Magazine of Concrete Research, 2008, 60, No.

5, June, 381–388
doi: 10.1680/macr.2008.60.5.381

Cracking in deep beams owing to shear loading.

Part 2: Non-linear analysis
R. Malm and J. Holmgren

Royal Institute of Technology (KTH)

In this paper, analyses based on laboratory tests of ten large deep beams with I-shaped cross-sections loaded to
failure are presented. All beams had the same geometry with a shear span-to-depth ratio of 1.25, but differed
in the amount of the vertical and horizontal web reinforcement. All beam tests resulted in shear failure, either
diagonal tensile failure or shear compressive failure, depending on the amount of reinforcement. The diagonal
tensile failure is generally considered to be the most difficult failure to treat numerically. In this study different
material models incorporated in commercial numerical analysis tools are studied. Material models based on
fracture mechanics with either rotated or fixed crack directions as well as a plasticity-based model are used in
the analyses. The analyses show that the plasticity-based model in Abaqus gives good agreement with the
experiments regarding crack pattern, load–displacement response and estimated crack widths. The models based
on fracture mechanics in Atena and Response tend to give too stiff behaviour in the load–displacement
response, but generally give a good estimation of the load capacity. The analyses performed with Atena gave
good estimations of the crack pattern, and the models with a fixed crack direction also gave good estimates of
the crack width.

Notation Introduction
Ec initial elastic modulus The structural safety of reinforced-concrete struc-
fcc compressive concrete cylinder strength tures is determined through analysis of the structural
fct concrete tensile strength failure, which requires accurate analyses of the ultimate
Gf fracture energy load. Reinforced-concrete structures are designed to
h length of crack band satisfy criteria of serviceability and safety. To ensure
Vcalc calculated ultimate strength the requirements in the serviceability state, cracking
Vexp experimental ultimate strength and deflections need to be predicted under service
wc crack opening displacement at which stress loads and should be limited to some values according
can no longer be transferred to Design Codes. Keeping this in mind, a non-linear
wcalc calculated crack width analysis is required to evaluate crack propagation and
wd plastic softening compression ultimate load-carrying capacity and the load–deforma-
wexp experimentally observed crack width tion response.1
w t, w t0 crack opening displacement and crack To study the capability to model the behaviour at
opening displacement at peak stress shear failure with different finite-element (FE)
åc0 compressive strain at peak stress packages, a few examples have been analysed. The
í Poisson’s ratio analyses were performed on flanged deep beams ex-
perimentally tested by Nylander and Holmgren2 and
internationally published by Malm and Holmgren.3 All
the tested beams had the same geometry and consisted
of an I-shaped cross-section but differed in the amount
Department of Civil and Architectural Engineering, Royal Institute of of the vertical and horizontal reinforcement in the
Technology (KTH), SE – 100 40 Stockholm, Sweden
webs. The purpose of the experiments was to determine
(MACR-D-07-00104) Paper received 23 August 2007; last revised 18 the effect of cracking, at different reinforcement ratios,
January 2008; accepted 14 February 2008 on the state of stress and the strength near the supports
381

www.concrete-research.com 1751-763X (Online) 0024-9831 (Print) # 2008 Thomas Telford Ltd

