Professional Documents
Culture Documents
Most Imp PDF
Most Imp PDF
Because of the properties of the material (concrete), computer simulations in the field of reinforced
concrete structures are pose a challenge. As opposed to steel, concrete when subjected to compression ex-
hibits nonlinearity right from the start. Moreover, it much quicker undergoes degradation under tension.
All this poses difficulties for numerical analyses. Parameters needed to correctly model concrete under
compound stress are described in this paper. The parameters are illustrated using the Concrete Damaged
Plasticity model included in the ABAQUS software.
Keywords: numerical modelling, concrete degradation, stress-strain relation, reinforced concrete struc-
tures, composite structures, Abaqus, concrete damaged plasticity
1. Introduction
The two main concrete failure mechanisms are cracking under tension and crushing
under compression. However, concrete strength determined in simple states of stress (uni-
axial compression or tension) radically differs from the one determined in complex states
of stress. For example, the same concrete under biaxial compression reaches strength of
between ten and twenty per cent higher than in the uniaxial state while in the hydro-static
state (uniform triaxial compression) its strength is theoretically unlimited. In order to de-
scribe strength with the equation for triaxial stress, its plane should be presented in a three-
dimensional stress space (since concrete is considered to be an isotropic material in a wide
range of stress). The states of stress corresponding to material failure are on this surface
while the states of safe behaviour are inside. Also the so-called plastic potential surface is
located inside this space. After the plasticity surface is crossed, two situations arise [9]:
• an increase in strain with no change in stress (ideal plasticity),
• material weakening (rupture).
energy and the boundary surface itself in the stress space assumes the shape of a cone.
The advantage of the use of this criterion is surface smoothness and thereby no com-
plications in numerical applications. The drawback is that it is not fully consistent with
the actual behaviour of concrete 0.
The CDP (Concrete Damaged Plasticity) model used in the ABAQUS software is
a modification of the Drucker–Prager strength hypothesis. In recent years the latter has
been further modified by Lubliner 0, Lee and Fenves 0. According to the modifi-
cations, the failure surface in the deviatoric cross section needs not to be a circle and it
is governed by parameter K c.
Another parameter describing the state of the material is the point in which the
concrete undergoes failure under biaxial compression. σb0/σc0 ( fb0 / fc0) is a ratio of the
strength in the biaxial state to the strength in the uniaxial state. The most reliable in
this regard are the experimental results reported by Kupler (1969). After their approxi-
mation with the elliptic equation, uniform biaxial compression strength fcc is equal to
1.16248 fc 0. The ABAQUS user’s manual specifies default σb0/σc0 = 1.16.
The last parameter characterizing the performance of concrete under compound
stress is dilation angle, i.e. the angle of inclination of the failure surface towards the
626 P. KMIECIK, M. KAMIŃSKI
hydrostatic axis, measured in the meridional plane. Physically, dilation angle ψ is in-
terpreted as a concrete internal friction angle. In simulations usually ψ = 36° 0,0 or ψ
= 40° 0 is assumed.
The unquestionable advantage of the CDP model is the fact that it is based on pa-
rameters having an explicit physical interpretation. The exact role of the above pa-
rameters and the mathematical methods used to describe the development of the bound-
ary surface in the three-dimensional space of stresses are explained in the ABAQUS
user’s manual. The other parameters describing the performance of concrete are deter-
mined for uniaxial stress. Table 1 shows the model’s parameters characterizing its per-
formance under compound stress.
Modelling of reinforced concrete structures and composite structures... 627
σc
ε 0elc = . (2)
E0
When transforming strains, one should consider from what moment the material
should be defined as nonlinearly elastic. Although uniaxial tests show that such be-
haviour occurs almost from the beginning of the compression process, for most nu-
merical analyses it can be neglected in the initial stage. According to Majewski, a lin-
ear elasticity limit should increase with concrete strength and it should be rather as-
sumed than experimentally determined. He calculated it as a percentage of stress to
concrete strength from this formula:
⎛− f ⎞
elim = 1 − exp⎜ c ⎟. (3)
⎝ 80 ⎠
628 P. KMIECIK, M. KAMIŃSKI
This ceiling can be simply arbitrarily assumed as 0.4 fcm. Eurocode 2 specifies the
modulus of elasticity for concrete to be secant in a range of 0–0.4 fcm. Since the basic
definition of the material already covers the shear modulus and the longitudinal
modulus of concrete, at this stage it is good to assume such an inelastic phase thresh-
old that the initial value of Young’s modulus and the secant value determined accord-
ing to the standard will be convergent. In most numerical analyses it is rather not the
initial behaviour of the material, but the stage in which it reaches its yield strength
which is investigated. Thanks to the level of 0.4·fcm there are fewer problems with so-
lution convergence.
