Received: 17 February 2021 Revised: 9 April 2021 Accepted: 14 April 2021

DOI: 10.1002/cend.202100004

FULL PAPER

Calculation of maximum crack width for practical design of

reinforced concrete

Dirk Schlicke1 | Eva Maria Dorfmann1 | Ekkehard Fehling2 | Nguyen Viet Tue1

1
Graz University of Technology, Institute of
Structural Concrete, Graz, Austria Abstract
2
University of Kassel, Institute of Structural Cracks are an essential characteristic of reinforced concrete (RC) construction. Service-
Engineering, Chair of Structural Concrete,
ability and durability, however, require a reasonable limitation of the maximum crack
Kassel, Germany
width. The prediction of the maximum crack width of RC, however, is not trivial and has
Correspondence
been much debated over the past decades. This article starts with a fundamental com-
Dirk Schlicke, Graz University of Technology,
Institute of Structural Concrete, Graz, Austria. parison of basic procedures for determining the crack width, namely using a mechanical
Email: [email protected]
or calibrated model. Following, the behavior of reinforced concrete during crack forma-
tion is thoroughly discussed with a focus on the bond stress-slip relation at the rein-
forcement-concrete interface and with regard to the particular crack stage. Based on
these fundamentals, a mechanical calculation model is then proposed for practical calcu-
lation of maximum crack width in RC members. The suitability of the proposed model is
demonstrated by the comparison of predicted and experimentally obtained maximum
crack widths for a database including 460 crack width measurements. It is shown that
the current state of knowledge enables a mechanically plausible calculation of the maxi-
mum crack width. Although the formulation based on the bond law is not trivial, the cal-
culation model can be prepared with appropriate simplifications in a practical way.

KEYWORDS
bond law, crack stage, crack width calculation, reinforced concrete

1 | I N T RO DU CT I O N members are most of all negatively influenced only by the cracks with
maximum crack width. Moreover, reliable limitation of maximum crack
Cracks are an essential characteristic of reinforced concrete width is of crucial importance for serviceability of reinforced concrete
(RC) construction. These cracks are usually irregularly distributed and structures with strict requirements on tightness, as outlined amongst
also have different widths along the member. As long as the crack others in References 4 and 5.
widths remain within an acceptable range these cracks do neither In structural design, the limitation of the crack width is done by
impair the serviceability or bearing capacity, nor the durability of the complying with design rules and limitation of the maximum crack
structure, see References 1-3. This means that both the number of width under the decisive load combination. The prediction of the max-
cracks and the mean value of the crack widths in a reinforced con- imum crack width of RC, however, is not trivial and has been much
crete member are not significantly important for its design. In con- debated over the past decades. Pioneering contributions on the phe-
trast, the maximum crack width in a RC member is of far greater nomena of cracking in RC are surely References6–13. Besides, a
importance. It is obvious that visual appearance and durability of RC numerous palette of experimental and theoretical investigations

This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any
medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made.
© 2021 The Authors. Ernst & Sohn Verlag für Architektur und technische Wissenschaften GmbH & Co. KG

Civil Engineering Design. 2021;3:45–61. wileyonlinelibrary.com/journal/cend 45

 2625073x, 2021, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cend.202100004 by Cochrane Netherlands, Wiley Online Library on [27/02/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
46 SCHLICKE ET AL.

looking closely into different aspects of cracking and crack width is difference between experimentally obtained mean and maximum
available. For the calculation of the maximum crack width itself a large crack spacing and crack width and its significance for maximum crack
variety of solutions had been proposed, as impressively shown in width under real structural conditions on-site, and so forth), which are
Reference 14. If one ignores for the time being the discussion on the then used to define and calibrate a basic calculation apparatus. This
significance of particular parameters for the resulting crack width, as second option, referred to as calibrated model hereinafter, is very
outlined in References 9, 15, and 19, two principle strategies for appealing as the agreement between prediction and under laid data
assessing the problem can be distinguished. The first strategy is the are efficiently controlled by the chosen form of the basic calculation
mechanically based development of a calculation apparatus on basis apparatus and the calibration parameters. However, as these experi-
of the bond stress-slip relationship at the reinforcement-concrete ments are naturally subject to a certain degree of scatter, the investi-
interface (bond law), whose results are validated against experimental gation and calibration is usually based on mean values for crack
data. This first option, referred to as mechanical model hereinafter, is spacing and crack width in the stabilized crack stage with a load level
undoubtedly not trivial as the calculation of maximum crack width on after which no new cracks occur. The difference between mean and
basis of the bond law is complex. Besides, the validation of such maximum crack width is then regarded by statistical modification for
approaches with experimental results is often subject to criticism as a an accepted probability of exceedance in order to predict a character-
certain disagreement between prediction and observation is unavoid- istic crack width for design, for example, according to the upper qua-
able due to system-immanent peculiarities of experiments and mea- ntile. For better illustration and a more detailed comparison, these
surements, as discussed in References 15, 17, 20 or 24. The second two principal strategies are contrasted in Table 1.
strategy is the detailed and systematic analysis of experiments on The mechanical model promises above all a mechanically sound
crack widths in RC in order to identify decisive correlations between solution on basis of the bond law, which should lead to results on the
test specimen properties and input parameters for crack width calcu- safe side due to conservative calculation assumptions. In addition,
lation (in particular: effect of concrete cover on crack spacing, the mechanical nature of the approach should enable straightforward

TABLE 1 . Comparison of principal strategies for calculation of (maximum) crack width

Mechanical model (strategy 1) Calibrated model (strategy 2)

Philosophy • Calculation on basis of bond stress-slip • Systematic analysis of experiments and
relationship at the reinforcement- calibration of a predefined calculation
concrete interface apparatus to it
• Safe side prediction of maximum crack • Prediction of characteristic crack width
width assuming maximum possible according to an accepted probability of
slipping length with regard to decisive exceedance
crack stage
Validation • Validation by comparison with • Validation within the calibration OR by
experiments (maximum crack width) comparison with an independent
verification database
Basic form w = sr (εsm
εcm)
σ s ds


With: sr ¼ 2 les ¼ 2dτssmf ctm
ρ ≤ 2 τsm sr ¼ βw kc c þ kϕ=ρ kfl kb fτctm ds
eff sm ρeff

Uncertainty to be addressed • Scatter of material properties addressed • Scatter of material properties addressed
by mean values (fctm, Ecm, τsm) by mean values (fctm, Ecm, τsm)
• Effective area of concrete in the cracking • Effective area of concrete in the cracking
process process
• Difference between mean and maximum
crack width
Key input • Ratio and diameter of reinforcement • Ratio and diameter of reinforcement
(ρeff, ds) (ρeff, ds)
• Concrete cover (c)
Result • Maximum crack width according to • Characteristic crack width according to
maximum slipping length with regard to upper quantile of experimental
crack stage observations
Relation to crack width on-site • Calculated crack width can be • No direct comparability
interpreted as the maximum surface
crack width which could be found
on-site
Application in standards • German and Austrian annex of EC2 • Basic document of EC2
• Model Code 1990 • Model Code 2010 and draft for Model
Code 2020
 2625073x, 2021, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cend.202100004 by Cochrane Netherlands, Wiley Online Library on [27/02/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
SCHLICKE ET AL. 47