Malm and Holmgren
150
of the deep beams. The FE analyses are made to com-
pare different material models for describing concrete 800
regarding failure, load capacity, cracking and deflec-
150
tions. The results of the FE analyses are compared to 200
the measured load–deformation curves to see how well 150
1225 1400 1225 500
the programs estimate the ultimate load and the struc-
tural behaviour. It is also interesting to study how the Fig. 1. Dimensions of the studied deep beams (in mm)
FE programs can be used to estimate the crack widths.
Several of the numerical programs calculate the crack
width since it is used to determine the shear transfer at
a crack. reinforcement, namely beam 1 and 5. One beam—beam
The analyses are performed with the material model 9—had horizontal and vertical web reinforcement with
concrete damaged plasticity in Abaqus (Abaqus/ a reinforcement spacing of 80 mm. The remaining
Explicit4 ) and the material model SBETA in Atena beams had various arrangements of shear reinforcement
(Cervenka et al.5 ). For comparison with the FE results within the shear span. The beams were subjected to
the program Response (Bentz6 ) has also been used. four-point bending where the load was applied through
This program is based on the modified compression a load-controlled system consisting of two hand-jacks
field theory (MCFT) developed by Vecchio and Col- producing approximately 75 kN each, at every load
lins.7,8 In the program a feature called ‘member re- increment. Between the load steps the load was main-
sponse’ is included where it is possible to obtain load– tained for 15 to 20 min while the cracks were marked
deformation curves for beams. The experimentally and measured. At the beginning and the end of each
tested beams used for comparison of the numerical loading increment the gauge readings were recorded.
methods only include one experiment per reinforcement The deflection of the bottom flange was measured with
solution since the test configuration was in some ways dial gauges below the loading plates.
different in the respective beams. If the experiments
were to be reproduced, it is likely that there would be a
scatter in the obtained results.
Material models for concrete
Verifications of non-linear FE material models for
concrete are usually performed based on experiments The material models used for analyses in this paper
with notched beams. For instance the extensive test treat cracked concrete as a continuum where the cracks
series performed by Petersson9 has been popular to are distributed over the element. The elements describe
reproduce and is successfully analysed with FE meth- both the elastic part of the uncracked concrete and the
ods by Rots et al.10 and de Borst,11 among others. inelastic deformation owing to cracking.
One problem, however, is that in a real case the crack In Atena, the constitutive material model for describ-
path may not be known in advance so the model ing the non-linear behaviour of concrete in a two-
must be also able to provide a good estimate of the dimensional (2D) plane stress state is called SBETA. It
crack pattern. In the current paper test results from is based on the concept of smeared cracks, damage and
deep beams are presented. The beams had a varied fracture mechanics. The concrete failure in both tension
degree of transverse reinforcement and the failure (cracking) and compression (crushing) causes disconti-
modes owing to shear forces resulted in either diag- nuities in the displacement fields which, according to
onal tensile failure or crushing in the web close to Jendele et al.,12 are in basic disagreement with the
the support. assumptions of continuum mechanics. On the macro
level, concrete failure exhibits itself as strain softening,
and is of a strongly localised nature. Special techniques
based on localisation limiters defined by Bažant and
Description of the experiments
Oh13 in the form of crack bands are therefore adopted
The experiments have been fully described in the to describe the post-peak behaviour of concrete in the
companion paper by Malm and Holmgren3 and in this FE model. In the constitutive model, no residual da-
section only a brief description of the experiments is mage occurs after unloading, which is typical for da-
made. The experiments consisted of ten beams with the mage models. Within the smeared crack approach
same geometry and loading conditions. The supports of adopted in SBETA two options are available for crack
the beams consisted of two columns which were cast in models: a fixed or a rotated crack direction. In both
concrete with higher strength to make sure that the models, cracks are formed when the principal stress
failure would occur in the beams. An illustration of the exceeds the tensile strength. The crack direction in the
experimentally tested beams is shown in Fig. 1. The fixed crack approach is determined by the principal
only difference between the beams was the reinforce- stress direction at the moment of crack initiation. Dur-
ment solutions in the web and the material properties ing further loading this direction is fixed. In the rotated
of the concrete. Two beams were made without web crack model, the crack direction changes so that it
382 Magazine of Concrete Research, 2008, 60, No. 5
 Cracking in deep beams owing to shear loading. Part 2: Non-linear analysis