Having defined the yield stress-inelastic strain pair of variables, one needs to de-
fine now degradation variable dc. It ranges from zero for an undamaged material to
one for the total loss of load-bearing capacity. These values can also be obtained from
uniaxial compression tests, by calculating the ratio of the stress for the declining part
of the curve to the compressive strength of the concrete. Thanks to the above defini-
tion the CDP model allows one to calculate plastic strain from the formula:
dc σ c
ε~cpl = ε~cin − , (4)
(1 − d c ) E0
where E0 stands for the initial modulus of elasticity for the undamaged material. Know-
ing the plastic strain and having determined the flow and failure surface area one can
calculate stress σc for uniaxial compression and its effective stress σ c .
(
σ c = (1 − d c )E0 ε c − ε~cpl , ) (5)
σc =
σc
(1 − d c )
(
= E0 ε c − ε~cpl . ) (6)
On the basis of uniaxial compression test results one can accurately determine the
way in which the material behaved. However, a problem arises when the person run-
ning such a numerical simulation has no such test results or when the analysis is per-
formed for a new structure. Then often the only available quantity is the average com-
pressive strength ( fcm) of the concrete. Another quantity which must be known in order
to begin an analysis of the stress-strain curve is the longitudinal modulus of elasticity
(Ecm) of the concrete. Its value can be calculated using the relations available in the lit-
erature 0:
Ecm = 22(0.1 f cm ) ,
0.3
(7)
Modelling of reinforced concrete structures and composite structures... 629
where:
fcm [MPa],
Ecm [GPa].
Other values defining the location of characteristic points on the graph are strain εc1
at average compressive strength and ultimate strain εcu 0:
The formulas (8–9) are applicable to concretes of grade C50/60 at the most.
On the basis of experimental results Majewski proposed the following (quite accu-
rate) approximating formulas:
Knowing the values of the above one can determine the points which the graph should
intersect.
The curve can be also plotted on the basis of the literature 0, 0, 0,0,0. The most
popular formulas are presented in Table 2, but the original symbols have been re-
placed with the uniform denotations used in Eurocode 2.
630 P. KMIECIK, M. KAMIŃSKI
⎡ 1 ⎛ ε c ⎞⎤
Madrid parabola σ c = Ecε c ⎢1 − ⎜⎜ ⎟⎥ σ c = f (Ec , ε c1 )
⎣⎢ 2 ⎝ ε c1 ⎟⎠ ⎦⎥
Desay Ecε c
σc =
σ c = f (Ec , ε c1 )
2
& Krishnan ⎛ε ⎞
1 + ⎜⎜ c ⎟⎟
formula ⎝ ε c1 ⎠
kη − η 2
σ c = f cm
1 + (k − 2)η
EN 1992-1-1 σ c = f (Ecm , f cm , ε c1 )
ε c1 ε
k = 1.05Ecm , η= c
f cm ε c1
σ c = Ecε c if σ c ≤ elim f cm
(elim − 2)2 ⎛⎜ ε c ⎞⎟ +
2
⎫
σ c = f cm ⎪
4(elim − 1) ⎝ ε c1 ⎠
⎜ ⎟ ⎪
Majewski ⎬ f cm
(2 − elim ) , σ c = f (Ec , f cm , ε c1 )
formula
− f cm
(elim − 2 ) ⎛ ε c ⎞
2
⎜ ⎟ + f cm
elim
2
⎪ Ec = εc
2(elim − 1) ⎝ ε c1 ⎠
⎜ ⎟ 4(elim − 1) ⎪⎭
if σ c > elim f cm
elim in formula (3)
⎡ ⎛ ε ⎞ ⎛ ε ⎞2 ⎤ εc
σ c = ζ f cm ⎢2⎜⎜ c ⎟ ⎜
⎟ − ⎜ c ⎟⎟ ⎥ if ζε ≤ 1
⎢⎣ ⎝ ζε c1 ⎠ ⎝ ζε c1 ⎠ ⎥⎦ c1
Wang & Hsu σ c = f ( f cm , ε c1 )
formula ⎡ ⎛ ε / ζε − 1 ⎞ 2 ⎤ εc
σ c = ζ f cm ⎢1 − ⎜⎜ c c1
⎟⎟ ⎥ if ζε > 1
2 / ζ −1 ⎠ ⎥
⎣⎢ ⎝
c1
⎦
εc
σc = ⎛ E , f , f cu , ⎞
Sáenz formula A + Bε c + Cε c + Dε c
2 3
σ c = f ⎜⎜ c cm ⎟⎟
⎝ ε c1 , ε cu1 ⎠
symbols in formula (12)
