Professional Documents
Culture Documents
Birtel PDF
Birtel PDF
V. Birtel, P. Mark
Ruhr-University Bochum, Universitätsstr. 150, 44780 Bochum, Germany
Institute for Reinforced and Prestressed Concrete Structures
Abstract: A spatial finite element model of reinforced concrete (RC) beams with rectangular cross
sections, typical side aligned stirrups and distributed or edge concentrated longitudinal
reinforcement is presented. It is parameterised in its properties of geometry, material,
discretisation and loads in biaxial directions. The concrete volume is discretisised into 8 or 20-
node solid elements. Truss elements discretely model each single reinforcement bar. They are
coupled to the concrete elements using the ″embedded modelling″ technique. The ″concrete
damage plasticity″ model of ABAQUS is used to describe the nonlinear material behaviour of
concrete. Suitable material functions and material parameters are derived and verified to
experimental data of (cyclic) uniaxial, biaxial or triaxial stress tests. Energy criteria and internal
length parameters ensure almost mesh independent results of the simulations. An elasto-plastic
material model with a gradually rising plastic branch is adopted for the reinforcing steel. The
parametric model is verified to experimental data of uniaxial shear tests taken from the literature.
Afterwards, it is used to establish a data base of biaxial shear resistances to check developed
biaxial shear design formulas that base on simple strut and tie models. More than 100 simulations
guarantee an extended and reliable verification that experiments – almost none of them are
available in the literature – are not able to give. Moreover, the arrangement of the stirrups is
optimised in dependence upon the distribution of the longitudinal reinforcement to minimise
reinforcement amounts and increase bearing capacities.
Keywords: Reinforced concrete, shear failure, parameterisation, embedded modelling
1. Introduction
The shear design of reinforced concrete beams bases on strut and tie models, first developed by
(Ritter, 1899) and (Mörsch, 1927), respectively. They realised that the complex inner states of
stresses can be idealised by tensile and compressive struts. Today, almost every design code – e. g.
Eurocode 2 – uses this basic principle. However, loads are assumed to act along one principle axis
of the cross section. There are no design rules for biaxial loadings with inclined resultant shear
forces that do not comply with vertical or horizontal directions.
Formulas for the resistances of the tensile and the compressive shear struts of beams under biaxial
shear forces have been developed by the coauthor (Mark, 2004; Mark 2005). They are valid for
beams with rectangular cross sections, typical side aligned stirrups, normal strength concrete and
arbitrary distributions of the longitudinal reinforcement. Moreover, there is no limitation for an
It is obvious that almost every variation in geometry, material properties as well as load or
reinforcement arrangement requires a complete revise of a finite element model. To avoid such
efforts, it is convenient to generate different similar models from just one parametric input file.
Then, the user has to specify only a few variable parameters (Table 1). Even the mesh generation
is included in the parameterisation.
The spatial finite element model idealises three-point-bending tests of RC beams under biaxial
loadings (Figure 2). Its boundary conditions and force applications are predefined and the
symmetries of geometry and load are utilised to halve the model structure and thus save
computing times.
Vy h (1)
αV =
Vz b
Vertical forces Vres = Vz a referred to by αV = 0, while a diagonally directed shear force Vres yields
αV = 1. It is indispensable to hold 0 ≤ αV ≤ 1, otherwise the notations of Vy, Vz, h and b have to be
exchanged.
″Compression only″ springs assure a numerically stable changeover of forces from the solid beam
to the steel supports. They also exclude unrealistic tensile bearing reactions. The concrete body is
discretised in nb x nh elements in the section plane and nl +8 (4) elements in the longitudinal
direction. They represent the ″host elements″ in the applied concept of ″embedded elements″. 8-
node C3D8 or 20-node C3D20 solid elements with linear or quadratic interpolation functions are
chosen from the ABAQUS element library. T3D2 truss elements – lying embedded in the concrete
The nonlinear set of equations is solved with the modified ″Static Riks″ arc-length method
(ABAQUS, 2003). Here, the method often achieved its best effectiveness with a limitation of the
arc-length increment to ∆lmin = 10-10 and no upper threshold value for ∆lmax.
3. Material models
The complex, nonlinear material behaviour of concrete is described by the elasto-plastic damage
model ″concrete damaged plasticity″ (ABAQUS, 2003) that was developed by (Lubliner et al.,
1989) and elaborated by (Lee & Fenves,1998). It uses a yield surface F in the space of effective
stresses σ of combined Drucker-Prager and Rankine type and assumes isotropic damage d as well
as non-associated flow. Its basic equations read:
{ }
σ = D 0 (ε − ε pl ) ∈ σ F (σ , ~ε pl ) ≤ 0 , σ = (1 − d (σ , ~ε pl ))σ (2)
~ε& pl = h(σ , ~ε pl ) ε&ˆ pl
∂G ( σ )
ε& pl = λ&
∂σ
A bilinear relation models the uniaxial stress-strain behaviour of the reinforcing steel. It is
characterised by the modulus of elasticity Es, the yield strength fy and the gradual rise of the
second branch (Es1).