follows the principal stress direction if it changes ow- interlocking, but will not able to transmit tensile forces.
ing to further loading. In the rotated crack model, no However, in the uncracked concrete between the cracks
shear strain occurs on the crack plane. To ensure the tensile stresses will be carried. The mean principal
coaxiality of the principal strain axes with the material stress capacity decreases as the mean principal strain in
axes a tangent shear modulus is calculated. the element increases. Cracked concrete is treated as a
In Abaqus there are three material models for de- new material with its specific stress–strain character-
scribing the non-linear behaviour of concrete. The mod- istics.
el used in this paper, which is most suited for the
analyses, is based on plasticity theory and is called
concrete damaged plasticity. It was developed by Lubli-
Simulations of the beams
ner et al.14 and includes the modifications that were
proposed by Lee and Fenves.15 In this model it is In the models made with Atena 2D the material
possible to define the material degradation in compres- definition SBETA was used, with either a fixed or a
sion as well as in tension. It can be defined so that its rotated crack direction. The Abaqus models were made
tensile softening behaviour is based on a crack-opening with the material definition concrete damaged plasti-
law and fracture energy. Damage is associated with the city, where the element can be damaged by either
failure mechanisms of the concrete (cracking and tension or compression. Since the damage is isotropic,
crushing) and therefore results in a reduction of the it means that the ‘cracks’ may be assumed to follow
elastic stiffness. Within the context of the scalar- the principal strain direction. The material parameters
damage theory, the stiffness degradation is isotropic used in the FE analyses with Abaqus and Atena are
and characterised by degradation variables for tension based on the measured cube strength of the respective
and compression respectively. beams. The other material properties are calculated as
All beams in this paper have also been analysed with a function of the cube strength with equations from
the program Response developed by Bentz.6 The pro- Model Code 9017 and Vos.18 To compare the FE pro-
gram is based on the modified compression field grams all input data have been made as similar as
theory developed by Vecchio and Collins.7,8 The differ- possible. All the numerical analyses are based on the
ence between MCFT and its preceding model compres- equations presented in Table 1. In all equations the
sion field theory,16 is that in MCFT the tension strength should be inserted in MPa and the units of
stiffening effect is included. MCFT is a smeared crack the obtained parameters are presented in the table. The
model with a rotated crack direction. The concrete uniaxial material curves used as input in the numerical
member will be capable of transmitting shear and com- analyses are illustrated in Fig. 2. The definition of the
pression at the crack surface owing to the aggregate transition from micro-cracks to macro-cracks in FE

Table 1. Equations for calculating the concrete material parameters

Parameter Equation Units

pffiffiffiffiffiffiffi
Initial elastic modulus Ec ¼ (6000
15:5 f cu ) f cu MPa
Poisson’s ratio í ¼ 0:2
Compressive cylinder strength f cc ¼ 0:85 f cu MPa
Strain at fc åc0 ¼ 2 f cc =Ec
Plastic softening compression wd ¼ 5 3 10
4 m
Tensile strength f ct ¼ 0:24 f cu 2=3 MPa
Fracture energy Gf ¼ 25 f ct Nm/m2
Crack opening displacement wc ¼ 5:14Gf = f ct m

σ Tension ⫺σ
Compression
fct fcc

Ec Gf

Ec
w ⫺ε
wt0 wt ⫽ wt0 ⫹ wc εc0 wd
εc ⫽ εc0 ⫹
h
(a) (b)

Fig. 2. Uniaxial material curves for (a) tension and (b) compression

Magazine of Concrete Research, 2008, 60, No. 5 383

Malm and Holmgren
700
analyses is performed by introducing a crack opening
law, which is often referred to as tension softening. The 600
crack opening law can normally be defined in terms of
fracture energy or a stress–strain or stress–displace- 500
ment relationship. The fracture energy Gf is a material