Modelling of reinforced concrete structures and composite structures... 631
The 2nd order parabola has this property that the tangent of the angle of a tangent
passing through a point on its branch, measured relative to the horizontal axis passing
through this point, is always double that of the angle measured as the inclination of the
secant passing through the same point and the extremum of the parabola, relative to
the same horizontal axis.
Fig. 8. Relation σc– εc for Madrid parabola depending on longitudinal modulus of elasticity
The consequence of this property of the parabola is either the exceedance of the
concrete’s strength for a correct initial modulus value or the necessity to lower the
value in order to reach a specific stress value in the extreme. Figure 8 shows relation
σc – εc for the Madrid parabola for grade C16/20 concrete. The following batch deno-
tations were assumed:
• Ecm – Ec = Ecm = 28608 MPa was assumed as the initial modulus, calculated ex-
tremum fcm = 26.81 MPa;
• Ec/Ecu = 2 – the doubled tangent of the angle of the secant passing through point
(εc1, fcm), amounting to Ec = 25602 MPa, calculated extremum fcm = 24.00 MPa (correct);
632 P. KMIECIK, M. KAMIŃSKI
• 0.4 fcm – the value of initial modulus Ec = 31808 MPa matched so that the curve
intersects point (εc, 0.4·fcm), calculated extremum fcm = 29.81 MPa;
• Ec = Ecm – a straight line describing the elastic behaviour of the concrete up to
(εc, 0.4·fcm).
As one can see, when initial modulus Ec is assumed to amount to Ecm, the strength of
the concrete is much overrated despite the fact that the initial modulus is still underrated
(numerically Ecm is not the highest value). In the case of parabolic relations one should
artificially lower modulus Ec in order for the graph to intersect the correct value fcm.
A sufficiently detailed description of relation σc – εc has been proposed by Sáenz. The
function with a 3rd order polynomial in the denominator (Table 2) depends on the vari-
ables:
1 P +P −2 ⎫
A= , B= 3 4 ⎪
Ec P3 f cm
⎪
2 P4 − 1 P4 − 1 ⎪
C=− , D=
P3 f cmε c1 P3 f cmε c1 ⎪⎪
⎬. (12)
ε cu f cm ⎪
P1 = , P2 =
ε c1 f cu ⎪
⎪
Eε P (P − 1) 1
P3 = c c1 , P4 = 3 2 2 − ⎪
f cm (P1 − 1) P1 ⎪⎭
Fig. 9. Comparison of curves σc-εc based on table 2 relations for grade C16/20 concrete
Modelling of reinforced concrete structures and composite structures... 633
The above notation allows one to shape the function graph so that it intersects
points: (εc1, fcm) and (εcu, fcu). The relation proposed by Wang and Hsu is an interesting
notation. These are two functions describing the curve’s ascending and descending
part. They also include coefficient ζ representing the reduction in compressive stress
of concrete resulting from locating reinforcing bars in the compressed zone. In figure
9 ζ = 1.0 (no reinforcement taken into account). It is worth noticing that Wang and
Hsu relation, the Majewski relation and the Madrid parabola almost coincide. The
same applies to the Desay and Krishanan relation and the Sáenz relation, but in the
latter case the same point (εcu, fcu) which followed from the Desay and Krishanan for-
mula was assumed since a lower value of function fcu would result in an improper
shape of the curve. The standard relation yields intermediate results.