The ″concrete damaged plasticity″ model requires the following material functions:
stress-strain relations for the uniaxial behaviours under compressive as well as tensile
loadings including cyclic un- and reloading,
functions for the evolutions of the damage variables dc and dt under compressive and tensile
loadings, respectively.
The stress-strain behaviour under sustained compressive loading is modelled in three phases
(Figure 3).
The first two sections describe the ascending branch up to the peak load fcm at εc1. Their
formulations are similar to the recommendations of the Model Code (CEB-FIB, 1993). The third
and descending branch takes account for its dependency on the specimen geometry (Vonk,1993;
Van Mier, 1984) to ensure almost mesh independent simulation results. Thus, σc(3) incorporates
within the descent function γc the constant crushing energy Gcl (Krätzig & Pölling, 2004) as a
material property in addition to an internal length parameter lc derived from the grid structure of
the element mesh.
σc εc1
σc [MPa]
Sinha, Gerstle
fcm loading path & Tulin (1964)
20 bc = 0,7
model
ε cpl = bcε cin
Ec (1 − d c ) 10
0,4 fcm bc = 0,3
Ec
0
ε cin σ c Ec−1 εc 0 2 4 6 8
ε pl
ε el
εc [‰]
c c
The evolution of the compressive damage component dc is linked to the corresponding plastic
strain εcpl which is determined proportional to the inelastic strain εcin = εc - σcEc-1 using a constant
factor bc with 0 < bc ≤ 1.
σ c Ec−1 (6)
dc = 1 −
ε (1 / bc − 1) + σ c Ec−1
c
pl
A value bc = 0,7 fits well with experimental data of cyclic tests (Figure 3, right). So, most of the
inelastic compressive strain maintains after unloading. Generally, unloading and subsequent
reloading up to the monotonic path occur linearly with no hysteretic loops.
using the principle of the ″Fictitious Crack Model″ (Hillerborg, 1983). Thus, a product of the
inelastic strain and an internal length parameter lt replaces the crack opening w to yield σt = σt(w =
ltεtin = lt(εt - σtEc-1)) and w is smeared over the average element length lt = Ve⅓. As intended, σt(εt)
then encloses the ratio of fracture energy GF and lt (Bazant & Oh,1983).
σ t / f ct σt [MPa]
1 fctm = 2,56 MPa, lt = 25 mm
wc = 180µm
σt(w) acc. Hordijk d = 16mm wc = 180 µm, dmax = 16 mm
max 2
(1992) Reinhardt, Cornelissen (1984)
≈ C30/37
½ experimental data bt = 0,1
Reinhardt, Cornelissen 1 σt(εt)
MC 90 (1984)
0
0
εt [‰]
0 40 80 120 0 2,4 4,8
w [µm]
Figure 4. Stress-crack opening and stress-strain relations for (cyclic) tensile
loading, experiments acc. (Reinhardt & Cornelissen, 1984).
Similar to (6) the damage dt depends on εtpl and an experimentally determined parameter bt = 0,1
(Figure 4, right). So, unloading is assumed to return almost back to the origin and to leave only a
small residual strain.
σ t Ec−1 (8)
dt = 1 −
ε tpl (1 / bt − 1) + σ t Ec−1
Damage d isotropically reduces the initial elastic stiffness parameters in D0 to gain D = (1-d)D0. It
arises from damage partitions associated with tensile and compressive loadings (dt and dc),
respectively.
1 − d = (1 − sc dt )(1 − st d c ) (9)
s = 1 − ½ r (σˆ ) , s = r (σˆ )
t c
σt
q/fc experiment
simulation
(4,7 | 5,5)
3
Calculated uniaxial, biaxial and triaxial strength results of normal strength concrete agree well
with experimental data taken from the literature (Kupfer & Gerstle,1973; Linse & Aschl,1976;
Van Mier,1984) (Figure 6). This holds true for tensile, compressive as well as combinations of
tensile and compressive loadings with low confining pressures. If high confinements occur,
resistances are often overestimated by the material model (cp. (ABAQUS, 2003)). The three
examples in Figure 6 indicate that these overestimations can have pronounced extents. But the
ratios of hydrostatic pressures p and the von Mises equivalent stresses q still fit well to the
experimental ones.
