Computational Methods in Reinforced Concrete

Computational Engineering for Reinforced Concrete
Structures
U. Huler-Combe
mailto:Ulrich.Haeussler-Combe@tu-dresden.de
Institute of Concrete Structures
http://www.tu-dresden.de/biwitb/mbau
Technische Universitt Dresden
http://www.tu-dresden.de
PRELIMINARY DRAFT
State
April 4, 2013
I
Introductory Remarks
The following notes
1
serve as accompanying study material for the lecture Computational
Engineering for Reinforced Concrete Structures of the Institute of Concrete Structures,
Technische Universitt Dresden.
The notes are not yet nished. Some existing chapters deserve a completion. More chap-
ters not yet included are under work. Thus, a Preliminary Draft is given.
The actual state has been worked out according to best knowledge. Nevertheless, it may
contain formal errors or wrong facts. Corresponding hints are highly appreciated by the
author (mailto:Ulrich.Haeussler-Combe@tu-dresden.de).
1
2013( Ulrich Huler-Combe. All rights preserved.
mailto:Ulrich.Haeussler-Combe@tu-dresden.de State April 4, 2013
Contents
1 Finite Elements in a Nutshell 1
1.1 Modeling Basics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Finite Element Basics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.3 Elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.4 Material Behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.5 Weak Equilibrium and Discretization . . . . . . . . . . . . . . . . . . . . . . . 7
1.6 Convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.7 Numerical Integration and Solution Methods for Algebraic Systems . . . . . . 14
2 Uniaxial Structural Concrete Behavior 20
2.1 Short Term Stress-Strain Behavior of Concrete . . . . . . . . . . . . . . . . . 20
2.2 Long Term Effects - Creep, Shrinkage and Temperature . . . . . . . . . . . . . 26
2.3 Strain-Rate Effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.4 Cracks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.5 Reinforcing Steel Stress-Strain Behavior . . . . . . . . . . . . . . . . . . . . . 33
2.6 Bond between Concrete and Reinforcing Steel . . . . . . . . . . . . . . . . . . 34
2.7 Reinforced Tension Bar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.8 Tension Stiffening for Reinforced Tension Bar . . . . . . . . . . . . . . . . . . 38
2.9 Cyclic Loading of Reinforced Tension Bar . . . . . . . . . . . . . . . . . . . . 40
3 2D Structural Beams and Frames 42
3.1 General Cross Sectional Behavior . . . . . . . . . . . . . . . . . . . . . . . . 42
3.1.1 Kinematics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.1.2 Linear elastic behavior . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.1.3 Cracked reinforced concrete behavior . . . . . . . . . . . . . . . . . . 45
3.1.3.1 Compressive zone and internal forces . . . . . . . . . . . . . 45
3.1.3.2 Linear concrete compressive behavior . . . . . . . . . . . . 47
3.1.3.3 Nonlinear concrete compressive behavior . . . . . . . . . . . 49
3.2 Equilibrium of Bars . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.3 Structural Beam Elements for 2D . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.4 System Building and Solution Methods . . . . . . . . . . . . . . . . . . . . . 60
3.5 Further Aspects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
3.5.1 Creep . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
3.5.2 Temperature and Shrinkage . . . . . . . . . . . . . . . . . . . . . . . 69
3.5.3 Tension Stiffening . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
3.5.4 Prestressing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
3.5.5 Shear stiffness for reinforced cracked concrete sections . . . . . . . . . 81
3.6 Application Case Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
3.6.1 Transient Dynamics of Beams . . . . . . . . . . . . . . . . . . . . . . 83
3.6.2 More Case Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
II
CONTENTS III
4 Strut-and-Tie Models 88
4.1 Linear Elastic Panel Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . 88
4.2 Truss Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
4.3 Computation of Plane Elasto-Plastic Truss Models . . . . . . . . . . . . . . . 90
4.4 Ideal Plastic Truss Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
4.5 Application Aspects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
5 Multiaxial Concrete Material Behavior 104
5.1 Scales . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
5.2 Some Basics of Continuum Mechanics . . . . . . . . . . . . . . . . . . . . . . 105
5.3 Basic Linear Material Behavior Description . . . . . . . . . . . . . . . . . . . 110
5.4 Basics of Nonlinear Material Behavior . . . . . . . . . . . . . . . . . . . . . . 114
5.4.1 Tangential Stiffness . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
5.4.2 Stress Limit States . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
5.4.3 Phenomenological Approach for Biaxial Anisotropic Stress-Strain Re-
lation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
5.5 Isotropic Damage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
5.6 Isotropic Plasticity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
5.7 Microplane . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
5.8 Localization and Regularization . . . . . . . . . . . . . . . . . . . . . . . . . 130
5.9 Long Term Behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
5.10 Short Term Behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
6 Deep Beams 133
6.1 Limit Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
6.2 2D Crack Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
6.3 2D Modeling of Reinforcement and Bond . . . . . . . . . . . . . . . . . . . . 145
6.4 Biaxial Concrete Material Laws . . . . . . . . . . . . . . . . . . . . . . . . . 151
6.5 Further Aspects and Application Case Studies . . . . . . . . . . . . . . . . . . 152
7 Slabs 154
7.1 Cross Sectional Behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
7.1.1 Kinematic Basics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
7.1.2 Linear elastic behavior . . . . . . . . . . . . . . . . . . . . . . . . . . 156
7.1.3 Reinforced cracked sections . . . . . . . . . . . . . . . . . . . . . . . 157
7.2 Equilibrium of slabs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
7.3 Structural Slab Elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
7.3.1 Area coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
7.3.2 A triangular Kirchhoff element . . . . . . . . . . . . . . . . . . . . . . 162
7.4 System Building and Solution Methods . . . . . . . . . . . . . . . . . . . . . 164
7.5 Reinforcement Design with linear elastic internal forces . . . . . . . . . . . . . 168
8 Shells 177
8.1 Preliminary Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
8.2 Approximation of Geometry and Displacements . . . . . . . . . . . . . . . . . 177
8.3 Approximation of Deformations . . . . . . . . . . . . . . . . . . . . . . . . . 179
8.4 Shell Stresses and Material Laws . . . . . . . . . . . . . . . . . . . . . . . . . 181
8.5 System Building . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
8.6 Slabs and Beams as a Special Case . . . . . . . . . . . . . . . . . . . . . . . . 185
8.7 Locking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
8.8 Reinforced Concrete Shells . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
IV CONTENTS
A ConFem 196
B Transformations of coordinate systems 197
C Linear regression analysis applications 199
C.0.1 Determination of moments in triangular slab elements . . . . . . . . . 199
D Numerical Integration of Elastoplastic Material Laws 201
E ACCESS lectures in summer term 2013 204
State April 4, 2013 mailto:Ulrich.Haeussler-Combe@tu-dresden.de
Chapter 1
Finite Elements in a Nutshell
1.1 Modeling Basics
See Fig. 1.1
Figure 1.1: Modeling Basics according to [CF08, Fig. 7.31]
Reality of interest
Conceptual model
Computational model
1.2 Finite Element Basics
Subdivision of area of interest into smaller units
elements
nodes
Interpolation of displacement elds elementwise
Trial functions
displacement compatibility
Integral equilibrium at nodes
1
2 1.3 Elements
Figure 1.2: FE basic idea
1.3 Elements
Basic concepts
Coordinates
*
global x = ( x y z )
T
*
local r = ( r s t )
T
*
element number I and local coordinate r in element
Displacement variables
*
translation u = ( u v w )
T
*
rotation = (
x

y

z
)
T
Deformation variables
*
1D strain (small strain)
=
u
x
(1.1)
*
2D strain (small strains Voigt notation)
=
_

x

y

xy
_
T
=
_
u
x
v
y
u
y
+
v
x
_
T
(1.2)
*
Curvature (small deections, 2D)
=

2
w
x
2
=

y
x
(1.3)
Generalized force variables, e.g.
*
with Cauchy stress tensor where dw = d gives a mechanical work
increment per unit volume in a body.
*
M with bending moment M where dw = M d gives a mechanical work
increment per unit length of a 2D bending bar.
Interpolation
*
Coordinates
x = N(I, r) x
I
(1.4)
with vector x
I
collecting global coordinates of nodes of element I and row
vector Ncollecting shape functions.
3
*
Displacements
u = N(I, r) u
I
(1.5)
with vector u
I
collecting global displacements of nodes of element I and row
vector Ncollecting shape functions.
*
Deformations, e.g. small 2D engineering strains
= B(I, r) u
I
(1.6)
with vector u
I
collecting global displacements of nodes of element I and matrix
B collecting derivatives of shape functions with respect to global coordinates.
Index I within (I, r) will be omitted in the following, as the functions N, B are the
same for all elements of one (isoparametric) element type.
Bar elements
Kinematic assumption
*
cross section displacements constant in longitudinal direction
Interpolation functions linear 1D, 2 nodes, 2 degrees of freedom
x(r) =
_
1
2
(1 r)
1
2
(1 +r)
_
x
I1
x
I2
_
u(r) =
_
1
2
(1 r)
1
2
(1 +r)
_
u
I1
u
I2
_
=
du
dx
=
_

1
2
1
2
dr
dx

_
u
I1
u
I2
_
=
_

1
2
1
2
2
L
I
_
u
I1
u
I2
_
(1.7)
with a bar length L
I
= x
I2
x
I1
and a Jacobian
J =
x
r
=
L
I
2
(1.8)
Interpolation functions linear 2D, 2 nodes, 4 degrees of freedom
_
x(r)
y(r)
_
=
_
1
2
(1 r) 0
1
2
(1 +r) 0
0
1
2
(1 r) 0
1
2
(1 +r)
_
_
_
_
_
x
I1
y
I1
x
I2
y
I2
_
_
_
_
_
u(r)
v(r)
_
=
_
1
2
(1 r) 0
1
2
(1 +r) 0
0
1
2
(1 r) 0
1
2
(1 +r)
_
_
_
_
_
u
I1
v
I1
u
I2
v
I2
_
_
_
_
(1.9)
Small strain is taken in the bar direction, i.e. in a rotated coordinate system. The
rotation angle (counterclockwise positive) and the transformation matrix are given
by
cos =
x
I2
x
I1
L
I
, sin =
y
I2
y
I1
L
I
, T =
_
cos sin
sin cos
_
(1.10)
with a bar length L
I
=
_
(y
I2
y
I1
)
2
+ (x
I2
x
I1
)
2
. Reusing Eq. (1.7)
3
we get
=
2
L
I
_

1
2
1
2
_
cos sin 0 0
0 0 cos sin
_
_
_
_
_
u
I1
v
I1
u
I2
v
I2
_
_
_
_
(1.11)
4 1.3 Elements
Spring elements
Kinematic assumption
*
Displacement difference between two nodes irrespective of their distance. The
two nodes share the same location as special but common case.
"Interpolation" function 1D
u =
_
1 1
_
u
I1
u
I2
_
(1.12)
"Interpolation" function 2D
_
u
v
_
=
_
1 1 0 0
0 0 1 1
_
_
_
_
_
u
I1
v
I1
u
I2
v
I2
_
_
_
_
(1.13)
2D continuum elements (e.g. plate with loading in plane panel, deep beam)
Kinematic assumptions
*
Displacements u(x, y), v(x, y) are continuous functions.
*
Lateral displacement w is w = 0 (plane strain) or determined from a condi-
tion
z
= 0 (plane stress)
Interpolation functions 4-node quadrilateral
x(r, s) =
4
J=1
N
J
x
IJ
, y(r, s) =
4
J=1
N
J
y
IJ
, u(r, s) =
4
J=1
N
J
u
IJ
, v(r, s) =
4
J=1
N
J
v
IJ
(1.14)
with
N
J
(r, s) =
1
4
(1 +r
J
r)(1 +s
J
s) (1.15)
with the local node coordinates r
J
, s
J
. For the following we need the Jacobian
J =
_
x
r
y
r
x
s
y
s
_
, J = det J (1.16)
which is easily derived from Eq. (1.14)
1,2
and relates the partial derivatives of a
function with respect to local and global coordinates
1
_

r
s
_
= J
_

x
y
_

_

x
y
_
= J
1
_

r
s
_
(1.17)
Small engineering strains with Eqns. (1.2), (1.14), (1.17)
2
_
_
x
(r, s)
y
(r, s)
xy
(r, s)
_
_
=
_
_
_
u
r
r
x
+
u
s
s
x
v
r
r
y
+
v
s
s
y
u
r
r
y
+
u
s
s
y
+
v
r
r
x
+
v
s
s
x
_
_
_
=
_
_
_
_
_
_
_
4
J=1
1
4
_
r
J
(1 +s
J
s)
r
x
+ (1 +r
J
r)s
J
s
x
u
IJ
4
J=1
1
4
_
r
J
(1 +s
J
s)
r
y
+ (1 +r
J
r)s
J
s
y
_
v
IJ
4
J=1
1
4
__
r
J
(1 +s
J
s)
r
y
+ (1 +r
J
r)s
J
s
y
_
u
IJ
+
_
r
J
(1 +s
J
s)
r
x
+ (1 +r
J
r)s
J
s
x
v
IJ
_
_
_
_
_
_
_
_
(1.18)
which after evaluation leads to a form like Eq. (1.6).
1
Closed forms of J and J
1
are available for the 4-node quadrilateral, see e.g. [???].
5
2D beam elements
*
Transverse displacement / deection w of reference axis
rotation angle of reference axis
*
Rotation angle of cross section
*
Plane cross sections remain plane
*
Bernoulli beam: rotation of reference axis = rotation of cross section
*
Timoshenko beam: Rotation of cross section decouples from rotation of refer-
ence axis and shear angle serves as independent variable
Interpolation functions Bernoulli beam, see Page 56.
Interpolation functions Timoshenko beam, see Page 58.
Plate elements (loading perpendicular to plane slab)
Kirchhoff plate (thin plate) small displacements, mid-surface without strains, ro-
tation of cross section can be derived from deection of mid-surface
Reissner-Mindlin plate (thick plate)
3D continuum elements
Additional aspects concerning elements
Continuity of displacements or deformations along element boundaries
Element locking
Independent interpolation of displacement, deformations, generalized forces
6 1.4 Material Behavior
1.4 Material Behavior
In the following, a material is considered as homogeneous.
Material behavior is constituted by a force response in reaction to a deformation.
Linear elastic material law
Uniaxial
x
= E
x
(1.19)
with Youngs modulus E.
plane strain
_
_
xy
_
_
=
E(1 )
(1 +)(1 2)
_
_
1

1
0
1
1 0
0 0
12
2(1)
_
_
_
_
xy
_
_
(1.20)
with Youngs modulus E and Poissons ratio . This is a subset of the triaxial
isotropic linear elastic Hooke law.
plane stress
_
_
xy
_
_
=
E
1
2
_
_
1 0
1 0
0 0
1
2
_
_
_
_
xy
_
_
(1.21)
with Youngs modulus E and Poissons ratio ensuring
z
= 0 for every combina-
tion
x
,
y
,
xy
2D bending
M = EJ (1.22)
with Youngs modulus E and cross-sectional moment of inertia J.
Eqns. (1.19)-(1.21) are a special case of
= C (1.23)
with the constant material stiffness matrix C.
Uniaxial perfect elasto-plastic (as a simple example for a nonlinear constitutive equation)
physical nonlinearity
x
=
_
E (
x

xp
) for
xp

x

xp
sign
x
f
y
otherwise
(1.24)
and
xp
=
x
for [
x
[ = f
y
(1.25)
with a yield stress f
y
and an internal state parameter
xp
. This value is the actual remain-
ing strain upon unloading, i.e.
x
= 0 for
x

xp
.
In case of nonlinear material equations at least an incremental form
d = C
T
d or = C
T
(1.26)
should exist, where the tangential material stiffness might depend on stress, strain, internal
state variables.
On occasion the exibility is needed, as counterpart of stiffness, i.e.
= D (1.27)
7
1.5 Weak Equilibrium and Discretization
Weak forms of equilibrium condition including dynamic parts
Preceding remark: given a point on a boundary either a kinematic boundary con-
dition or a force boundary condition (zero force is also a condition!) has to be
prescribed for this point.
Body (continuum) in space with a volume V , a specic mass , prescribed dis-
tributed loads p, prescribed displacements u on the surface part A
u
, prescribed trac-
tions

t on surface part A
t
(A
u
together with A
t
give the whole surface A), accelera-
tion u, virtual displacements u and associated virtual strains
_
V
T
dV +
_
V
u
T
udV =
_
V
u
T
pdV +
_
A
t
u
T
t dA (1.28)
under the conditions
u = u on A
u
, u = 0 on A
u
(1.29)
Unixial bar along 0 x L
_
L
x
Adx +
_
L
u uAdx =
_
L
u p
x
dx + [u
t]
L
0
(1.30)
under the conditions
u
0
= u
0
, u
0
= 0 or u
L
= u
L
u
L
= 0 (1.31)
with a cross section area A and a load per length p
x
, whereby the formulation of
the last term indicates the boundary term of a partial integration where traction is
prescribed at either x = 0 or x = L (or none, but not both at the same time).
2D Bernoulli beam along 0 x L
_
L
w w mdx +
_
L
M dx =
_
L
w p dx [M]
L
0
+ [wQ]
L
0
(1.32)
with a distributed mass m per length and a distributed load p per length. with cor-
responding pairs (, M) and (w, Q). Only one quantity out of a pair can be pre-
scribed at a boundary. Furthermore, at least two kinematic boundary conditions
should be given with at least one deection w
0
and/or w
L
.
All these forms are variations of the principle of virtual displacements and immediately
allow for physical nonlinearities.
8 1.5 Weak Equilibrium and Discretization
Discretization
The following is performed using Eq. (1.28). Other weak formulations use the same
procedure. The general approach of GALERKIN is followed here.
Small displacements are used, if not otherwise stated.
Steps of discretization
1. Interpolation approach for displacements in the spatial domain, e.g. with Eq. (1.14)
3,4
*
An innite number of degrees of freedom is reduced to a nite number of
nodal values of degree of freedom (discretization).
*
This leads to discretized strains e.g. with Eq. (1.18) and further to stresses
with a constitutive law, e.g. Eq. (1.20) or (1.21) in the linear elastic case.
2. Interpolation approach for virtual displacements
*
The same interpolation Eq. (1.14)
3,4
is used as for the displacement (
approach of BUBNOV-GALERKIN).
*
This leads to discretized virtual strains with Eq. (1.18).
3. Evaluation of integrals
*
This is performed on a element by element base
_
V
I
T
dV = u
T
I
f
I
, f
I
=
_
V
I
B
T
dV
_
V
I
u
T
udV = u
T
I
M
I
u
I
, M
I
=
_
V
I
N
T
NdV
_
V
I
u pdx = u
T
I
p
I
, p
I
=
_
V
I
N
T
pdV
_
A
tI
u
T
t dA = u
T
I

t
I
,

t
I
=
_
A
tI
N
T
t dA
(1.33)
with an element index I. For integration methods see Section 1.7.
Figure 1.3: Schematic ow of nonlinear calculation
4. Assembling of element contributions to the whole system, see Figure 1.2.
9
*
Regarding, e.g. internal nodal forces f , it consists of entries for every de-
gree of freedom of every global node. On the other hand, every discretiza-
tion should have a table, which connects every element to the global nodes
belonging to it.
*
This table relates the position of the entries of f
I
to a position in f .
*
As a node generally gets contributions from more than one element, the
value of an entry in f
I
has to be added to the corresponding entry in f . This
is symbolically described by
_
V
T
dV = u
T
f =
I
u
T
I
f
I
(1.34)
*
The same argumentation holds for u
I
( u), u
I
( u), M
I
( M),
p
I
( p), t
I
(t).
*
Regarding arbitray values of u a spatially discretized system
M u +f (u) = p +
t (1.35)
nally results. This is a system of ordinary differential equations of sec-
ond order in time t. It might be nonlinear due to nonlinear dependence of
internal nodal forces f on nodel displacements u.
The linear case = C leads to internal nodal forces
f
I
=
_
V
I
B
T
C BdV u
I
= K
I
u
I
(1.36)
see Eqns. (1.33)
1
, (1.6), with a constant element stiffness matrix K
I
. Assembling
leads to a system stiffness matrix K
f (u) = K u (1.37)
and regarding Eq. (1.35) to
M u +K u = p +
t (1.38)
which embodies a system of linear ordinary differential equations of second order
in time t.
The system Eqns. (1.35) or (1.38) should be supported with appropriate kinematic
boundary conditions to prevent rigid body displacements.
Tangential stiffness
The system tangential stiffness matrix is needed for the solution of the system and
furthermore reveals characteristic properties of a system, i.e. in particular its stabil-
ity properties.
Element tangential stiffness
df
I
=
f
I
u
I
du
I
= K
TI
du
I
or

f
I
= K
TI
u
I
(1.39)
with
K
TI
=
_
V
I
B
T

u
I
dV =
_
V
I
B
T
C
T
BdV (1.40)
see Eqns. (1.33)
1
, (1.26) (1.6), and a system tangential stiffness K
T
df = K
T
du or

f = K
T
u (1.41)
10 1.6 Convergence
1.6 Convergence
The linear elastic, quasistatic small displacement case is regarded in the following. Given
a linear material law
= C (1.42)
weak, integral equilibrium Eq. (1.28) for a structural problem can be written as
_
V
T
C dV =
_
V
u
T
pdV +
_
A
t
u
T
t dA (1.43)
with appropriate displacement boundary conditions preventing rigid body motions, given
values for p,
t and arbitrary virtual displacements u.

