur centrally between the existing cracks, so that under certain circumstances, regions
with higher strength values are also affected by crack formation, i.e., “late” cracks
Appl. Sci. 2021, 11, x FOR PEER REVIEW can
14 of 19
emerge under relatively high stresses.

Figure
Figure 13. CPM
CPM case
case example
example showing
showing the
the stress
stress distributions
distributions and
and crack
crack sequences
sequences along
along the
the model
model axis
axis for
for different
different
stress
stress levels.
levels.

5. Comparison with Experiments

The performance of the CPM will be demonstrated in the following using a series of
tests from [36]. The test series included three identical large-scale tests. The focus of the
Appl. Sci. 2021, 11, 785 14 of 19

The individual crack spacing and crack widths for each stress level can then be read
out from the CPM calculation results. If several calculation runs are carried out with the
newly generated specimen tensile strength distributions along the model axis in each case,
extensive distributions of crack spacing and crack widths can be generated as functions of
the stress level and a large data basis can be obtained. In contrast to the stages defined in
Figure
Figure 13. CPM case example 1, thethe
showing switch
stressbetween crack
distributions formation
and stages along
crack sequences and the
thestabilized
model axiscracking stage is
for different
stress levels. quite smooth. Hence, stress regions causing fast crack formation can be identified.

5. Comparison with Experiments

The performance
performance of of the
theCPM
CPMwillwillbe
bedemonstrated
demonstratedininthe thefollowing
following using
using a series
a series of
of tests from [36]. The test series included three identical large-scale tests.
tests from [36]. The test series included three identical large-scale tests. The focus of the The focus
of thewas
tests testsonwastheon generation
the generation of many
of many cracks
cracks andinvestigation
and investigationofofthe
the crack spacing
spacing
distribution.
distribution. For Forthis
thispurpose,
purpose,a very highhigh
a very reinforcement ratio of
reinforcement ρ = 8%
ratio of ρwith a bar
= 8% diameter
with a bar
of 32 mmof
diameter was
32 chosen.
mm wasAchosen.
tensileAforce of 900
tensile forcekNofwas
900 required to reach to
kN was required a stress
reach level of
a stress
= 280 N/mm 2 Figure2 14 shows the geometrical parameters of the test specimens
level
σ s,max of σs,max = 280 . N/mm . Figure 14 shows the geometrical parameters of the test
used in [36].used in [36].
specimens

Figure 14. Details of the axial tensile specimen (dimensions shown in mm) [36].

The material parameters of the concrete were determined in [36] using compression
tests and splitting tensile tests (no data available on scattering). The maximum aggregate
size was 16 mm. The parameters used in the CPM are listed in Table 3. A correction of the
tensile strength according to Section 3.2 was not performed, as splitting tensile tests were
conducted (here f ct,f,m = f ct,m ).

Table 3. CPM parameters for the test series conducted by Naotunna [36].

Geometry
Reinforcement ratio ρ 8%
Diameter of reinforcing bar ∅ 32 mm
Length of model l 2000 mm
Element length le 10 mm
Material
Concrete
Mean value of compressive strength of concrete f cm 35 N/mm2
Mean value of tensile strength of concrete f ct,m 3.2 N/mm2
Mean value of modulus of elasticity of concrete Ecm 32,000 N/mm2
Reinforcement: B500
Modulus of elasticity of reinforcing steel Es 200,000 N/mm2
Bond slip relationship
Model Code 2010 (good bond conditions)
Bond stress τb (s) = 14.79 · (s/1.0)0.4
Reduction factor λ( x ) = 0.5 · x/32 ≤ 1.0
Scattering
Tensile strength of concrete 18 distributions (in each case)
Coefficient of variation vk 10% (15%)
Correlation length Lx 20 mm (10 mm)
Appl. Sci. 2021, 11, 785 15 of 19

