Reliability assessment of international design concepts for punching

shear retrofitting by use of post-installed shear links

P. Mowlavi, N. Mellios & P. Spyridis
Faculty of Architecture and Civil Engineering, TU Dortmund, Dortmund, Germany
A. Strauss
Institute of Structural Engineering (IKI), University of Natural Resources and Life Sciences, Vienna, Austria

ABSTRACT: Flat slabs have often been used in construction in the last decades since they allow for higher
height clearance and more versatile architectural plans, compared to beam-columns solutions. For such types
of structures punching shear failure is often a critical stability item. Since safety requirement of an existing
structure may increase, e.g. due to change of use, ageing of materials, or emergence of more stringent design
standards, retrofitting against punching shear is often required. This contribution aims to showcase a method to
compare international design methods for punching shear retrofitting, and to present the respective results for
6 characteristic design cases. The design solutions presented in this paper rely on the design requirements of
the American Concrete Institute, the Canadian Standards Association, and the European Eurocode 2. Also the
Critical Shear Crack Theory is considered. The equations used to calculate resistances account for the contri-
bution of the concrete slab itself combined with the resistance of the post-installed shear links assumed to be
installed into this slab. The variables with the highest impact are identified and used as random input variables
in a self developed design program. The code is written in MATLAB and processed with a full Monte-Carlo
simulation, considering variations of the material and geometrical characteristics. The structural reliability in-
dexes and failure probabilities for the different design situations are compared. The different outcomes of re-
sistances and variances give evidence about contradictions in the inherent safety concepts of various design
standards worldwide.

1 INTRODUCTION - PROBLEM STATEMENT be compared and a rational conclusion on the inherent

safety of various codes can be explained. Further-
For an economical retrofit of systems endangered by more, the output results are evaluated as distributions
punching shear, it must first be found out which fac- (mainly normal and lognormal distributions), in order
tors play a decisive role for the punching shear re- to assess the influence of uncertainties on the punch-
sistance of a flat slab. Different codes such as the ing resistance of a flat slab. Based on this approach,
ones of the American Concrete Institute (ACI), the the real resistance of a component can be estimated in
Canadian Standards Association (CSA) and the Euro- an efficient and economical way.
pean Eurocode 2 (EC2), as well as the Critical Shear
Crack Theory (CSCT) are taken into account. In the
following relevant variables are selected based on a 2 THEORETICAL BACKGROUND
preliminary literature review. Considering these vari-
ables, a MATLAB-Code that calculates the re- 2.1 Previous investigations
sistances of flat slabs and their post-installed rein- Punching shear failure occurs when thin plates are
forcement is briefly presented, which further supported punctually by columns or similar elements.
incorporates uncertainties in the strengths of the con- When the system is overloaded, the plate fails within
struction materials used. a distance around the column, so that the column
Finally, the equations for flat slabs without shear punches through the plate. (Novacek & Zich, 2016)
reinforcement selected from the ACI, CSA, EC2, and The loads are highly localized at column support
CSCT are probabilistically elaborated, for 6 charac- points, and failure is very brittle, without visible signs
teristic cases of details strengthened against punching prior to failure. The unpredictability of its occurrence
shear. The main criterion considered tis the probabil- makes it a particularly critical and dangerous phe-
ity of failure of each calculation according to the nomenon. Moreover the redistribution of the inner
aforementioned codes. The target probability is gen- forces is extremely limited during the failure. This of-
erally to be used as a benchmark, thus the results can ten leads to a progressive collapse of the structure (see
e.g. Paul, 2014, and Obeng-Ankamah, 2015, and
Fiugres 1 and 2). Retrofitting existing structures is a
topic that gets more and more important during the
years. The older a building gets, the higher is the ne-
cessity of maintenance and monitoring. Maintenance
should be in a fast and non-destructive way, to ensure
a soon reopening and also an appearance with no or
at least minimal changes.
For the design of such measures, the work of Fer-
nández Ruiz & Muttoni,(2009, 2010) and Muttoni et
al. (2008) is taken at first into consideration, which is
based on the CSCT, specifically introduced for the
problems of punching shear strengthening with epoxy Figure 1. Collapse of Sampoong Department Store in South-Ko-
bonded anchors under a 45 degree angle (Figure 3). rea (1995) due to punching shear in the area of the top floor.
(Obeng-Ankamah, 2015).
These study the impact of varying loads and different
positioning of the reinforcements on the punching
shear resistance of a flat slab. The method is con-
firmed by an experimental program, showing the in-
crease of the punching shear resistance and of the de-
formation capacity of the slabs. The increase of
punching shear resistance ensures sufficient strength
– and the progressive collapses can be avoided due to
the higher deformation capacity, which leads to a bet-
ter redistribution of internal forces.
The investigations of Polak & Bu (2013), in con-
trast to the other approaches, uses shear bolts drilled
through and anchored at both sides of the slab with
headed ends or nuts and washers, instead of inserts Figure 2. Collapse of the car park Pipers Row in England (1997),
(Paul, 2014)
anchored within concrete. According to the authors,
the retrofitted flat slabs with this method still have the
same behaviour as properly cast-in-place retrofitted 2.2 Significant variables
flat slabs.
The method of Walkner at al. (2017) is partly based Based on a critical review of the available research,
on the CSCT, and it provides for vertically drilled an- the variables and factors with the highest influence on
chors and mechanical interlock with concrete screws, the resistance of slabs against punching shear have
since epoxy adhesives lose strength in case of fire. been identified. The factors that appear to have one of
Overall, the proposed retrofitting scheme based on the most important influence on the resistance of the
mechanical interlock is well predicted with the CSCT slab, are the height (h) and the compressive strength
for inclined bonded anchors. of the concrete used (fc), and the yielding strength of
The scientific work of Dondrea & Bayrak. (2019) the anchors (fyv). Furthermore, according to (Fernán-
examines retrofitting flat slabs with undercut and dez Ruiz et al, 2010) the amount and yielding
grouted anchors and testing on different reinforced strength of the flexural reinforcement (As, fyt𝑐𝑐𝑚𝑚2 )
specimens in diamond cored boreholes perpendicular must be considered, because it has an high impact on
to the slab. This study highlights that members with the deflection of the slab and therefore on the capacity
cast-in-place transverse reinforcement have a very of the member. In addition to this aspect, another im-
similar behaviour to members with post-installed re- portant factor would be the used epoxy adhesive per-
inforcement, especially the grouted anchors have a formance, as bonded anchors are dependent on the
good agreement due to their continuous bond along quality and strength of the bonding material, which is
the length of the anchors. It is also noted however that mainly reflected in the bond strength of the material
preparation of the holes for grouted anchors are criti- (τb). With reference to the investigations of (Walkner
cal due to the strong dependency of the intervention’s et al. 2017), the embedment depth of the anchors
performance on bond between the adhesive-to-con- plays a decisive role, too. The anchors should be in-
crete connection. The necessity of considering alter- stalled up to the layer of flexural reinforcement in or-
nate modes of anchor failure, thus the failure modes der to have a reliable force transmission. This aspect
concrete breakout, bond failure and local anchorage will be ensured by having the embedment of the an-
failure, is also mentioned. chor dependent on the position of flexural reinforce-
ment inside the slab in the equations; represented by
considering in the effective depth of the slab
𝑑𝑑 𝑖𝑖𝑖𝑖 [𝑚𝑚](d) for the calculation of the effective height
of the anchor. In reference to the paper of (Dondrea tension in (7.2.1.5); bond strength of an anchor in ten-
& Bayrak, 2019), the anchorage conditions need to be sion as in (7.2.1.9.2), where α1 and α2 are calculated
accounted for as well as the installation angle (bi𝛽𝛽𝑖𝑖 ). acc. to DIN EN 1992-1-1:2004, 8.4.4.