results which originate from a transparent and verifiable calculation cracks, too. And on the other hand, the crack spacing in the stabilized
method. crack stage is not constant but changes with increasing stress level.
The calibrated model, on the other hand, suggests a certain refer- The calibration basis requires therefore an adjustment to the stress
ence to reality by being derived from experiments, whereby the pre- level in the experiment — which is usually insufficiently addressed by
diction accuracy (predominantly in regard to the difference of mean an assumed mean bond stress depending on the tensile strength.
and maximum crack width) is to be addressed by statistical modifica- Overall, different design results may be obtained depending on
tion. But this solution brings with it some complications. First of all, whether a mechanical or a calibrated model is used. Nevertheless,
the size of the test specimens in the laboratory investigation differs both models are found in different standards and pre-standards, and
considerably from real member sizes so that the scattering in the test in a broad range of applications they also provide comparable results.
specimen may not represent the variation of the crack width in real Outside of this broad application range, however, there are clear dif-
members. This means that the composition and correct interpretation ferences, especially in cases of restraint stressing and prestressed con-
of the database for calibration is very important for the derivation of crete as well as in cases with greater concrete cover.
design parameters. Secondly, the various calibration factors and the For the benefit of transparent and verifiable service life design of
accepted probability of exceedance complicate the practical applica- RC members, a mechanical model is therefore proposed. The apparatus
tion. In certain applications, a tolerable exceedance on-site is not a is based on the solutions presented in References 13, 21, and 23. It
serious problem (especially durability, see Reference 22); in other focuses on the calculation of maximum possible crack width in a mem-
cases, such an exceedance is the starting point for protracted disputes ber within a range of 0.2 to 0.4 mm which is commonly agreed as rele-
— even if the crack width on-site lies within the tolerable range (espe- vant for service life of RC. Lower crack widths are usually increasingly
cially visual appearance and tightness). Thirdly, and irrespective of the affected by an interplay of larger scatter of bond behavior for small steel
statistical considerations, the calibrated model is also subject to the strains as well as transmission of tensile stresses over the concrete in
meaningfulness of the experiment(s) and it is not necessarily immune the crack, see for example, see Reference 24. Larger crack widths are
from leading to unexpected results in practical application that cannot excluded as the bond stress-slip relationship is excessively affected by
be confirmed by practical experience. In addition to these criticisms, it internal damage on the reinforcement-concrete interface at fairly high
can be stated that the fundamental constitution of the calibrated steel strains. Within the focused range, however, the presented solution
model on basis of the mean crack spacing in the stabilized crack stage yields viable results with strict regard of the cracking stage of the mem-
is very doubtful. On the one hand, the stabilized crack stage is not ber. The great advantage of an explicit consideration of the crack stage
meaningful for a large part of practical cases with requirements on the expresses itself in a realistic representation of the member conditions
crack width, namely cracking due to restraint stressing or cracks in under restraint stressing or in case of prestressed concrete — both of
prestressed concrete both of which are usually in the condition of which are significantly influenced by the reinforcement required to limit
stand-alone single cracks. For the purpose of better illustration, Fig- the crack width. It shall further be noted that this solution has served
ure 1 shows such examples of stand-alone single cracks in RC mem- well with some crucial simplifications over the past 20 years in the for-
bers in service. In addition, typical building slabs with low mer German design code DIN 1045-125 as well as following in the Ger-
reinforcement degrees are often in the condition of stand-alone single man and Austrian Annexes to Eurocode 226,27.

F I G U R E 1 Examples of stand-alone single cracks in RC members in service.

Source: iBB TU Graz
 2625073x, 2021, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cend.202100004 by Cochrane Netherlands, Wiley Online Library on [27/02/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
48 SCHLICKE ET AL.

It is emphasized again, that this model always calculates a maxi- symmetrically reinforced tie. The tie contains adequate robustness
mum crack width. This means that the result indicates the crack width reinforcement so that yielding of reinforcement is excluded during the
to be expected under worst conditions. Observations of smaller crack formation of cracks.
widths in experiments or in real buildings should therefore not be Before cracking, the representative tie behaves comparably stiff
seen as inaccuracies of the model. In addition, it should be said that and the force-deformation relation can be described with Δl/l = F/
the outcome of this model is also not directly applicable in the frame- [EAc (1 + αEρs)]. The first crack can be assumed to occur as soon as the
work of nonlinear calculation models (usually non-linear finite element concrete stresses reach the lower limit of the tensile strength of
analyses which, due to their nature, basically reflect only average the concrete (fctk,0.05 = 0.7 fctm). With regard of the reinforcement,
component behavior, as outlined amongst others in Reference 28). crack formation begins at a global force of:
The article is structured as follows. At first, the general behavior of
RC in tension is discussed as usual with the behavior of a representative Fcr,0 ¼ 0:7 f ctm Ac ð1 þ αE ρs Þ ð1Þ
RC tie. The focus is hereby set on the context of crack formation, strain
conditions, and crack opening with regard to the mechanical sound con- with fctm = mean tensile strength, Ac = cross-sectional area of the
sideration of the bond stress-slip relationship at the reinforcement-con- member, αE = ratio of elastic moduli (αE = Es/Ec) with Es = elastic
crete interface. Building on these basic explanations, the calculation modulus of steel and Ec = elastic modulus of concrete, as well as ρs =
approach is then presented. The meaningfulness of the presented solu- reinforcement ratio (ρs = As/Ac) with As = cross-sectional area of pro-
tion is following demonstrated by means of validation against 460 experi- vided reinforcement.
mental observations from the literature and own experiments (169 in From that force level onwards, further cracks will form subse-
pure tension and 291 in flexure). Finally, the article concludes with a pro- quently by only slightly increasing the force. This so-called crack for-
posal for practical application of the presented solution. mation process causes stand-alone single cracks along the member,
Another important aspect for determining the reinforcement which are initially not necessarily interacting with each other. Figure 3
required to limit the crack width is whether the member under consider- illustrates this fundamental mechanism by crack patterns of ties. It can
ation is a thin or thick member. In this context, it is also of great impor- clearly be seen that the first cracks occur rather randomly along the
tance whether the limitation of the crack width is due to load or imposed tie and even at a later stage, new cracks are forming between first and
and restrained deformations. However, this question exceeds the possible third respectively second and third crack.
scope of this article and will be dealt with in a subsequent article of the As not only confirmed by these selected experimental observa-
authors on calculation of minimum reinforcement for crack control. tions but also by many others in the cited works of References 29, 30
or 31, 32, the distance between the first single cracks is often large so
that new cracks may well appear between them. This is particularly
2 | GENERAL BEHAVIOR OF RC MEMBERS important because it indicates undoubtedly that the occurrence of
IN TENSION new cracks in the crack formation process depends strongly on both
the scatter of tensile strength and on the scatter of bond behavior
2.1 | Force-deformation-behavior and crack along the member.
formation Overall, two types of cracks are distinguished in the crack forma-
tion stage: (a) stand-alone single cracks, and (b) cracks which develop
To begin with, Figure 2 illustrates the fundamental behavior of ten- in such close proximity to the existing single cracks that they hence-
sioned RC members by the force-deformation relation of a forth influence each other. The width of the latter cracks is generally

F I G U R E 2 Force-strain relation of a
symmetrically reinforced concrete tie
 2625073x, 2021, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cend.202100004 by Cochrane Netherlands, Wiley Online Library on [27/02/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
SCHLICKE ET AL. 49