The calculations were first carried out for a coefficient of variation of vk = 10% and
a correlation length of 20 mm (~1.25 dg , cf. Equation (12)). In a second calculation run, a
correlation length of 10 mm was applied for comparison and to consider the small crack
spacings. To account for the uncommonly high reinforcement ratio and large reinforcement
diameter in combination with the relatively small crack spacing, a coefficient of variation
of vk = 15% was also applied. Since in [36] the crack spacing was recorded on 3 sides at
the depths of each reinforcement layer (3 specimens × 3 surfaces × 2 bars = 18 sets), 18
generated distributions were recorded for each calculation run. To take the disturbed stress
region at loading points into account, both the values presented in [36] and the CPM did
not take into account the last crack spacing on both sides of the test specimens.
The distribution of experimental crack spacing and the calculated crack spacing with
the CPM is shown in Figure 15a for vk = 10% and in Figure 15b for vk = 15%. The statistical
parameters are summarized in Table 4. A calculation with twice as many elements (le = 5
Appl. Sci. 2021, 11, x FOR PEER REVIEW
mm) showed similar results, with the exception of small deviations in the distribution 16 ofof
19
the crack spacings due to the randomly generated tensile strength distributions.

(a) (b)
Figure
Figure15. 15.Distribution
Distributionof
ofexperimental
experimentaland
andcalculated
calculatedcrack
crackspacings
spacingsininthe
thetest
testconducted
conductedby
byNaotunna
Naotunna[36]: (a)vkvk==10%;
[36]:(a) 10%;
(b)vvkk== 15%.
(b) 15%.

Table4.4.The
Table Thestatistical
statisticalparameters
parametersfrom
fromFigure
Figure15.
15.

Naotunna [36] Naotunna [36]CPM: vk = 10%

CPM: vk = 10% CPM: CPM: vk = 15%
vk = 15%
Lx = 20 mm L x = 20 mm
Lx = 10 mmLx = 10 mm Lx = 20 mm
Lx = 20 mm Lx L = 10
=x10 mmmm
Total number n 1 230 214 217 215 224
Total number n 1 230 214 217 215 224
Mean s(section)131–140
Mean (section) sr,m mm 131–140 mm
141–150 mm 141–150 mm
141–150 mm141–150 mm 141–150
141–150 mm mm 131–140 mmmm
131–140
r,m
Median (section)
Median (section) 121–130 mm 121–130 mm
131–140 mm 131–140 mm
131–140 mm131–140 mm 131–140
131–140 mm mm 121–130
121–130 mmmm
95th
95thpercentile
percentile (section) sr,0.95mm 191–200 mmmm 191–200 mm mm 191–200 mm 191–200 mm 181–190
181–190
191–200 191–200 191–200 191–200 mm mmmm
(section) sr,0.95
sr,0.95/sr,m ~1.5 ~1.4 ~1.4 ~1.4 ~1.4
sr,0.95 /sr,m ~1.5 ~1.4 ~1.4 ~1.4
1 Recorded crack spacings for 3 specimens, 3 surfaces, and 2 bars each (6 sets per specimen).
~1.4
1 Recorded crack spacings for 3 specimens, 3 surfaces, and 2 bars each (6 sets per specimen).
Thecomparison
The comparisonof of thethe experimental
experimental and and calculated
calculated crack spacing
crack spacing shows
shows the the
follow-
following
ing results:results:
• In Ingeneral,
general,thethetest
testresults
resultscan
canbebereproduced
reproducedvery
verywell
wellwith
withthe
theCPM
CPM(see(seeFigure
Figure15 15
and Table
and Table 4);4);
• Varying
Varyingthethe coefficient
coefficient of variation
of variation and correlation
and correlation length to the
length to determine determine the
distribution
of the tensile strength along the model axis has only a limited effect on the results; on
distribution of the tensile strength along the model axis has only a limited effect
• the range
The results;of very small crack spacings (sr ≤ 100 mm) is underestimated for v = 10%
k
 and Thevrange of very small crack spacings (𝑠r ≤L100
k = 15%, as well as for Lx = 20 mm and
mm) is underestimated for vk = 10%
x = 10 mm (see Figure 15);
and vk = 15%, as well as for Lx = 20 mm and Lx = 10 mm (see Figure 15);
 The experimental value sr,0.95/sr,m = 1.5 is estimated with adequate accuracy by the
CPM (sr,0.95/sr,m = 1.4) (see Table 4);
 The comparison of the median and mean values of the experimental results show a
positive skew (skewed to the right) (see Table 4) indicating a log-normal distribution
of the crack spacing sr.
Appl. Sci. 2021, 11, 785 16 of 19