2.3.3 Equations of CSA-A23

2.3 Design-oriented equations in current
The equations referred to from ACI 318 are as fol-
international standards lows: Maximum shear stress resistance from (13.3.4),
The equations briefly presented below in terms of assuming critical shear area 0.5*d around the column,
code references are used to calculate the contribution normal weight concrete, and factor for column spac-
of the flat slab without shear reinforcement and the ing αs = 4 for interior columns, per 13.3.4.1; yielding
contribution of the anchors used as post-installed re- strength of an anchor in tension as in (D.6.1.2); pull-
inforcement. In codes, the contribution of reinforce- out Resistance of an anchor in tension (D.6.3.1).
ment is assumed in accordance with cast-in shear
links. From these, the total resistance of the system 2.3.4 Equations of of the CSCT
can be obtained. The equations used herein are in accordance to (Mut-
toni et al. 2008).
2.3.1 Equations of ACI 318 Concrete resistance against punching shear (Eq. (3)
The equations referred to from ACI 318 are as fol- and (4)) in kN:
lows: Concrete resistance for non-prestressed mem- 2 𝑢𝑢∗𝑑𝑑∗�𝑓𝑓𝑐𝑐𝑐𝑐
𝑉𝑉𝑅𝑅𝑅𝑅,𝑑𝑑 = 3𝛾𝛾 ∗ 𝜓𝜓𝑑𝑑 ∗𝑑𝑑 (3)
bers without axial forces given in Equation 22.5.5.1, 𝑐𝑐 1+20∗
𝑑𝑑𝑔𝑔0 +𝑑𝑑𝑔𝑔
p. 354); the nominal strength of an anchor in tension
(17.4.1.2); pull-out strength in tension of a single Load-Rotation-relation:
headed bolt (17.4.3.4), accounting for the mean com- 1.5
0.22∗𝑙𝑙 𝑓𝑓𝑦𝑦𝑦𝑦,𝑑𝑑 𝑉𝑉
pressive concrete strength; bond strength of a single 𝜓𝜓𝑑𝑑 = 𝑑𝑑
∗ ∗ �𝑎𝑎∗𝑚𝑚𝑑𝑑 � (4)
adhesive anchor (17.4.5.1a) (no edge near the anchor 𝐸𝐸𝑠𝑠 𝑅𝑅𝑅𝑅