The strains , are derived from the displacements u, u by a differential oper-
ator depending on the type of the structural problem under consideration.
The interpolation functions u, , u, belong to an appropriate function space
2
H
dened over the problem domain V (space occupied by the the structure).
The boundary A of V is composed of A
u
and A
t
, i.e. A = A
t
A
u
and A
t

A
u
= 0. u is prescribed on A
u
(Dirichlet conditions) and

t on A
t
(Neumann
conditions) whereby

t = n with the boundarys normal n.
The mathematical model of the structural problem Eq. (1.43) can be written as
a(u, v) = (f , v) v H (1.44)
with a symmetric, bilinear a(, ), a linear (f , ) and v formally replacing u .
Symmetry
a(u, v) = a(v, u) (1.45)
Bilinearity
a(
1
u
1
+
2
u
2
, v) =
1
a(u
1
, v) +
2
a(u
2
, v)
a(u,
1
v
1
+
2
v
2
) =
1
a(u, v
1
) +
2
a(u, v
2
)
(1.46)
Linearity
(f ,
1
v
1
+
2
v
2
) =
1
(f , v
1
) +
2
(f , v
2
) (1.47)
A norm maps a function v into a non-negative number. Sobolev norms [[v[[
i
of order i
are used in this context [Bat96, 4.3.4,(4.76)]. It is assumed that i = 1 is appropriate for
the following. It can then be shown that a has the properties
Continuity
3
M > 0 : [a(v
1
, v
2
)[ M |v
1
|
1
|v
2
|
1
v
1
, v
2
H (1.48)
Ellipticity
> 0 : a(v, v) |v|
2
1
v H (1.49)
where M, depend on problem type and material values but not on v
1
, v
2
, v.
Due to Eq. (1.49) a(v, v) 0, i.e. a is a norm and may be physically interpreted as
energy, i.e. it is twice the internal strain energy.
2
Square integrable functions fullling the kinematic boundary conditions, see, e.g., [Bat96, 4.3.4].
3
means: it exists a (real number).
11
It can be shown that the problem Eq.(1.44) i.e. determine a function u H such that
Eq. (1.44) is fullled for all v H has a unique solution u, see, e.g., [Bat96, 4.3]. This
solution is exact for the mathematical model of the structural problem.
Discretization uses functions u
h
, v
h
H
h
of a subset H
h
H based upon the concept
of meshes with elements and nodes, see Section 1.3.
A uniform mesh
4
of elements is assumed and h is a mesh size parameter, e.g. a
diameter or length of a generic element.
The approximate solution u
h
H
h
of Eq. (1.44) is determined by
a(u
h
, v
h
) = (f , v
h
) v
h
H
h
(1.50)
The difference between approximate and exact solution gives the error
e
h
= u u
h
(1.51)
The approximation u
h
is known for H
h
given, the error e
h
has to be estimated.
Some properties of the approximate solution
Orthogonality of error, see [Bat96, (4.86)]
a(e
h
, v
h
) = 0 v
h
H
h
(1.52)
Energy of approximation is smaller than exact energy, see [Bat96, (4.89)]
a(u
h
, u
h
) a(u, u) (1.53)
Energy of error is minimized, see [Bat96, (4.91)]
a(e
h
, e
h
) a(u v
h
, u v
h
) v
h
H
h
(1.54)
Combining Eqs. (1.49), (1.54), (1.48) leads to
|e
h
|
2
1
= |u u
h
|
2
1
a(e
h
, e
h
) = inf
v
h
H
h
a(uv
h
, uv
h
) M inf
v
h
H
h
|u v
h
|
2
1
(1.55)
where inf is inmum, the largest lower bound
5
. This is rewritten as
|u u
h
|
1
c d(u, H
h
), d(u, H
h
) = inf
v
h
H
h
|u v
h
|
1
, c =
_
M/
(1.56)
d is a "distance" of functions in H
h
to the exact solution u, c depends on the struc-
tural problem type and the values of its parameters, but not on H
h
.
Convergence means u
h
u or |u u
h
|
1
0 with mesh size h 0.
This can generally be reached with an appropriate selection of function spaces H
h
whereby reducing the distance d(u, H
h
), see Eq. (1.56).
A more precise statement is possible utilizing interpolation theory. It introduces the
interpolant
6
u
i
H
h
of the exact solution u.
4
The following considerations may be transferred to non-uniform meshes, see [Bat96, 4.3.5].
5
u v
h
1
, v
h
H
h
is a subset of real numbers. inf
v
h
H
h
u v
h
1
is the largest number less or equal to
the numbers in this subset.
6
u and u
i
coincide at nodes, but generally not apart from nodes. Generally it is u
i
= u
h
.
12 1.6 Convergence
Complete polynomials
7
of degree k are used for discretization and interpolation,
respectively. Interpolation theory estimates the interpolation error with
|u u
i
|
1
c h
k
|u|
k+1
(1.57)
with the mesh size h and a constant c which is independent of h, see [Bat96, (4.99)].
|u|
k+1
is the k + 1-order Sobolev norm of the exact solution.
On the other hand it must be inf
v
h
H
h
|u v
h
|
1
|u u
i
|
1
as u
i
H
h
. Using
this and Eqs. (1.56), (1.57) yields
|u u
h
|
1
c c h
k
|u|
k+1
(1.58)
The value c c can be merged to c, which depends on the structural problem type and
the values of its parameters, but not on h. A further merger of c and |u|
k+1
leads
to the well known formulation
|u u
h
|
1
c h
k
(1.59)
where c depends on the structural problem type, the values of its parameters and the
norm of the exact solution.
Conditions for convergence, see also [Bat96, 4.3.2]
A basic prerequisite is integrability of all quantities. This leads to requirements
for the integrands of the energy a and the Sobolov norms
8
, which are u
h
, v
h
, u or
derivatives thereof.
*
It corresponds to the requirement of compatibility of nite element interpolation
functions along inter element borders.
According to Eq, (1.59) a sequence of approximate solutions u
h
with h 0 will
converge
9
with respect to |u u
h
|
1
if k 1.
*
The case k = 1 is covered by the so called patch test, i.e. the ability to model
elds with constant rst derivatives of nite element interpolation functions in
arbitrary element congurations, see, e.g., [BLM00, 8.3.2]).
The convergence rate will be higher for larger values of k, i.e. if the nite element
interpolation has a higher order of completeness.
Limitations
The coefcient c may become so large under certain conditions that accepable solu-
tions, i.e. a sufciently small values [[u u
h
[[, cannot be reached with realizable
values h small enough.
*
A particular occurence is given with locking of approximate solutions with in-
compressible or nearly incompressible materials.
Extended weak forms
7
A polynomial in x, y is complete of order 1 if it includes x, y, complete of order 2 if of order 1 and including
x
2
, xy, y
2
, complete of order 3 if complete of order 2 and including x
3
, x
2
y, xy
2
, y
3
and so on.
8
Sobolev norms are built from integration of squares of functions and squares of their derivatives.
9
Converge with respect to rst order Sobolev norm u u
h
1
may not be sufcient if generalized strains are
derived from higher derivatives of displacements, e.g. with beams, slabs, shells. The theory has to be extended for
this case.
13
Eqs. (1.43,1.44) are weak forms of displacement based methods, as a solution is
given by a displacement eld. Strains and stresses are derived from this solution.
Extended weak forms allow to involve elds for stresses and strains as independent
solution variables. Most prominent are the principles of Hu-Washizu and Hellinger-
Reissner [Bat96, 4.4.2].
*
An abstract extended problem denition analogous to Eq. (1.50) given by, see
[Bat01, (16)]
a(u
h
, v
h
) +b(
h
, v
h
) = (f , v
h
) v
h
H
h
b(w
h
, u
h
) c(
h
, e
h
) = 0 w
h
W
h
(1.60)
in which a, c are symmetric bilinear forms, b is a bilinear form, f is a linear
form, H
h
, W
h
are appropriate functions spaces, u
h
H
h
,
h
W
h
are the
approximate solutions.
In most cases
h
stands for an independent eld of strains or stresses.
*
The foregoing theory concerning convergence has to be extended, see [Bat01].
Such an extension includes the widely referenced inf-sup condition.
These approaches may solve, e.g., locking problems.
Nonlinear problems
The foregoing considerations relate to linear problems. They cannot be strictly
applied to nonlinear probems physically nonlinear and/or geometrically nonlin-
ear. But the conclusions to be drawn regarding element selection and discretization
should also be considered for nonlinear problems.
14 1.7 Numerical Integration and Solution Methods for Algebraic Systems
n
i

i

i
0 0.0 2.0
1 0.57735 02691 89626 1.0
2 0.77459 66692 41483 0.55555 55555 55556
0.0 0.88888 88888 88889
3 0.86113 63115 94053 0.34785 48451 37454
0.33998 10435 84856 0.65214 51548 62546
.
.
.
.
.
.
.
.
.
Table 1.1: Sampling points and weights for Gaussian numerical integration (up to 15 digits
shown)
1.7 Numerical Integration and Solution Methods for Algebraic Sys-
tems
Integration of Eqs. (1.34) for the two-dimensional case as an example
Integration of isoparametric elements
_
V
I
f(x, y) dV =
_
y
2
y
1
_
x
2
(y)
x
1
(y)
f(x, y) tdxdy =
_
+1
1
_
+1
1
f(r, s) J(r, s) tdrds
(1.61)
with the determinant J of the Jacobian, see Eq. (1.16)
2
, and a constant thickness t.
Numerical integration
_
+1
1
_
+1
1
f(r, s) J(r, s) tdrds = t
n
i
i=0
n
i
j=0
j
f(
i
,
j
) J(
i
,
j
) (1.62)
with integration order n
i
, sampling points and weighting factors .
Gaussian scheme
*
Sampling points and weighting factors see Table 1.1. Generally we have
n
i
i=1
i
=
2. Also for other integration schemes.
*
Integration accuracy: an integration order n
i
gives exact (within the scope of
numerical accuracy) results for polynoms of order 2n
i
+ 1, e.g. a uniaxial
integration of order 1 with two sampling points integrates exactly a polynom of
order 3.
Other schemes: Simpson, Newton-Cotes, Lobatto
Integration of triangular elements
*
Remains to be added.
15
Solution method for quasistatic problems
Regarding Eq. (1.35) we have to solve
r(u) = f (u) p = 0, p = p +
t (1.63)
with internal nodal forces f depending on displacements u and external nodal load
p, which are assumed to be independent of u.
The general case is nonlinear dependence of f on u. So the solution of Eq. (1.63)
has to be determined with an iteration with a sequence u
(0)
. . . u
()
.
Looking at an arbitrary iteration step () we have r(u
()
) ,= 0 and seek for a cor-
rection u. A linear Taylor-expansion is used as a basic approach
r(u
()
+u) r(u
()
) +K
()
T
u, K
()
T
=
r
u
u=u
()
=
f
u
u=u
()
= 0
(1.64)
for the Newton-Raphson method with the tangential stiffness matrix K
T
. This leads
to
u = [K
()
T
]
1
r(u
()
)
u
(+1)
= u
()
+u
(1.65)
and an (hopefully) improved value u
(+1)
. Iteration may stop if |r(u
(+1)
)| 1
and |u| 1 with a suitable norm | |. The method generally has a fast conver-
gence, but is relatively costly. In every step () the tangential stiffness matrix has to
be computed and a LU-decomposition has to be performed.
Other iteration methods use variants of the iteration matrix (modied Newton Raph-
son, secant methods like BFGS)
Iterative methods like Newton-Raphson generally are embedded in an incrementally
iterative scheme, so loading is given as a history: p = p(t). Often we choose
0 t 1 for the load history time
10
, without restrictions to generality.
*
We look at discrete time values t
i
, t
0
= 0 and have p
i
= p(t
i
). This is known
before a solution.
*
Accordingly, we have u
i
= u(t
i
), f
i
= f (u
i
). These are not known before a
solution, except u
0
, f
0
. We assume r
0
= f
0
p
0
= 0.
*
The solution starts with t
1
, where u
1
has to be determined. This is done with
an iteration sequence u
(0)
1
. . . u
()
1
, e.g. with the Newton-Raphson method and
u
(0)
1
= u
0
.
*
A converged u
1
is used as a base for t
2
and so on.
10
Load history time is different to real time.
Solution method for transient problems, in particular creep problems
Extended constitutive law
= C
T
+ (1.66)
with the tangential material stiffness C
T
, see Eq. (1.26), and an additional term
depending on stress and strain, see e.g. Section 2.2. We have to solve for (t),
whereupon (t), (t) are given as functions of time.
A numerical method is used with the trapezoidal rule for derivatives of stress and
strain
i+1
=
i
+ t [
i+1
+ (1 )
i
]
i+1
=
i
+ t [
i+1
+ (1 )
i
]
(1.67)
with time discretization parameters t, . Eqs. (1.66), Eq. (1.67)
1
lead to
i+1
=
i
+ (1 ) t
_
C
T,i

i
+
i
_
+t
_
C
T,i+1

i+1
+
i+1
_
(1.68)
The parameters , t rule stability and accuracy of the temporal discretization.
Regarding numerical methods for systems of rst order differential equations see
[BSMM00, 19.4], [Hug00, 8].
In the following, two approaches will be considered: (1) the implicit Euler approach
with = 1, (2) the explicit Euler approach with = 0. Regarding Eq. (1.67)
2
, the
expicit approach leads to
i+1
=
i
+C
T,i
(
i+1

i
) + t
i
(1.69)
and the implicit to
i+1
=
i
+C
T,i+1
(
i+1

i
) + t
i+1
(1.70)
The latter may be a nonlinear equation in case depends on and/or .
A particular case is given with
= V W (1.71)
with a constant Kelvin viscosity V and a constant Maxwell viscosity W, see e.g.
Section 2.2. Thus,
i+1
= V
i+1
W
i+1
and the implicit scheme is
i+1
=
i
+C
T,i+1
(
i+1

i
) + t V
i+1
t W
i+1
= [I + t W]
1
i
+C
T,i+1
(
i+1

i
) + t V
i+1
= [I + t W]
1
i
+ [C
T,i+1
+ t V] (
i+1

i
) + t V
i
(1.72)
Internal nodal forces according to Eq. (1.33)
1
f
i+1
=
_
V
B
T

i+1
dV = f
i
+K
T,i+1
u +

f
i+1
(1.73)
with
f
i
=
_
V
B
T

i
dV
f
i+1
= t
_
V
B
T

i+1
dV
K
T,i+1
=
_
V
B
T

C
T,i+1
BdV
(1.74)
This is embedded in a time stepping scheme, compare Eq. (1.63)
r
i+1
= f
i+1
p
i+1
= 0 K
T,i+1
u = p
i+1
f
i

f
i+1
(1.75)
Starting with initial conditions f
0
,
f
0
and a given load p(t) displacements may be
determined time step by time step. But Eq. (1.75) might be nonlinear in a time step
and need an iteration, see Eq. (1.65).
17
Solution methods for dynamic problems
Based on Eq. (1.35) we have in analogy to Eq. (1.63)
r = M u +f p(t) = 0, p(t) = p(t) +
t(t) (1.76)
where the loading p is a prescribed function of the time t. The displacements u(t)
and all derived values (velocities u(t), accelerations u(t), internal nodal forces f )
are unknown before solution. Eq. (1.76) is discretized in the spatial domain, but not
yet in the time domain, i.e. it is system of ordinary differential equations of 2nd
order in time.
This problem needs appropriate boundary conditions and initial conditions for the
displacements u
0
= u(0) and velocities u
0
= u(0).
A solution may be computed with a discretization in time via a difference scheme.
A widespread approach is given with the Newmark method
u
i+1
= u
i
+ t
_
u
i+1
+ (1 ) u
i
_
u
i+1
= u
i
+ t u
i
+ t
2
_
u
i+1
+ (
1
2
) u
i
_
(1.77)
with u
i+1
= u(t
i+1
), u
i+1
= u(t
i+1
), u
i+1
= u(t
i+1
) a time step length t =
t
i+1
t
i
and integration parameters , . Eqns. (1.77) are solved for the acceleration
and velocity in time step i + 1. We get
u
i+1
=
1
t
2
[u
i+1
u
i+1
] (1.78)
with an auxiliary quantity
u
i+1
= u
i
+ t u
i
+
t
2
2
(1 2) u
i
(1.79)
and the velocity
u
i+1
=

t
[u
i+1
u
i
] +
_
1

_
u
i
+ t
_
1

2
_
u
i
(1.80)
Finally, dynamic equilibrium Eq. (1.76) is applied for time step i +1 with the accel-
eration according to Eq. (1.78):
r =
1
t
2
M [u
i+1
u
i+1
] +f
i+1
p
i+1
= 0 (1.81)
With given parameters , , t, a given previous state u
i
, u
i
, u
i
, given mass matrix
M and load P
i+1
, Eq. (1.81) has to be solved for u
i+1
, where the dependence of
f
i+1
on u
i+1
is crucial and might be nonlinear.
We apply the Newton-Raphson method Eq. (1.65). The Jacobian matrix is given
with
A
()
T
=
1
t
2
M+
f
u
u=u
()
i+1
=
1
t
2
M+K
()
T
(1.82)
leading to an iteration scheme
u
(+1)
i+1
= u
()
i+1
+
_
A
()
T
_
1
_
p
i+1
f
()
i+1

1
t
2
M
_
u
()
i+1
u
i+1
_
_
(1.83)
In the linear case with
f
()
i+1
= K u
()
i+1
, A
()
T
= A =
1
t
2
M+K (1.84)
the Eq. (1.83) simplies to
u
i+1
= A
1
_
p
i+1
+
1
t
2
M u
i+1
_
(1.85)
with no iteration necessary, compare [Bat96, 9.2.4].
Numerical integration parameters , rule consistency and numerical stability of
the method.
*
Stability means that an amount of error introduced in a certain step due to a
nite time step length t is not is not increased in the subsequent steps.
*
Consistency means that the iteration scheme converges to the differential equa-
tion for t 0.
Stability and consistency are necessary to ensure that the error of the numerical
method remains within some bounds for a nite time step length t
Generally a choice =
1
4
, =
1
2
is reasonable for the Newmark method to reach
consistency and stability.
Bibliography
[Bat96] BATHE, K.J.: Finite Element Procedures. Englewood Cliffs, New Jersey : Prentice
Hall, 1996
[Bat01] BATHE, K.J.: The inf-sup condition and its evaluation for mixed nite element
methods. In: Computers and Structures 79 (2001), S. 243252
[BLM00] BELYTSCHKO, T. ; LIU, W.K. ; MORAN, B.: Nonlinear Finite Elements for Con-
tinua and Structures. Chichester : John Wiley & Sons, 2000
[BSMM00] BRONSTEIN ; SEMENDJAJEW ; MUSIOL ; MUEHLIG: Taschenbuch der Mathe-
matik. 5. Auage. Frankfurt/Main : Verlag Harri Deutsch, 2000
[CF08] CEB-FIP: Practitioners guide to nite element modelling of reinforced concrete
structures. Bd. Bulletin Nr. 45. Lausanne : International Federation for Structural
Concrete p, 2008
[Hug00] HUGHES, T.R.: The Finite Element Method - Linear Static and Dynamic Finite
Element Analysis. Mineola, New York : Dover Publications, Inc., 2000
19
Chapter 2
Uniaxial Structural Concrete Behavior
2.1 Short Term Stress-Strain Behavior of Concrete
Short term behavior means, that loading is applied within a few minutes, hours, or days,
and that one regards the immediate reaction of the structural material.
Time-dependent behavior, e.g. creep and shrinkage, is not considered.
Unconstrained compression
Unconstrained spatially homogeneous loading
Stress-strain behavior under monotonic loading, see Fig. 2.1a
*
Linear part
*
Nonlinear hardening part
*
Nonlinear softening part
Figure 2.1: a) Uniaxial compressive stress strain behavior b) mesoscale view
*
Analytical description
Approach by Saenz, see [CS94, 8.8.1]
=
E
c0
1 +
_
E
c0
E
c1
2
_

c1
+
_

c1
_
2
(2.1)
with initial Youngs modulus E
c0
of concrete, its secant modulus E
c1
=
f
c
/
c1
at compressive strength f
c
(unsigned), strain
c1
(signed) at strength.
With =
c1
Eq. (2.1) leads to = f
c
.
20
21
The derivative with respect to ( tangential material stiffness) is given
with
E
t
=
d
d
=
E
c0
_
1