The term cracking strain ε~tck is used in CDP model numerical analyses. The aim is
to take into account the phenomenon called tension stiffening. Concrete under tension
is not regarded as a brittle-elastic body and such phenomena as aggregate interlocking
634 P. KMIECIK, M. KAMIŃSKI
in a crack and concrete-to-steel adhesion between cracks are taken into account. This
assumption is valid when the pattern of cracks is fuzzy. Then stress in the tensioned
zone does not decrease sharply but gradually. The strain after cracking is defined as the
difference between the total strain and the elastic strain for the undamaged material:
εt
ε 0elt = . (15)
Ec
Plastic strain ε~t pl is calculated similarly as in the case of compression after defin-
ing degradation parameter dt.
In order to plot curve σt – εt one should define the form of the weakening function.
According to the ABAQUS user’s manual, stress can be linearly reduced to zero, start-
ing from the moment of reaching the tensile strength for the total strain ten times
higher than at the moment of reaching fctm. But to accurately describe this function the
model needs to be calibrated with the results predicted for a specific analyzed case.
Fig. 11. Modified Wang & Hsu formula for weakening function at tension stiffening for concrete C16/20
The proper relation was proposed by, among others, Wang and Hsu [11]:
σ t = Ecε t if ε t ≤ ε cr ⎫
0.4 ⎪
⎛ε ⎞ ⎬, (16)
σ t = f cm ⎜⎜ cr ⎟⎟ if ε t > ε cr ⎪
⎝ εt ⎠ ⎭
Modelling of reinforced concrete structures and composite structures... 635
where εcr stands for strain at concrete cracking. Since tension stiffening may consid-
erably affect the results of the analysis and the relation needs calibrating for a given
simulation, it is proposed to use the modified Wang & Hsu formula for the weakening
function:
n
⎛ε ⎞
σ t = f cm ⎜⎜ cr ⎟⎟ if ε t > ε cr , (17)
⎝ εt ⎠
5. Conclusion
Problems with solution convergence may arise when full nonlinearity of the material
(concrete) with its gradual degradation under increasing (mainly tensile) stress is assumed.
Simple FE techniques, consisting in reducing the size of stress increment or increasing the
maximum number of steps when solving the problem by means of the Newton-Raphson
approach, may prove to be insufficient. Therefore the CDP model uses viscosity parameter
μ which allows one to slightly exceed the plastic potential surface area in certain suffi-
ciently small problem steps (in order to regularize the constitutive equations). Viscoplastic
adjustment consists in choosing such μ > 0 that the ratio of the problem time step to μ ap-
proaches infinity. This means that it is necessary to try to match the value of μ a few times
in order to find out how big an influence it has on the problem solution result in
ABAQUS/Standard and to choose a proper minimum value of this parameter.
The CDP model makes it possible to define concrete for all kinds of structures. It is
mainly intended for the analysis of reinforced concrete structures and concrete-con-
crete and steel-concrete composite structures. However, it is recommended that before
an analysis of the structure one should test the behaviour of the material itself, e.g. by
carrying out a numerical analysis of cylindrical samples under compression, in order
to compare it with the given stress-strain relation. Because of the character of concrete
failure, some quantities can be rather assumed than determined in laboratory tests.
Therefore the assumptions should be verified by comparing other parameters, e.g. the
deflection of the modelled structural component. This means that the model parame-
ters often need to be calibrated several times in the course of the numerical analysis.
Wrocław Centre for Networking and Supercomputing holds a licence for the Abaqus software
(http://www.wcss.wroc.pl), grant No. 56.
References
[1] ABAQUS: Abaqus analysis user's manual, Version 6.9, 2009, Dassault Systèmes.
[2] Eurocode 2: Design of concrete structures. Part 1-1: general rules and rules for buildings,
Brussels, 2004.
636 P. KMIECIK, M. KAMIŃSKI