Load: kN
parameter that describes the amount of energy that is 400

needed to open a unit area of a crack to obtain a stress-

300
free crack. In these analyses the exponential function Measured curve
derived by Cornelissen et al.19 is used to describe the 200
Ultimate load
Extrapolated deflection
tensile behaviour during the fracture process zone. The Atena SBETA fixed crack
ascending branch in tension is assumed to be linear, 100 Atena SBETA rotated crack
Abaqus damaged plasticity
and in compression it is assumed to be a second-degree Response
parabola. 0
0 2 4 6 8 10
The strain at strain hardening and the rupture strain Displacement: mm
of the reinforcement were not measured in this experi-
Fig. 3. Load–displacement curves for beam 7
ment; therefore these have been estimated with results
for the same reinforcement types tested at the same
laboratory and at the same time period presented in
Nylander et al.20 Atena obtained a stiffer load–displacement response
The experimental loading procedure has two effects compared to the test results. In calculations of all the
when comparing the experimental results with the beams, except beams 3 and 10, conservative results
numerical analyses. First, the unloading part after the regarding the ultimate load capacity were obtained.
peak load could never be recorded. Second, while a The mean values of the ratio of the calculated di-
displacement-controlled loading system would result in vided by the experimental ultimate load were 0.898
a drop in the load–displacement curve after initiation of and 0.949 for the fixed and the rotated models re-
a crack, a continuous loading system would have a crack spectively. The standard deviation was 0.092 and
plateau. In the experiments there were discrete readings 0.084 for the fixed and the rotated models respec-
of the load level. These will probably not show the tively.
maximum load and therefore the displacement at failure The results obtained with Response were very similar
is not available from the experimental result. The static to the results from Atena. For most beams, the analyses
FE analyses in Atena have been performed as deforma- gave load–displacement curves that coincided well
tion controlled, with a forced displacement introduced with those obtained from Atena. The ultimate load
in the loading plate. The FE analyses in Abaqus have capacity calculated with Response was mainly slightly
been made with Abaqus/Explicit, which is a dynamic higher than that obtained with Atena and overestimated
analysis method. In this case a static solution is of the load capacity compared to the experiment for six of
interest, hence care must be taken that the beam is the ten beams. The mean value of the ratio of the
loaded slowly enough to eliminate significant inertia calculated load with Response and the ultimate load
effects. For problems involving brittle failure, this is obtained from the tests was 0.903 with a standard
especially important since the sudden drop in load- deviation of 0.291. The reason for the high value of the
carrying capacity that normally accompanies the brittle standard deviation obtained from the Response analyses
failure generally leads to increases in the kinetic energy is that the estimates for the two beams without web
content of the response. Therefore, all beams analysed reinforcement—beams 1 and 5—are rather poor. This
with Abaqus/Explicit are loaded with a small velocity, can be seen in Fig. 4: the results for two of the beams
which keeps the kinetic energy low and thereby ensuring with Response ended prematurely at the crack plateau,
a quasi-static solution. The member response analyses approximately at 40% of the measured ultimate load.
performed with Response can only be made as load- The calculations yield failure at the load that corre-
controlled, which means that a drop in the load-carrying sponds to the change in inclination of the load and
capacity will never be obtained. deflection curves for these beams. This may be a result
The load–displacement responses obtained from of the calculations being based on a load-controlled
the tests and the numerical analyses are illustrated for calculation and the convergence criteria being too
beam 7 in Fig. 3. The trends that can be seen in this strict. One other reason could be that the program or
figure clearly illustrate the trends of the numerical the material model is not suitable for calculating the
analyses of all the analysed beams. The calculations behaviour of unreinforced deep beams or unreinforced
with the fixed or rotated crack model in Atena beams in general with a low shear span-to-depth ratio.
showed only a small difference in the load and defor- This problem in Response has also been observed in
mation curve. The rotated model consequently gave a previous research by Malm21 where a beam experimen-
slightly higher load-carrying capacity than the fixed tally tested by Leonhardt and Walter22 was evaluated.
crack model in Atena. The models analysed with This beam was subjected to four-point bending and had
384 Magazine of Concrete Research, 2008, 60, No. 5
 Cracking in deep beams owing to shear loading. Part 2: Non-linear analysis
1·0
and Abaqus is illustrated respectively. In the sketch of
0·8
the experimental crack pattern, the crack resulting in
Deviation in load Vcalc /Vexp ⫺ 1