FIGURE 3 Sequence of crack formation in tensioned bars

smaller than the width of a stand-alone single crack. Up to a certain The cracking stage of flexural members can be assessed on the basis
force level, it is therefore enough from theoretical as well as practical of the cracking moment. In analogy to Equation (2) and with regard to
point of view to regard only the widths of stand-alone single cracks. the reinforcement on the tensioned side the force limit in flexural
Only if the force is increased further, the entire member goes into the members whether the maximum crack width results from conditions
conditions of a stabilized crack stage. in the single crack stage or stabilized crack stage is:
In the sense of the model presented, stabilized crack stage means
  !
that most cracks along the member have formed, whereby the steel d1 2
Mcr,n ¼ 1:3 Mcr ¼ 1:3 f ct,fl W c 1 þ αE ρs  3 1
2 ð3Þ
strains exceed the concrete strains over the entire member length, h
and thus, stand-alone single cracks no longer exist. However, this defi-
nition of stabilized crack stage must not be confused with a stabilized with fct,fl = flexural strength of concrete and d1 = surface distance of
crack pattern as new cracks may still be formed in the stabilized crack the centroid of the reinforcement.
stage as a result of further increased force and accordingly higher
steel strains, see References21,23. Other theoretical predefinitions of
the stabilized crack stage, as for example given in Reference 18 where 2.2 | Crack formation and strain distributions
a fixed crack pattern is deemed to occur in the stabilized crack stage
over the entire member, are not pursued further hereinafter. For the purpose of illustration, Figure 4 sketches possible crack pat-
The force level above which the stabilized crack stage prevails terns and corresponding strain distributions in the viewed tie for the
can be determined on basis of the condition that tensile stresses along following different force levels:
the member can nowhere be greater than the concrete tensile
strength. With regard to the scatter of concrete tensile strength in a 1. Directly after initial cracking when few cracks occurred but strains
member the force level above which the stabilized crack stage prevails of concrete and steel are equal over large areas of the tie.
can be given in accordance with Reference 33 as follows: 2. After slight increase of the force so that a last crack is still sub-
jected to the conditions of the single crack stage.
Fcr,n ¼ 1:3 F cr ¼ 1:3 f ctm Ac ð1 þ αE ρs Þ: ð2Þ 3. In an early stadium of the stabilized crack stage (steel strains
exceed the concrete strains along the entire length of the tie but
Beyond the force level of Fcr,n, the member stiffness is dominated not all possible cracks have formed).
by the opening of existing cracks and can be compared to the stiffness 4. In an advanced stadium of the stabilized crack stage (further
of plain steel. Equation (2) is given with regard of the contribution cracks have formed with increased force, even though steel
of steel in the uncracked concrete, which is often neglected. In case strains already exceeded the concrete strains along the entire
of experimental analysis, however, this contribution shall be regarded. length).
2625073x, 2021, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cend.202100004 by Cochrane Netherlands, Wiley Online Library on [27/02/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
SCHLICKE ET AL.

F I G U R E 4 Crack formation
and strain distributions for
different force levels
50
 2625073x, 2021, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cend.202100004 by Cochrane Netherlands, Wiley Online Library on [27/02/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
SCHLICKE ET AL. 51

The illustrations in Figure 4 are created on basis of the following The length lslip in Figure 4 describes the specific length belonging
generally accepted premises: to a particular crack over which the concrete slips away on the steel
to both sides of the viewed crack. This length is often equated with
• The first cracks occur rather randomly along the bar as a result of the the crack spacing (sr). This comparison is somewhat inadequate, as the
scatter of the tensile strength along the bar (these cracks are not nec- length lslip would only correspond to the physical crack spacing in a
essarily caused by bond action starting from existing cracks or the stabilized crack stage with uniform crack pattern. In fact, lslip complies
bond behavior is not important for the occurrence of these cracks). during crack formation with twice the transfer length (le), which
• Equilibrium exists in each cross section along the tie describes the length needed to transfer the force from the concrete in
(Fs(x) + Fc(x) = F) and this is independent of the present crack stage. the uncracked part into the steel section at the crack by bond. In the
• Steel strains are maximum directly in the crack while concrete stabilized crack stage, however, lslip cannot be determined directly. In
strains are zero at this location. detail, the final distribution of cracks along the member is subject to
• Bond action between steel and concrete decreases the steel strains the location of first cracks (formed under the influence of the varia-
with increasing distance from the crack whereas concrete tion of tensile strength along the member) and the possibility that
strains are increased: maximum concrete strains occur either after additional cracks may occur in between by building up the cracking
full transmission of the bond force (single crack condition) or at force of the effective concrete area (for further increase of load and if
half-length between two cracks in stabilized condition, whereas the distance is large enough). More appropriate designations for lslip
steel strains are minimum at this location. Note that the physical are therefore slipping length or integration length which can be gener-
location of minimum steel strain in stabilized crack stage is not nec- ally used irrespectively of the crack stage.
essarily exactly at half-length between two cracks but depends on The local variations of steel and concrete strains in the slipping
the sequence of crack formation. As this effect is not decisive for length are directly related to the distribution of bond stresses at the
maximum crack width it is neglected hereinafter. reinforcement-concrete interface. Tests on the bond stresses
• Maximum steel strains are always subject to the global force for between reinforcement and concrete indicated that the distribution
equilibrium with εs = F/(Es As). of bond stresses (τs(x)) along a deformed bar in concrete can be
• The variation of steel strains is always subject to the local equilib- described for static loading in the SLS by a function of the slip, as
rium (ΔFs + ΔFc = 0 ! Δεs). given in References 23 or 32. It reads in the general form:
• Maximum concrete strains depend at initial cracking on the global
force but between two cracks the concrete strains cannot exceed τs ðxÞ ¼ C sα ðxÞ ð4Þ
the ultimate strain according to the mean tensile strength (scatter
of fct is neglected in the transfer length; εc,max = F/(Ec Ac)/ with C = factor depending on steel surface and compressive concrete
(1 + αE ρs) ≤ fctm/Ec). strength, s(x) = slip along the deformed bar and α = factor depending
• New cracks occur as long as concrete stresses between two exis- on steel surface and bond quality; for conventional reinforcement
ting cracks can reach the tensile strength again, the following C = 0.31 fc,cube and α = 0.3 with fc,cube = cube compressive strength
hypothesis is formulated: of the concrete.
 the location of the first cracks is predominantly affected by The bond law is usually derived from standardized pull-out tests
the scatter of tensile strength along the member length which must be seen as a simplification in comparison to real condi-
until incompatibility between concrete and steel strains tions within the slipping length. Common points of discussion are
exists, and hereby the stress state of concrete during the pull-out test as well as
 areas with incompatibility between concrete and steel strains the integrated measurement of bond force in the experiment without
may still produce a new crack if the spacing between the exis- detailed insights on the real bond stress distribution in the experi-
ting cracks is large enough to build up the cracking force of the ment. In view of the shown SLS behavior of RC members with crack
effective concrete area in between. widths within a range of 0.2 to 0.4 mm, however, both discussion
points are rather academic aspects. If the bond law is accepted one
The sketched crack patterns and associated strain distributions can illustrate the context of the local distributions of slip and bond
show the behavior of a RC tie in which the load is successively stresses together with the resulting variations of concrete and steel
increased. The cracks initially form randomly along the tie, see strains for both crack stages, as done in Figure 5. The variations of
Figure 4A,B, until a so-called stabilized crack stage occurs (Figure 4C). steel and concrete strains depend in detail on the local variation
From this load level onwards, further cracks can be formed between of the bond force by:
the existing cracks until a very narrow crack spacing occurs
(Figure 4D). The maximum crack width always occurs where there is πds τs ðxÞ dx ¼
dεs Es As ¼ dεc Ec Ac : ð5Þ
still a last single crack or at the location with the largest crack spacing
in the stabilized crack pattern. Due to the continued crack formation Overall, the mathematical description of the interplay between
in the stabilized crack stage, the location with the largest crack width slip, bond stress, and local strain variations is not trivial. Using a mean
can change in the course of the increase in force. bond stress in the slipping length, as indicated in Figure 5, offers a
 2625073x, 2021, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cend.202100004 by Cochrane Netherlands, Wiley Online Library on [27/02/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
52 SCHLICKE ET AL.