2
2
c1
_
_
1 +
_
E
c0
E
c1
2
_

c1
+

2
2
c1
_
2
(2.2)
which leads to E
t
= E
c0
for = 0, furthermore E
t
= 0 for =
c1
and
E
t
< 0 for <
c1
, [[ > [
c1
[.
Alternatives by Modelcode 90 [Com93, 2.1.4.4.1] and DIN 1045-1 [din08,
9.1.5, 8.5.1].
These approaches may all be classied as phenomenological,
i.e. free parameters of a polynomial form are chosen such, that character-
istic points of measured stress-strain behavior are reproduced.
Mesoscale view with a distinction of
*
aggregates,
*
mortar,
*
interface layer between aggregates and mortar,
*
highly inhomogeneous stress distribution within a specimen with internal
lateral tensile stresses locally.
Failure modes
*
Lateral splitting / cracking with respect to compressed longitudinal direction
*
Volume expansion
*
Diffuse failure
Unconstrained tension
Figure 2.2: a) Uniaxial tensile stress strain behavior b) mesoscale cracking
Homogeneous end loading of a uniform bar
Stress-strain behavior, see Fig. 2.2a
*
Linear part
*
Nonlinear hardening part
*
Nonlinear softening part
22 2.1 Short Term Stress-Strain Behavior of Concrete
Mesoscale view, see Fig. 2.2b, with successive states
*
micro cracking,
*
crack bridges,
*
macro cracking.
Figure 2.3: Scheme of localization
Localization and crack band
*
Weak cross section within specimen where tensile strength is reached
*
Narrow band of high strains crack band
*
Strain distribution
c
(x) across crack band with x
1
x x
2
and a crack band
width b
w
= x
2
x
1
*
Softening of crack band increasing strains with decreasing stresses
*
Unloading of specimen areas beyond crack zone
Snap back behavior of whole bar depending on ratio of crack band width
to total length
Localized failure in contrast to diffuse failure
*
Process band ending up in macro crack with zero stress and zero strain beyond
crack band.
*
The whole process may be unstable under quasistatic conditions.
Fictitious crack width
w =
_
x
2
x
1
c
(x) dx = b
w

c
(2.3)
with mean strain
c
in the crack band.
*
In the following,
c
const. is assumed within the crack band for simplica-
tion.
*
Accordingly, the ctitious crack width, see Eq. (2.3), is given with
w = b
w

c
(2.4)
*
Furtmermore, this assures a
c
-relation in the softening part of Fig. 2.2a and
a critical strain
cr
with (
cr
) = 0.
*
Regarding Eq. (2.4), a
c
-relation may be transformed into a w-relation
with a ctitious critical crack width w
cr
with (w
cr
) = 0.
23
Crack energy
*
Energy in a material element within crack band depending on
c
or w
g =
_

c
1
(
) d
=
1
b
w
_
w
0
(w
) dw
(2.5)
with a lower bound e.g.
1
=
ct
and (
) according to Fig. 2.2a.

*
With
c
=
cr
or w = w
cr
Eq. (2.5) leads to the specic crack energy g
f
(related
to volume), which corresponds to the shaded area in Fig. 2.2a.
*
Sum along crack band leads to crack energy (related to surface)
G
f
= b
w
g
f
= b
w
_

cr
1
(
) d
=
_
w
cr
0
(w
) dw
(2.6)
*
G
f
measures energy dissipation due to creation of new surfaces.
*
G
f
is assumed as (constant) material parameter, so that Eq. (2.6) leads to a
constraint for a w-relation or
c
-relation, respectively.
Unloading to stress free state with a reduction of stresses and strains
Reduction of stiffness damage
Remaining strains in stress free state plasticity
Unloading is hard to realize for softening structural elements, as it requires total displace-
ment control of softening areas.
Example 2.1 Simple concrete tensile bar with localization
Figure 2.4: Example 2.1 a) geometry scheme b) force-displacement curve
Geometry and Discretization, see Fig. 2.4a
Bar Length L = 0.5 m, cross sectional area A
c
= 0.1 0.1 m
2
Discretization with N
e
= 500 1D-bar elements with two nodes, see Eq. (1.7).
Material
Concrete grade C40 according to [Com93, 2.1]
24 2.1 Short Term Stress-Strain Behavior of Concrete
Initial Youngs modulus E
c
= 36 000 MN/m
2
, tensile strength f
ct
= 3.5 MN/m
2
Material model
Damage model with gradient damage, see [HC07].
Boundary conditions
Left end with zero displacements,right end with prescribed displacement.
Nonlocal damage assumed as zero at both ends (failure in mid point).
Solution method
Incrementally iterative with arc length control of load step size and Newton-Raphson
iteration within each increment.
Results
Figure 2.5: Example 2.1 strain distribution a) load step A b) load step B
Load displacement curve see Fig. 2.4b.
*
Initial elastic part followed by increasing damage
*
Limit load where the stresses reach the tensile strength of the material
*
Softening part with increasing displacements and decreasing load
Bar has two types of material behavior in its longitudinal direction. Hard-
ening and softening tangential material stiffness. In the softening part the
bar elongates with decreasing stresses ( localization) in the hardening
part it becomes shorter due to reduced stresses.
*
Snap back part with decreasing displacements and decreasing load
Totally, it shortens in the example case, but this depends on the ratio of
length of softening part to hardening part.
Longer bars show a more pronounced snap back behavior.
Strain distribution along bar
*
Before limit load see Fig. 2.5a. Due to prescribed zero nonlocal damage on
both ends strain moderately increases in the mid range of the bar. Stress is
= 2.90 MN/m
2
and mean strain = 0.94 10
4
.
*
After limit load in the snap back range see Fig. 2.5b. Strains strongly increase
in a short mid range and otherwise decrease due to the load decrease. Stress is
= 1.64 MN/m
2
and mean strain = 1.08 10
4
.
25
end example 2.1
Remarks
In a real specimen the localization will not center exactly in the mid point of a bar,
but in the weakest cross section. This cross section will arise due to the stochastic
variation of material strength. Its cross-sectional strength and location cannot be
exactly determined, but only with statistical parameters.
26 2.2 Long Term Effects - Creep, Shrinkage and Temperature
2.2 Long Term Effects - Creep, Shrinkage and Temperature
Besides the immediate reaction of a structural material, its deferred reaction upon an ap-
plied loading is regarded. This is typically creep or relaxation, respectively. The following
approach is chosen for the uniaxial strain depending on time t
(t) =
_
t
0
J(t, ) () d, 0 t (2.7)
for a creeping material exposed to a uniaxial load history (). The compliance function
J(t, ) is specic for every material type. Eq. (2.7) describes linear creep.
The following form has proven to be appropriate for the compliance function
J(t, ) =
N
=1
J
(t, ), J
(t, ) =
1
E
()
_
1 e
[y
()y
(t)]
_
, y
() = (/
)
q
(2.8)
with material parameters
, q
and material functions E
(). The parameter
has a
dimension of time and the function E
a dimension of stress.
The case N = 1, E
1
= E = const.,
1
0 gives J(t, ) = 1/E with Eq. (2.8),
and with Eq. (2.7) a linear elastic law (t) = C (t).
Combining Eqns. (2.7), (2.8) leads to
(t) =
N
=1
(t),
(t) =
_
t
0
J
(t, ) () d (2.9)
Determination of J
literature
Figure 2.6: a) Kelvin-Voigt element, Maxwell element b) Simple chains
Simplications are used in the following, i.e. E
= const., q
= 1. Using Eqns. (2.8),

(2.7) leads to
(t) =
(t)
E
_
t
0
() e
d
(2.10)
with

(t) =
(t)
E
+
1
_
t
0
() e
d
e
() e
=
1
_
t
0
() e
d
(2.11)
27
Thus, the strain Eq. (2.10) fullls the differential equation
(t) +E
(t) = (t),
= E
(2.12)
with an initial condition
(0) = 0. Eq. (2.12) describes a Kelvin-Voigt element with a

spring and a damper in parallel, see Fig. 2.6a.
An alternative is given with the Maxwell element with a spring and a damper in
series, see Fig. 2.6a.
Eq. (2.9)
1
yields a Kelvin-Voigt chain, i.e. a number of Kelvin-Voigt elements in series.
These systems of springs and dampers add up to rheological models.
A simple Kelvin-Voigt chain N = 2, = 0 . . . 1,
0
= 0 is considered
E
0
0
(t) = (t),
1

1
(t) +E
1
1
(t) = (t) (2.13)
see Fig. 2.6b, with
(t) =
0
(t) +
1
(t), (t) =
0
(t) +
1
(t) (2.14)
First of all, we have
1
(t) = (t)
(t)
E
0
,
1
(t) = (t)
(t)
E
0
(2.15)
Thus, Eq. (2.13)
2
after a few calculations leads to
(t) +
E
0
+E
1
1
(t) = E
0
(t) +
E
0
E
1
1
(t) (2.16)
We introduce a nal stiffness 1/E = 1/E
0
+1/E
1
, consider E
0
as an initial stiffness, and
introduce a creep coefcient dened by E = E
0
/(1 + ). This leads to E
1
= E
0
/
and Eq. (2.16) may be reformulated as
(t) +
(1 +)E
0
1
(t) = E
0
(t) +
E
2
0
1
(t) (2.17)
Thus, prescribing a constant stress (t) =
0
, = 0 with an initial strain (0) =
0
/E
0
leads to a strain varying with time
(t) =

0
E
0
_
1 +
_
1 e
t
__
, =
E
0
1
(2.18)
The asymptotical nal value is
asym
= (1+)
0
/E
0
, the creep portion is
0
/E
0
. This
provides a method to describe
1
, as the creep portion has a characteristic time t
upon a
part with 0 < 1 of the asymptotic creep part. This leads to
1 e
t
= =
ln(1 )
t
(2.19)
The values E
0
, t
, , may be determined based on experiments. This results in

1
and
a differential visco-elastic material law Eq. (2.17) to describe creep.
28 2.2 Long Term Effects - Creep, Shrinkage and Temperature
Incremental material law
See Section 1.4, Eq. (1.26). This has to be extended with respect to viscous parts,
like e.g. Eq. (2.17). For following uniaxial applications the form
(t) = E
0
_
(t) + (t)
_
1 +
(t), =
E
0
1
=
ln
t
(2.20)
is chosen.
Shrinkage and temperature
Temperature and shrinkage impose strains independent from loads. Uniaxial tem-
perature strains, e.g., are given by
T
=
T
T (2.21)
with thermal expansion coefcient
T
and a temperature change T. Concrete
shrinkage strains
cs
mainly depend on time, humidity conditions and ratio of sur-
face to volume, see [din08, 9.1.4] or [Com93, 2.1.6.4.4].
Stress is induced by total, measured strain less imposed strains. Thus, in the uniaxial
linear elastic case stress is given by
= E (
I
) ,
I
=
T
+
cs
(2.22)
This is transferred to the case including linear creep with Eq. (2.20)
(t) = E
0
_
(t)
I
(t) +
_
(t)
I
(t)
_
1 +
(t)
= E
0
_
(t)
I
(t)
+ E
0
_
(t)
I
(t)
_
1 +
(t)
(2.23)
or
(t) = C
T
_
(t)
I
(t)
+V
_
(t)
I
(t)
W (t) (2.24)
with
C
T
= E
0
, V = E
0
, W =
_
1 +
(2.25)
where
I
(t),
I
(t) should be a known function of time.
A transient problem arises rst of all without inertial terms , which needs some
care regarding integration of systems equations in time, see Section (1.7).
Example 2.2 Simple concrete tensile bar with creep and imposed strains
Geometry and Discretization
Bar Length L = 1.0 m, cross sectional area A
c
= 1.0 m
2
. Discretization with
N
e
= 5 1D-bar elements with two nodes, see Eq. (1.7).
As a homogeneous state along longitudinal direction is considered in this exam-
ple, the length of the bar is irrelevant and one element would be sufcient.
Material
Concrete with an initial Youngs modulus E
0
= 30 000 MN/m
2
and a tensile strength
f
ct
= 3.5 MN/m
2
.
29
Material creep is assumed with = 2.0 and t
= 100 [d] for = 0.5, i.e. half of

total creep occurs after 100 days for a constant stress load. With Eqns. (2.20), (2.25)
this leads to
= 0.006931 1/d, C
T
= 30 000 MN/m
2
, V = 207.94 MN/m
2
d, W = 0.020794 1/d
(2.26)
Boundary conditions
Displacement of left node is prescribed with zero in all following cases.
Solution method
Incrementally iterative with Newton-Raphson iteration within each increment, if
neccessary. The method is described on Page 16.
Following time discretization values are chosen: implicit and t = 10 days. A
period of 500 days is regarded.
Figure 2.7: a) Strain depending on time b) Stress depending on time
Case 1: validation with stress loading
0
= 3.0 MN/m
2
constant in time
Computed strain depending on time see Fig. 2.7a. An exact solution is given with
Eq. (2.18) for this case. Differences between exact solution and numerically com-
puted solution are small and are not visible in the Figure.
Case 2: prescribed immediate strain
0
= 3.0/30 000 = 0.0001 constant in time
Computed stress depending on time see Fig. 2.7b.
In exact technical terms, this is not creep anymore but relaxation.
Case 3: prescribed imposed contraction
A slow contraction of 0.15 is prescribed over a period of t = 100 d linear in
time and then hold constant. As total strain is prescribed with zero, a tensile
stress is induced, which would lead to tensile failure without relaxation. Stress with
relaxation is shown in Fig. 2.7b.
end example 2.2
While analytical, exact solutions are availabe for cases 1, 2, the numerical approach is
necessary for arbitrarily prescribed loads or displacements. Furthermore, more complex
creep models, see e.g. [Mal69, 6.4], [JB01, 28, 29], can only be solved with numerical
models.
30 2.3 Strain-Rate Effects
2.3 Strain-Rate Effects
31
2.4 Cracks
This is strongly connected to the notions of localization, crack band and ctitious crack
width, see Page 21.
We consider an instant with a ctitious crack width w given according to Eq. (2.3). Due
to equilibrium reasons a constant cross sectional stress occurs along the crack band,
which gives a pair , w. Considering all instants of tension softening a relation (w) can
be derived, i.e. a stress across crack band depending on ctitious crack width.
Figure 2.8: Constitutive law for cohesive crack model
Cohesive crack model (general spatial view)
Fictitious crack
*
Implies a geometric plane corresponding to one of the two crack surfaces.
*
This plane supports a reference system, where the distance between the crack
surfaces is measured by a normal crack width component w
x
and two sliding
crack width components w
y
, w
z
.
A high idealization in the area of e.g. crack bridges.
Crack-tractions t
x
, t
y
, t
z
Cauchy stress tensor projected onto crack plane
*
Crack-tractions t
x
, t
y
, t
z
are transferred across crack plane.
Material law connecting crack-traction and crack-width
*
A common format is
t
x
= f
n
(w
x
), t
y
= f
s
(w
y
), t
z
= f
s
(w
z
) (2.27)
with different laws f
n
for the normal component and f
s
for the sliding compo-
nent.
*
This material law uses crack tractions as force variable and crack widths as
deformation variable, compare page 2
*
All material frameworks like plasticity, damage etc. may be used.
32 2.4 Cracks
A common simplied approach: Set f
s
= 0. f
n
is given by a cut off and a scaling
of Fig. 2.2a, which results in Fig. 2.8. This reects the uniaxial case.
The cohesive crack model is a macroscopic model for the complex crack develop-
ment process.
Smeared crack model for the simplied approach
Approach for strain of a uniaxial black box element
1
with a length L
c
with one crack
=
1
L
c
[(L
c
b
w
)
m
+b
w

c
] = (1 )
m
+
c
, =
b
w
L
c
(2.28)
with the strain
m
of the uncracked material, see Fig. (2.9). For
c
, b
w
see Eq. (2.4).
Figure 2.9: Smeared crack concept
Connection to stresses / forces
*
The strain part
u
is connected to a homogeneous material, see Section 1.4,
leading to material stresses.
*
The strain part
c
or w is connected to a crack or discontinuity, leading to crack
stresses, see Eq. (2.27)
1
.
*
Both stress parts are connected through equilibrium, e.g. have the same value
in an uniaxial element.
If element length is adjusted to crack band width b
w
, i.e. L
c
= b
w
from Eqns. (2.4),
(2.28) we get a crack width
w = L
c
(2.29)
In case of = 0,
u
= 0 adjusting of element length is not necessary, i.e. w = L
c
for any selection of L

c
.
The smeared crack model is a model for the ctitious crack preserving continuity
of displacements, which is important for ordinary nite elements. There may be
alternative models for the ctitious crack, i.e. for the implemention of the cohesive
crack model.
1
We cannot see inside.
33
2.5 Reinforcing Steel Stress-Strain Behavior
Uniaxial Compression / Tension, see 2.10a, b
Linear range
Transition range
Yielding range
Unloading
Reloading
Cyclic loading
Energy dissipation
Figure 2.10: a) Uniaxial stress-strain behavior for reinforcing steel b) hardening behavior
Simple uniaxial constitutive law with isotropic hardening
Stress-strain law
s
=
_
E
s
(
s

p
) if
p

f
y
E
s

p
+
f
y
E
s
sign f
y
else
(2.30)
with Youngs modulus E
s
of reinforcing steel, its yield stress f
y
, the sign function
sign and the actual plastic strain
p
as internal state parameter.
Evolution law for internal state parameter

p
=
s
f
y
= E
T
[
s
[
_
if
s

s
> 0 and [
s
[ = f
y
(2.31)
with a hardening or tangential modulus E
T
. Eqns. (2.30), (2.31) cover loading,
unloading and reloading for tension and compression for uniaxial ideal elastoplastic
behavior. The yield stress f
y
may increase due to hardening.
Hardening, see Fig. 2.10b
Isotropic
Kinematic
Codes
DIN 1045-1 [din08, 9.2, 9.3], Modelcode 90 [Com93, 2.2, 2.3].
34 2.6 Bond between Concrete and Reinforcing Steel
2.6 Bond between Concrete and Reinforcing Steel
Bond mechanisms, see Fig. 2.11a, with
reinforcement ribs,
conical struts,
circumferential ties,
complex triaxial problem.
Figure 2.11: a) Triaxial bond mechanisms b) Bond law
Modeling and model variables, compare page 2
Uniaxial orientation along rebar direction.
Kinematic variable slip s with a dimension of length.
Force variable bond stress with a dimension of stress.
Bond law = f
(s).
*
All constitutive frames like plasticity, damage etc. may basically be used.
Bond ow t = U with a rebar cirumference U and dimension force per length.
Bond laws
Monotonous loading.
*
See e.g. [Com93, 3.1.1].
*
A smoothed version = f
(s), see Fig. 2.11b.