assumed); basic bond-strength of a single adhesive Where:

anchor (17.4.5.2). 𝑙𝑙 =span of the slab (set as 5m for every calculation)
In this case, da is the diameter of the anchor in mm,
dh is the diameter of the anchor plate in mm, hef is the 𝑑𝑑 = 𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒 ℎ𝑒𝑒𝑒𝑒𝑒𝑒ℎ𝑡𝑡 𝑜𝑜𝑜𝑜 𝑡𝑡ℎ𝑒𝑒 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠
effective anchoring depth of the anchor and τuncr is the 𝑓𝑓𝑦𝑦𝑦𝑦 = yielding strength of the flexural reinforcement
minimum characteristic bond stress in uncracked con- 𝐸𝐸𝑠𝑠 = Modulus of elasticity
crete according to Table 17.4.5.2, but it is calculated 𝑎𝑎 = coefficient for the position of the column (here:
in MPa instead of psi (for dry indoor
τuncr = 1000 psi = 6.89 MPa). a=8 for interior columns)
The code programmed for the investigations pre- 𝑚𝑚𝑅𝑅𝑅𝑅 = 𝜌𝜌 ∗ 𝑑𝑑 ∗ 𝑓𝑓𝑦𝑦𝑦𝑦𝑦𝑦 ∗ 𝑧𝑧 , bending-resistance of slab
sented herein calculates each of these resistances ac- 𝐴𝐴
𝜌𝜌 = 𝑑𝑑∗𝑏𝑏𝑠𝑠 , flexural reinforcement ratio
cording to the anchorage and uses the lowest as the 𝑤𝑤
decisive resistance for one anchor. This resistance is 𝑧𝑧 = 0.9 ∗ 𝑑𝑑 , inner lever arm
called Rres𝑅𝑅𝑟𝑟𝑟𝑟𝑟𝑟 . The resistance of all the anchors com- Strength activated due to the deflection in kN:
bined is calculated per Equation 1. Finally, the total
resistance of the slab is being calculated by adding the
strength of the anchors to the strength of the concrete 𝑁𝑁𝑠𝑠,𝑒𝑒𝑒𝑒,𝑑𝑑 = 𝐾𝐾𝑎𝑎𝑎𝑎 ∗ �∆𝜓𝜓𝑑𝑑 ∗ (0.5 ∗ 𝑠𝑠) (5)
slab as in Equation (2)
0.22∗𝑙𝑙 𝑓𝑓𝑦𝑦𝑦𝑦 𝑉𝑉 1.5
𝑉𝑉𝑎𝑎𝑎𝑎𝑎𝑎ℎ𝑜𝑜𝑜𝑜,𝐴𝐴𝐴𝐴𝐴𝐴 = 𝑛𝑛 ∗ 𝑅𝑅𝑟𝑟𝑟𝑟𝑟𝑟 (1) ∆𝜓𝜓𝑑𝑑 = 𝜓𝜓𝑑𝑑 − 𝑑𝑑
∗ ∗ �𝑎𝑎∗𝑚𝑚𝑠𝑠𝑠𝑠𝑠𝑠 � (6)
𝐸𝐸𝑠𝑠 𝑅𝑅𝑅𝑅