• The experimental value sr,0.95 /sr,m = 1.5 is estimated with adequate accuracy by the
CPM (sr,0.95 /sr,m = 1.4) (see Table 4);
• The comparison of the median and mean values of the experimental results show a
positive skew (skewed to the right) (see Table 4) indicating a log-normal distribution
of the crack spacing sr .
The high number of very small crack spacings (sr ≤ 100 mm) can be attributed to the
fact that complete separation cracks do not always form in testing and that overlapping
cracks can occur. Both phenomena can be observed in [36]. Furthermore, the effects of
shrinkage strains have not been taken into account in the calculation with the CPM, as the
respective input values were not documented in [36]. However, the model can consider
shrinkage by adding shrinkage strains as an additional load component to reduce the
effective tensile concrete strength [6,7]. As a result, the crack spacing would be reduced.

6. Discussion and Conclusions

In the present paper, a simulation model of the random crack propagation in reinforced
concrete elements is presented. The model consists of two uniaxial bars (concrete and
reinforcing steel) divided into finite elements, which are connected with springs. For the
random cracking, random concrete tensile strength values are assigned to the element.
Several features and advantages of this model are introduced in the paper, the main
objective of which is to discuss the major role of the assumed tensile strength and its
distribution, as well as the element size considerations in the CPM.
The most important parameter to simulate random cracking is the uniaxial tensile
strength along the model axis. It was pointed out that a normal distributed concrete tensile
strength can be used for a common range of coefficients of variation (vk < 20%). This
was confirmed by investigations in [4], according to which the assumed distribution type
of the tensile strength has only a minor effect on the calculated crack widths. In CPM,
different coefficients of variation can be assigned to the normal distribution. The model
also distinguishes between the mean value of the fracture tensile strength f ct,f and the
mean value of specimen tensile strength f ct,m . If several tests using a test method where the
fracture plane develops randomly at the weakest point of the specimen (e.g., direct tensile
test without a notch) are available, then the test results represent only the distribution of
the weakest tensile strength values f ct,f (here this is the fracture tensile strength). To capture
the tensile strength distribution in an entire bar along the longitudinal axis with weak
and strong parts (here this is the specimen tensile strength), increasing the mean value of
specimen tensile strength f ct,m seems logical. For this purpose, an approach was proposed
in Section 3.2, setting f ct,m as equal to the 95th percentile of fracture tensile strength values
f ct,f based on [3]. This approach requires further experimental evidence; however, there
are no common alternative approaches available in the literature. Since the approach was
applied for direct tensile tests (without a notch), an application for splitting tensile tests (as
in experiments in [36]) is not recommended without further research.
For numerical models, the results are supposed to be mesh-independent. Since the
number of tensile strength values along the model axis in the CPM is dependent on the
number of elements (and the respective element lengths), tensile strength values are not
supposed to scatter randomly between adjacent elements. For this purpose, a dependency
between adjacent elements was defined in the model (autocorrelation). The correlation
length—present in the correlation function—specifies the scope of this dependency. The
correlation length is a mathematical value, which when combined with mechanical values,
enables a mesh-independent and realistic distribution of the tensile strength along the
longitudinal axis of a specimen. Since realistic data on the correlation length for tensile
strength are scarce [4], for each problem different correlation length values are available
in the literature [37]. Therefore, the representative length was introduced in this paper
with a mechanically based value considering the aggregate size in order to stay within
the macroscale.
Appl. Sci. 2021, 11, 785 17 of 19