F I G U R E 5 Distribution of steel and concrete strain, bond stresses, mean bond stress in the vicinity of the maximum crack as well as slip over
the slipping length

significant facilitation. In theory, the mean bond stress can be calcu-

lated with:

ð le
C
τsm ¼ sα ðxÞ dx: ð6Þ
le 0

The solution of Equation (6) is challenging since the explicit form of

the slip distribution in the transfer length (slip function) is not known
beforehand. As shown in Reference 21 the mean bond stress can
therefore only be calculated analytically for the single crack state. A
general solution for the mean bond stress in both crack stages, how-
ever, can only be found by approximation of the slip function. A well-
founded approximation for the slip function according to the bond
law was proposed in Reference 34 by:
F I G U R E 6 Mean bond stress according to the crack width for
 
C
wα
single and stabilized crack stage, according to Equation (7)
w x λ
sðxÞ ¼ ! τsm ¼ ð7Þ
2 le 1þλα 2
given concrete strength class and utilization in the stabilized crack
with λ = initially unknown exponent for the course of slip in the trans- stage. For the same crack width, the mean bond stress in the stabi-
fer length; according to Reference 34 for single crack stage: λ ¼ 1
α
2
lized crack stage is greater than in the single crack stage. Moreover,
1þα
σsr 1:5
and for stabilized crack stage: λ ¼ 1 þ1
α σ s2 with σ s2 = steel the mean bond stress in the stabilized crack stage depends also on the
stress in the viewed load condition and σ sr = steel stress in the load level, expressed with the ratio of steel stresses at crack formation
cracked section under the load condition of Ncr and Mcr, respectively. and present load level (σ cr/σ s). The reason is the increase of strain dif-
Overall, this solution yields that the mean bond stress is not a ference between steel and concrete in the middle between two cracks
constant but depends on the integrated slip over the transfer length, and so the slip does.
which corresponds to the crack width w, expressed with w/2. In par- Although the solution presented indicates the necessity of a case-
ticular, this approximation constitutes that the slip increases faster specific consideration of the mean bond stress with regard to crack
and reaches higher absolute values with increasing utilization in the width criteria and crack stage, a simplification of τsm = 1.8 fctm is
stabilized crack stage. In accordance with the bond law also the bond established in practical engineering, independent of the crack width
stresses increase. Figure 6 pictures the analysis of Equation (7) for a criteria in the SLS. Strictly speaking, this simplification of
 2625073x, 2021, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cend.202100004 by Cochrane Netherlands, Wiley Online Library on [27/02/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
SCHLICKE ET AL. 53

τsm = 1.8 fctm only represents the combination of a specific crack the calculation of maximum crack width, namely the ones according
width in a specific crack stage, whereas all other combinations are vio- to single crack stage and the ones according to stabilized crack stage.
lating the equilibrium condition for the underlying bond law. This
aspect was also taken up in Reference 18, whereby here, under the
premise that the mean bond stress can no longer change at a given 3 | C A L C U L A T I O N OF M A X I M U M
crack spacing, the maintenance of equilibrium is theoretically CRACK W IDTH
explained via a local decrease in the local bond stress in the vicinity of
the crack. In any case, a broader application of the simplification 3.1 | General procedure and assumptions
τsm = 1.8 fctm in the practical application of the mechanical model by
References 26, 27 did not lead to significant mispredictions and has The equilibrium of forces and the global compatibility of deformations
become state of the art. applies over crack formation until stabilized crack stage and it is as
According to Figure 4, the mean steel strain differs significantly in follows:
both crack stages. In the single crack stage, the mean steel strain is
smaller than 50% of the steel strain in the crack. In the stabilized crack F ¼ Fs ðxÞ þ F c ðxÞ ð9Þ
stage, however, the mean steel strain increases simultaneously with
the strain at the crack. The shape factor kt for calculating the mean and
strain is usually assumed for short-term loading with kt = 0.6,
ðl ðl X
n
whereby a detailed determination is achieved with the approximated
Δl ¼ εs ðxÞdx ¼ εc ðxÞdx þ wi ð10Þ
slip distribution in Equation (7): 0 0 i¼1

1þβα with Δl = total length change of the member, l = length of the mem-
kt ¼ : ð8Þ
2þβα
ber, εs(x) = distribution of steel strains along the member, εc(x) = dis-
tribution of concrete strains along the member, wi = crack width of
With regard to bond-creep under sustained loading, however, a the individual cracks, F = external force, Fs(x) = steel force at the posi-
redistribution of stresses from concrete to steel occurs in the slipping tion x along the member, Fc(x) = concrete force at the position x along
length. This redistribution can be regarded with a 33% reduction of the member.
the shape factor (kt,long
term = 0.4). From mechanical perspective, the width of a crack represents the
It should be noted that internal cracking at the reinforcement- integrated strain difference between concrete and reinforcement within
concrete interface of deformed bars, as firstly reported by Reference the slipping length of the viewed crack. This fundamental relation is inde-
11, reduces the bond stresses close to the crack in comparison to the pendent of whether the stressing results from loading or imposed defor-
ones according to the theoretical bond law, especially in the case of mations. Finally, the general equation for calculating the maximum crack
wider cracks. This behavior is simplifying considered in some calcula- width is derived from translating the global compatibility in Equation (10)
tion models with a bond-free length or, as mentioned earlier with ref- to the strain conditions in the slipping length of the decisive crack:
erence to Reference 18, considered via local decrease in the local
ð lslip,max
bond stress at the crack. The existence of such a bond-free length is
wmax ¼ εs ðxÞ
εc ðxÞ dx ¼ lslip,max ðεsm
εcm Þ ð11Þ
generally known and its significance for very large crack widths 0

is agreed. In the case of the desired crack width criteria for SLS
designs, however, the significance of a possible bond-free length has with εsm = mean steel strain and εcm = mean concrete strain over the
not yet been clearly proven. Even if the local support of the concrete maximum slipping length lslip,max.
“corbel” at the first rib might be limited, as for example shown in Ref- A common point of discussion is the question of whether the
erence 35, it is not reasonable from the author´s point of view to con- determined crack width is the width at the reinforcing bar or at
sider explicitly a possible bond-free length in the apparatus for the the surface of the member, as a significant difference between them is
viewed SLS behavior of RC members with crack widths within a shown in various experiments, see References 37 or 38, but also indi-
branch of 0.2 to 0.4 mm. Therefore, the illustrated strain distributions cated by on-site observations. In principle, the result of a mechanical
in Figure 4 as well as the given parameters for bond action in the slip- model with adequate consideration of the bond stress-slip relationship
ping length (mean bond stress and shape factors for mean strains) can be interpreted as a global crack width (wglobal) which — in theory —
neglect this behavior consciously. neither belongs to the real crack with at the level of reinforcement nor
Overall, Figure 4 indicates that the maximum crack width in RC to the real one at the surface of the member. As schematically illus-
members depends significantly on the slipping length according to the trated in Figure 7, the real crack width in RC is additionally influenced
crack stage. Especially cases in the transition between single crack by the partial closure of the crack from surface to reinforcement due to
stage and stabilized crack stage are wrongly assessed if the slipping internal cracking along the transfer length and the potential increase at
length is generally assumed according to the stabilized crack stage. It the surface due to shear lag deformations originating from an eccentric
is therefore appropriate to distinguish two basic strain conditions for transfer of the bond force into the concrete cover. Both effects are
 2625073x, 2021, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cend.202100004 by Cochrane Netherlands, Wiley Online Library on [27/02/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
54 SCHLICKE ET AL.

FIGURE 7 Conceptual distinction of crack opening due to global compatibility (wglobal) and additional effects due to internal cracking and
shear lag

usually neglected in the conception of a mechanical model. The partial according to Reference 17. For the calculation of maximum crack width
closure towards the reinforcement can be neglected on the safe side in the range of 0.2 to 0.4 mm it was found by means of own volumetric
anyway, whereas the importance of the shear lag is assumed as insignif- FEM simulations that the increase of surface crack width compared to
icant for typical steel strains under SLS conditions. Unfortunately, there wglobal is smaller than 10%. Details will only be given in a future publica-
is no solid evidence for such assumption in the literature, except a tion of the authors. For the aimed calculation of maximum crack width
statement in Reference 16 which attaches great importance to the with the mechanical model, however, the shear lag effect is not further
shear lag effect, whereas this effect is expected to be negligible pursued from this point onwards.
 2625073x, 2021, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cend.202100004 by Cochrane Netherlands, Wiley Online Library on [27/02/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
SCHLICKE ET AL. 55