𝑉𝑉𝑟𝑟,𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡,𝐴𝐴𝐴𝐴𝐴𝐴 = 𝑉𝑉𝑐𝑐,𝐴𝐴𝐴𝐴𝐴𝐴 + 𝑉𝑉𝑎𝑎𝑎𝑎𝑎𝑎ℎ𝑜𝑜𝑜𝑜,𝐴𝐴𝐴𝐴𝐴𝐴 (2) ∆𝜓𝜓𝑑𝑑 = difference between the deflection at the time
of strengthening and after the strengthening process
𝑉𝑉𝑠𝑠𝑠𝑠𝑠𝑠 = strength at the time of strengthening (set as
2.3.2 Equations of Eurocode 2 (EC2) 0.5𝑉𝑉𝑑𝑑 )
The equations referred to from EC2 are as follows: Yielding strength of an anchor in tension in kN acc.
the characteristic resistance of a slab without punch- To Equation 7:
ing shear reinforcement (6.4.4, Eq. 6.47) – assuming
mean concrete strength, and factor CRdc per the Ger-
man National Annex DIN EN 1992-1-1NA, 6.4.4. 𝑁𝑁𝑠𝑠,𝑝𝑝𝑝𝑝,𝑑𝑑 = 𝐴𝐴𝑖𝑖 ∗ 𝑓𝑓𝑦𝑦𝑦𝑦 (7)
The anchor capacities are based on EC2 Part 4
(DIN EN 1992-1-4, 2019): Yielding strength of an an- Bond strength of an anchor in tension in kN, accord-
chor as in (7.2.1.3); pull-out failure of an anchor in ing to Equation 8:
 ℎ𝑒𝑒𝑒𝑒 iables are assumed normally distributed. The varia-
𝑁𝑁𝑠𝑠,𝑏𝑏,𝑑𝑑 = 𝐴𝐴𝑖𝑖 ∗ 4 ∗ 𝜏𝜏𝑏𝑏𝑏𝑏 ∗ 𝜋𝜋 ∗ (8)
𝑑𝑑𝑎𝑎 tions of the variables are intended to capture variabil-
Concrete breakout resistance of an anchor in tension ities and inaccuracies mainly due to material and
in kN according to Equation 9: workmanship, and model uncertainties are not specif-
ically targeted. To do so, the characteristic values of
the abovementioned variables are converted into
0.360 ℎ𝑒𝑒𝑓𝑓 1.5 𝑑𝑑
𝑁𝑁𝑠𝑠,𝑝𝑝,𝑑𝑑 = 𝐴𝐴𝑖𝑖 ∗ 𝛾𝛾𝑐𝑐
∗ �𝑓𝑓𝑐𝑐𝑐𝑐 ∗ ∗ �1 + ℎ ℎ � (9) mean values and an appropriate standard deviation is
𝑑𝑑𝑎𝑎 𝑒𝑒𝑒𝑒
attributed to each one. The distributions of the com-
pressive strength of the concrete and the bond
strength of the adhesive are calculated according to
3 METHODOLOGY the Eurocode 0, assuming a factor 𝑘𝑘𝑛𝑛 =1.64 for a 5%-
quantile of the distribution. The failure probability is
As mentioned in the chapters above, six scenarios of calculated based on the reliability principles dis-
flat slabs with post-installed bonded anchors are cussed by Zilch & Zehetmaier (2006).
tested. The anchors used in this work have a diameter The coefficient of variation from the bond strength
of 20 mm and an anchoring plate with a diameter of of the adhesive τb is 0.10 and was chosen based on the
60mm and is used in every case. The longitudinal re- investigations of (Eligehausen et al. 2004). Even
inforcement is chosen as B500 reinforcing steel bars though the coefficient of variation for the yield
(fyv,k = 500 MPa). The same material strength is con- strength of steel according to the International Feder-
sidered for the shear reinforcement. The material ation for Structural Concrete (2016) is 0.05, a deci-
characteristics, the height of the slab and the number sion to use 0.15 was used in order to account for pos-
of anchors is changed in the scenario to see the differ- sible ageing and deterioration for the existing slabs.
ent impacts on the strength of the members. The ma- The variation of the angle of the reinforcement de-
terial characteristics considered are being the com- pends on the chosen angle, because the standard de-
pressive strength of the concrete, the effective depth viation chosen is 3° assuming this deviation is visible
of the slab and the yielding strength of both the trans- for the human eye (DIBt, 2010). The cov used for both
verse as the longitudinal reinforcement, just as the the concrete and shear/post-installed reinforcement
bond strength of the adhesive. Furthermore, the strength was taken as 0.15 based on (Braml, Fischer,
amount of flexural reinforcement is also be varied, the Keuser, & Schnell, 2009). The coefficient of variation
angle of the anchors as well. Table 1 shows the above- of d is not noted, as the standard deviation is set to 10
mentioned scenarios and Figure 3 shows the geomet- mm, according to the Probabilistic Model Code (Joint
rical situation and explains the variables. In this way, Committee on Structural Safety, 2001). Dimensional
we get matrices that have the dimension of the num- variability was also excluded. Table 2 summarises the
ber of samples chosen. When, for example, the num- assumed variation coefficients.
ber of samples is 1000, the random generator creates
matrices with one column and 1000 rows, where Table 1. Cases 1-6 tested for the reliability analysis.
every row includes a value within the range that is set Case h d fck n βi [°] τb,k As
by the mean value and the standard deviation of a nor- [m] [m] [MPa] [-] [MPa] [cm
2
mal distribution. Thus, the resistance of the slab with- ]
out punching shear reinforcement is calculated with 1 0.35 0.32 26.53 20 90 7.45 20
these variables, which are matrices. Different anchor 2 0.35 0.32 39.79 15 90 7.45 20
failure types are considered to find the minimum
3 0.30 0.27 26.53 15 90 7.45 30
strength, that an anchor can bear – if this strength is
exceeded, one of the considered failure types occurs. 4 0.30 0.27 53.05 20 45 7.45 30
Therefore, this strength is set as the decisive maxi- 5 0.25 0.22 33.16 26 45 7.45 35
mum capacity of one anchor. By adding the strength
6 0.25 0.22 39.79 26 45 7.45 35
of the concrete and the anchors together, the total re-
sistance of each the slab in the different scenarios is
obtained.
The six different configurations re calculated deter-
ministically according to the codes to yield the shear
resistance that each code refers to the specimens – one
time as characteristic resistances, the second time
with consideration of the safety factors, which the
codes demand to get the design resistances. In the
next step, the variables of the Cases 1 to 6 that were
used in the deterministic calculation are elaborated in
the probabilistic calculation as well. The random var-
Figure 3. Geometry of the problem (see also Table 1)
Table 2. Coefficients of variation of the variables 4.2 Mean values and standard deviations for
used in the probabilistic calculation differing sample orders
d fc fyt fyv βi τb,k As
The total resistances noted in Table 4 include the
CoV - 0.15 0.15 0.15 0.03 0.10 0.05 contribution of the concrete Vc and the contribution
of the shear reinforcement anchors Vanchor. First, the
To obtain the design value of the load, here assumed differences in the results of the Eurocode for the sce-
as exclusively dead loads, an inverse calculation was narios are examined. The coefficient of variation in-
carried out in accordance to the semi-probabilistic creases from Case 2 to 3 and then again from Case 4
concept of Eurocode 0 (DIN EN 1990, 2010), and the to Case 5 for both sample sizes. The Cases 2 and 3
use of Equations 10 to 12, leading to Equation 13. The differ in the flexural reinforcement ratio, the height of
coefficient of variation of 0.10 is assumed for shear the slab and the compressive strength of the concrete.
loads (International Federation for Structural Cases 4 and 5 have the same differences. The com-
Concrete, 2016). The load variable was assigned a pressive strength from Case 1 to Case 2 increases sig-
normal distribution with 𝐿𝐿𝑚𝑚 as the mean value and nificantly, the coefficient of variance, however, re-
0.10 as the coefficient of variation, calculated for mains stable. Thus, it can be assumed, that the higher
each case individually, considering every variable. flexural reinforcement ratio - due to the lower height
The main measures extracted from the probabilistic of the slab and the higher amount of flexural rein-
calculations where the failure probability and the re- forcement - induce variability in the resistance of the
liability index in line with established methodologies anchors and therefore have higher scattering. Also,
as the ones presented by Schneider (2006) and Müller the variation of the two sample sizes differ. The sig-
(2010). nificantly higher sample size caused about 2% more
variation than the lower sample size, which allows as-
𝑅𝑅𝑘𝑘 suming that there is little to no sensitivity from the
𝑅𝑅𝑑𝑑 = ≥ 1.35 ∗ 𝐿𝐿𝑘𝑘 = 𝐿𝐿𝑑𝑑 (10) sample size, at least as regards the main structure of
1.5
the resistance probability distribution functions.
𝑅𝑅𝑘𝑘 = 𝑅𝑅5% = 𝑅𝑅𝑚𝑚 − 1.64 ∗ 𝑐𝑐𝑐𝑐𝑣𝑣𝑅𝑅 ∗ 𝑅𝑅𝑚𝑚 (11) The results of the ACI-based calculations display
𝐿𝐿𝑘𝑘 = 𝐿𝐿95% = 𝐿𝐿𝑚𝑚 + 1.64 ∗ 𝑐𝑐𝑐𝑐𝑣𝑣𝐿𝐿 ∗ 𝐿𝐿𝑚𝑚 (12) that the standard deviations increased significantly,
but the mean values remain consistent. This happens
𝐿𝐿𝑚𝑚 = 0.354 ∗ 𝑅𝑅𝑚𝑚 (13) from Case 3 to Case 4, where the number of post-in-
stalled anchors and the concrete compressive strength
increased. As already seen in the results of the Euro-
4 RESULTS code above, the coefficient of variation dis not af-
fected by the increase of the concrete compressive
4.1 Deterministic assessments strength, therefore it can be assumed, that it does not
cause additional variation in the results. Even though
Table 3 shows the results of two deterministic calcu- the amount of flexural reinforcement is kept constant
lations (design value and characteristic values) from Case 3 to Case 4, due to the increase of the num-
against the 5% - quantile of the probabilistic calcula- ber of anchors used, it can be assumed, that the calcu-
tion The 5%-quantiles of the probabilistic calculated lation of the anchors causes higher variation. This
resistances are seen as equivalent to the characteristic phenomenon can be observed for the CSA as well.
resistances out of the deterministic values. The 5%- The CSCT assessment however, has as a different
quantiles of the Eurocode 2 are 30 to 40% higher than outcome. Here, the coefficient of variation increases,
the characteristic resistances. Comparing the design together with the increases in the compressive
resistances with the 5%-quantiles, the design re- strength, and the variation becomes lower by higher
sistances are as expected significantly lower. In addi- amounts of flexural or anchor- reinforcement.
tion to that, the ratio between the resistances remain Moreover, differences have been investigated by
stable for the different Cases – the 5%-quantiles of the comparison of the coefficients of variation be-
the Eurocode are nearly double of the design re- tween the codes among each other. The resistances of
sistance in the most Cases, which corresponds to the the Eurocode 2 and the ACI are by far the highest in
product of the material and action safety factors. every calculated Case, the resistances of the CSA and
While EC2 has clearly higher 5%-quantiles than the CSCT on the other hand, are significantly lower;
design values, the 5%-quantiles of the ACI and the in the most Cases the only sum up to approximately
design values do not exhibit this deviation; the char- 65% of the respective EC2 and ACI-values. In addi-
acteristic resistances are generally higher than the cal- tion to that, the coefficients of variation of the ACI
culated quantiles. Calculations per the CSA have a and the CSA are significantly higher than the varia-
similar pattern. On the contrary, the CSCT has higher tions of the other codes. On the contrary, the re-
design values than 5%-quantiles; instead, these are in sistances according to the CSCT only show variation
the same range as the characteristic values. of 5 to 10%. Finally, the calculation with a 10 million
 Table 3. Total resistances in design, characteristics and 5%-quantiles
Case 1 Case 2 Case 3 Case 4 Case 5 Case 6
Vr,total,EC2,d 3007.60 2436.00 2023.20 2665.40 2582.20 2608.20
EC2 Vr,total,EC2,k 3925.70 3253.80 3111.50 3173.10 3744.40 3798.10
Vr,total,EC2,0.05 5249.00 5074.70 4530.10 4969.50 4769.30 4646.50
Vr,total,ACI,d 5512.20 4643.70 4234.80 4419.50 4905.40 4984.20
ACI Vr,total,ACI,k 7349.60 6191.60 5646.40 5892.70 6790.60 6919.40
Vr,total,ACI,0.05 5815.20 5006.20 4476.30 4414.40 4952.80 4780.70
Vr,total,CSA,d 3600.30 2921.80 2729.10 2751.60 3212.10 3243.80
CSA Vr,total,CSA,k 4896.10 4012.90 3716.50 3778.60 4492.40 4554.60
Vr,total,CSA,0.05 3818.70 3182.50 2898.80 2753.90 3175.70 3097.10
Vr,total,CSCT,d 2782.10 2798.70 2025.60 2632.70 2205.40 2322.20
CSCT Vr,total,CSCT,k 3634.20 3795.60 2816.90 3598.30 3247.60 3463.50
Vr,total,CSCT,0.05 4055.00 4074.30 3139.00 3786.40 3690.60 3367.60