Table 2 summarises the material parameters of the concrete model. It distinguishes between
parameters related to the uniaxial and to the multiaxial behaviour. Strength and stiffness
parameters are taken from European Standards. But of course, this choice is variable.
The following parameters are adopted for the reinforcing steel of stirrups and longitudinal bars:
Es = 200.000, Es1 = 1111, fy = 500 (all values given in [MPa]).
The parametric model is verified to the data of two well documented uniaxial shear experiments
(αV = 0), both currently carried out with a three-point-bending test set-up, rectangular cross
sections and normal strength concretes. The first was accomplished by the coauthor (Mark, 2004),
the second by (Toongoenthong & Maekawa, 2005).
Figure 7 shows specimen details and reinforcement arrangements of the first test by (Mark, 2004).
All dimensions are given in [mm]. The girder spans over leff = 3000 mm. For the sake of
simplicity, standard strength values according to Eurocode 2 were adopted for concrete and steel
during the numerical simulations, although the actual test values slightly differed: compressive
concrete strength 38 MPa (test 35,8 MPa), steel yield strength 500 MPa (test 570 MPa).
The global load-displacement relation of the test is well met by the simulations, as the element
grids are sufficiently fine. This applies to both isoparametric solid element types with linear or
quadratic interpolation functions. Thus, the linear element is selected for further investigations to
limit computational efforts that rapidly increase, when using the 20-node solid. Initially, the
stiffness of the girder is overestimated. However, the deviations of about 5% in the peak loads are
mainly due to the assumed reduced yield strength of stirrups and longitudinal bars. They almost
vanish, if fy is corrected to its actual experimental value.
Figure 8 shows similar simulation results for the recalculation of the test data of (Toongoenthong
& Maekawa, 2005), where leff = 2000 mm. Peak load and even the descending branch acceptably
agree with the experimental ones, if the element mesh does not get too course. The overestimation
of initial stiffness properties remains.
Figure 9 proofs that the numerical model is also capable to describe local test results like cracking,
the formation of shear struts or the redistribution of forces onto the vertical stirrup legs, at least in
Generally, the FE model is appropriate to describe the shear failure mechanism of RC beams close
to reality. This includes the typical stirrup yielding, the branched cracking and the redistributions
of tensile stresses to stirrups and longitudinal bars. Furthermore, peak loads (or shear resistances)
only slightly differ from experimental ones. So, those numerically determined resistances are well
suited to reliably verify practical design formulas for shear resistances of RC beams, even in cases
of biaxial loadings.
6. Biaxial shear
For the verification of the biaxial shear design formulas more than 100 single simulations are
performed with the parametric model varying the basic parameters in ranges that typically occur in
practical applications. The variations cover concrete properties (24 ≤ fcm ≤ 58 MPa, normal
strength concrete), the aspect ratio (1 ≤ h/b ≤ 2,5), the shear load inclination (0 ≤ αV ≤ 1), the
mechanical reinforcement ratios of stirrups and longitudinal bars as well as different typical edge
concentrated or side aligned distributions of the longitudinal reinforcement. Of course, usual
detailing rules (ENV 1992-1-1, 1992) like minimum distances of stirrups and bars, minimum
Numerically and experimentally determined ratios fall into the same scatter range. So the design
formulas describe uniaxial and biaxial shear resistances with almost the same accuracy and the
well known decreasing relationship (Reineck, 2001) between Vsim/exp/VR3 and ωw – governed by the
″concrete partition″ Vc for small ωw – emerges. Furthermore, most of the ratios – especially the
ones derived from biaxial loadings – exceed unity, so they conservatively underestimate actual
resistances.
7. Conclusions
Spatial finite element models with concrete solids and embedded truss elements – modelling
stirrups and longitudinal bars – are very suitable to numerically simulate the load bearing
behaviour of RC beams. On the one hand, they reliably evaluate global parameters like ultimate
loads or deformations close to reality. This even holds true for complex loading conditions like
biaxial ones. On the other hand, they allow extended variations of basic parameters that
experiments – due to their demand on time and costs – are not able to give. Parametric input files
reduce the user’s effort for such variations to a minimum.
Moreover, simulations open the view to the inside of RC girders. Here, the discrete modelling of
stirrups and longitudinal bars is especially appropriate, as then their results can be separately
8. References
9. Acknowledgment
The authors thank the German Research Foundation (Deutsche Forschungs-Gemeinschaft DFG,
http://www.dfg.de) for the financial support of the project "Experimental and numerical
investigations of reinforced concrete girders under biaxial shear forces".