In addition, a theoretical difference between surface crack width εcm ¼ kt εc,max ð15Þ
of ties and flexural members is conceived if an additional crack open-
ing over the cover is derived from the linear strain profile over the
height, for example, as proposed in Reference 33 with the extrapola- In total, the maximum crack width in case of single crack stage is:
tion of (h
x)/(d
x). However, such extrapolation is only meaningful
if the basis for extrapolation refers to the crack width at the level of σ s ds σs σ 2s ds ð1
kt Þ
wk ¼ 2 le ðεsm
εcm Þ ¼ 2 ð1
kt Þ ¼ :
reinforcement — which is not the case for the crack width due to 4 τsm ð1 þ αE ρs Þ Es 2 τsm Es ð1 þ αE ρs Þ
ð16Þ
global compatibility. In addition, the assumption of a linear strain pro-
file is rather incorrect for the concrete in the cracked cross
section because of the local effect due to load transfer from the rein-
forcing bar to the concrete by bond. The calculation of maximum 3.2.2 | Stabilized crack stage
crack width with adequate regard of the bond stress-slip relation can
therefore directly be interpreted as surface crack width for ties and An important characteristic of the stabilized crack stage is the incom-
flexural members without any further modifications based on the patibility of concrete strains and steel strains along the entire member
strain plane. length. Thus, the force transfer between reinforcement and concrete
Another much-debated influence parameter of maximum crack within the slipping length is generally unknown. In fact, the crack
width is the effect of unrestrained shrinkage in the slipping length width cannot be determined directly only on mechanical consider-
which will be discussed in addition to the basic value of crack opening ations. A solution on the safe side is obtained when the maximum
due to global compatibility (wglobal) below. possible concrete strain between two cracks is considered equal to
the cracking strain of the concrete (εcr = fctm/Ec) and the maximum
concrete force in this stage is considered with regard to the so-called
3.2 | Maximum crack width due to global effective area Ac,eff. Under these assumptions, the maximum transfer
compatibility length in the stabilized crack stage is as follows:

3.2.1 | Single crack stage ds f ctm

Ac, eff f ctm ¼ τsm us le ! le ¼ ð17Þ
4 τsm ρeff

In the single crack stage, the strain variation occurs over the transfer
length of the concrete force. At the end of the transfer length, the where Ac,eff = effective concrete area and ρeff = reinforcement ratio
steel strain is again equal to the strain of the concrete. Important char- of the effective concrete area (ρeff = As/Ac,eff).
acteristics of the single crack stage are the existence of zones Mean steel and concrete strains in the maximum slipping length
between the single cracks, in which concrete and steel strains are in the stabilized crack stage are:
equal, and the involvement of the whole cross-section in the crack
formation process. Due to these conditions, the crack width can be σs Ac,eff f ctm
εsm ¼
kt ð18Þ
Es As Es
determined directly based on mechanical considerations. The transfer
length at each side of the crack is:
and
Fs σ s ds
¼ τsm us le ! le ¼ ð13Þ
1 þ αE ρs 4 τsm ð1 þ αE ρs Þ f ctm
εcm ¼ kt : ð19Þ
Ec

with τsm = mean bond stress in the transfer length depending on the
crack width and bond condition, us = circumferential length of Finally, the maximum crack width in the stabilized crack stage results
the provided reinforcement, ds = diameter of reinforcement and σ s= with the assumption that the slipping length is twice the transfer
steel stress in the crack (σ s = Fs/As). length in the stabilized crack stage. It reads:
With regard to the shape of the strain variation (shape factor kt)
 
the mean steel strain in the transfer lengths to both sides of the crack ds f ctm σ s f ctm
wk ¼ 2 le ðεsm
εcm Þ ¼
kt ð1 þ αE ρeff Þ : ð20Þ
2 τsm ρeff Es Es ρeff
(slipping length) is:

 
σs σs The calculated crack width according to Equation (20) depends
εsm ¼
kt
εc,max : ð14Þ
Es Es strongly on the conceptual model of the effective concrete area. This
conceptual model of the effective concrete area is often reinterpreted
In analogy to the mean steel strain, the mean concrete strain in for the purposes of the respective model. In the herewith presented
the slipping length amounts: model, Ac,eff represents the spreading of the bond force from the
 2625073x, 2021, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cend.202100004 by Cochrane Netherlands, Wiley Online Library on [27/02/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
56 SCHLICKE ET AL.

FIGURE 8 Determination of the steel force during crack formation

F I G U R E 9 Strain distributions due to

subsequent shrinkage after load-induced
cracking

reinforcement location into the member, whereby new macroscopic Ac,eff if the member is clearly in the stabilized crack stage. Moreover,
cracks are formed as soon as the bond force exceeds the cracking stirrup spacing is not regarded for the determination of the decisve
force of the effective concrete area (Fcr,eff = Ac,eff  fctm). In relevant slipping length in accordance with References 17, 20 and 29.
member conditions, that is, bending or thick concrete members in ten- Another aspect of the crack pattern in the stabilized crack stage is
sion with common reinforcement location near the surface, these the potential increase of crack spacing and crack widths in members
cracks form always beginning from the outer fiber of the cross with overly large concrete cover compared with the member thick-
section and must therefore strongly be distinguished from the repre- ness. This effect was demonstrated with several experiments, presented
sentation that they predominantly originate from internal cracks (also in References 17, 19 or 30, in which the cover of the specimens was
called Goto-cracks). In this model understanding, the determination of increased while maintaining the cross-sectional area as constant.
Ac,eff according to MC201033 and EC2-136 usually provides satisfac- Although the mean crack spacing increases noticeably in ties with cross-
tory predictions for typical configurations of RC members. For the sectional configurations of cnom  h/4, the prediction of maximum crack
determination of Fcr,eff in practical design, however, it shall be noted width in such members using Equation (20) is not significantly shifted
that Fcr,eff is always limited to the steel force resulting from a cross towards the unsafe side. On the contrary, the calculation of maximum
sectional analysis for the cracking force, as illustrated in Figure 8. crack width for the flexural specimens with overly large covers pres-
It must further be kept in mind that the conceptual model of Ac,eff ented in Reference 17 would lead to a significant underestimation if the
can lead to a significantly smaller slipping length in comparison to the definition of Ac,eff for flexural members (hc,eff = (h
x)/3) is applied.
conditions of a single crack - and this is only on the safe side for mem- However, these specimens with abnormal covers do not comply with
bers being certainly in the stabilized crack stage. The deceptive the typical conditions of flexural members, namely that the surface dis-
assumption that the stabilized crack stage must already be present as tance from the application point of the tensile force in the cross-section
soon as the steel force is greater than the cracking force of the effec- is greater in the uncracked state than after cracking. In these cases with
tive concrete area does therefore not necessarily lead to results on d1 > h/6, the stress distribution in the effective concrete area is rather
the safe side. Only the reverse conclusion is permissible, namely that comparable with the one of ties and good agreement for the calculated
slipping length and tension stiffening can be calculated on the basis of crack width can only be found when using the respective definition of
 2625073x, 2021, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cend.202100004 by Cochrane Netherlands, Wiley Online Library on [27/02/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
SCHLICKE ET AL. 57

hc,eff = 2.5 d1. Overall, the general consideration of a stress-independent due to subsequent shrinkage after load-induced cracking. Overall,
cover term in the prediction of the slipping length in the stabilized crack unrestrained shrinkage in the slipping length does not cause additional
stage, as proposed in References 33 or 36, appears therefore to be inad- crack opening in the model condition of maximum slipping length as
equate. Current regulations in References 26, 27 are therefore nullifying any shrinkage will lead to shorter slipping length for the same crack
the influence of the cover term in the crack width calculation. width or cause a new crack at smaller distance, respectively.