Table 4. Mean values µ and standard deviations 𝜎𝜎 of the total resistances Vr,total according to the codes [kN]
Case kn Vr,total,EC2 Vr,total.ACI Vr,total,CSA Vr,total,CSCT
µ 𝝈𝝈 µ 𝝈𝝈 µ 𝝈𝝈 µ 𝝈𝝈
1 1 thousand 6551.73 664.75 7519.93 945.03 4962.18 656.59 4496.00 235.00
10 million 6585.67 812.61 7559.59 1060.48 5000.80 718.99 4521.00 290.00
2 1 thousand 6267.11 584.51 6419.14 744.36 4121.26 513.95 4730.00 320.00
10 million 6299.85 744.83 6451.65 878.78 4143.16 584.05 4770.00 440.00
3 1 thousand 5644.94 557.58 5793.33 723.36 3735.56 485.93 3541.00 211.00
10 million 5675.47 696.32 5824.73 819.81 3802.05 549.13 3563.00 268.00
4 1 thousand 6382.28 707.76 6139.35 932.822 3897.59 644.74 4510.00 350.00
10 million 6420.93 882.38 6178.83 1066.60 3923.66 711.179 4550.00 480.00
5 1 thousand 5964.40 783.10 6698.86 1162.32 4420.97 807.58 3599.00 200.00
10 million 6346.28 958.73 7079.41 1292.87 4609.99 872.01 4280.00 360.00
6 1 thousand 6157.14 797.03 6818.99 1168.71 4478.90 811.841 3818.00 221.00
10 million 6196.61 942.41 6864.65 1266.98 4510.32 859.19 3843.00 295.00