0·6 failure is illustrated with a thicker line width. The

0·4 rotated crack model in Atena gives a larger number of
cracks compared to the corresponding Atena model
0·2
with fixed crack direction. The physical crack is repre-
0 sented by several parallel cracks in the Atena model
⫺0·2 with the rotated crack direction. In the Response model
⫺0·4 the left support has been defined as ‘column top’ and
therefore the illustration of that support is shown as a
⫺0·6 Atena SBETA fixed crack
Atena SBETA rotated crack fixed support. The crack pattern obtained from the
⫺0·8 Abaqus damaged plasticity Abaqus model is visualised as the elements damaged in
Response
⫺1 tension where elements in white are undamaged and
⫺1 ⫺0·5 0 0·5 1
Deviation in deflection wcalc /wexp ⫺ 1
elements in black are fully damaged. Generally, all
models give a good representation of the crack pattern,
Fig. 4. Deviations in load and deflection of the numerical but the best representation is obtained with the fixed
analyses compared to the experiments crack model in Atena.
The experimentally observed crack widths have been
compared with the crack widths obtained from the FE
analyses. In Atena, the crack width is calculated for an
a rectangular cross-section with two flexural reinforce- element by multiplying the crack opening strain after
ment bars and a shear span-to-depth ratio of 2. complete stress release with the characteristic length of
The calculations with the damaged plasticity model the element. In the results from Abaqus the crack width
in Abaqus often yielded results that were softer in the has been calculated afterwards in the post-processor in
load–displacement response than the experiments. The the same way as in Atena. The crack width is calcu-
results are conservative for all analysed beams made lated in the same way in Response, where the mean
with Abaqus, except for beams 4 and 7. For these two principal strain is multiplied with the crack spacing. In
beams there is only a slight overestimation of the load Figs 6 and 7 the load and crack width curves are shown
capacity compared to the tests. The mean value of the for beam 5 and beam 9 respectively, where beam 5 is
ratio between the calculated and the experimental ulti- without web reinforcement and beam 9 is the beam
mate load was 0.906 with the standard deviation 0.094. with the highest ratio of web reinforcement. The results
Figure 4 illustrates the deviations of the predicted obtained with the rotated crack model show much
ultimate load and corresponding deflection obtained smaller crack widths than those obtained with the fixed
with the three programs compared to the tests. The crack model and compared to the experiments for the
experimental deflection at the ultimate load has here beams with the lower amount of web reinforcement.
been extrapolated from the load step prior to failure. In For the beams with higher web reinforcement ratio the
this figure it can be seen that the results from the fixed crack widths obtained with these two models gave
and the rotated model in Atena and the results from similar results, but not very close to the measured crack
Response are concentrated in an area left of the origin, width. The crack widths obtained from Response were
that corresponds to a too stiff prediction of the result. similar to the results obtained from the rotated crack
The results from Abaqus are almost centred in the model in Atena where the crack widths were generally
horizontal direction, which implies that the predictions much smaller than those obtained in the experiments.
of the stiffness correspond well with the results from The crack widths calculated from the plastic strains
the experiments. All analyses had similar mean accu- obtained from Abaqus generally showed the best agree-
racy in the prediction of the ultimate load as mentioned ment with the crack widths measured in the experi-
above. The best agreement considering both the experi- ments. This is because the Abaqus models showed
mental load and the corresponding deflection was de- better agreement with the experimental deflections than
termined with the least square method and the most Atena and the Response models.
accurate result was obtained with the analyses per- In the rotated crack models the physical crack was
formed with the Abaqus damaged plasticity model. often represented by several smaller cracks in adja-
The crack patterns obtained from experiments and cent elements. Fig. 8 is a typical illustration of the
analyses of beam 5 at the ultimate load are illustrated distribution of crack opening displacement in adja-
in Fig. 5. In Fig 5(a) and (b) a sketch of the observed cent elements that cross the crack path. To obtain the
crack pattern from the experiments of the left and right best possible estimation of the crack width, the
side of the beam respectively is presented and in Figs widths of all the closely aligned parallel cracks
5(c) and (d) the results from Atena with the rotated and should be added since all of them describe one
the fixed crack models are shown respectively. In Figs physical crack that illustrates the crack pattern, as
5(e) and (f) the crack pattern obtained from Response seen in the figure.
Magazine of Concrete Research, 2008, 60, No. 5 385
Malm and Holmgren

(a) (b)

(c) (d)

(e) (f)

Fig. 5. Crack patterns from experiments and the numerical analyses at failure of beam 5: (a) experimental left end;
(b) experimental right end; (c) Atena SBETA rotated crack; (d) Atena SBETA fixed crack; (e) Response; ( f ) Abaqus damaged
plasticity