The comparison with the experimental test in [36] shows a good reproduction of the
results with the CPM, although small crack spacings are slightly underestimated by the
CPM. The main reason for this underestimation is probably the influence of shrinkage.
The investigations in [38] confirm a high influence of shrinkage strains, especially for high
reinforcement ratios. However, no information on shrinkage strain is available in [36].
Nevertheless, the tests in [36] were selected for the comparison, as detailed information on
crack spacing distributions is very rare in recent studies (the same applies for crack widths,
which were not recorded in [36]).
The CPM also enables investigation of the entire sequential cracking process, which is
expected to show a higher dependency of the results on the tensile strength distribution
under low load levels. In further investigations, the CPM will be used to generate load
Appl. Sci. 2021, 11, x FOR PEER REVIEW
deformation curves of reinforced concrete ties (as in [39]). The results can be used 18 of to
19
define a damage propagation index based on an increasing number of cracks or decreasing
crack spacing (compare Figure 16) (cf. [40]). Based on the index, the load-level-dependent
progress of
progress ofthe
thedamage
damagecancanbebederived,
derived,including
including the
the statistical
statistical distribution
distribution of the
of the dam-
damage
age parameters.
parameters.

Figure
Figure 16.
16. Crack
Crack spacing
spacing and
and number
number of
of cracks
cracks versus
versus the
the load
load level.
level.