sample size caused somewhat higher variations, prob- of failure increased not only for the Eurocode, but
ably because the higher number of different situations also for the other codes as well. This could be due to
lead to a higher spectrum of values. In addition, the the fact, that more variables influence each other for
degree of fitting of the distributions of the resistances the calculation of the resistance of reinforcing mem-
concerning the normal and the lognormal distribution bers and thus have greater variation in the resistances.
are examined. The used algorithm is programmed to This phenomenon is due to the increased coefficient
output a result that is either 0 or 1, depending on the of variation with a rising amount of shear and flexural
adequacy of curve fitting (“0” means, that the curve reinforcement as mentioned above. The EC2 and the
is fitting). Furthermore, a degree of fitting was calcu- ACI have acceptable failure probabilities and thus re-
lated, that is between 0 and 1. If the fitting is accepta- liability indices above 4.0 and in some cases even
ble, this reference value is around 0.50, which allows around 4.6. On the contrary, the CSCT and especially
for the assessment of the sample adequacy. the CSA, have β-values lower than 4.0 – the CSA
even lower than 3.0.
4.3 Probability of failure and reliability index Table 5. Calculated failure probability (pf) and relia-
For this assessment only the calculation of 10 mil- bility index (β).
lion samples was accounted for. that the load applied EC2 ACI CSA CSCT
on the samples was based on the approach of the Eu- pf (β) pf (β) pf (β) pf (β)
rocodes and was therefore kept the same for every 1 5.0 ∙ 10-7 (4.9) 1.3 ∙ 10-6 (4.7) 2.5 ∙ 10-4 (3.5) 8.1 ∙ 10-5 (3.8)
sample. Moreover, the codes deliver different re- 2 1.0 ∙ 10-7 (5.2) 2.8 ∙ 10-6 (4.5) 1.2 ∙ 10-3 (3.0) 6.3 ∙ 10-5 (3.8)
sistances for the same Case based on varying correla- 3 3.0 ∙ 10 (5.0)
-7
4.7 ∙ 10 (4.0)
-6
1.2 ∙ 10 (3.0)
-3
1.0 ∙ 10-4 (3.7)
tions of the variables. As shown in Table 5, the prob- 4 6.0 ∙ 10 (4.9)
-7
4.0 ∙ 10 (3.9)
-5
1.0 ∙ 10 (2.3)
-2
2.1 ∙ 10-4 (3.5)
abilities and the β-values differ significantly. First, 5 2.2 ∙ 10-6 (4.6) 1.6 ∙ 10-5 (4.2) 2.4 ∙ 10-3 (2.8) 6.5 ∙ 10-4 (3.2)
the failure probabilities of the first four Cases of the 6 2.7 ∙ 10 (4.5)
-6
2.0 ∙ 10 (4.1)
-5
2.5 ∙ 10 (2.8)
-3
6.3 ∙ 10-4 (3.2)
Eurocode are even smaller than 1∙10-6 (the probability
generally demanded by the codes).
With the growing contribution of the reinforcement
until Case 6 – the height of the slab becomes smaller,
the amount of reinforcement higher – the probability
 5 DISCUSSION