Characterized by maximum bond stress
max
, a corresponding slip s
1
, a
residual bond stress
f
and a corresponding slip s
2
.
This is composed of a quadratic, cubic and linear polynom with continuous
derivatives at the connection points.
*
As long as monotonous loading is considered, approaches with plasticity and
damage approaches are dispensable.
Unloading and Reloading
*
35
2.7 Reinforced Tension Bar
A nite element model
Concrete
*
Element type: 1D linear bar element, see Eq. (1.7).
*
Constitutive law: uniaxial linear elastic, see Eq. (1.19) with a limited tensile
strength f
ct
=
_
E
c
,
f
ct
E
c
0, else
(2.32)
Concrete with Cracks
*
Smeared crack approach, see Page 32.
Reinforcing steel
*
Element type: 1D linear bar element, see Eq. (1.7).
*
Constitutive law: uniaxial elastoplastic, see Eqns. (2.30), (2.31).
Bond
*
Element type: 1D spring element, see Eq. (1.12), with slip s instead of u.
*
Constitutive law: bond law, see Fig. 2.11b.
Geometry scheme see Fig. 2.12a.
Boundary conditions / loading
*
Base with xed zero displacement on one end.
*
Prescribed displacement on the other end.
Figure 2.12: a) Geometry scheme of reinforced tension bar b) force-displacement curve of
example 2.3
Solution methods
Incrementally iterative. Equilibrium iteration within each loading increment with
Newton-Raphson method, see Page 15.
Computation of crack width
See Eq. (2.29) and the following remark. As cracks are assumed as stress free imme-
diately after cracking we have w = L
I
with the length L
I
of the cracked element
and its computed strain .
36 2.7 Reinforced Tension Bar
Example 2.3 Simple reinforced concrete tension bar
Model input values
Bar length L = 1.0 m, cross sectional area of concrete A
c
= 0.1 0.1 m
2
, rein-
forcement 1 16, A
s
= 2.01 cm
2
, circumference U
s
= 5.02 cm.
Discretization with two-node bar elements, see Eq. (1.7), length L
I
= 0.01 m.
*
100 Elements for concrete with 101 nodes, 100 Elements for reinforcement
with 101 nodes, 101 bond elements connecting concrete nodes and reinforce-
ment nodes.
*
Totally 202 nodes.
Material parameters, see Table 2.1
concrete
Youngs modulus E
c
MN/m
2
35 000
tensile strength f
ct
MN/m
2
3.5
reinforcing steel
Youngs modulus E
s
MN/m
2
200 000
yield strength f
sy
MN/m
2
500
bond
strength
max
MN/m
2
6.0
slip at strength mm 0.1
residual strength
res
MN/m
2
3.0
slip at residuum mm 1.0
Table 2.1: Material parameters of RC tensile bar example 2.3
*
Concrete material model see Eq. (2.32).
*
Reinforcement material model see Eqns. (2.30), (2.31).
*
Bond model see see Fig. 2.11b. This corresponds to a material behavior per
unit surface of bond. To gain the corresponding values for a whole spring, a
multiplication by U
s
L
I
is neccessary.
Boundary conditions
*
Zero displacement reinforcement node on left boundary, prescribed displace-
ment u
R
= 2.4 mm for reinforcement node on right side incrementally applied
in 100 steps corresponding to a mean strain
mean
= 2.4 .
Results: monotonic loading
Load-displacement curve see Fig. 2.12b
*
state I: Uncracked
*
state IIa: Ongoing cracking. Each sudden load decrease corresponds to a crack.
Stiffness is reduced after each crack.
*
state IIb: Final cracking state before rebar yielding.
*
state III: Limit state with rebar yielding with a limit load P
u
= A
s
f
sy
=
0.1005 MN at a displacement u
R
= 2.14 mm
37
Figure 2.13: Example 2.3 a) concrete stresses b) steel stresses
*
Stiffness of bare reinforcement EA = 40 MN. Limit load of P
u
= 0.1005 MN
applied on steel alone leads to a displacement u
R
= 2.5 mm compared to the
actual displacement of u
R
= 2.14 mm (tension stiffening effect).
Stress distributions along bar at beginning limit state (u
R
= 2.14 mm).
*
Concrete stresses see Fig. 2.13a with zero stresses at a crack. Three cracks
according to three peaks in the load-displacement curve, see Fig. 2.12b.
*
Reinforcement stresses see Fig. 2.13b with peak stresses at a crack. Yield
strength is reached at each crack.
*
In between cracks forces are transferred among concrete and reinforcement by
bond stresses, see Fig. 2.14a. Bond stresses have maximum absolute values in
cracks and change sign across a crack.
Figure 2.14: Example 2.3 a) bond stresses b) displacements
Displacements at beginning limit state
*
See Fig. 2.14b. Displacements of concrete and reinforcement are different at
a bars cross section due to exible bond. The difference results in slip, which
connected to bond stresses, see Section 2.6.
Crack widths by Eq. (2.29), where element length L
I
and crack band width are
chosen to be equal. Thus, in a cracked concrete element crack width is given by
the displacement difference of the end nodes. According to Fig. 2.14b here we get
typically w 0.4 mm in the beginning limit state.
end example 2.3
38 2.8 Tension Stiffening for Reinforced Tension Bar
2.8 Tension Stiffening for Reinforced Tension Bar
Mean strain of tension bar
There is some contribution of concrete between cracks with reduced reinforcement
stresses between cracks, see Fig. 2.13b.
It is assumed that reinforcement has not yet reached its yield limit. Therefore, rein-
forcement strain has the same shape as the reinforcement stress.
We consider the mean strain of the reinforcement which corresponds to the mean
strain of the tension bar.
*
Mean strain refers to some reference length in the center area of the tension bar
spanning several cracks. It results fromthe integration of strains in the reference
length (displacement of reference length) divided by reference length itself.
Stress depending on mean strain
Stress of tension bar is given by reinforcement stress in cracks. This value also
corresponds to the force of the tension bar.
Stress depending on mean strain corresponds to the course of Fig. 2.12b. It shows a
higher stiffness compared to the stiffness of the reinforcement alone. The difference
is called tension stiffening effect.
Stabilized cracking is distinguished from ongoing cracking with single cracks in the fol-
lowing. For a discussion of crack types see e.g. [HCH09].
Figure 2.15: Models for a) cracking b) tension stiffening
Estimation of tension stiffening effect for stabilized cracking
The mean value of reinforcement strain between cracks may be estimated with
sm
=
1
E
s
(
sr

t

s
) (2.33)
with reinforcement stress
sr
in a crack, reinforcement stress difference
s
from
crack to minimum value between cracks, and a parameter
t
for the shape of re-
inforcement stress distribution, see Fig. 2.15a. For a discussion of
t
see e.g.
[HCH09]. Common values are assumed in the range 0.4
t
0.6.
39
In case of stabilized crack pattern the stress difference is given by
s
=
f
ct
eff
(2.34)
with the concrete tensile strength f
ct
and the effective reinforcement ratio
eff
.
Eqns. (2.33), (2.34) lead to
sr
= E
s
sm
+
t
f
ct
eff
(2.35)
Eqn. (2.35) leads to a shift of the pure reinforcement stiffness to the left, see Fig. 2.15a,
i.e. a given mean strain has a higher stress with tension stiffening. This is limited by
the yielding plateau of the reinforcement.
Ongoing cracking with single cracks
In case of single cracks a suitable reference length is hard to determine, as crack
spacing is irregular and areas affected by cracks alternate with such not affected by
cracks which have a full utilization of concrete.
The envelope
2
of the a stress-strain curve is relatively at. Starting with the initial
reinforcement stress
sr,i
= f
ct
/
eff
in a crack immediately after cracking and a
corresponding mean strain f
ct
/E
c
immediately before cracking an assumption
sr
= k
_
sm

f
ct
E
c
_
+
f
ct
eff
(2.36)
is made with a free parameter k. This line should meet the line Eq. (2.35) at a stress
value
sr,i
leading to
k = E
c
1

t
n
eff
, n =
E
s
E
c
(2.37)
whereby experience shows that 1.3 can be assumed.
The strain
sm
belonging to the meeting point can be determined with Eq. (2.35),
which leads to
sm
=
1
E
s
f
ct
eff
(
t
) (2.38)
A simplied version might straighten the initial kink, see Fig. 2.15b.
This leads to
sr
=

sr,i
sm
sm
(2.39)
in the initial range.
Further references: [Com93, 3.2], [DAf03, zu 8.5.1].
2
Envelope to saw tooth behavior during ongoing cracking.
40 2.9 Cyclic Loading of Reinforced Tension Bar
2.9 Cyclic Loading of Reinforced Tension Bar
Reinforcement should not yield in case of cyclic loading. Thus, linear elastic behavior of
reinforcement will be assumed. Furthermore, a stabilized cracking state is assumed.
First unloading
Initially along loading path of stabilized crack state
Upon further unloading crack closure resistance follows as rough crack surfaces do
not t exactly together.
Residual displacements / strains remain compared to uncracked state in stress free
state.
Figure 2.16: Model for cyclic loading
Reloading
Degradation of bond initially lower reloading stiffness compared to nal unload-
ing stiffness.
After full utilization of concrete same tangential stiffness as reinforcement. But the
connection point / area to this tangential stiffness is shifted compared to that of the
previous unloading path.
Further unloading-reloading cycles follow the same pattern with lower initial reloading
stiffness compared to previous nal unloading stiffness. But this effect becomes less
pronounced.
After a large number of loading cycles the stress-strain behavior of tensions bar should
converge against stress-strain behavior of reinforcement alone, with some horizontal shift
due to crack closure effect.
A corresponding stress-strain relation is given by
sr
= E
s
sm

s,c
(2.40)
see Fig. 2.15b, but the value
s,c
is hard to estimate.
Total crack closure w = 0 is connected with ein eigenstress state, i.e. concrete com-
pressive stresses and reinforcement tensile stresses which upon integration equili-
brate and have no resulting tensile bar force and stress, respectively.
See also Fig. (2.16) and [Com93, ?].
Bibliography
[Com93] COMITE EURO-INTERNATIONAL DE BETON: CEB-FIP Model Code 1990. London
: Thomas Telford, 1993
[CS94] CHEN, W.F. ; SALEEB, A.F.: Constitutive Equations for Engineering Materials,
Volume 1: Elasticity and Modeling. 2. Auage. Amsterdam : Elsevier Science B.V.,
1994
[DAf03] DAFSTB: Erluterungen zu DIN 1045-1 / Deutscher Ausschuss fr Stahlbeton.
Berlin, 2003 (Heft 525). Forschungsbericht
[din08] DIN 1045-1: Tragwerke aus Beton, Stahlbeton und Spannbeton. Teil 1: Bemessung
und Konstruktion. : DIN 1045-1: Tragwerke aus Beton, Stahlbeton und Spannbeton.
Teil 1: Bemessung und Konstruktion, August 2008
[HC07] HUSSLER-COMBE, U.: Zur Verwendung von Stoffgesetzen mit Entfestigung in
numerischen Rechenverfahren. In: Bauingenieur 82 (2007), S. 286298
[HCH09] HUSSLER-COMBE, U. ; HARTIG, J.: Rissbildung von Stahlbeton bei Zwang-
beanspruchungen. In: Bauingenieur 84 (2009), S. 546556
[JB01] JIRAZEK, M. ; BAZANT, Z.: Inelastic Analysis of Structures. 1. New York : John
Wiley & Sons, 2001
[Mal69] MALVERN, L. E.: Introduction to the Mechanics of a Continuous Medium. 1. Au-
age. Englewood Cliffs, New Jersey : Prentice-Hall, 1969
41
Chapter 3
2D Structural Beams and Frames
3.1 General Cross Sectional Behavior
3.1.1 Kinematics
Straight reference axis along beam direction
Coordinate systemx, z with x in the longitudinal direction and z in the transverse
direction, see Fig. 3.1.
Bernoulli-hypothesis (cross sections remain plane during deformation)
Displacements
w(x, z) = w(x, 0)
= w(x)
u(x, z) = u(x) z (x)
= u(x) z
_
w(x)
x
(x)
_
(3.1)
with a cross section rotation angle (x), a shear angle (x) and a displacement
u(x), w(x) of the reference axis.
u(x,z)
z
1
z
2
x,u
w
w(x,z)
z,w

w
Figure 3.1: Beam kinematics
Strains, compare Eqns. (1.1)-(1.3)
x
(x, z) =
u
x
=
u
x
z
_
2
w
x
2

x
_
xz
(x, z) =
u
z
+
w
x
=
w
x
+ +
w
x
=
(3.2)
42
3.1.1 Kinematics 43
In the following a notation /x =
,
2
/x
2
=
is used for abbreviation.

The overbars on u, w will be omitted in the following. To simplify the notation u, w will
be written instead.
Considering shear deformations, deformations are advantageously dened as
(x) =
u
x
, (x) =
= w
(3.3)
With Eq. (3.2) this leads to strains
x
(x, z) = (x) z (x) (3.4)
linearily varying along the beam height with extremal values at the top and bottom of the
cross section.
In the following , , are chosen as deformation variables.
strain of the reference axis
curvature of deformed cross sections
1
shearing angle of deformed cross sections relative to reference axis
To describe material behavior, deformation variables have to be connected to force vari-
ables, which are moment M, normal force N and shear force Q in case of 2D beams.
Generally, the following dependencies are assumed
M = M(, ), N = N(, ), Q = Q() (3.5)
Basics of beam theory look quite simple, but there are some little inconsistencies.
A shear strain
xz
, which is constant over the cross section leads to non-vanishing
shear stress at the lower and upper side of a beam. But this contradicts the local
equilibrium conditions.
On the other hand, a parabolic or other nonlinear course of shear stresses according
to equilibrium conditions with linear normal stresses leads to a curved course of
shear strains with vanishing values on top and bottom sides.
These contradictions can be resolved with plate theory. Plane beam theory is its
limiting case or a very useful approximation, respectively.
1
This is different compared to the curvature of the deection w of the reference axis. Both are related by Eq. (3.3).
44 3.1 General Cross Sectional Behavior
3.1.2 Linear elastic behavior
We consider Eq. (1.21) with
y
= 0, i.e.
y
=
x
, and gain as normal stresses
x
(x, z) = E
x
= E [(x) z (x)] (3.6)
and as shear stresses with the shear modulus G
xz
(x) = G
xz
, G =
E
2(1 +)
(3.7)
Internal forces
N =
_
z
2
z
1
x
bdz = E
_
z
2
z
1
bdz (x) E
_
z
2
z
1
z bdz (x)
M =
_
z
2
z
1
x
z bdz = E
_
z
2
z
1
z bdz (x) +E
_
z
2
z
1
z
2
bdz (x)
Q =
_
z
2
z
1
xz
bdz = G
_
z
2
z
1
xz
bdz = G
_
z
2
z
1
bdz (x)
(3.8)
The shear correction factor is used to compensate the difference between mean
shearing strain/stress over the cross section and the shearing strain/stress , G in
the reference point with z = 0.
In case of a rectangular cross section shape with a reference axis through the mid
points it is = 5/6. This is only valid in the linear elastic case. Other cross section
shapes have other values .
Section properties with cross-sectional area A, sectional modulus S and second moment
of area J
A =
_
z
2
z
1
bdz, S =
_
z
2
z
1
z bdz, J =
_
z
2
z
1
z
2
bdz (3.9)
Generally it is assumed, that the reference axis x coincides with the centre of area, which
has a condition
S =
_
z
2
z
1
z bdz = 0 (3.10)
which formally simplies a lot in the linear elastic case, but is not mandatory.
Finally, in the linear elastic case under the condition of Eq. (3.10) we get the well known
relations
2
N = EA, M = EJ , Q = GA (3.11)
or material stiffness which equals the tangential material stiffness
C = C
T
=
_
_
EA 0 0
0 EJ 0
0 0 GA
_
_
(3.12)
or in case of a theory neglecting shear deformations
C = C
T
=
_
EA 0
0 EJ
_
(3.13)
which altogether is the simple linear elastic form of Eqns. (3.5).
2
Sign for moment is different compared to classical structural beam. The difference results from a different
orientation of the z-axis, see Fig. 3.1.
3.1.3 Cracked reinforced concrete behavior 45
3.1.3 Cracked reinforced concrete behavior
3.1.3.1 Compressive zone and internal forces
For cross sections of reinforced concrete structures the linear elastic relations
3
hold only
until the tensile strength of the concrete is reached.
In the following cracked sections are treated, i.e. beam strains
x
exceed the tensile limit
strain
ct
of concrete.
More kinematic items
The reference axis is placed in the centre of the cross section, thus the bottom side
has a coordinate z = z
1
= h/2 and the top side z = z
2
= h/2.
We look at the strain
1
at the bottomside z
1
= h/2 and
2
at the top side z
2
= h/2
with the cross section height h. Eq. (3.4) leads to
1
= z
1
= +
h
2
,
2
= z
2
=
h
2
(3.14)
and further to
=

1

2
h
, h = z
2
z
1
(3.15)
Correct signs for strains have to be considered, e.g.
1
> 0,
2
< 0 in bending with
compression on the top side.
The coordinate of a line (in width direction) with a given strain
x
=
is determined
with Eq. (3.4)
z
(3.16)
e.g. the zero line
= 0 is determined with
z
0
=

(3.17)
The vertical z-coordinates of the concretes tensile or compressive limit strain may
be determined in the same way.
Strains in cracked concrete sections are ctitious values regarding concrete.
The lower reinforcement has a coordinate z
s1
= h/2+d
1
und the upper reinforce-
ment z
s2
= h/2 d
2
, where d
1
, d
2
give the edge distances of both reinforcements.
Eq. (3.4) leads to reinforcement strains
s1
=
_
h
2
d
1
_
,
s2
= +
_
h
2
d
2
_
(3.18)
Perfect bond is implicitely assumed.
3
The reinforcement part has then to be considered with a weighting factor n = E
S
/E
c
.
Concrete compressive zone
Edge strains according to Eq. (3.14) and zero line Eq. (3.17) determine the size of
the concrete compressive zone:
z
0
< h/2 and
1
< 0 cross section totally under compression
z
0
< h/2 and
1
0 totally under tension
h/2 z
0
h/2 and
2
< 0 upper bending compressive zone
h/2 z
0
h/2 and
1
< 0 lower bending compressive zone
z
0
> h/2 and
2
< 0 totally under compression
z
0
> h/2 and
2
0 totally under tension
Cases with compression or tension of total cross section may also include bending
moments and normal forces. Concrete contribution is assumed within a cross section
range z
c1
, z
c2
:
cross section totally under compression z
c1
= z
1
, z
c2
= z
2
totally under tension no concrete contribution
upper bending compressive zone z
c1
= z
0
, z
c2
= z
2
lower bending compressive zone z
c1
= z
1
, z
c2
= z
0
To consider a concretes restricted compressive limit strain or an tensile limit strain
larger zero due to a tensile strength the value z
0
has to be replaced according to Eq. (
3.16) with appropriate values chosen for
.
Resulting internal forces
Normal force
N = A
s1
s1
+A
s2
s2
+
_
z
c2
z
c1
c
bdz (3.19)
where uniaxial reinforcement stresses
s1
,
s2
and conrete stresses
c
are functions
of
x
, see Eq. (3.4). Correct stress signs have to be considered.
Bending moment
M = A
s1
s1
z
s1
A
s2
s2
z
s2