Discussion
tained with the FE models corresponded well with the
The analyses with the different numerical programs crack pattern obtained in the experiments.
all showed good agreement with the measured ultimate The performed analyses also showed that crack
load. One exception was the beams without web rein- widths may be calculated with FE analyses with reason-
forcement that could not be analysed accurately with able accuracy. Since all the studied FE material models
Response. The fixed and the rotated models in Atena are continuum based, there will not be one discrete
and the Response model gave deflections that were crack. Instead the crack will be represented by inelas-
smaller than those experimentally obtained, resulting in tic/plastic strains within the elements. From the per-
too stiff behaviour. As expected, the FE analyses formed analyses it can be seen that the material model
showed better agreement regarding the ultimate load based on a rotating crack approach gave crack widths
than the strut-and-tie and truss models presented in that were several times smaller than those obtained in
Malm and Holmgren.3 When compared with results of the experiments or obtained with the plasticity-based
the method developed by Zhang and Tan,23 however, approach and the fixed crack approach for beams with
the difference was rather small. The crack pattern ob- a smaller amount of reinforcement. Despite this, the
386 Magazine of Concrete Research, 2008, 60, No. 5
 Cracking in deep beams owing to shear loading. Part 2: Non-linear analysis
350
rotated models gave rather good estimations of the load
300
and deflection response. This is because there are often
several closely spaced cracks in the rotated crack mod-
250 els and the representation of the physical crack thereby
becomes distributed over several minor cracks. In shear
200 cracks there can be some elements in a row perpendi-
Load: kN

cular to the direction of the crack that will be affected.

150 The crack widths from these elements should therefore
be summarised if the crack widths in these elements
100
Experimental
evidently represent one physical crack. If the crack
Atena SBETA fixed crack opening displacement is summed up over these close
50 Atena SBETA rotated crack and parallel cracks, the result from the rotated crack
Abaqus damaged plasticity
Response model instead shows good agreement with the experi-
0
0 1 2 3 4 5 mentally observed crack widths. For the analyses made
Crack width/COD: mm in Abaqus and in Atena with the fixed crack model it
was basically one element that exhibited a large major-
Fig. 6. Load and crack width (crack opening displacement ity of the crack opening strain, in these analyses usually
(COD)) for the experiments and numerical analyses of beam 5 90%. The other elements in the same row perpendicular
to the crack direction therefore gave rather negligible
crack widths. In the rotated crack model in Atena the
crack width obtained in a few elements in the same row
700 perpendicular to the crack direction was almost of the
same size as that seen in Fig. 8. This effect appeared in
600
all models, regardless of the mesh size.
One big advantage with Atena is that for the analyses
500
in this paper the program had no problems finding
Load: kN

400 convergence to describe the response of the concrete

beams, not even after that the ultimate load was
300 reached. This is usually a problem in static analyses.
Even for tests with a rather coarse mesh, the diagonal
200 Experimental shear crack could successfully be captured. In the
Atena SBETA fixed crack
Atena SBETA rotated crack
quasi-static analyses performed with Abaqus, finding a
100
Abaqus damaged plasticity converging solution did not prove problematic either. If
Response these analyses had been performed as static, conver-
0
0 0·5 1 1·5 2 gence difficulties would have ended several calcula-
Crack width/COD: mm
tions prematurely unless damping or some other
Fig. 7. Load and crack width (crack opening displacement technique to prevent numerical instability was intro-
(COD)) for the experiments and numerical analyses of beam 9 duced.
COD at the elements

COD at the elements

Rotated crack model Fixed crack model

Elements Elements

Fig. 8. Sketch of the distribution of crack opening displacement in adjacent elements in models with rotated and fixed crack
model respectively
Magazine of Concrete Research, 2008, 60, No. 5 387
Malm and Holmgren