Examining the 5%-quantiles of the codes in Table 4,

the ACI and the EC2 have the highest resistances,
whereas the resistances of the CSA are higher than
the results from the CSCT. This should intuitively
lead to the conclusion, that the probability of failure
of the CSA is lower than the failure probability of the
CSCT – but that is not the case. Considering that the
CSA has by far the highest coefficient of variation up
to 0.19, shows that even though the peak of the nor-
mal distribution – and thus the resistances – may be
high, the standard deviation leads to an extended
graph and thus to a greater overlap with the distribu-
tion of the load. It can be assumed that in the case of
the CSA the scatters of the variables influence each
other in a way that the failure rate is not acceptable.
Consequently, there will be more failures than in the
case of the CSCT, where the mean value and 5%-
quantile may be lower, but the low coefficient of var-
iation under 0.10 leads to a small distribution width.
This also underlies the fact that the resistances of the
Eurocode 2 and the ACI are up to 60% higher, and it
easily can be assumed that the probabilities of failure
of these two codes are lower.
Finally considering every result, the approach of
the Eurocode and the ACI have very reasonable out-
comes in the reliability assessment. The probability
Total resistance acc. to the ACI (kN) that a failure occurs is close to the generally de-
manded values and therefore this approach seems to
be a good design approximation of the design of ret-
rofitting existing flat slabs and thus ensuring enough
safety against punching shear. As a consequence, the
design would be more economical and thus construc-
tion material can be saved.
The comparison of the results of the codes among
each other points out, that the resistances of the CSA
generally contain the larger safety reserves. The
CSCT however, lies closer to the results per the ACI
or the EC2 codes. In the case of the CSCT, the defor-
mation of the slab also has a great impact on the
strength and should be taken into account also in a
reliability assessment.

6 REFERENCES

ACI Committee 318. (2014). Building Code Requirements for

Structural Concrete (ACI 318-14) and Commentary. IHS.
Beutel, R. R. (2002). Durchstanzen schubbewehrter Flachde-
cken im Bereich von Innenstützen. Aachen, Northrhine-
Westphalia, Germany.
Braml, T., Fischer, A., Keuser, M., & Schnell, J. (2009). Beur-
teilung der Zuverlässigkeit von Bestandstragwerken hin-
sichtlich einer Querkraftbeanspruchung. In K. Bergmeister,
F. Fingerloos, & J.-D. Wörner, Beton-Kalender 2019 (S.
799-812). Berlin: Ernst & Sohn Verlag.
Canadian Standards Association, (2007). Design of Concrete
Structures CSA-A23.3-04. Standards Council of Canada.

Figure 4. Indicative probabilistic analyses results for Case 1.