Author Contributions: Conceptualization, M.E., J.C. and S.J.; methodology, M.E. and J.C.;
Author Contributions: Conceptualization, M.E., J.C. and S.J.; methodology, M.E. and J.C.; analyses,
analyses, J.C. and M.E.; writing—original draft preparation, J.C.; writing—review and editing, S.J.
J.C. and M.E.; writing—original draft preparation, J.C.; writing—review and editing, S.J. and M.E.;
and M.E.; supervision, M.E.; project administration, M.E. and S.J.; funding acquisition, M.E. All
supervision, M.E.; project administration, M.E. and S.J.; funding acquisition, M.E. All authors have
authors have read and agreed to the published version of the manuscript.
read and agreed to the published version of the manuscript.
Funding: The investigations presented in this paper are part of a research project conducted at the
Funding: The investigations presented in this paper are part of a research project conducted at the
Institute of Building Materials, Concrete Construction and Fire Safety(iBMB), Division of Concrete
Institute of Building Materials, Concrete Construction and Fire Safety(iBMB), Division of Concrete
Construction of the Technische Universität Braunschweig as part of Research Training Group 2075
Construction of the Technische Universität Braunschweig as part of Research Training Group 2075
project “Modeling the Constitutive Evolution of Building Materials and Structures with Respect to
project “Modeling the Constitutive Evolution of Building Materials and Structures with Respect to
Aging”, funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation),
Aging”, funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation),
project number 255042459. We acknowledge support by the German Research Foundation and the
project number 255042459. We acknowledge support by the German Research Foundation and the
Open Access Publication Funds of Technische Universität Braunschweig.
Open Access Publication Funds of Technische Universität Braunschweig.
Conflicts of Interest: The authors declare no conflict of interest. The funders had no role in the
Conflicts of Interest: The authors declare no conflict of interest. The funders had no role in the
design of the study, in the collection of analyses and in the interpretation of data; in writing of the
design of the study, in the collection of analyses and in the interpretation of data; in writing of the
manuscript, and in decision to publish the results.
manuscript, and in decision to publish the results.
References
References
1.
1.
Base, G.D.; Read, J.B.; Beeby, A.W.; Taylor, H.P.J. An Investigation of the Crack Control Characteristics of Various Types of Bar in
Base, G.D.; Read, J.B.; Beeby, A.W.; Taylor, H.P.J. An Investigation of the Crack Control Characteristics of Various Types of Bar in
Reinforced
Reinforced Concrete Beams, Report
Concrete Beams, Report 18,
18, Part
Part 1;
1; Cement
Cement and
and Concrete
Concrete Association
Association Research:
Research: London,
London, UK,
UK, 1966.
1966.
2.
2. Rüsch,
Rüsch, H.; Rehm, G. Versuche mit Betonformstählen, Teil I-III. In Heftreihe des Deutschen Ausschuss für Stahlbeton,
H.; Rehm, G. Versuche mit Betonformstählen, Teil I-III. In Heftreihe des Deutschen Ausschuss für Stahlbeton, 1st
1st ed.;
ed.; Beuth:
Beuth:
Berlin, Germany, 1963; Volume 140, pp. 160–165.
Berlin, Germany, 1963; Volume 140, pp. 160–165.
3.
3. Empelmann,
Empelmann, M. M. Zum
Zumnichtlinearen
nichtlinearenTrag-
Trag-undund Verformungsverhalten
Verformungsverhalten von von Stabtragwerken
Stabtragwerken aus Konstruktionsbeton
aus Konstruktionsbeton unter
unter beson-
besonderer Berücksichtigung von Betriebsbeanspruchungen. Ph.D. Thesis, RWTH Aachen,
derer Berücksichtigung von Betriebsbeanspruchungen. Ph.D. Thesis, RWTH Aachen, Aachen, Germany, 1995.Aachen, Germany, 1995.
4.
4. Häußler-Combe,
Häußler-Combe, U.; U.; Hartig,
Hartig, J.
J. Rissbildung
Rissbildung vonvon Stahlbeton bei streuender
Stahlbeton bei streuender Betonzugfestigkeit.
Betonzugfestigkeit. Bauingenieur
Bauingenieur 2010,
2010, 85,
85, 460–470.
460–470.
5.
5. Empelmann, M.; Cramer, J. Rissbreiten an biegebeanspruchten Bauteilen. Beton Stahlbetonbau 2018,
Empelmann, M.; Cramer, J. Rissbreiten an biegebeanspruchten Bauteilen. Beton Stahlbetonbau 2018, 113, 291–297. [CrossRef] 113, 291–297,
doi:10.1002/best.201700092.
6. Empelmann, M.; Cramer, J. Modell zur Beschreibung der zeitabhängigen Rissbreitenentwicklung in Stahlbetonbauteilen. Beton
Stahlbetonbau 2019, 114, 327–336, doi:10.1002/best.201800109.
7. Empelmann, M.; Cramer, J. Time-dependent development of crack widths in reinforced concrete structures. In Concrete
Innovations in Materials, Design and Structures, Proceedings of the fib Symposium, Kraków, Poland, 27–29 May 2019; International
Appl. Sci. 2021, 11, 785 18 of 19