_
z
c2
z
c1
c
z bdz (3.20)
for sign conventions see Fig. 3.3.
A variable cross section width b(z) may be regarded within a numerical integration of
Eqns. (3.19), (3.20), (3.43), (3.44).
3.1.3.2 Linear concrete compressive behavior
A linear stress-strain behavior is assumed for concrete in its compressive zone
= E
c
, E
c
=
_
E
c
0
0 else
(3.21)
excluding a tensile strength. A limited tensile strength may easily be incorporated.
A cross section with height h is considered. A beam strain , given, a compressive zone
is determined according to Page 46. The compressive zone has an extension z
c1
z
z
c2
with a height h
c
= z
c2
z
c1
and edge strains
ce
= B
(3.22)
with
ce
=
_

c1
c2
_
, B
=
_
1 z
c1
1 z
c2
_
, =
_

_
(3.23)
This leads to edge stresses
ce
= E
ce
= E
c
B
,
ce
=
_

c1
c2
_
(3.24)
Following cross sectional values will be used
A
c
=
_
z
c2
z
c1
bdz, S
c
=
_
z
c2
z
c1
z bdz, J
c
=
_
z
c2
z
c1
z
2
bdz (3.25)
Extremal stresses at z
c1
, z
c2
are given by
c1
,
c2
with a linear course in between accord-
ing to Eq. (3.21). They are interpolated with
c
(z) =

c1
z
c2

c2
z
c1
z
c2
z
c1
+

c2

c1
z
c2
z
c1
z (3.26)
Thus, concrete contributions to internal forces are given by
N
c
=
_
z
c2
z
c1
c
(z) bdz =

c1
z
c2

c2
z
c1
h
c
A
c

c1

c2
h
c
S
c
M
c
=
_
z
c2
z
c1
c
(z)z bdz =
c1
z
c2

c2
z
c1
h
c
S
c
+

c1

c2
h
c
J
c
(3.27)
see also Eq. (3.8), or
c
= A

ce
= A B

ce
(3.28)
with
c
=
_
N
c
M
c
_
, A =
_
A
c
S
c
S
c
J
c
_
, B
=
1
h
c
_
z
c2
z
c1
1 1
_
(3.29)
whereby
c
and
ce
have to be clearly distinguished in the following!
Eq. (3.21) is applied to stresses
ce
,
ce
at z
c1
, z
c2
, i.e. leading to a material stiffness
c
= C (3.30)
with
C = E
c
A B
= E
c
A = E
c
_
A
c
S
c
S
c
J
c
_
(3.31)
In case of rectangular cross sections with width b and a height h this evaluates to A
c
=
b(z
c2
z
c1
), S
c
=
b
2
(z
2
c2
z
2
c1
), J
c
=
b
3
(z
3
c2
z
3
c1
), in case of z
c1
= h/2, z
c2
= h/2 to
A
c
= bh, S
c
= 0, J
c
= bh
3
/12. Contrarily to Eq. (3.11) the Eq. (3.30) couples a normal
force N
c
to and a bending moment M
c
to .
As the compressive zone limits z
c1
, z
c2
may change due to change of in , the relation
Eq. (3.30) is basically nonlinear. To consider this effect Eq. (3.28) is written as
_

N
c
M
c
_
=
_
N
c
c1
N
c
c2
M
c
c1
M
c
c2
_
_

c1

c2
_
+
_
N
c
z
c1
N
c
z
c2
M
c
z
c1
M
c
z
c2
_
_
z
c1
z
c2
_
(3.32)
or

c
= A

ce
+A
z
z
c
(3.33)
with A
according to Eqs. (3.28,3.29). To simplify A

z
a linear variation of width b is
assumed with b
1
= b(z
c1
), b
2
= b(z
c2
). This yields
A
z
=
1
6
_
_
c1
b
2
2
c1
b
1
2
c2
b
2

c2
b
1

c1
b
2
+ 2
c1
b
1
+ 2
c2
b
2
+
c2
b
1
(b
2
z
c1
+z
c2
b
2
+z
c1
b
1
)
c2
+
(b
2
z
c1
z
c2
b
1
+ 3z
c1
b
1
)
c1
(b
2
z
c1
3z
c2
b
2
z
c2
b
1
)
c2
(z
c1
b
1
+z
c2
b
2
+z
c2
b
1
)
c1
_
_
(3.34)
The values z
c1
, z
c2
stand for zero lines given by
z
0
=

, z
0
=

2
(3.35)
see Eq. (3.17), or upper or lower edges of the concrete compression zone. The following
cases have to be considered:
1. Dominating bending with lower compression zone z
c1
= h/2, z
c2
= z
0
< h/2
and z
c1
= 0,
c2
= 0,
c2
= 0 and
_
z
c1
z
c2
_
=
_

h
2
_
,
_
z
c1
z
c2
_
=
_
0 0
1
2
_
_

_
(3.36)
2. Dominating normal forces with fully compressed cross section z
c1
= h/2, z
c1
=
0, z
c2
= h/2, z
c2
= 0, h
c
= h and
_
z
c1
z
c2
_
=
_

h
2
h
2
_
,
_
z
c1
z
c2
_
=
_
0 0
0 0
_
_

_
(3.37)
3. Dominating bending with upper compression zone z
c1
= z
0
> h/2, z
c2
= h/2
and z
c2
= 0,
c1
= 0,
c1
= 0 and
_
z
c1
z
c2
_
=
_

h
2
_
,
_
z
c1
z
c2
_
=
_
1
2
0 0
_
_

_
(3.38)
Anyway, we set
z
c
= B
z
(3.39)
We use Eqs.(3.22,3.39) and (3.24) to formulate the edge strains and stresses

ce
= B
z
c
= B
B
z
= (B
B
z
)

ce
= E
c

ce
= E
c
(B
B
z
)
(3.40)
This yields a tangential material stiffness together with Eq. (3.33)

c
= E
c
A
(B
B
z
) +A
z
B
z

=
_
E
c
A
(B
B
z
) +A
z
B
z

= C
T

(3.41)
also considering the change of the extension of the concrete compressive zone.
A reinforcement can be superposed.
3.1.3.3 Nonlinear concrete compressive behavior
Material behavior
Concrete uniaxial stress
c
see Eq. (2.1) or alternatives Modelcode 90 [Com93,
2.1.4.4.1] and DIN 1045-1 [din08, 9.1.5, 8.5.1].
Reinforcement uniaxial stress
s
see Eq. (2.30) or alternatives Modelcode 90 [Com93,
2.2.4] and DIN 1045-1 [din08, 9.2.3, 8.5.1]
Tangential material stiffness for bending moment and normal force
Strain dependence on deformation variables according to Eq. (3.4) leads to
= 1,

x
= z (3.42)
With Eq. (3.19)
N
= A
s1
s1
+A
s2
s2
+
_
z
c2
z
c1
bdz
= A
s1
s1
x
+A
s2
s2
x
+
_
z
c2
z
c1
x
bdz
N
= A
s1
s1
+A
s2
s2
+
_
z
c2
z
c1
bdz
= A
s1
s1
x
z
s1
A
s2
s2
x
z
s2

_
z
c2
z
c1
x
z bdz
(3.43)
With Eq. (3.20)
M
= A
s1
s1
z
s1
A
s2
s2
z
s2

_
z
c2
z
c1
z bdz
= A
s1
s1
x
z
s1
A
s2
s2
x
z
s2

_
z
c2
z
c1
x
z bdz
=
N
= A
s1
s1
z
s1
A
s2
s2
z
s2

_
z
c2
z
c1
z bdz
= A
s1
s1
x
z
2
s1
+A
s2
s2
x
z
2
s2
+
_
z
c2
z
c1
x
z
2
bdz
(3.44)
Basically the derivatives of the longitudinal stresses
s1
/
x
,
s2
/
x
,
c
/
x
are needed, the latter varying with z in the limits z
c1
, z
c2
, see Fig. (3.1), whereby
dependency of integration limits z
c1
, z
c2
on , has been neglected.
The tangential material stiffness, see Eq. (1.26), is given with
C
T
=
_
N
_
(3.45)
which again couples normal force to and moment to .
Linear elastic, symmetric system as special case with
s1
/
x
=
s2
/
x
=
E
s
,
c
/
x
= E
c
and
_
bdz = A,
_
z bdz = 0,
_
z
2
bdz = J.
To include shear forces Eq. (3.45) has to be extended with
C
T
=
_
_
N
_
(3.46)
Calculation types
Specication of strain , curvature calculation of M, N
*
Normal force with Eq. (3.19), moment with Eq. (3.20). Integration of concrete
stresses analytically or numerically.
Specication of curvature , normal force N calculation of M,
*
With a given N, the nonlinear equation f(, ) N = 0 has to be solved for
. This may efciently be done with a Newton-Raphson iteration, i.e.
(+1)
=
()
1
f
=
()
_
f(
()
, ) N
_
(3.47)
see Eq. (1.65). A starting value is
(0)
= 0. f/ is given with Eq. (3.43)
1
.
*
With given and calculated, M = f(, ) can be determined.
Specication of moment M, normal force N calculation of ,
*
From Eqns. (3.8,3.10) a nonlinear algebraic system arises for unknowns ,
0
after evaluation of integrals of
s1
,
s2
,
c
. This may principally be solved with
the Newton-Raphson method, see Page 15.
Example 3.1 Prescribed curvature , normal force N calculation M,
Parameters
Cross section h = 0.4 m, b = 0.2 m. Concrete C30/37, f
cR
= 21.675 MN/m
2
,
E
c0
= 31 900 MN/m
2
. Uniaxial concrete behavior with DIN 1045-1 [din08, 9.1.5,
8.5.1].
Upper and lower reinforcement each 4 20, A
s2
= A
s1
= 12.57 cm
2
, d
2
= d
1
=
5 cm. Reinforcement behavior with DIN 1045-1 [din08, 9.2.3, 8.5.1].
Figure 3.2: Moment-curvature curve Example 3.1
Initial bending stiffness concrete alone EJ
c
= 34.0 MN/m
2
. Initial bending stiffness
reinforcement alone EJ
s
= 11.3 MN/m
2
. Total initial stiffness EJ = 45.3 MN/m
2
.
Given values of normal forces N
Ed
= 0, 1, 2 MN.
Curvature is varied within certain bounds. Therefore, N, serve as input values. The
computed moments are of primary interest and lead to moment-curvature curves, see
Fig. 3.2.
The initial stiffness is ruled by the following factors: (1) initial height of the com-
pression zone, (2) initial values of concrete stress, higher compression leads to lower
tangential stiffness due to the nonlinearity of the stress-strain curve, see Fig. 2.1.
The upper kinks result from the yielding of reinforcement.
The signed end points roughly mark the state, when ultimate concrete compressive
stress of
c1u
= 0.0035 is reached.
The behavior of the uncracked, linear elastic cross section is shown as reference, further-
more the case with reinforcement alone.
End Example 3.1
Generalization Fiber models
Every line along the beam axis cut by the cross section may be regarded as a ber.
Each ber is strained according to beam kinematics a leads to a stress.
With the numerical integration of stresses to result in internal forces any type of a
uniaxial material law may be used for bers.
Efciency of ber models is inuenced by numerical integration methods.
*
Commonly used methods like trapezoidal rule or Simpson rule may be used.
*
Gauss integration is not optimal, as the important upper and lower edges are not
captured. Lobatto integration is an alternative.
52 3.2 Equilibrium of Bars
3.2 Equilibrium of Bars
Preliminaries
We regard a loading p
x
(x, t), p
z
(x, t) varying with place x and time t. Inertial ef-
fects will be taken into consideration. Therefore, the bars (inertial) mass per length
m and the 2nd moment of inertia have to be regarded.
In the following a notation /t = ,
2
/t
2
= is used for abbreviation.
To simplify the formulation another kinematic variable
= w
= (3.48)
is used in the following, see Eq. (3.3),which describes the cross section rotation due
to bending.
Figure 3.3: Beam equilibrium
Strong differential equilibrium of an innitesimal bar element is described by
p
x
+N
= m u
p
z
+Q
= m w
Q+M
=
(3.49)
according to Newtons law force = mass acceleration with the accelerations u, w, , =
w
.
The linear elastic case
From Eq. (3.11) we have
N = EA = EAu
, M = EJ = EJ
, Q = GA
(3.50)
with A
= A.
m u EAu
= p
x
m w GA
= p
z
( w
) EJ (w
) = GA
(3.51)
Eq. (3.51)
1
represents the one dimensional wave equation
4
.
4
We can set m = A with the specic mass and p
x
= 0 to gain the familiar form.
53
In case of a normal, slender beam the shear rotation is small compared to the
total rotation w
. The same holds for the derivatives. Thus, we have
and
w
.
Furthermore, the term ( w
) is neglected as 1.
Combining the derivative of Eq. (3.51)
3
and Eq. (3.51)
2
nally leads to
m w +EJ w
= p
z
(3.52)
representing the equation of dynamic beam bending.
Weak integral equilibrium
Weak equilibrium formulations utilize test functions or so called virtual displace-
ments u, w, , = w
. These test functions have to be kinematically

compatible, i.e. the rst derivative of u, w, , should exist.
An equivalent to the strong differential equilibrium formulation Eq. (3.49) is given
by a weak integral equilibrium formulation regarding a bar of length L
L
_
0
um u dx
L
_
0
uN
dx +
L
_
0
wm wdx
L
_
0
wQ
dx +
L
_
0
dx
L
_
0
M
dx
=
_
L
0
u p
x
dx +
_
L
0
w p
z
dx +
L
_
0
(w
) Qdx
(3.53)
Generally, the solutions of Eq. (3.53) solve also Eq. (3.49) and vice versa.
Those terms with derivatives of internal forces are partially integrated:
_
L
0
uN
dx =
_
u(L) N(L) u(0) N(0)

_
L
0
N dx
_
L
0
M
dx =
_
(L) M(L) (0) M(0)

_
L
0
M dx
_
L
0
wQ
dx =
_
w(L) Q(L) w(0) Q(0)

_
L
0
w
Qdx
(3.54)
with virtual deformations = u
, =
. Combining Eq. (3.53) leads to

_
L
0
um udx +
_
L
0
dx +
_
L
0
wm wdx+
+
_
L
0
N dx +
_
L
0
M dx +
_
L
0
Qdx
=
_
L
0
u p
x
dx +
_
L
0
w p
z
dx +
_
uN
L
0
+
_
M
L
0
+
_
wQ
L
0
(3.55)
where boundary terms have been abbreviated. The last equation may be interpreted
as virtual work principle and has the following parts:
*
Inertial forces with the rst three terms of the left hand side.
*
Internal forces with the last three terms of the left hand side.
*
Distributed loading with the rst two terms of the right hand side.
*
Finally, the boundary terms with the forces at the beams ends with the last
three terms of the right hand side.
54 3.2 Equilibrium of Bars
A generalizing matrix notation of Eq. (3.55) is given with
_
L
0
u
T
m udx +
_
L
0
T
dx =
_
L
0
u
T
pdx +U
T
t (3.56)
with
u =
_
u w
_
T
, p =
_
p
x
p
z
0
_
T
, m=
_
_
m 0 0
0 m 0
0 0
_
_
T
=
_

_
T
, =
_
N Q M
_
T
U =
_
u
0
w
0

0
u
L
w
L

L
_
T
t =
_
N
0
Q
0
M
0
N
L
Q
L
M
L
_
T
(3.57)
with
L
= (L),
0
= (0) and = w
, =
= w
.
*
Actually Eqn. (3.54) requires a formulation
t =
_
N
0
Q
0
M
0
N
L
Q
L
M
L
_
T
, but regarding nite element
discretization a global sign convention is more appropriate: internal forces on
the left cross section are assumed as positive with the same directions as on the
right cross section.
As to the beams ends, i.e. its boundary conditions, they have to be prescribed with
one element out of each the pairs (u, N), (w, Q), (, M) for every end. In case of
prescribed forces, they are marked with an overbar, i.e. N

N . . .
Disregarding shear deformations (Bernoulli-Beam)
*
In case of a normal, slender beam the shear rotation is small compared to the
total rotation w
. The same holds for the derivatives and it is set

= w
, =
= w
(3.58)
*
With = 0 a contribution Q vanishes in Eq. (3.55). Thus, we set
=
_

_
T
, =
_
N M
_
T
(3.59)
Shear force Q is determined via Eq. (3.49)
3
, which also holds for = 0.
Disregarding of rotational inertia
*
The inertial mass moment generally is relatively small. Thus, we may neglect
the corresponding contributions and set
u =
_
u w
_
T
, p =
_
p
x
p
z
_
T
, m=
_
m 0
0 m
_
T
(3.60)
The quasistatic case: m= 0 in Eq. (3.56).
Boundary conditions
Kinematic boundary conditions: It is easy to choose displacement trial functions
such that kinematic boundary conditions are fullled by them. Thus, test functions
or virtual displacement may be set to zero along boundaries with kinematic bound-
ary conditions. This will simplify Eqns. (3.55) or (3.56).
Force boundary conditions: Force boundary conditions may arise with the term

t in
Eq. (3.57). It may be shown that the solutions of the integral equation (3.56) fulll
the differential equation (3.49) and prescribed force boundary conditions, under the
assumption that kinematic boundary conditions are fullled by the trial function
choice.
55
3.3 Structural Beam Elements for 2D
In the following, specic forms will be chosen for the interpolation of displacement vari-
ables u, w, with Finite Elements, see Sections 1.3, 1.5.
The beam with a coordinate range 0 x L is subdivided into N
e
elements. In a
rst approach each element has two nodes. Totally, there are N = N
e
+ 1 nodes.
An element I, I = 1 . . . N
e
, has global nodal coordinates x
I1
, x
I2
and a length
L
I
= x
I2
x
I1
.
Moreover, an element I has a local coordinate r. The relation between local and
global coordinates is given with
x =
_
1
2
(1 r)
1
2
(1 +r)
_
x
I1
x
I2
_
(3.61)
with x(1) = x
I1
, x(1) = x
I2
. The inverse relation is r = (2x x
I2
x
I1
)/L
I
leading to a Jacobian J
J =
r
x
=
2
L
I
(3.62)
which is used for numerical integration, see Eq. (1.61).
To interpolate displacement variables within an element polynomial forms
y(r, t) =
n
i=0
a
i
(t) r
i
(3.63)
are chosen, where the coefcients a
i
are functions of time t, while y may stand
for the displacement variables u, w and . A particular formulation yields a trial
function.
Test functions are selected in the same way as the trial functions, i.e.
y(r, t) =
n
i=0
a
i
(t) r
i
(3.64)
(Bubnov-Galerkin-approach).
Deformation variables
Deformation variables , , are dened from displacement variables according to
Section 3.1.1. With a given trial function they are derived on base of the derivatives
of Eqns. (3.63), (3.62).
Due to the required integrability of Eq. (3.56), see also Section 1.6 and the compatibility
requirement, some constraints for trial and test functions arise, as their derivatives should
be nite across elements.
In case of Bernoulli kinematics = w
and = w
, i.e. deformation results from

2nd derivative of displacements. Thus, rst derivations of test functions have to be
continuous across elements (C
1
-continuity).
In case of Timoshenko kinematics =
and and w are decoupled by the shear

deformation . Deformations result from 1st derivatives of displacements and con-
tinuity of test functions across elements is sufcient (C
0
-continuity).
56 3.3 Structural Beam Elements for 2D
2D Bernoulli beam element
For Bernoulli beam see Page 54. u, w, = w
remain as variables. The latter

enforces a coupling of w and . A reasonable approach is given with
u(r, t) =
_
1
2
(1 r)
1
2
(1 +r)
_
u
I1
(t)
u
I2
(t)
_
w(r, t) =
_
r
3
4

3r
4
+
1
2
L
I
r
3
8

L
I
r
2
8

L
I
r
8
+
L
I
8

r
3
4
+
3r
4
+
1
2
L
I
r
3
8
+
L
I
r
2
8

L
I
r
8

L
I
8
_
_
_
_
_
w
I1
(t)
I1
(t)
w
I2
(t)
I2
(t)
_
_
_
_
w
(r, t) =
w
r
r
x
=
2
L
I
_
3r
2
4

3
4
3L
I
r
2
8

2L
I
r
8

L
I
8

3r
2
4
+
3
4
3L
I
r
2
8
+
2L
I
r
8

L
I
8
_
_
_
_
_
w
I1
(t)
I1
(t)
w
I2
(t)
I2
(t)
_
_
_
_
(3.65)
yielding u(1, t) = u
I1
(t), u(1, t) = u
I2
(t) and w(1, t) = w
I1
(t), w(1, t) =
w
I2
(t) and w
(1, t) =
I1
(t), w
(1, t) =
I2
(t).
Eq. (3.65) is abbreviated with
u(r, t) = N(r) u
I
(t) (3.66)
with
u(r, t) =
_
u(r, t) w(r, t)
_
T
N(r) =
_
1
2
(1 r) 0 0
0
r
3
4

3r
4
+
1
2
L
I
r
3
8

L
I
r
2
8

L
I
r
8
+
L
I
8

_
u
I
(t) =
_
u
I1
(t) w
I1
(t)
I1
(t) u
I2
(t) w
I2
(t)
I2
(t)
_
T
(3.67)
Deformation variables are derived with Eqns. (3.3), (3.58), (3.62), (3.65)
(r, t) =
u
r
r
x
=
1
L
I
_
1 1
_
u
I1
(t)
u
I2
(t)
_
(r, t) =
w
r
r
x
=
4
L
2
I
_
6r
4
6L
I
r
8