3.3 | Effect of free shrinkage in the slipping length 4 | VALIDATION

The effect of additional opening of load-induced cracks due to subse- 4.1 | Data basis and analysis
quent concrete shrinkage is thoroughly discussed in References 20,
39, 40 and currently addressed in References 33 by inclusion of The proposed solution was validated by comparison of computational
shrinkage in the strain difference in the stabilized crack stage under results with experimental observations from the literature, in particu-
sustained loading. In fact, this additional crack opening must also be lar from References 6, 7, 10, 17, 19, 29-32, 44-49. In contrast to the
assessed with regard to the load-induced concrete stresses in the par- usual approach of assessing selected studies, the herewith aimed vali-
ticular cracking stage. This is because as soon as the concrete stresses dation was made upon a rather broad database containing results of
between two cracks reach the tensile strength already due to load, different studies with different conditions. In detail, 460 experimental
any additional shrinkage will rather cause the formation of a new observations from the literature and own experiments (altogether 169
crack than any additional opening of an existing crack. in pure tension and 291 in flexure) were assessed. In principal, the val-
The effect of subsequent shrinkage after load-induced cracking idation is a comparison of predicted and experimentally obtained max-
can be regarded by a combination of the load-induced stressing imum crack widths. The prediction of maximum crack width was
together with a self-equilibrated stress state between reinforcement performed with the presented approach with both a slip-depending
and concrete due to shrinkage of concrete. The equilibrium of forces determination of mean bond stress τsm and shape factor for stress dis-
and the compatibility of deformations are affected as follows: tribution kt according to Equation (7) and Equation (8) as well as a sim-
plified approach with τsm = 1.8 fctm and kt = 0.6 or 0.4, respectively.
Fc ΔF c Fs ΔF s Apart from these differences, the applied equations were fully
εcs þ ð1 þ φÞ þ ð1 þ ρφÞ ¼ þ ð21Þ
Ec Ac Ec Ac Es As Es As
identical.
Regarding the experimentally obtained crack widths, maximum
and crack widths were used whenever reported. Unfortunately, some of
the result reports of experimental observations did not reveal maxi-
ΔFc þ ΔF s ¼ 0 ð22Þ mum crack widths but only mean values. The conversion from mean
to maximum crack widths was therefore performed computationally.
with εcs = shrinkage strain, φ = creep factor, ρ = relaxation factor, If the pure material properties are solely considered via mean values
ΔFc = change of concrete force and ΔFs = change of steel force in (neglecting scatter of material properties) according to the basic
the uncracked zone. assumption for calculation of crack width, any crack in the condition
By solving Equation (21) into Equation (22) and Equation (9) the of stand-alone single crack shows the maximum crack width, and thus
increase of the concrete force can be determined with: wmean = wmax in the single crack stage, while the particular crack wid-
ths in the stabilized crack stage are influenced by the — even for
αE ρs φ
εcs Es As þ F1þα E ρs
extremely low scatter of material properties unavoidable - variation of
ΔFc ¼
: ð23Þ
1 þ αE ρs ð1 þ ρφÞ crack spacings between le and 2 le along the member, and thus wmean <
wmax in the stabilized crack stage. Since the crack pattern is not fixed
Along with the changed distributions of concrete and steel strains due in the condition of stabilized crack stage, as illustrated in Figure 4C,D,
to subsequent shrinkage, the increase of the concrete force causes an but changes even further with further load increase, the difference
increase of the transfer length. Figure 9 pictures the conditions. between mean and maximum values must be defined with regard to
Although the analysis of Equation (23) indicates a rather small the utilization of the member.23 The ratio of wmean/wmax is hereby not
increase of concrete stresses due to subsequent shrinkage after load- changing equally to the ratio of smean/sr,max. The reason is that any
induced cracking, it can be concluded that additional crack opening is new crack in the stabilized crack stage decreases smean by division of
only expected in the case of single crack stage in which the concrete an existing crack spacing into two halves. At the same time, the mean
stresses due to loading are significantly smaller than the tensile strain difference between steel and concrete increases which miti-
strength. In other cases, such as single crack stage with σ c ≈ fctm or gates the resulting decrease of wmean. The ratio of wmean/wmax must
stabilized crack stage, the maximum crack width according to Equa- therefore decrease slower than the ratio of smean/sr,max. The theoreti-
tion (16) and Equation (20), respectively, is not expected to increase cal difference between mean and maximum crack width following to
 2625073x, 2021, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cend.202100004 by Cochrane Netherlands, Wiley Online Library on [27/02/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
58 SCHLICKE ET AL.

TABLE 2 Approach for conversion of experimentally obtained crack widths (index “test”) for comparison with theoretical maximum crack
widths

Single crack stage Stabilized crack stage

Pure tension or flexure wtest ffi wmax wtest,max ffi wmax if provided, otherwise
" !#
wmax ¼ wtest,mean 1 þ 0:3 1
k
0:33 ð3r
r2 Þ

1
r 1 k
0:67
1

FIGURE 10 Comparison between calculated and experimentally observed maximum crack width

the solution of Reference 34 is given in References 14, 23 or 50. It and r = ratio of mean and maximum slipping length for the viewed
reads generally: load condition (r = smean/sr,max).
The ratio of mean and maximum slipping length for the viewed
 
wmean k1
0:33 3r
r 2 load condition can also be determined according to the solution in
¼r ð24Þ
wmax k1
0:67 Reference 34 as given in References 14, 23 or 50:

vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi


with k1 = factor for utilization of the member (k1 = σ s2/σ sr) with σ s2 = u
u   1þ 2:5
smean t 1 ρ s ½% 
steel stress in the viewed load condition and σ sr = steel stress in the ¼ 1
0:5 1
ð25Þ
sr,max k1
cracked section under the load condition of Ncr and Mcr respectively;
 2625073x, 2021, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cend.202100004 by Cochrane Netherlands, Wiley Online Library on [27/02/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
SCHLICKE ET AL. 59