6. Empelmann, M.; Cramer, J. Modell zur Beschreibung der zeitabhängigen Rissbreitenentwicklung in Stahlbetonbauteilen. Beton
Stahlbetonbau 2019, 114, 327–336. [CrossRef]
7. Empelmann, M.; Cramer, J. Time-dependent development of crack widths in reinforced concrete structures. In Concrete Innovations
in Materials, Design and Structures, Proceedings of the fib Symposium, Kraków, Poland, 27–29 May 2019; International Federation for
Structural Concrete: Lausanne, Switzerland, 2019.
8. Liao, W.-C.; Chen, P.-S.; Hung, C.-W.; Wagh, S.K. An Innovative Test Method for Tensile Strength of Concrete by Applying the
Strut-and-Tie Methodology. Materials 2020, 13, 2776. [CrossRef] [PubMed]
9. Kim, J.J.; Taha, M.R. Experimental and Numerical Evaluation of Direct Tension Test for Cylindrical Concrete Specimens. Adv. Civ.
Eng. 2014, 2014, 1–8. [CrossRef]
10. Joint Committee for Structural Safety (JCSS). Probabilistic Model Code—Part III. 2002. Available online: https://www.jcss-lc.
org/jcss-probabilistic-model-code (accessed on 9 November 2020).
11. Mirza, S.A.; Hatzinikolas, M.; MacGregor, J.G. Statistical description of strength of concrete. ASCE J. Struct. Div. 1979, 105,
1021–1037.
12. Fédération Internationale du Béton (Fib). Fib Model Code for Concrete Structures 2010; Fédération Internationale du Béton: Lausanne,
Switzerland, 2013. [CrossRef]
13. DIN EN 1992-1-1:2004 + AC:2010. In Eurocode 2: Design of Concrete Structures—Part 1-1: General Rules and Rules for Buildings;
Beuth: Berlin, Germany, 2011.
14. Voigt, J. Beitrag zur Bestimmung der Tragfähigkeit Bestehender Stahlbetonkonstruktionen auf Grundlage der Systemzuverläs-
sigkeit. Ph.D. Thesis, University of Siegen, Siegen, Germany, 28 February 2014.
15. Rüsch, H.; Sell, R.; Rackwitz, R. Statistische Analyse der Betonfestigkeit. In Heftreihe des Deutschen Ausschuss für Stahlbeton, 1st ed.;
Beuth: Berlin, Germany, 1969; Volume 206.
16. Tue, N.V.; Schenck, G.; Schwarz, J. Eine kritische Betrachtung des zusätzlichen Sicherheitsbeiwertes für hochfesten Beton.
Bauingenieur 2007, 82, 39–46.
17. Schwennicke, A. Zur Berechnung von Stahlbetonbalken und -Scheiben im gerissenen Zustand unter Berücksichtigung der
Mitwirkung des Betons zwischen den Rissen. Ph.D. Thesis, Technische Universität Berlin, Berlin, Germany, 1983.
18. Koch, R. Verformungsverhalten von Stahlbetonstäben unter Biegung und Längszug im Zustand II, auch bei Mitwirkung des
Betons zwischen den Rissen. Ph.D. Thesis, University of Stuttgart, Stuttgart, Germany, 1976.
19. Heilmann, H.G. Beziehungen zwischen Zug- und Druckfestigkeit des Betons. Beton 1969, 19, 68–70.
20. Blaschke, F.; Losekamp, C.; Mehlhorn, G. Zugtragverhalten von Beton nach langandauernder statischer sowie schwellender
Zugvorbelastung. Beton Stahlbetonbau 1994, 89, 204–205. [CrossRef]
21. Duda, H. Bruchmechanisches Verhalten von Beton unter monotoner und zyklischer Zugbeanspruchung. In Heftreihe des Deutschen
Ausschuss für Stahlbeton, 1st ed.; Beuth: Berlin, Germany, 1991; Volume 419.
22. Bažant, Z.P.; Oh, B.H. Crack band theory for fracture of concrete. Mater. Struct. 1983, 16, 155–177. [CrossRef]
23. Jahn, M. Zum Ansatz der Betonzugfestigkeit bei den Nachweisen zur Trag- und Gebrauchsfähigkeit von unbewehrten und
bewehrten Betonbauteilen. In Heftreihe des Deutschen Ausschuss für Stahlbeton, 1st ed.; Beuth: Berlin, Germany, 1983; Volume 341.
24. Ord, J.K.; Getis, A. Local Spatial Autocorrelation Statistics: Distributional Issues and an Application. Geogr. Anal. 1995, 27,
286–306. [CrossRef]
25. Ahrens, A. Ein stochastisches Simulationskonzept zur Lebensdauerermittlung von Stahlbeton- und Spannbetontragwerken und
seine Umsetzung an einer Referenzbrücke. Ph.D. Thesis, Ruhr-Universität Bochum, Bochum, Germany, 3 December 2010.
26. Vanmarcke, E. Random Fields: Analysis and Synthesis; Rev. and expanded new ed.; World Scientific Publication: Singapore, 2010.
[CrossRef]
27. Real, M.D.V.; Filho, A.C.; Maestrini, S.R. Response variability in reinforced concrete structures with uncertain geometrical and
material properties. Nucl. Eng. Des. 2003, 226, 205–220. [CrossRef]
28. Tepfers, R.; Achillides, Z.; Azizinamini, A.; Balázs, G.; Bigaj-Van-Vliet, A.; Cabrera, J.; Cairns, J.; Cosenza, E.; Uijl, J.D.; Eligehausen,
R.; et al. Fib Bulletin 10, Bond of reinforcement in concrete. In Fib Bulletins; Fédération Internationale du Béton: Lausanne,
Switzerland, 2000; Volume 10.
29. Farra, B. Influence de la résistance du béton et de son adhérence avec l’armature sur la fissuration. Ph.D. Thesis, Ecole
Polytechnique Fédérale de Lausanne, Lausanne, Switzerland, 1995.
30. König, G.; Tue, N.V. Grundlagen und Bemessungshilfen für die Rissbreitenbeschränkung im Stahlbeton und Spannbeton. In
Heftreihe des Deutschen Ausschuss für Stahlbeton, 1st ed.; Beuth: Berlin, Germany, 1996; Volume 466.
31. Noakowski, P. Nachweisverfahren für Verankerung, Verformung, Zwangbeanspruchung und Rissbreite, Kontinuierliche Theorie
der Mitwirkung des Betons auf Zug. Rechenhilfen für die Praxis. In Heftreihe des Deutschen Ausschuss für Stahlbeton, 1st ed.; Beuth:
Berlin, Germany, 1988; Volume 394.
32. Huang, Z.; Engstrom, B.; Magnusson, J. Experimental and Analytical Studies of the Bond Behaviour of Deformed Bars in High Strength
Concrete, Preceedings of the 4th International Symposium on the Utilization of High Strength/High Performance Concrete, Paris, France,
29–31 May 1996; Pr. Ponts et Chaussées: Paris, France, 1996; Volume 3, pp. 1115–1124.
33. Eligehausen, R.; Popov, E.P.; Bertero, V.V.; Earthquake Engineering Research Center. Local Bond Stress-Slip Relationships of Deformed
Bars under Generalized Excitations; Report No. UCB/EERC-83/23; University of California: Berkeley, CA, USA, 1983.
Appl. Sci. 2021, 11, 785 19 of 19