2L
I
8

6r
4
6L
I
r
8
+
2L
I
8
_
_
_
_
w
I1
(t)
I1
(t)
w
I2
(t)
I2
(t)
_
_
_
_
(3.68)
This is abbreviated with
(r, t) = B(r) u
I
(t) (3.69)
with
(r, t) =
_
(r, t) (r, t)
_
T
B(r) =
1
L
I
_
1 0 0 1 0 0
0
6r
L
I
3r 1 0
6r
L
I
3r + 1
_
(3.70)
Potential locking risks
*
Longitudinal strains are constant within this element, while curvature is linear.
Thus, articial constraints arise to the coupling of normal forces to curvature
and moments to longitudinal strains.
57
Enhanced 2D two-node Bernoulli beam element
As mentioned before the 2D two-node Bernoulli beam element holds some locking
risks. This should be resolved by an enhanced two-node Bernoulli beam element to
a large degree.
The enhancement is performed with the parabolic extension of longitudinal dis-
placement with a third interior node at a local position r = 0. This node is not
used for interpolation of geometry and lateral deections and carries only one de-
gree of freedom.
The additional degree of freedom leads to a linear longitudinal strain within an ele-
ment, corresponding to the linear curvature.
Extension of Eq. (3.67)
u(r, t) =
_
u(r, t) w(r, t)
_
T
N(r) =
_
1
2
r(r 1) 0 0 1 r
2

0
r
3
4

3r
4
+
1
2
L
I
r
3
8

L
I
r
2
8

L
I
r
8
+
L
I
8
0
_
u
I
(t) =
_
u
I1
(t) w
I1
(t)
I1
(t) u
I2
(t) u
I3
(t) w
I3
(t)
I3
(t)
_
T
(3.71)
and Eq. (3.70)
(r, t) =
_
(r, t) (r, t)
_
T
B(r) =
1
L
I
_
2r 1 0 0 4r 2r + 1 0 0
0
6r
L
I
3r 1 0 0
6r
L
I
3r + 1
_
(3.72)
This type of element may not be found in every nite element package.
58 3.3 Structural Beam Elements for 2D
2D Timoshenko beam element
Trial function choice for displacement variables
u(r, t) =
_
1
2
(1 r)
1
2
(1 +r)
_
u
I1
(t)
u
I2
(t)
_
w(r, t) =
_
1
2
(1 r)
1
2
(1 +r)
_
w
I1
(t)
w
I2
(t)
_
(r, t) =
_
1
2
(1 r)
1
2
(1 +r)
_

I1
(t)
I2
(t)
_
(3.73)
This may again be abbreviated with
u(r, t) = N(r) u
I
(t) (3.74)
with
u(r, t) =
_
u(r, t) w(r, t) (r, t)
_
T
N(r) =
_
_
1
2
(1 r) 0 0
1
2
(1 +r) 0 0
0
1
2
(1 r) 0 0
1
2
(1 +r) 0
0 0
1
2
(1 r) 0 0
1
2
(1 +r)
_
_
u
I
(t) =
_
u
I1
(t) w
I1
(t)
I1
(t) u
I2
(t) w
I2
(t)
I2
(t)
_
T
(3.75)
Deformation variables with Eq. (3.3), (3.48), (3.62)
(r, t) =
u
r
r
x
=
1
L
I
_
1 1
_
u
I1
(t)
u
I2
(t)
_
(r, t) =

r
r
x
=
1
L
I
_
1 1
_

I1
(t)
I2
(t)
_
(r, t) =
w
r
r
x
=
1
L
I
_
1
L
I
2
(1 r) 1
L
I
2
(1 +r)
_
_
_
_
w
I1
(t)
I1
(t)
w
I2
(t)
I2
(t)
_
_
_
_
(3.76)
This may be abbreviated with
(r, t) = B(r) u
I
(t) (3.77)
with
(r, t) =
_
(r, t) (r, t) (r, t)
_
T
B(r) =
1
L
I
_
_
1 0 0 1 0 0
0 0 1 0 0 1
0 1
L
I
2
(1 r) 0 1
L
I
2
(1 +r)
_
_
(3.78)
Potential locking risks
*
In case of pure bending or slender beams with a low bending stiffness compared
to shear stiffness it is 0. Thus, Eq. (3.76)
3
imposes a constraint on the nodal
variables, which does not result from physics but from discretization. For an
analysis of Timoshenko beam locking see ???.
*
A remedy is given with reduced integration, see Pages 60, 63.
59
Enhanced 2D Timoshenko beam element
Trial function choice
Extension of Eq. (3.75)
u(r, t) =
_
u(r, t) w(r, t) (r, t)
_
T
N(r) =
_
_
r(r1)
2
0 0 1 r
2
0
r(1+r)
2
0 0
0
r(r1)
2
0 0 1 r
2
0
r(1+r)
2
0
0 0
(1r)
2
0 0 0 0
(1+r)
2
_
_
u
I
(t) =
_
u
I1
(t) w
I1
(t)
I1
(t) u
I2
(t) w
I2
(t) u
I3
(t) w
I3
(t)
I3
(t)
_
T
(3.79)
and Eq. (3.78)
(r, t) =
_
(r, t) (r, t) (r, t)
_
T
B(r) =
1
L
I
_
_
2r 1 0 0 4r 0 2r + 1 0 0
0 0 1 0 0 0 0 1
0 2r 1
L
I
2
(1 r) 0 4r 0 2r + 1
L
I
2
(1 +r)
_
_
(3.80)
60 3.4 System Building and Solution Methods
3.4 System Building and Solution Methods
As a key element the weak equilibrium condition, see Eqn. (3.56), has to be evaluated.
This performed element by element, see Page 8 and Eqn. (1.33).
For elementwise integral evaluation see Eqn. (1.61). In case of structural beam elements
as described before the Jacobian is given by Eq. (3.62).
Internal nodal forces, e.g., are determined by
f
I
=
_
L
I
B
T
(x) (x) dx =
L
I
2
_
1
1
B
T
(r) (r) dr (3.81)
where Bis given according to the element type chosen and according to Eq. (3.57)
or (3.59)
5
. The integration is performed numerically, see Eq. (1.62)
f
I
=
L
I
2
n
i
i=0
i
B
T
(
i
) (
i
) (3.82)
The same procedure is applied to gain the tangential element stiffness matrix
K
TI
=
L
I
2
n
i
i=0
i
B
T
(
i
) C
T
(
i
) B(
i
) (3.83)
see Eq. (1.40), with the tangential material stiffness according to Eqs. (3.45) or
(3.13) in case of Bernoulli beams, or (3.46), (3.12) in case of Timoshenko beams.
Furthermore the element mass matrix
M
I
=
L
I
2
n
i
i=0
i
N
T
(
i
) m N(
i
) (3.84)
see Eqn. (1.33)
2
, with maccording to Eq. (3.57) and the element distributed loading
p
I
=
L
I
2
n
i
i=0
i
N
T
(
i
) p(
i
) (3.85)
see Eqn. (1.33)
3
, with p according to Eq. (3.57). The elements boundary
6
loading
can be directly taken from Eq. (3.57)
7
t
I
=
_

N
I1

Q
I1

M
I1

N
I2

Q
I2

M
I2
_
T
(3.86)
in case that any nodal loads are prescribed for a particular element.
Required integration order n
i
, see Eq. (1.62), to have an exact numerical integration
depends on polynomial degree of trial functions for deformation variables.
*
The Bernoulli beam element with strains according to Eq. (3.70) yields a stiff-
ness matrix K
TI
with highest polynomial degree 2 in case of constant material
stiffness within an element. This requires n
i
= 1 and two sampling points with
the Gaussian integration scheme, see Page 14
*
The same argument bascially holds for the Timoshenko beam element, see
Eq. (3.78), due to the shear deformation parts. Reduced integration n
i
= 1
neglects the linear contribution of shear deformations but might solve the lock-
ing problem in case of slender beams, see Page 63.
5
Internal forces dimension has to match to row dimension of B.
6
Boundaries are embodied by nodes for one-dimensional elements.
61
A few words about components. According to components of nodal displacements,
see Eq. (3.67)
3
, the vector of internal nodal forces has components
f
I
=
_
N
I1
Q
I1
M
I1
N
I2
Q
I2
M
I2
_
T
(3.87)
in case of the two-node elements, which corresponds to number of rows of the ma-
trix B
T
. Moreover, the nodal normal force N
Ii
, shear force Q
Ii
and bending mo-
ment M
Ii
correspond to the nodal longitudinal displacement u
Ii
, the nodal lateral
displacement w
Ii
and the nodal rotation
Ii
.
In a similar way Eq. (3.85) leads to
p
I
=
_

N
p
I1
Q
p
I1
M
p
I1
N
p
I2
Q
p
I2
M
p
I2
_
T
(3.88)
where

N
p
I1
. . . should not be confused with

N
I1
. . . from

t
I
, see Eq. (3.86).
According to the rules of matrix multiplication the element stiffness matrix K
TI
and
the element mass matrix M
TI
have a dimension 6 6 for 2D two node elements,
with appropriate dimensions for C
T
and m.
In case of the enhanced two-node Bernoulli element an additional component arises
according to Eq. (3.71)
3
with a corresponding nodal force N
I
for the central node.
Transformation to global system
Figure 3.4: Beam orientation in 2D space
Up to now we assumed that the longitudinal axis or local xaxis of a beam element
and the global x-axis have the same direction. But a 2D structural beam may have
an orientation in 2D space. Thus, we have to consider a transformation of vectors in
2D cartesian coordinate systems.
The direction of a straight element is assumed with the rst element node as start
and the last element node as end. A rotation angle is assumed starting from global
xdirection to element direction, see Fig. 3.4.
Regarding Eq. (B.2) and considering two nodes and a perpendicular rotation direc-
tion of the displacement, at rst the following relations hold
u
I
= T u
g
I
, u
g
I
= T
T
u
I
, f
I
= T f
g
I
, f
g
I
= T
T
f
I
(3.89)
with
T =
_
_
cos sin 0 0 0 0
sin cos 0 0 0 0
0 0 1 0 0 0
0 0 0 cos sin 0
0 0 0 sin cos 0
0 0 0 0 0 1
_
_
(3.90)
for two-node elements, where the upper index g indicates that the nodal displace-
ments and internal nodal forces are measured with respect to the global coordinate
system and
T
denotes the transpose of . The Eqns. (3.89) are also valid for incre-
ments du
I
, du
g
I
, df
I
, df
g
I
and for loadings p
I
, p
g
I
,

t
I
,

t
g
I
.
The tangential element stiffness matrix is at rst dened in the local system of a
beam by
df
I
= K
TI
du
I
(3.91)
see Eq. (1.41). Using the transformation rules of Eq. (3.89) and considering T
1
=
T
T
this leads to
T df
g
I
= K
TI
T du
g
I
df
g
I
= T
T
K
TI
T du
g
I
(3.92)
and nally results in a transformation rule for the tangential element stiffness matrix
K
g
TI
= T
T
K
TI
T (3.93)
It has to be pointed out, that this rule is valid only in connection with the sign
conventions of Fig. 3.4. Similar arguments lead to the transformation rule for the
element mass matrix
M
g
TI
= T
T
M
TI
T (3.94)
In case of the enhanced two-node Bernoulli element the transformation matrix T is
given with
T =
_
_
cos sin 0 0 0 0 0
sin cos 0 0 0 0 0
0 0 1 0 0 0 0
0 0 0 1 0 0 0
0 0 0 0 cos sin 0
0 0 0 0 sin cos 0
0 0 0 0 0 0 1
_
_
(3.95)
The central 1 belongs to the longitudinal degree of freedomof the interior node. This
value should not be rotated as it does not have a lateral component by denition.
Beyond the change in T, all transformations remain unchanged.
Assembling of all element contributions
See Page 8. The procedure is basically the same for all element types. Regarding
data organization it has to be considered that different nodes may have a different
number of degrees of freedom with the beam elements, e.g. in case of enhanced
elements.
Regarding a particular node, which is shared by several elements, the elements con-
tributions to nodal forces and nodal loads basically have to sum up to zero with
respect to equilibrium.
63
Consideration of kinematic boundary conditions
Kinematic boundary conditions are applied by prescribing nodal values of u
Ii
, w
Ii
,
Ii
.
To simplify the procedure the quasistatic linear case
K u = p (3.96)
is considered, see Eqns. (1.37), (1.63).
We consider, e.g., the transverse displacement w
Ii
of a node i with a prescribed value
w. If necessary, the value has to be transformed to the global system according to
Eq. (3.89). The global index of this degree of freedom be k, see Page 8. Let N be
the total number of degrees of freedom. To apply the particular boundary condition
one may set
p
i
:= p
i
K
ik
w i = 1 . . . k 1, k + 1 . . . N
p
k
= w
K
kk
= 1
K
kj
= 0 j = 1 . . . k 1, k + 1 . . . N
K
ik
= 0 i = 1 . . . k 1, k + 1 . . . N
(3.97)
There must be enough boundary conditions to prevent rigid body motions.
The particular degree of freedom is released from balancing equilibrium by sum-
ming up nodal force contributions from elements and loadings. The corresponding
internal nodal forces give a support reaction.
The same procedure may principally be used in nonlinear and dynamic cases.
Integration aspects for Timoshenko beam elements
Regarding Eq. (3.83) with n
i
= 1 this leads to a stiffness matrix K
TI
of dimension
6 6 with rank three, as B has a dimension 3 6 acc. to Eq. (3.78) and C
T
has a
dimension 3 3 acc. to, e.g. Eq. (3.12). Thus, an analysis of eigenforms leads to 3
rigid body modes and 3 deformation modes, which makes sense.
The portion of the shear deformation mode is ruled by the parameter k = GAL
2
/EI,
where L is a measure of the total beam span.
Adding another integration sampling point, i.e. n
i
= 2 increases the rank of K
TI
.
This introduces non-physical constraints with restriction of rigid body displace-
ments. This restriction is proportional to k.
Solution method
Cracked reinforced concrete leads to nonlinear models due to the nonlinear relations
between moment, normal force, curvature and longitudinal strain.
Generally, an incremental iterative method is used to solve nonlinear systems, i.e.
a loading history is regarded where a real time ( dynamic or transient case) or a
pseudo time ( quasistatic case) is used to control the load. Thus, increments are
dened by time steps.
Regarding quasistatic problems, the Newton-Raphson method may be an appropri-
ate method for equilibrium iteration within every time step, see Eq. (1.65).
Example 3.2 Simple reinforced concrete beam
Geometry, discretization an boundary conditions
Span L = 5.0 m, cross section width b = 0.2 m, height h = 0.4 m. Discretization
with N
e
= 10 enhanced two-node Bernoulli beam elements, see Page 57.
Hinge bearing of left and right node, i.e. lateral displacements are zero but rotations
are not restricted. Longitudinal displacement of left node is zero, right node not
restricted in longitudinal direction.
Material and loading
Concrete without tensile strength with an initial Youngs modulus E
c0
= 31 900 MN/m
2
in the compressive range. Concrete compressive strength is assumed with f
cd
=
21.675 MN/m
2
with
c1
= 0.0023,
c1u
= 0.0035 according to DIN 1045-1
[din08, 9.1.5, 8.5.1].
Reinforcing steel 4 20, A
s1
= 12.57 cm
2
, d
1
= 5 cm. Uniaxial reinforcing steel
behavior acc. to DIN 1045-1 [din08, 9.2.3, 8.5.1]. No compression reinforcement.
Altogether an ultimate moment M
u
0.18 MNm is given with N = 0 leading to
uniform ultimate load of q
u
= 8M
u
/L
2
= 58 kN/m. A load q
u
= 40 kN/m is
chosen for the computation.
Solution method
Incrementally iterative with Newton-Raphson iteration within each loading incre-
ment as described on Page 16. A loading step t = 0.1 is chosen for time dis-
cretization with the nal target 1.
Figure 3.5: Example 3.2 a) System b) Deection curve (scaled by 100)
Results
As a statically determined system is given the course of the bending moment can
be computed with M =
q
2
(l x)x and a midspan maximum value M
max
=
ql
2
8
=
0.125 MNm in this particular case. Linear statics can only be approximately applied
in case of statically indeterminate systems as the bending stiffness is not constant
anymore but depending on internal forces.
Deections of cracked concrete beams even for statically determinate systems
cannot be determined with linear elastic statics and require the methods as described
65
before. Fig. 3.5b shows the computed deection curve. The uncracked linear elastic
case with a bending stiffness EJ = 31 900 0.2 0.4
3
/12 = 34.03 MNm is shown
as comparison. The difference roughly amounts to a factor of 2.4.
A special property of cracked reinforced cross sections is given with the fact, that the
reference axis changes its length even if there is no resulting normal force. In case
of this example the reference axis and with it the beam becomes slightly longer after
loading, see Fig. 3.5b. This is connected with longitudinal strains of the reference
axis, see Fig. 3.6a. It is caused by a coupling of Normal forces to curvature, i.e.
Figure 3.6: Example 3.2 a) Strain of reference axis b) Upper concrete strain and lower rein-
forcement strain
N = N(, ), see Section 3.1.3, which leads to ,= 0 for ,= 0 even in case
N = 0.
Finally, the computed strains of the upper compressed concrete edge and the strains
of the reinforcement, as they can be determined on base of Eq. (3.4), are shown in
Fig. 3.6b. They have to be compared to the concrete strain corresponding to strength
c1
= 0.0023 and to steel strain corresponding to yield
y
= f
y
/E
s
= 0.0025.
end example 3.2
66 3.5 Further Aspects
3.5 Further Aspects
3.5.1 Creep
Uniaxial material laws regarding creep, see Section 2.2, can be directly incorporated in
the ber model. We start with collecting previous items. Eq. (2.20) is applied to concrete

c
= E
c

x
+
c
E
x

c
c
,
c
=
_
1 +
c
c
(3.98)
with E
c
according to Eq. (3.21). Reinforcing steel is treated as usual

s
= E
s

x
(3.99)
with longitudinal strains
x
. Kinematics is ruled by Eq. (3.4)
x
= z ,
x
= z (3.100)
Strain of reinforcement is considered with
s1
= z
s1
,
s1
= z
s1
,
s2
= z
s2
,
s2
= z
s2
(3.101)
according to Eq. (3.18). Finally, regarding Eqs. (3.19,3.20) internal forces are
N
c
=
_
z
c2
z
c1
c
bdz
N
s
= A
s1
s1
+A
s2
s2
M
c
=
_
z
c2
z
c1
c
z bdz
M
s
= A
s1
s1
z
s1
A
s2
s2
z
s2
(3.102)
and N = N
c
+ N
s
, M = M
c
+ M
s
. The coordinates z
c1
, z
c2
indicate the range of the
concretes compression zone.
Internal forces of concrete
c
= ( N
c
M
c
)
T
will be connected to the concrete edge
stresses of the compression zone
cedge
= (
c1

c2
)
T
. Eqs. (3.98,3.22) yield rates of
edge stresses for xed coordinates z
c1
, z
c2
_

c1

c2
_
= E
c
_
1 z
c1
1 z
c2
_
_

_
+
c
E
c
_
1 z
c1
1 z
c2
_
c
_

c1
c2
_
(3.103)
or

cedge
= E
c
B
+
c
E
c
B
cedge
(3.104)
They are connected to rates of internal forces by Eq. (3.33)

c
= A

cedge
+A
z
z
c
(3.105)
with z
c
= ( z
c1
z
c2
)
T
.
Combining edge stresses Eq. (3.104), internal forces Eq. (3.105) and regarding the relation
Eq. (3.39) between rates of z
c
and generalized beam strains leads to

c
= A

_
E
c
B
+
c
E
c
B
ce
_
+A
z
B
z

=
_
E
c
A
+A
z
B
z
_
+
c
E
c
A

c
A

ce
(3.106)
3.5.1 Creep 67
whereby A
, B
, B
z
are functions of z
c1
, z
c2
or and A
z
is a function of z
c1
, z
c2
,
c1
,
c2
.
To consider the reinforcement, Eqs. (3.99,3.101,3.102) give

s
= C
T,s
(3.107)
with
C
T,s
= E
s
_
A
s1
+A
s2
A
s1
z
s1
A
s2
z
s2
A
s1
z
s1
A
s2
z
s2
A
s1
z
2
s1
+A
s2
z
2
s2
_
(3.108)
Combining all together yields
=
_
C
T,s
+E
c
A
+A
z
B
z
+
c
E
c
A

c
A

ce
(3.109)
Eqns. (3.104), (3.109) form a coupled system of ordinary differential equations of rst
order for ,
ce
depending on t, while (t), (t) are given as a function of time.
Eq. (3.104), e.g., is a specication of Eq. (1.66)
= C
T
+ (3.110)
with
=
_

c1

c2
_
T
=
_

_
T
C
T
= E
c
B
=
c
E
c
B
(3.111)
Applying Eq. (1.70) yields
i+1
=
i
+E
c
B
,i+1

_
i+1

i
_
+ t
_
c
E
c
B
,i+1

i+1

c
i+1
_
(3.112)
or
i+1
=
1
1 + t
c
_
i
+E
c
B
,i+1

_
i+1

i
_
+ t
c
E
c
B
,i+1

i+1
_
(3.113)
This is embedded into system equations Eqs. (1.73)-(1.75).
Example 3.3 Creep deformations of reinforced concrete beam
Geometry, boundary conditions and discretization is adopted from Example 3.2, see Fig-
ure 3.5.
Material
Concrete without tensile strength with a Youngs modulus E
c0
= 31 900 MN/m
2
in the compressive range. Creep properties with = 2.0 and t
= 100 [d] for