With regard to the use of Equation (24) in the conversion of the mea- flexure. Further statistical parameters for the ratio of predicted versus
sured values, it must be kept in mind that the theoretical ratio of experimental results, such as mean value, coefficient of variation or
wmean/wmax is only applicable for the condition of very long members. maximum and minimum values of the above quotient, as usually
The experimentally determined mean and maximum values of the claimed by the advocates of calibrated models, are not given here.
crack width, that is, the mean and maximum values from individual The reason is that such detailed statistical analysis cannot be per-
results of a test, therefore must not necessarily correspond to the the- formed for the calculation of maximum crack width as this calculation
oretical mean and maximum values. Especially in case of experiments result marks the worst case which must not necessarily be found in
with specimens of limited length (and therefore a decreasing signifi- experiments. In turn, it should be pointed out that such detailed statis-
cance of the scatter of tensile strength and bond properties along the tical analyses always become worthless, even in the case if mean
member), it is particularly typical to obtain higher mean values than values predicted from calibrated models are regarded, if the database
the theoretical ones. The stubborn application of Equation (24) may is not adjusted for the respective load level in the experiments.
therefore lead to false assessments. For an appropriate assessment, Furthermore, the results also indicate that the common assump-
the experimental results were interpreted according to Table 2. tion of a mean bond stress independent of the crack width (slip-inde-
When performing such comparisons of predictions and test pendent τsm) and thus a fixed crack spacing in the stabilized crack
results it must further be noted that the reported maximum crack stage, is a very gross simplification which only inaccurately reflects
width in the test does not necessarily have to be similar to the theo- the actual member behavior. In addition to calculation inaccuracies in
retically calculated maximum crack width. From such a comparison, mechanical models, the simplification of slip-independent τsm will
one can only conclude that the calculation approach is on the safe above all worsen the calibration and prediction quality of calibrated
side as long as the calculated maximum crack width is similar or models. In detail, the simplification represents an incorrect basic
greater than the obtained maximum crack width in the experiment. assumption that cannot be transferred after calibration to the bound-
Besides, a certain disagreement is caused by individuality of experi- ary conditions of a practical case without making a new error — even
mental setups, measuring techniques and result interpretation. after good fitting of the calibrated model to the database. The reverse
Further information on these artifacts in the course of crack argumentation, that a slip-independent τsm would be necessary to
width measurements are outlined in References 15, 17, 20 or 24. The maintain equilibrium in the stabilized crack stage, as outlined in Refer-
results of the validation are given in Figure 10. ence 18, is only the consequence of an incorrect basic assumption
about a fixed crack pattern deemed to occur as soon as the member is
in the stabilized crack stage, which can already be refuted with the
4.2 | Discussion illustrations in Figure 4C,D.
Nevertheless, the well-established simplification of τsm = 1.8 fctm
In the first part of the article, it was shown that the calculation of the and kt = 0.6 or 0.4, respectively, is still expected to provide a reason-
maximum crack width requires an appropriate consideration of able basis for design of RC members using a mechanical model as long
the crack stage with the related slipping length. A general calculation as the limitation of crack width is no strict requirement. In cases with
of the maximum crack width on basis of a crack spacing deemed to special requirements, that is, for tightness and visual appearance, the
occur in the stabilized crack pattern, as proposed in References 32 or more conservative calculation of τsm and kt according to Equation (7)
35, leads therefore to a clear false assessment for all RC members and Equation (8) should be performed. It shall further be noted that
which are in the relevant SLS condition not clearly in the stabilized the often-debated effect of excessive crack opening at the surface in
crack stage with the respective stress level. Especially members with case of large concrete covers is well addressed within the validation
strict requirements on crack width limitation or prestressed RC mem- by the database and a reasonable definition of Ac,eff.
bers, which are usually in the single crack stage, are therefore put at
the risk of underestimated maximum crack width.
Moreover, the comparison of prediction results for the database, 5 | PROPOSAL FOR PRACTICAL DESIGN
shown in Figure 10, underlines the importance of an appropriate con-
sideration of the bond law for crack width calculation. Aiming predic- As shown before, crack width calculation of RC members is not trivial
tions on the safe side (wmax,calc ≥ wmax,test), a distinct determination of and subject to many influencing factors. For the design, however, a
τsm and kt is to be given preference, for example, on basis of Equa- practicable and efficient solution is required. The least requirement for
tion (7) and Equation (8). In particular, the application of Equation (7) this solution is a suitable consideration of the member's cracking stage.
and Equation (8) for τsm and kt indicates a satisfying safe side predic- It is important to distinguish between single and stabilized crack stages
tion of the experimental observations in the database with 93% pre- as the two input values, that is, the strain difference (between the con-
diction on the safe side for cases in pure tension and 74% in flexure. crete and reinforcement) and the length of the zone with slippage
Contrarily, the application of τsm = 1.8 fctm and kt = 0.6 or 0.4, respec- between the concrete and reinforcement (slipping length), develop in
tively, leads to a significant computational underestimation of the opposite directions. In the single crack stage, the strain difference is
maximum crack width for the present database with only 40% predic- smaller than in the stabilized stage while the slipping length in the single
tion on the safe side for cases in pure tension and even only 16% in crack stage becomes larger than in the stabilized crack stage.
 2625073x, 2021, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cend.202100004 by Cochrane Netherlands, Wiley Online Library on [27/02/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
60 SCHLICKE ET AL.

It is further important to provide a solution with appropriate reinforcement, as proposed in References 51, are not permissi-
impact of the input variables on the design result. An excessively ble. Although it is of little importance for practical application,
strong weighting of the influence of the concrete cover, as for exam- but for scientific work the application requirement of h/6 > d1
ple suggested in References 33 or 36, may lead to unwanted results. should be fulfilled before using hc,eff = (h
x)/3 for flexural
Even the partial reduction of this weighting if the crack width limita- members, as illustrated in Figure 8.
tion is only due to durability, as proposed in Reference 17, must be
viewed skeptically. Structural engineers might be misdirected to use OR CID
smaller cover to reduce crack width — and this would be counterpro- Dirk Schlicke https://orcid.org/0000-0003-3543-7459
ductive for durability since sufficient cover of low permeability is the
most important parameter for durability. RE FE RE NCE S
Moreover, additional crack opening due to imposed deformations, 1. Leonhardt F. Zur Behandlung von Rissen in Beton in den deutschen
Vorschriften. Beton- und Stahlbetonbau 7/1985. Berlin: Wilhelm
such as shrinkage, is to be assessed with a sense of proportion. In case
Ernst & Sohn Verlag, 1985;p. 179–184, and 8/1985, pp. 209-215
of externally unrestrained conditions, relevant crack openings are only (somewhat similar to: Cracks and Crack Control in Concrete Struc-
to be expected if they are caused by subsequent shrinkage and if the tures, PCI Journal/July-August 1988: 124-145).
member is still in the single crack stage with σ c fctm. For the model 2. Schießl, P. Zum Einfluß der Rißbreite und der Betondeckung auf die
Korrosion von Stahl in Beton. Deutscher Ausschuss für Stahlbeton,
condition of maximum slipping length, however, additional crack
Heft 255. Wilhelm Ernst und Sohn; 1976.
opening due to unrestrained shrinkage in the slipping length must be 3. Beeby AW. Corrosion of reinforcing steel in concrete and its relation
neglected as any shrinkage will lead to shorter slipping length for the to cracking. Struct Eng. 1978;1978:77–81.
same crack width or cause a new crack at smaller distance, respec- 4. EN1992-3:2006, EC2 – Design of Concrete Structures - Part 3 Liquid
tively. In case of externally restrained imposed deformations without retaining and containment structures.
5. DAfStb-Richtlinie. Wasserundurchlässige Bauwerke aus Beton.
combination of loading, an appropriate determination of the crack
Deutscher Ausschuss für Stahlbeton, 2017.
width and an efficient crack control is only provided by a deforma- 6. Rüsch, H., Rehm G. Versuche mit Betonformstählen. Deutscher Aus-
tion-based design, as outlined for example in References 41, 42. In schuss für Stahlbeton, Heft 140, 1963; 182.
case of externally restrained imposed deformations in combination 7. Broms BB. Crack width and crack spacing in reinforced concrete
members. ACI J Proc. 1965;62(10):1237–1256.
with loading, the load-induced crack opening is to be assessed with
8. Broms BB, Lutz LA. Effects of arrangement of reinforcement on crack
regard to the absorption of imposed deformations, as further width and spacing of reinforced concrete members. ACI J Proc. 1965;
explained in References 43. 62(11):1395–1410.
For the practicable case of RC members under externally unre- 9. Ferry-Borges, J. Cracking and deformability of reinforced concrete
beams. Association International des Ponts et Charpentes, 1966;26.
strained conditions and cracking due to loading, the equations for cal-
10. Gergely, P., Lutz, L.A. Maximum crack width in reinforced flexural
culating the crack width can be combined in a single apparatus to members. Causes, mechanism and control of cracking in concrete.
facilitate the application, as implemented in References 26, 27. Alto- SP20, American Concrete Institute; 1968; 87-117.
gether it reads: 11. Goto Y. Cracks formed in concrete around deformed tension bars.
ACI J Proc. 1971;68(4):244–251.
12. Beeby AW. The prediction of crack widths in hardened concrete.
wk ¼ lslip,max ðεsm
εcm Þ ð26Þ
Struct Eng. 1979;57A(1):9–17.
13. König, G. and Fehling, E. Zur Rissbreitenbeschränkung im
σ s ds
with lslip,max ¼ 2 τdsm ρeff ≤ 2 τsm
s f ct
and εsm
εcm ¼ Eσss
Stahl-betonbau. Betonund Stahlbetonbau 83, H. 6: 161–167 and H.
σs 7: 199–204; 1988.
kt Es ρ ð1 þ αE ρeff Þ ≥ Es ð1
kt Þ.
f ct
eff
14. Borosnyo  i A, Balazs GL. Models for flexural cracking in concrete: The
The values of τsm and kt should generally be determined with
state of the art. Struct Concr. 2005;6:53–62.
regard to cracking stage and crack width, for example, on basis of 15. Beeby AW. The influence of the parameter φ/ρ eff on crack widths.
Equation (7) and Equation (8) according to References 34, which can Struct Concr. 2004;4:155–165.
easily be included in the design where the crack width criterion is 16. Beeby AW. Discussion on “the influence of the parameter φ/ρ eff on
crack widths”. Struct Concr. 2004;5:71–83.
known beforehand. The well-established simplification of
17. Caldentey AP, Peiretti HC, Iribarren JP, Soto AG. Cracking of RC
τsm = 1.8 fctm and kt = 0.6 or 0.4, respectively, is only conceivable in members revisited: Influence cover, φ/ρs, ef and stirrup spacing –
cases of crack width limitation for durability reasons but without An experimental and theoretical study. Struct Concr. 2013;14:
requirements on the visual appearance or tightness. 69–78.
18. Debernardi PG, Taliano M. An improvement to Eurocode 2 and fib
Another aspect is the determination of the effective con-
model code 2010 methods for calculating crack width in RC
crete area. In this model understanding, Ac,eff represents the structures. Struct Concr. 2016;17:365–376. https://doi.org/10.
spreading of the bond force from the reinforcement location into 1002/suco.201500033.
the member, whereby new macroscopic cracks are always 19. Gribniak V, Caldentey AP, Kaklauskas G, Rimkus A, Sokolov A. Effect
of arrangement of tensile reinforcement on flexural stiffness and
formed as soon as the cracking force of the effective area is
cracking. Eng Struct. 2016;124:418–428.
built up by bond action. The determination of Ac,eff according to 20. Windisch A. Crack control: An advanced calculation model – Part I:
MC201033 and EC2-137 is suitable for this purpose while other Review of classic tests: The model. Conc Struct. 2016;41–48. Part II:
definitions depending predominantly on the diameter of the The model. Conc Struct, 2017;40-47.
 2625073x, 2021, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cend.202100004 by Cochrane Netherlands, Wiley Online Library on [27/02/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
SCHLICKE ET AL. 61