References 13. Bažant Z. and Oh B. Crack band theory for fracture of con-
crete. Materials and Structures, RILEM, 1983, 16, No. 93, 155–
1. Kwak H. G. and Kim D. Y. FE analysis of RC shear walls 177.
subject to monotonic loading. Magazine of Concrete Research, 14. Lubliner J., Oliver J., Oller S. and Onate E. A plastic-
2004, 57, No. 7, 387–403. damage model for concrete. International Journal of Solids and
2. Nylander H. and Holmgren J. Upplagshållfasthet vid hög Structures, 1989, 25, No. 3, 299–326.
balk förbunden med bjälklagsplatta. Royal Institute of Technol- 15. Lee J. and Fenves G. Plastic-damage model for cyclic loading
ogy, Stockholm, Sweden, 1975, Technical report 113, pp. 1–21. of concrete structures. Journal of Engineering Mechanics,
3. Malm R. and Holmgren J. Cracking in deep beams owing to ASCE, 1998, 124, No. 8, 892–900.
shear loading. Part 1: Experimental study and assessment. Ma- 16. Mitchell D. and Collins M. Diagonal compression field
gazine of Concrete Research, 2008, 60, No. 5, 371–379. theory—a rational model for structural concrete in pure torsion.
4. Abaqus/Explicit 6.6-1. Theory manual. Simulia, Providence, RI, Journal of the American Concrete Institute, 1974, 71, No. 8,
USA, 2006. 396–408.
5. Cervenka V., Jendele L. and Cervenka J. Atena Program 17. Comité Euro-International du Béton. CEB-FIP Model
Documentation. Part 1: Theory. Cervenka Consulting, Prague, Code 1990. Design Code, 6th edn. Thomas Telford, London,
Czech Republic, 2005. 1993.
6. Bentz E. Response 2000: 2D sectional analysis of beams and 18. Vos E. Influence of Loading Rate and Radial Pressure on Bond
columns, 2000. See www.ecf.utoronto.ca/~bentz/r2k.htm (last in Reinforced Concrete. PhD thesis, Delft University of Technol-
accessed April 2008). ogy, the Netherlands, 1983.
7. Vecchio F. and Collins M. The modified compression-field 19. Cornelissen H., Hordijk D. and Reinhardt H. Experimental
theory for reinforced concrete elements subjected to shear. Jour- Determination of Crack Softening Characteristics of Normal
nal of the American Concrete Institute, 1986, 83, No. 2, 219– Weight and Lightweight Concrete. Heron, Delft, the Netherlands,
231. 1986, 31, No. 2.
8. Vecchio F. and Collins M. Predicting the response of rein- 20. Nylander H., Kinnunen H. and Andersson J. Armerad be-
forced concrete beams subjected to shear using modified com- tong. Royal Institute of Technology, Stockholm, Sweden, 1971.
pression field theory. ACI Structural Journal, 1988, 85, No. 3, Report.
258–268. 21. Malm R. Shear Cracks in Concrete Structures Subjected to In-
9. Petersson P. E. Crack Growth and Development of Fracture plane Stresses. Licentiate thesis, Royal Institute of Technology
Zones in Plain Concrete and Similar Materials. Division of (KTH), Sweden, 2006.
Building Materials, University of Lund, Sweden, 1981, Report 22. Leonhardt F. and Walter R. Schubversuche an einfeldrigen
No. tvbm-1006. Stahlbetonbalken mit und ohne Schubbewehrung. Deutscher
10. Rots J., Nauta P., Kusters G. and Blaauwendraad J. Ausschuss Für Stahlbeton, Berlin, Germany, 1962, No. 151, pp.
Smeared Crack Approach and Fracture Localization. Heron, 1–83.
Delft, the Netherlands, 1985, Vol. 30, No. 1. 23. Zhang N. and Tan K.-H. Direct strut-and-tie model for single
11. de Borst R. Non-linear Analysis of Frictional Materials. PhD span and continuous deep beams. Engineering Structures, 2007,
thesis, Delft University of Technology, the Netherlands, 1986. 29, No. 11, 2987–3001.
12. Jendele L., Cervenka J., Saouma V. and Pukl R. On the
choice between discrete or smeared approach in practical struc-
tural FE analyses of concrete structures. Proceedings of the 4th
International Conference on Analysis of Discontinuous Defor- Discussion contributions on this paper should reach the editor by
mation, Glasgow, UK, 2001. 1 December 2008

388 Magazine of Concrete Research, 2008, 60, No. 5