= 0.5, i.e. half of total creep occurs after 100 days for a constant stress load.
With Eqns. (2.20), (3.98)
c
= 0.006931 1/d,
c
= 0.020793 1/d.
Reinforcing steel 4 20, A
s1
= 12.57 cm
2
, d
1
= 5 cm. Uniaxial reinforcing steel
behavior acc. to DIN 1045-1 [din08, 9.2.3, 8.5.1]..
Concrete compressive strength is assumed with f
cd
= 21.675 MN/m
2
with
c1
=
0.0023,
c1u
= 0.0035 according to DIN 1045-1 [din08, 9.1.5, 8.5.1]. Regard-
ing the reinforcement an ultimate bending moment M
u
0.18 MNm with N = 0
is given.
Loading and characteristics of material behavior
A total uniform load of q
u
= 8M
u
/L
2
58 kN/m can be sustained in the ultimate
limit state.
Roughly a third of this load is assumed to occur in the service state, q
s
= 20 kN/m.
This loading leads to a moment M
s
= 0.0625 MNm with a maximum concrete
strain
c
= 0.39 in the cracked cross section.
Thus, a linear visco-elastic behavior may be assumed for the concretes compressive
zone.
Solution method
Incrementally iterative with Newton-Raphson iteration within each loading or time
increment as described on Page 16. A time step t = 10 days is chosen for time
discretization. A period of 500 days is regarded.
Figure 3.7: Example 3.3 a) System b) Mid span deection during time
Results
Fig. 3.7 shows the computed midspan deection in the course of time. The short
term uncracked linear elastic deection value (EJ = 31 900 0.2 0.4
3
/12 =
34.03 MNm) is given for comparison to cracked short term and long term deec-
tion values.
Inuence of concrete creeping can be characterized as follows: concrete contrac-
tion becomes larger with a (roughly) constant concrete stress, while reinforcement
extension (roughly) remains the same. This leads to an increasing curvature, see
Eq. (3.15), and an increasing deection. But increase factor is less than 1 + due
to bending and reinforcement.
As a variation, the inuence of a compressive reinforcement 2 20, A
s2
= 6.28 cm
2
,
d
2
= 5 cm is regarded. It can be seen, that a compressive reinforcement further re-
duces long term creep deections, as concrete deformations are constrained to some
extent.
end example 3.3
3.5.2 Temperature and Shrinkage 69
3.5.2 Temperature and Shrinkage
Uniaxial considerations for temperature and shrinkage, see Section 2.2 Page 28, can be
directly incorporated in the ber model. Increments of measurable strain are given by

x
=
1
C
T

x
+
tem
x
(3.114)
leading to

x
= C
T
_

x

tem
x
_
(3.115)
According to Eq. (3.4) a linear variation of imposed longitudinal strains is assumed over
cross section height
tem
x
(z) =
tem
z
tem
(3.116)
Imposed strain of the reference axis
I
and imposed curvature
I
can be determined with
prescribed imposed strains
tem
1
,
tem
2
of lower and upper edges
tem
=

tem
1
+
tem
2
2
,
tem
=

tem
1

tem
2
h
, h = z
2
z
1
(3.117)
Thus,
tem
x
(h/2) =
tem
1
,
tem
x
(h/2) =
tem
2
.
Eq. (3.115) can be directly integrated in the linear case C
T
= E
0
= const., i.e.
x
= E
0
_
x

tem
x
_
(3.118)
Using kinematics according to Eq. (3.4), internal forces relations Eq. (3.8) and cross sec-
tion properties Eq. (3.9) yield
= C
_

tem
_
(3.119)
with
=
_
N
M
_
, C = E
0
_
A
c
S
c
S
c
J
c
_
, =
_

_
,
t
=
_

tem
tem
_
(3.120)
Eq. (3.119) is sufcient for implementation in programs. The effect is as follows: if
Eq. (3.119) is regarded in Eq. (3.81) a split of internal nodel forces like
f
I
=
_
L
I
B
T
dx =
_
L
I
B
T
C dx
_
L
I
B
T
C
tem
dx = f
I
f
tem
I
(3.121)
occurs. As imposed strains
tem
are prescribed as, e.g., function of time, the part f
tem
I
of
nodal forces is principally known and automatically included as part of the load vector p,
see Eq. (1.63).
Example 3.4 Effect of temperature actions on a reinforced concrete beam
We refer to Example 3.2 with essentially the same system with the exception of boundary
conditions
7
and increased upper reinforcement.
Left and right node are totally constrained, i.e. lateral and longitudinal displace-
ments and rotations are set to zero.
Upper and lower reinforcement with A
s1
= A
s2
= 12.57 cm
2
, d
1
= d
2
= 5 cm.
Thermal expansion coefcient is chosen with
T
= 1 10
5
K
1
both for concrete
and reinforcement.
A corresponding linear elastic case with E = 31 900 MN/m
2
and the same geometry is
regarded as a reference case.
Figure 3.8: Computed bending moments for Example 3.4 a) Linear elastic b) RC
Two load cases are investigated:
1. Service loading with q = 20 kN/m as in Example 3.3.
2. Service loading + Temperature loading
8
on lower / upper edge with T
1
= 10 K, T
2
=
10 K.
As serviceability states occur with these loadings a linear concrete behavior can be as-
sumed in the compressive range. Concrete tensile strength is neglected.
Solution method
Incrementally iterative with BFGS-method for equilibrium iteration in each load
increment.
Results
Fig. 3.8a shows the bending moments for the linear elastic reference case, where the
rReinforcement is neglected.
7
Statically determinate systems will have no imposed forces in case of temperature actions.
8
In case of physical nonlinearities all loadings have to considered together.
3.5.2 Temperature and Shrinkage 71
*
Load 1 case has a xed end moment M
s
= 0.0416 MNm and a eld moment
M
f
= 0.0208, which accords to the analytic solution qL
2
/12 and qL
2
/24,
respectively.
*
The temperature load case can be superposed with a constant M
tem
= EJ
tem
with
tem
=
T
(T
1
T
2
)/h = 0.5 10
3
leading to M
tem
= 0.0170.
Fig. 3.8b shows the bending moments for the case of reinforced concrete (RC) with-
out tensile strength for concrete.
*
First of all, distribution of moments required for equilibrium in a statically in-
determintate system depends on stiffness relations. Principally, those depend
on loading in case of RC.
*
For RC bending stiffness is generally lower compared to the linear elastic case,
but the stiffness relations do not differ very much. Thus, we get M
s
= 0.0406, M
f
=
0.0219 in load case 1 for RC, i.e. the eld area is somehow stiffer and attracts
bending moments.
Computed mid span deection is w
max
= 1.26 cm in load case 1 for RC
and is considerably higher compared to the linear elastic case with w
max
=
qL
4
/384EJ = 0.96 cm due to overall reduced stiffness.
*
Superposing is not allowed for load case 2 due to the physical nonlinearities of
RC. Computed total moments are M
s
= 0.0278, M
f
= 0.0333 leading to
imposed moments M
tem
s
= 0.0406 0.0278 = 0.0128, M
tem
f
= 0.0219 +
0.0333 = 0.0114. The additional temperature moment is lower compared to
the linear elastic case and not constant along the beam.
Acomputation with temperature loading alone without other loading would
lead to a constant M
tem
= 0.0227, i.e. a superposition would be denitely
wrong.
Figure 3.9: Example 3.4 RC a) computed curvature b) computed strain of reference axis
Fig. 3.9 shows the deformation state for RC for load case 2.
*
Curvature along the beam see Fig. 3.9a. The course is not anymore analog to
the bending moment course, as it would be in the linear elastic case.
*
Fig. 3.9b shows the strain of the reference or central axis, respectively. Such
strains arise in contrast to the linear elastic case, as cracked RC beams tend to
elongate without longitudinal displacement restrictions.
*
An overall elongation is not allowed in the example due to boundary conditions,
i.e. a normal compressive force is induced on one hand, on the other hand strain
values occur depending on M/N-ratio. But integral of strains must sum up to
zero.
end example 3.4
3.5.3 Tension Stiffening 73
3.5.3 Tension Stiffening
For basics of tension stiffening see Section 2.8.
Tension stiffening is a matter of tensile stress transfer between reinforcement and
concrete via bond.
Experimental observations and detailed calculations show, that the affected concrete
area is restricted to a neighbourhood of the the main tensile reinforcement.
Concept of area of action of reinforcement, see DIN 1045-1 [din08, 11.2.3], or
effective reinforcement ratio
eff
respectively.
These considerations have to be modied, as there is not a homogeneous behavior in a
cross section in case of bending. Thus, the tension zone of bending is regarded separately
from the compression zone and treated as an isolated tension bar, whereby the concept of
effective concrete cross section is applied.
Tension stiffening may be quantitatively described as a reduction of reinforcement strains
between cracks, i.e. reinforcement strains in cracks are replaced by reinforcement mean
strains while reinforcement stresses in cracks are kept.
This leads to the following modied stress-strain relations for the reinforcement, see
Eqns. (2.35), (2.40)
s
=
_
_
f
ct
eff
s
for
s

s
E
s
s
+
t
f
ct
eff
s
<
s

y
f
y
+E
t
(
s

y
)
y
<
(3.122)
with the reinforcement yield stress f
y
and
s
=
1
E
s
f
ct
eff
(
t
),
y
=
1
E
s
_
f
y
+
t
f
ct
eff
_
(3.123)
and with k according to Eq. (2.37). For discussion of the parameters
t
, see Sec-
tion 2.8. A reasonable choice is
t
= 0.4, = 1.3.
These relations are controlled by
t
, e.g.
t
= 0 leads to
s
= f
ct
/
eff
,
y
=
f
y
/E
s
and retains the relation
s
= E
s
s
.
Eq. (3.122) may replace
s
= E
s
s
in Eqs. (3.19), (3.20) and other of the similar type.
The tangential material stiffness has to be adjusted in a corresponding way to reach con-
vergence in equilibrium iterations.
Example 3.5 Effect of tension stiffening on a reinforced concrete beam exposed to external
loading and temperature loading
We refer to Example 3.3 with the same system. The inuence of tension stiffening on
displacements will be examined.
Additionally to the material values of Example 3.3 the following values are assumed to
model tension stiffening
f
ct
= 3.0MN/m
2
, A
c,eff
= 0.02 m
2
,
t
= 0.4, = 1.3 (3.124)
leading to
eff
=
0.1257 10
2
0.02
= 0.063,
sm
= 0.2147971360 10
3
(3.125)
with
sm
according to Eq. (2.38). The modied stress-curve for the tensile reinforcement
Figure 3.10: Example 3.5 a) Moded stress-strain law for tensile reinforcement b) bending
moments in case of gravity + temperature loading with and without tension stiffening
is shown in Fig. 3.10. The difference to the original curve is relatively low due to the high
effective reinforcement ratio.
Results 1
The computed immediate deformation at midspan with tension stiffening is w
0
=
0.87 cm compared to 0.97 cm without tension stiffening, compare Fig. 3.7b. This
difference of 0.1 cm is maintained during the 500 day creep phase, i.e. w
500
= 1.33
cm versus 1.43 cm.
As beam stiffness is increased due to tension stiffening an inuence on imposed forces
may be supposed. To examine this item we refer to Example 3.4.
All system and loading parameters are kept but with tension stiffening included.
Load case 2 combined gravity and temperature is regarded.
Results 2
3.5.3 Tension Stiffening 75
Fig. 3.10b shows the computed bending moments for the cases with and without
tension stiffening. The difference is quite low, but this may not be generalized to
other systems.
end example 3.5
3.5.4 Prestressing
Originally, generalized stress is formulated as a function of generalized deformations
, e.g.
= C (3.126)
in the most simple linear elastic case. For C see, e.g., Eq. (3.120).
This concept is extended with respect to prestressing, i.e. an additional part is assigned to
generalized stresses resulting from prestressing tendons
= C +
p
(3.127)
This additional part depends on the tendon prole, i.e.
p
=
_
N
p
M
p
_
= F
p
_
cos
p
z
p
cos
p
_
(3.128)
compare Eq. (3.59), with the prestressing force F
p
, the height coordinate or lever arm z
p
of the tendon and the inclination
p
= dz
p
/dx of the tendon against the beam reference
axis. Internal shear forces are neglected in a rst approach.
The prestressing force may vary due to loss of prestress from friction of the tendon
in a conduit.
Tendon prole parameters z
p
,
p
may vary with the beam coordinate x according to
prestressing design.
Furthermore, a beam deformation may lead to a change in the tendon prole after
application and xing of prestressing. Two cases have to be considered:
1. Prestressing without bond: total length of the tendon changes. This leads to a
global change of the prestressing force.
2. Prestressing with bond: length of the tendon changes locally to keep the geo-
metric compatibility with the concrete. This leads to locally varying changes in
the prestressing force.
Eq. (3.127) may be generalized to nonlinear material laws.
An additional remark:
The very basic approach to prestressing of beams is based on Eq. (3.49). We con-
sider the quasistatic case, split internal forces into a part
from beam deformation,

a part
p
from prestressing and eliminate shear forces
N
N
p
= p
x
M
M
p
= p
z
(3.129)
Furthermore, Eq. (3.128) is used leading to
N
= p
x
+ (P cos
p
)
= p
z
+ (z
p
P cos
p
)
(3.130)
Eqs. (3.127), (3.128) can be used in the weak forms of Page 53 to gain weak forms
corresponding to Eq. (3.130).
A common approximation is F
p
const., cos
p
1 resulting in M
= p
z

z
p
F
p
, wherein z
p
F
p
is a redirection force from curvature z
p
.
3.5.4 Prestressing 77
Regarding Eq. (3.127) in Eq. (3.81), a split of internal nodel forces like
f
I
=
_
L
I
B
T
dx =
_
L
I
B
T
C dx +
_
L
I
B
T

p
dx = f
I
+f
p
I
(3.131)
occurs. The part f
p
I
may be regarded as a further contribution to the load vector, see
Eq. (1.63). Two stages have to be considered:
Application of prestressing with prescribed prestressing force including friction losses.
The period after xing the tendon with changes of the prestressing force according
to the enumeration above. Different approaches are necessary to describe the global
or local change of the tendon geometry for this stage.
Tendon geometry may be described for each nite beam element as follows:
An approach analog to the Bernoulli beam shape function Eq. (3.65) is chosen
z
p
=
_
r
3
4

3r
4
+
1
2
L
I
r
3
8

L
I
r
2
8

L
I
r
8
+
L
I
8

r
3
4
+
3r
4
+
1
2
L
I
r
3
8
+
L
I
r
2
8

L
I
r
8

L
I
8
_
_
_
_
_
z
p,I1
p,I1
z
p,I2
p,I2
_
_
_
_
P
=
dz
p
dx
=
dz
p
dr
dr
dx
= z
p
(3.132)
with the element length L
I
, z
p,I1
,
p,I1
at the start node and z
p,I2
,
p,I2
at the end
node, and the local element coordinate in the range 1 r 1. This approach
reproduces z
p
(1) = z
p,I1
, z
p
(1) =
p,I1
and z
p
(1) = z
p,I2
, z
p
(1) =
p,I2
.
Geometric length of tendon in an element I, see [BSMM00, 8.2.2.2]
L
P
I
=
L
I
2
_
1
1
_
(x
p
)
2
+ (z
p
)
2
dr (3.133)
where also the derivative of the tendon position x
p
in the longitudinal direction has
to be regarded. Eq. (3.133) has to be integrated numerically for each element, e.g.
with a Gaussian quadrature, see Page 14.
Nominal tendon geometry according to design
x
p
=
dx
p
dx
= 1
_
z
p,I1

p,I1
z
p,I2

p,I2
_
T
=
_
z
p0,I1

p0,I1
z
p0,I2

p0,I2
_
T
(3.134)
with prescribed values z
p0,I1
,
p0,I1
, z
p0,I2
,
p0,I2
.
Tendon geometry in deformed beam
x
p
= 1 +
_
z
p,I1

p,I1
z
p,I2

p,I2
_
T
=
_
z
p0,I1
+w
I1

p0,I1
+
I1
z
p0,I2
+w
I2

p0,I2
+
I2
_
T
(3.135)
with the longitudinal strain of the beams reference axis and the beams nodal dis-
placements w
I1
,
I1
, w
I2
,
I2
, see Eq. (3.65).
Summing up all element contributions gives the total length L
P
of a tendon.
A prestressing force F
p
0
is given by denition with the application of prestressing.
Regarding prestressing without bond, tendon length can be determined separately for ap-
plication of prestressing, lets say L
P
, and for the states following the xing of tendons,

lets say L
P
, using Eqs. (3.133), (3.135) with the appropriate deformations or displace-

ments, respectively. Regarding Eq. (3.128) this leads to a prestressing force
F
p
=
L
P
L
P
F
p
0
(3.136)
in the states following the xing of tendons.
Regarding prestressing with bond a tendon gets a local elongation after xing of prestress-
ing due to bond. This local elongation is ruled by the beams deformation kinematics
Eq. (3.4), i.e. the additional strain of the tendon is given by
p
(x) = (x) z
p
(x) (3.137)
with the changing , of the beam deformations after xing of prestressing. This
leads to to a prestressing force
F
p
(x) = F
p
0
+E
p
A
p

p
(x) (3.138)
with the Youngs modulus E
p
of the prestressing steel, the cross section A
p
of prestressing
and F
p
(x) used in Eq. (3.128).
The procedure can be summarized as follows:
Dene tendon geometry and prestressing force.
Compute internal forces from prestressing.
Compute nodal loads from prestressing.
Compute system reaction including prestressing loads.
Iterate if neccessary to capture changing in prestressing force.
Example 3.6 Prestressed reinforced concrete beam
We refer to Example 3.2 with principally the same system, but the span is doubled to L =
10 m. Thus, the load bearing capacity is strongly reduced with the given dimensioning.
Prestressing is used to increase the bearing capacity.
System parameter items
Concrete cross section b = 0.2, h = 0.4, compressive strength f
cd
= 21.675 MN/m
2
,
and lower and upper reinforcement A
s1
= A
s2
= 12.57 cm
2
, d
1
= d
2
= 5 cm
result in an ultimate bending moment M
u
0.19 MNm with N = 0, see Exam-
ple 3.3. This corresponds to a uniform loading q
u
= 8M
u
/L
2
= 15.2 kN/m. The
question is, whether this loading can be increased by prestressing.
A nominal uniform concrete prestressing stress of
c
0
= 10 MN/m
2
is chosen in
a rst approach leading to F
p
0
= 0.8 MN. Nominal tendon geometry of the whole
beam is given with a parabola starting and ending in the centerline with a downward
camber h
p
. This is described by
z
p
= 4h
p
_
x
2
L
2