34. Kreller, H. Zum nichtlinearen Trag- und Verformungsverhalten von Stahlbetonstabtragwerken unter Last- und Zwangeinwirkung.
In Heftreihe des Deutschen Ausschuss für Stahlbeton, 1st ed.; Beuth: Berlin, Germany, 1990; Volume 409.
35. Comite Euro-International Du Beton. CEB-FIP Model Code 1990. In CEB Bulletin; Thomas Telford Ltd.: London, UK, 1993;
Volume 213–214. [CrossRef]
36. Naotunna, C.N.; Samarakoon, S.S.M.M.K.; Fosså, K.T. Experimental and theoretical behavior of crack spacing of specimens
subjected to axial tension and bending. Struct. Concr. 2020, 1–18. [CrossRef]
37. Meinhardt, M.; Keuser, M.; Braml, T. Numerical Simulation of Crack Propagation Using Spatially Varying Material Parameters,
Proceedings of the 2018 Fib Congress, Melbourne, Australia, 7–11 October 2018; International Federation for Structural Concrete:
Lausanne, Switzerland, 2019.
38. Kaklauskas, G.; Gribniak, V. Effects of shrinkage on tension stiffening in RC members. In Structural Concrete and Time, Proceedings
for the 2005 Fib Symposium, La Plata, Argentina, 28–30 September 2005; International Federation for Structural Concrete: La Plata,
Argentina, 2005.
39. Rimkus, A.; Gribniak, V. Experimental investigation of cracking and deformations of concrete ties reinforced with multiple bars.
Constr. Build. Mater. 2017, 148, 49–61. [CrossRef]
40. Mobasher, B.; Stang, H.; Shah, S. Microcracking in fiber reinforced concrete. Cem. Concr. Res. 1990, 20, 665–676. [CrossRef]