21. Krips, M. Rissbreitenbeschränkung im Stahlbeton und Spannbeton. 38. Borosnyo  i A, Sno

bli I. Crack width variation within the concrete cover
[PhD Thesis], TU Darmstadt, in German; 1984. of reinforced concrete members, Építöanyag. J Silicate Based Compos
22. Schießl, P. Grundlagen der Neuregelung zur Beschränkung der Mater. 2010;62:70–74.
Rißbreite. Deutscher Ausschuss für Stahlbeton, Heft 400, Berlin, 39. Bischoff P. Effects of shrinkage on tension stiffening and cracking in
Beuth Verlag; 1989;157–175. reinforced concrete. Can J Civ Eng. 2001;28:363–374.
23. König, G., Tue, N. V. Grundlagen und Bemessungshilfen für die 40. Kaklauskas G, Gribniak V, Bacinskas D, Vainiunas P. Shrinkage influ-
Rissbreitenbeschränkung im Stahlbeton und Spannbeton, DAfStb ence on tension stiffening in concrete members. Eng Struct. 2009;
Heft 466, Berlin, Beuth Verlag; 1996. 131:1305–1312.
24. Eckfeld, L., Schröder, S., Lindorf, A., Lemnitzer, L., Hamdam, A., 41. Schlicke D, Tue NV. Minimum reinforcement for crack width control
Curbach, M. Verbesserung der Vorhersagequalität von sehr kleinen in restrained concrete members. Struct Concr. 2016;16:221–232.
Rissbreiten. Schlussbericht des Forschungsvorhabens im Auftrag des 42. Tue NV, Schlicke D. Zwangbeanspruchung und
DIBt. Fraunhofer IRB Verlag, T3219, Dresden; 2009. Rissbreitenbeschränkung in Stahlbetonbauteilen auf Grundlage der
25. DIN 1045–1:2001–07: Concrete, reinforced and prestressed con- Verformungskompatibilität. in Beton Kalender 2020. Wasserbau -
crete structures - Part 1: Design and construction, in German, in ser- Konstruktion und bemessung, band 109–2: 831–887. Ernst & Sohn
vice until introduction of EN 1992–1-1. GmbH & Co. KG, Berlin, 2020.
26. DIN EN 1992–1-1/NA:2013–04: National Annex to EC2 Design of 43. Schlicke, D., Tue, N. V., Gomes, J., Sousa, C., Azenha, M. Analytical
concrete structures - Part 1 1: General rules and rules for buildings. assessment of restraint forces and crack widths in end-restrained
27. OENORM B 1992–1-1:2018-01-01: National Annex to EC2 Design of building slabs Engineering Structures. 2020;224.
concrete structures - Part 1 1: General rules and rules for buildings. 44. Gilbert, R., Nejadi, S. An experimental study of flexural cracking in
28. Rimkus, A., Cervenka, V., Gribniak, V., Cervenka, J. Uncertainty of the reinforced concrete members under short term loads. Report No. R-
smeared crack model applied to RC beams Engineering Fracture 434, University of NSW, Sydney, Australia; 2004.
Mechanics. 2020; 233. 45. Clark AP. Cracking in reinforced concrete flexural members. ACI J
29. Farra, B., Jaccoud, J.-P. Influence du Béton et de l'Armature sur la Proc. 1956;27(8):851–862.
Fissuration des Structures en Béton. Rapport des Essais EPFL IBAP 46. CUR Dutch Centre for Civil Engineering, Research and Codes Report
publication No 140. 1993;436 p. no. 37 (1994), out of print.
30. Rimkus A, Gribniak V. Experimental investigation of cracking and 47. Hognestad (1962): Journal of PCI Research and Development
deformations of concrete ties reinforced with multiple bars. Constr Laboratories.
Build Mater. 2017;148:49–61. 48. García, R., Pérez Caldentey, A. Cracking of Reinforced concrete: Ten-
31. Turner, K. Ganzheitliche Betrachtung zur Ermittlung der sion vs. Flexure: Report on cracking tests of large ties. Technical Uni-
Mindestbewehrung für fugenlose Wasserbauwerke. [PhD thesis], versity of Madrid (UPM); 2018.
Graz University of Technology, in German; 2017. 49. Caldentey, A. P., García, R., Gribniak, V., Rimkus, A. Tension vs. Flex-
32. Schlicke, D., Hofer, K., & Nguyen, V. T. Adjustable Restraining Frames ure: Reasons to modify the formulation of MC 2010 for cracking;
for Systematic Investigation of Cracking Risk and Crack Formation in 2019.
Reinforced Concrete Under Restrained Conditions. in Advanced 50. Tue, N. V., Zur Spannungsumlagerung im Spannbeton bei der
Techniques for Testing of Cement-Based Materials 2020: Springer Rißbildung unter statischer und wiederholter Belastung. Deutscher
Nature; 2020. Ausschuss für Stahlbeton, Heft 435; 1993.
33. fib Model Code 2010: fib Model Code for Concrete Structures 2010; 51. Pérez Caldentey A, García R, Gribniak V, Rimkus A. Tension versus
2013 flexure: Reasons to modify the formulation of MC 2010 for cracking.
34. Tue NV, König G. Calculating the mean bond and steel stress in Struct Concr. 2020;21:2101–2123. https://doi.org/10.1002/suco.
reinforced and prestressed concrete members. Darmstadt Conc. 202000279.
1992;6:77–86.
35. Eckfeld, L. In-depth analysis on bond of reinforcement in cracked rc
structures. In: Proceeding of 4th international Symposium: Bond in
concrete 2012: Bond, Anchorages, Detailing. General Aspects of
Bonds -Vol.1, Brescia; 2012. How to cite this article: Schlicke D, Dorfmann EM, Fehling E,
36. EN 1992–1-1:2004 + AC:2008: Eurocode 2: Design of concrete Tue NV. Calculation of maximum crack width for practical
structures - Part 1–1: General rules and rules for buildings. design of reinforced concrete. Civil Engineering Design. 2021;3:
37. Husain SI, Ferguson PM. Flexural crack width at the bars in reinforced
45–61. https://doi.org/10.1002/cend.202100004
concrete beams. Research Report 102-1F. Austin: Center for High-
way Research, The University of Texas at Austin; 1968.