DIBt (2010) Hinweise für die Montage von Dübelverankerun- Schneider, J. (2006). Introduction to Safety and Reliability of
gen (Guidelines for the installation of anchorages), Deut- Structures. Zürich, Switzerland: IABSE-AIPS-IVBH.
sches Institut für Bautechnik, 2010 Walkner, R., Spiegl, M., & Feix, J. (2017). A new method for
DIN EN 1990, (2010). Eurocode: Basis of structural design; post-installed punching shear reinforcement. In A.-f. I. Sym-
German version EN 1990:2002 + A1:2005 + posium, Bulletin 81 - Punching shear of structural concrete
A1:2005/AC:2010. Beuth Verlag GmbH. slabs (pp. 337-351). University of Innsbruck, Austria: Amer-
DIN EN 1992-1-1, (2011). Eurocode 2: Design of concrete ican Concrete Institute.
structures - Part 1-1: General rules and rules for buildings; Zilch, K., & Zehetmaier, G. (2006). Bemessungen im konstruk-
German version EN 1992-1-1:2001 + AC:2010. Beuth Ver- tiven Betonbau. Springer.
lag GmbH.
DIN EN 1992-1-1/NA, (2013). National Annex - Nationally de-
termined parameters - Eurocode 2: Design of concrete struc-
tures - Part 1-1: General rules and rules for buildings. Beuth
Verlag GmbH.
DIN EN 1992-1-4, (2019). Eurocode 2 - Design of concrete
structures - Part 4: Design of fastenings for use in concrete;
German version EN 1992-4:2018. Beuth Verlag GmbH.
Dondrea, A., & Bayrak, O. (2019, May). Undercut and Groutes
Anchors as Post-Installed Shear Reinforcement. ACI Journal
No. 116-S51, pp. 27-37.
Eligehausen, R., Appl, J. J., Bernhard, L., Meszaros, J., & Fuchs,
W. (2004). Tragverhalten und Bemessung von Befestigun-
gen mit Verbunddübeln unter Zugbeanspruchung. In K.
Bergmeister, F. Flingerloos, & J.-D. Wörner, Beton-Kalen-
der 2019 (S. 561-571). Ernst & Sohn.
Fernández Ruiz, M., & Muttoni, A. (2009). Applications of Crit-
ical Shear Crack Theory to Punching of Reinforced Concrete
Slabs with Transverse Reinforcement. ACI Structural Jour-
nal V. 106, No.4, pp. 485-494.
Fernández Ruiz, M., & Muttoni, A. (2010). Performance and de-
sign of Punching Shear Reinforcing Systems. Lausanne,
Switzerland.
Fernández Ruiz, M., Muttoni, A., & Kunz, J. (2010). Strength-
ening of Flat Slabs Against Punching Shear Using Post-In-
stalled Shear Reinforcement. ACI Structural Journal, V.107,
No.4, pp. 434-441.
International Federation for Structural Concrete. (2016). A.2
Derivation of probabilistic models. In M. Prieto, P. Tanner,
& e. al., fib Bulletin 80 - Partial safety factor method for ex-
isting concrete structures. Germany: Document Competence
Center Siegmar Kästl e.K.
Jain, V. (2019). The Constructor - Civil Engineering Home. Re-
trieved from Flat Slabs - Type of Flat Slab Design and its
Advantages: https://theconstructor.org/structural-engg/flat-
slab-types-design-advantages/13919/
Joint Committee on Structural Safety, . (2001). JCSS Probabil-
istic Model Code Part 3: Resistance Models. In J. C. Safety,
JCSS Probabilistic Model (p. 3).
Müller, H. S. (2010). The role and tools of lifetime management
of civil concrete structures. In Non-destructive Evaluation of
Reinforced Concrete structures (pp. 94-113). Karlsruhe, Ger-
many: Woodhead Publishing.
Muttoni, A., Fernandez Ruiz, M., & Kunz, J. (2008). Nachträg-
liche Durchstanzbewehrung zur Verstärkung von Stahlbe-
tonflachdecken. In S. V. Verlag, Bauingenieur, Springer
VDI Verlag, 83, 503-511 (p. 9). Germany: Springer VDI
Verlag.
Novacek, J., & Zich, M. (2016). Study of Flat Slabs Strengthen-
ing against Punching Shear. Brno, Czech Republic.
Obeng-Ankamah, N. (2015). ResearchGate. Retrieved from
https://www.researchgate.net/figure/Sampoong-department-
store-collapse_fig3_306323560
Paul, A. (2014). Civildigital. Retrieved from https://civildig-
ital.com/punching-shear-punching-shear-flat-slabs/
Polak, M. A., & Bu, W. (2013). Design Considerations for Shear
Bolts in Punching Shear Retrofit of Reinforced Concrete
Slabs. ACI Structural Journal No. 110-S02, pp. 15-25.
RiskNet. (2019). RiskNet - The Risk Management Network. Re-
trieved from https://www.risknet.de/wissen/rm-
methoden/monte-carlo-simulation/