x
L
_
(3.139)
A value h
p
= 0.15 m is chosen in this example.
3.5.4 Prestressing 79
Figure 3.11: Example 3.6 a) system b) midspan load-deection curve
Tendon properties with cross section A
p
= 6 cm
2
, strength f
p0,1
= 1600 MN/m
2
, f
p
=
1800 MN/m
2
and Youngs modulus E
p
= 200 000 MN/m
2
. Nominal initial steel
stress is
p
0
= 1 333 MN/m
2
with a strain
p
0
= 6.67 .
Self weight loading is assumed with q
g
= 5 kN/m.
Nonlinear uniaxial concrete behavior acc. to DIN 1045-1 [din08, 9.1.5, 8.5.1], see
also Fig. 2.1a.
Loading is applied in two steps: (1) Application of prestressing and self weight, (2) xing
of prestressing and additional application of a service load q
p
= 25 kN/m. Frictional
losses are neglected to simplify this example. Two cases are distinguished
Prestressing without bond
Prestressing with bond
Solution method is incrementally iterative with BFGS iteration within increments.
Results for prestressing without bond
The computed increase in prestressing force after load step 2 according to Eq. (3.136)
is minimal with F
p
/F
p
0
= 1.002. This results from the low ratio h
p
/L = 1/67.
Displacements, see Fig. 3.11. Deection starts with an uplift during application of
prestressing. Final mid span deection is quite large with 0.145 m ( 1/70 of the
span), but the load carrying capacity is not yet exhausted with an upper midspan
concrete compressive strain of -3.17 (limit strain is -3.5 ). Serviceability is
obviously not given due to high slenderness (1/25).
Bending moment and normal force in reinforced concrete (RC) cross section, see
Fig. 3.12. Total moment moment from self weight and service load is M
q
= 0.03
10
2
/8 = 0.375 MNm. The computed RC midspan contribution is M
c
= 0.255 and
the contribution from prestressing M
p
= 0.12. The increased RC moment compared
to the initial estimation results from the compressive normal force.
Results for prestressing with bond
In case of bond the tendon gets a local additional strain due to the locally varying de-
formation of the beam, see Eq. (3.137). This lead to an additional prestressing force,
see Eq. (3.138) and Fig. 3.12b, and in the end to higher contribution of prestressing
to load bearing capacity and a higher total load bearing capacity.
Figure 3.12: Example 3.6 a) RC bending moment b) RC normal force under total loading
Detail results are given with nal midspan deection 0.12 m, see Fig. 3.11b, RC
moment contribution M
c
= 0.220, see Fig. 3.12a, contribution from prestressing
M
p
= 0.155.
The comparison between prestressing with and without bond in this example is somehow
academic, as in practice prestressing with bond is exposed to more non-mechanical ef-
fects, which might lead to some restrictions to fully utilize the load carrying capacities of
the prestressing steel. Details are ruled in codes.
end example 3.6
Relaxation of prestressing steel follows basically the approach given in Section 2.2, which
may be transferred to other materials as concrete.
Details remain to be added.
Prestressing may be superposed with
creep (Section 3.5.1),
temperature and shrinkage (Section 3.5.2),
tension stiffening (3.5.3).
All desribed approaches are principally compatible and may be used in any combination.
Stresses
s
of reinforcing steel in the tensile regime are generally computed under the
assumption of a state II, i.e. a cracked cross section occurs. These values can directly be
used for a crack width calculation, see e.g. [din08, 11.2.4].
3.5.5 Shear stiffness for reinforced cracked concrete sections 81
3.5.5 Shear stiffness for reinforced cracked concrete sections
Figure 3.13: Shear stiffness
The following considerations base on the truss model for shear.
Simple truss with shear deformations only
Undeformed, deformed length of concrete strut 1
l
1u
=
h
sin
, l
1d
=
_
h
tan
_
2
+
_
h +
h
tan
_
2
= l
1u
_
1 + 2 sin cos
(3.140)
where
2
has been neglected. Bar strain is given with
1
=
l
1d
l
1u
l
1u
=
_
1 + 2 sin cos 1 (3.141)
The root term is expanded with a Taylor series. As 1, this results in
1
= sin cos (3.142)
Force in bar 1 with linear elastic behavior with Youngs modulus E and a bar cross
section height h
1
and a beam width b
F
1
= bh
1
E
1
= bh
1
E sin cos (3.143)
Summing up all struts intersected by a cross section leads to
F = nF
1
=
h
h
1
/ cos
F
1
= bhE sin cos
2
(3.144)
Due to equilibrium reasons the shear force Q and the total concrete strut forces are
related by Q = F sin . This leads to
Q = bhE sin
2
cos
2
(3.145)
and a shear stiffness
Q
= EAsin
2
cos
2
(3.146)
Reinforcement ties basically follow the same argumentation. A strut 2 is considered.
It is
l
2u
=
h
sin
, l
2d
=
_
h
tan
_
2
+
_
h
h
tan
_
2
= l
2u
_
1 2 sin cos
(3.147)
and
2
=
l
2d
l
2u
l
2u
=
_
1 2 sin cos 1 sin cos (3.148)
and
F
2
= A
s2
E
s
2
= A
s2
E
s
sin cos (3.149)
Cross section spacing s
c
and longitudinal spacing of rebars s are related by s
c
/s =
tan . This leads to n = h/s
c
= h/(s tan ) rebars in a cross section. Summing
up all ties intersected by a cross section leads to
F =
h
s
c
/ tan
F
2
= h
A
s2
s
E
s
sin
2
cos = ha
s2
E
s
sin
2
cos
(3.150)
Q = F sin nally leads to
Q = ha
s2
E
s
sin
3
cos (3.151)
We consider = /4 as special case. This leads to a shear stiffness EA/4 = GA/2
for = 0, see Eq. (3.7) for the shear modulus G. Now, the linear elastic case is
compared with Eq. (3.11)
3
. In the current argumentation concrete tensile struts are
disregarded due to the limited tensile strength of concrete and shear reinforcement
was implicitely assumed as vertical stirrups, which do not directly contribute to Q.
Thus, the current argumentation and the linear elastic case conrm for = /4.
Concrete shear strut angle
There is some margin to choose the concrete shear strut angle, see also [Com93,
6.3.3.1], [din08, 10.3.4], roughly in the range 20 45. As a rst estimation,
the same strut angle should be used as for the design of the stirrups.
83
3.6 Application Case Studies
3.6.1 Transient Dynamics of Beams
To begin with, the idea of inertial mass has to be discussed with respect to beams. Basics
are given with Eqs. (3.49), (3.56), which introduce the inertial mass m per unit length of
a beam and the inertial mass moment . These are given with
9
m = A, = J (3.152)
with the materials specic mass , the cross section area A and the second moment of
area J. Some care should be given to the units as specic mass has to be distinguished
from specic weight, i.e. a given specic weight has to be divided by earth acceleration.
Contributions connected with the inertial mass moment are often neglected, as
they are relatively small.
Regarding dynamics, appropriate methods have already been laid down. A basic approach
is given with Eq. (1.76)
r = M u +f p(t) = 0 (3.153)
with element contributions from
element mass matrices M
I
according to Eq. (3.84),
element nodal displacements according to Eq. (3.67)
3
or Eq. (3.71)
3
,
element internal nodal forces according to Eq. (3.82),
element nodal loads according to Eq. (3.85).
Nodal loads p(t) are generally prescribed as a function of time t. Eq. (3.153) forms
a system of ordinary differential equations of 2nd order in time t. Furthermore, initial
conditions have have to be given for the nodal displacements at a time t = 0.
The additional inertial term M u may have a strong inuence on the structural behavior
in case of loads with a rapid change of loading in time
10
.
Two fundamental solution approaches are given:
Modal decomposition
Integration in time
A well known method for numerical integration in time is given with the Newmark-
method, see Section 1.7 Eq. (1.83). This may immediately be applied to Eq. (3.153).
9
Has to be modied, if the centre line of area does not coincide with the beams reference line.
10
A reference value of rapidness may be given by the largest natural period of a system.
84 3.6 Application Case Studies
Example 3.7 Beam under impact load
We refer to Example 3.3 with the same geometry and boundary conditions. A sudden
concentrated single point load (impact) is applied in midspan. Alinear elastic behavior
is assumed in a rst approach to demonstrate basic characteristics of dynamic behavior
under impact.
Youngs modulus is assumed with E = 31 900 MN/m
2
.
Furthermore, the specic weight is taken as 25 kN/m
3
. With an earth acceleration
g 10 m/s
2
this leads to a specic mass = 0.025/10 = 2.5 10
3
MNs
2
/m
4
and with the cross section A = 0.2 0.4 m
2
of this example to a beam mass m =
0.2 10
3
MNs
2
/m
2
.
With the given parameters the longest natural period is given with T =
2L
2
_
m
EJ
=
0.0386 s and a frequency = 25.9 Hz.
The point load is characterized by magnitude and a time variation function
P(t) = P
0
f(t) (3.154)
The time function is chosen as a step function with limited duration
f(t) =
_
1 for t t
d
0 t > t
d
(3.155)
Thus, loading is characterized by the parameters P
0
, t
d
.
Discretization with N
e
= 20 enhanced two-node Bernoulli beam elements, see Page 57,
time step t = 0.001s
Results
Figure 3.14: Example 3.7 a) Deection time curve b) course of moments in 10 time steps until
1st maximum
Values P
0
= 0.07 MN and t
d
= 0.1 s are used for a linear elastic reference case.
Fig. (3.14)a shows the mid span deection in the course of time. Following items
are characteristic: (1) a cosine shaped oszillation, (2) a doubled maximum deection
0.0106 m compared to the quasistatic value with 0.0053 m.
3.6.1 Transient Dynamics of Beams 85
In same way maximum internal forces are doubled compared to the quasistatic case.
Fig. (3.14)b shows the bending moments along the beam in certain time steps up to
the time 0.02 s when rst midspan maximum is reached.
*
Moment does not immediately follow the load due to inertial effects.
*
There is no triangular course anymore due to a moment wave effect. In the
beginning moments are initiated in the impact point, while the support areas
are unaffected. In the following period moment waves travel along the beam
and bring the whole beam into action.
Figure 3.15: Example 3.7 a) related maximum deection depending on related load duration
time b) RC deection time curve
Impact loads often have a short duration. Thus, a small parameter study is performed
with varying load duration t
d
and constant load amplitude P
0
= 0.07 MN.
Fig. 3.15a shows the computed maximum midspan deections related to the qua-
sistatic deection 0.0053 m depending on t
d
related to the longest natural period
0.0386 s.
*
Very short loadings are compensated by inertia and only partially result in in-
ternal forces. If we consider a very short load duration time, e.g. t
d
= 0.001 s,
the beam gets only roughly 20 % of the quasistatic moment, or in other words,
it may sustain ve times the original load to have the same internal forces.
Finally, we consider the original reference case P
0
= 0.07 MN and t
d
= 0.1 s but
with a nonlinear reinforced concrete section (RC) instead of linear elastic behavior.
Material properties and reinforcement are chosen as in Example 3.6.
*
Fig. 3.15b shows the mid span deection in the course of time. Due to the
reduced stiffness the period of the oscillation grows, compare Fig. 3.14a. Max-
imum midspan deection grows to 0.019 m, roughly a doubling occurs com-
pared to the linear elastic case.
*
Fig. 3.16a shows the bending moments along the beam in 10 time steps up to
the time 0.03 s when rst midspan maximum is reached. Roughly the same
moments occur with some time shift compared to the linear elastic case, see
Fig. 3.14a.
*
In case of RC some normal forces arise without normal force loading, see
Fig. 3.16b. This is caused by the beams movement in the longitudinal direction
due to cracking, which is constrained by the beams inertia.
end example 3.7
86 3.6 Application Case Studies
Figure 3.16: Example 3.7 RC a) course of bending moments in 10 time steps until 1st maximum
b) course of normal forces in 10 time steps until 1st maximum
3.6.2 More Case Studies
Heavy multi-span RC-beam with prestressing (post-tension) under truck load
Light single span RC-beam with prestressing (post-tension) under pedestrian load
Prefabricated long span binder with prestressing (pre-tension)
Damaged RC girder with harmonic loading to identify progress of damage
remain to be appended.
Bibliography
[BSMM00] Bronstein, Semendjajew, Musiol, and Muehlig. Taschenbuch der Mathematik. Ver-
lag Harri Deutsch, Frankfurt/Main, 5. auage edition, 2000.
[Com93] Comite Euro-International de Beton. CEB-FIP Model Code 1990. Thomas Telford,
London, 1993.
und Konstruktion, August 2008.
87
Chapter 4
Strut-and-Tie Models
4.1 Linear Elastic Panel Solutions
Typical example
Single-span panel / deep beam with large opening, see e.g. Fig. 4.1a.
FE modeling
2D continuum elements, e.g. 4-node quadrilateral, see page 4.
Linear elastic constitutive law, see Page 6.
Figure 4.1: a) Deep beam with opening b) principal stresses in deep beam with opening
Prinicipal stress / force ow, see Fig. 4.1b
Plane stress state
x
,
y
,
xy
principal stress state
1
,
2
, .
stress trajectories compression force ow & tensile force ow.
*
How to determine a stress trajectory: let y(x) be the function describing the
course of a stress trajectory in the x, y-plane and x
1
, y
1
with y
1
= f(x
1
) a
starting point. Then, dy/dx = = f(
x
(x, y),
y
(x, y),
xy
(x, y)) provides
an ordinary differential equation, which at least can be solved numerically for
y(x).
*
Practically, an estimation of trajectories using a principal stress visualization is
sufcient.
88
89
Compression force ow concrete struts.
Tensile force ow reinforcement ties.
Lattice truss model for a panel with
concrete struts,
reinforcement ties,
nodes connecting struts and ties
forms a so called strut-and-tie model.
4.2 Truss Modeling
System determination
Short force ow path from load to support.
Minimization of total potential energy with elastic internal energy (minimization
of displacement of loading point).
Approximation of stress trajectories by polygons.
Basic rule: any change of a force direction needs needs an extra diversion force.
Property determination
Ties: reinforcement cross section, Youngs modulus of reinforcement steel (tension
stiffening!)
Struts: a constructed cross section along system line according to a fan out and fan
in of concrete stresses. There is some margin of discretion in the determination of
strut cross sections.
Nodes: area of nodes is determined by adjoining struts and supports.
Statically determinate systems vs. indeterminate systems
Internal forces of determinate systems can be calculated by equilibrium conditions
alone.
Indeterminate system need bar stiffness values for internal forces calculation.
90 4.3 Computation of Plane Elasto-Plastic Truss Models
4.3 Computation of Plane Elasto-Plastic Truss Models
We consider a plane truss with m bars and n nodal degrees of freedom.
Initially, there are two degrees of freedom per node, but some nodes should be sup-
ported and degrees of freedom constrained to prevent a rigid body movement of the
system.
It is assumed, that the system is stable, i.e. an arbitrary nodal load can be transferred
to the supports.
Kinematic compatibility
Bar k with end nodes i, j has an orientation angle from i to j and node displace-
ments u
j
k
, u
i
k
in the longitudinal bar direction. We dene a strain of bar k as
1
e
k
= u
j
k
u
i
k
(4.1)
Displacements in the longitudinal bar direction have to be transformed into the
global system. This leads to global displacements of nodes i, j
u
i
=
_
u
i
x
u
i
y
_
=
_
cos
k
sin
k
_
u
i
k
, u
j
=
_
u
j
x
u
j
y
_
=
_
cos
k
sin
k
_
u
j
k
(4.2)
resulting in
u
i
k
=
_
cos
k
sin
k
_
u
i
, u
j
k
=
_
cos
k
sin
k
_
u
j
(4.3)
Combining Eqns. (4.1), (4.3) gives the strain of bar k
e
k
=
_
cos
k
sin
k
_
u
i
+
_
cos
k
sin
k
_
u
j
. (4.4)
Collection of all bars in a vector leads to a formulation
_
_
_
_
.
.
.
e
k
.
.
.
_
_
_
_
=
_
_
.
.
.
.
.
.
.
.
.
.
.
.
cos
k
sin
k
cos
k
sin
k

.
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
_
_
.
.
.
u
i
x
u
i
y
.
.
.
u
j
x
u
j
y
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
_
(4.5)
or in a matrix notation
e = B u, B R
mn
, u R
n
, e R
m
(4.6)
where the vector e collects all m bar strains and u all n nodal displacements. There
is a similarity compared to Eq. (1.6).
1
This differs from the conventional denition of strain, but is more convenient for the following argumentation.
91
Equilibrium
A bar k has end nodes i, j and a bar force s
k
(tension positive, compression nega-
tive). Bar k contributes forces both to the nodes i, j, see Fig. 4.2a
f
i
k
=
_
f
i
k,x
f
i
k,y
_
= s
k
_
cos
k
sin
k
_
, f
j
k
=
_
f
j
k,x
f
j
k,y
_
= s
k
_
cos
k
sin
k
_
(4.7)
Figure 4.2: a) Truss equilibrium b) Example truss system
Local equilibrium condition for nodes i, j with external nodal loads p
i
, p
j
k
f
i
k
= p
i
,
k
f
j
k
= p
j
(4.8)
Global equilibrium condition through assembling of all nodes
_
_
.
.
.
cos
k

sin
k

.
.
.
cos
k

sin
k

.
.
.
_
_
_
_
_
.
.
.
s
k
.
.
.
_
_
_
_
=
_
_
_
_
_
_
_
_
_
_
_
_
_
.
.
.
p
i
x
p
i
y
.
.
.
p
j
x
p
j
y
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
_
(4.9)
or in a matrix notation
B
T
s = p, B
T
R
nm
, s R
m
, p R
n
(4.10)
where the vector s collects all m bar forces and the vector f all n nodal loads. Here
we see again matrix B of Eq. (4.6), but now it has been transposed. There is a
similarity compared to Eq. (1.33)
1
.
A case m = n indicates a statically determinate and m > n a statically indetermi-
nate system. Bar forces may be determined directly from Eq. (4.10) for a statically
determinate system. This is not possible for an indeterminate systems as the number
of unknowns m is larger as the number of equations n.
Linear constitutive law and linear equations of systems
To begin with, a uniaxial linear elastic material behavior is assumed, see Eq. (1.19).
A
I
is used for the cross sectional area of a bar I, L
I
as its length and E
I
as its
Youngs modulus. With s
I
= A
I

xI
, where
xI
is stress of bar I, and
xI
= e
I
/L
I
,
where
xI
is the conventional strain of bar I, bar forces are given by
s
I
= C
I
e
I
, C
I
=
E
I
A
I
L
I
, I = 1 . . . m (4.11)
In a matrix notation this can be written as
s = C e, C R
mm
(4.12)
where the material matrix C is diagonal with coefcients C
I
.
Combination of Eqs. (4.10), (4.12) and (4.6) leads to
K u = p, K = B
T
C B, K R
nn
(4.13)
with a constant symmetric stiffness matrix K. There is a similarity of Eq. (4.13)
compared to Eq. (1.36).
Example 4.1 Deep beam with strut and tie model
We refer to the problem given with Fig. 4.1a and linear elastic principal stresses shown in
Fig. 4.1b. This problem is also discussed in [CF08, 8.8].
Figure 4.3: Example 4.1 a) member stresses b) proposed reinforcement scheme
The chosen strut and tie model is shown in Fig. 4.2b.
*
The system has m = 36 bars and 21 nodes. Seven nodal degrees of freedom
are restrained by boundary conditions leading to n = 2 21 7 = 35. Thus,
we have nearly a statically determinate system. Internal forces are inuenced
by the stiffness of the members to a small extent.
Obviously there is some effort to circumvent the central rectangular hole.
The resulting total load has been distributed to upper nodes according to the loading
scheme of Fig. 4.1a.
93
A preliminary computation reveals a distinction between struts ( compressive,
concrete) and ties (tensile, reinforcing steel). The following reference values for
Youngs modulus and cross section area are chosen for the ties: E
s
= 200 000 MN/m
2
, A
s
=
1010
3
m
2
, and for the struts: E
c
= 30 000 MN/m
2
, A
c
= 0.12 m
2
. The last value
corresponds to a deep beam thickness t = 0.6 m and an assumed strut width of 0.2
m.
Results
Computed stresses of struts and ties are shown in Fig. 4.3a. Negative values with
small amount correspond to struts, positive values with relatively large amount cor-
respond to ties. According to each individual stress value the reference value of
cross section area may be increased or reduced for the corresponding member to
gain some target stress.
Target stress values may be derived from strength values in codes, see e.g. [din08]
Design of reinforcement
A reinforcement scheme as proposed in [CF08, Fig. 8.49] is shown in Fig. 4.3b.
end example 4.1
Nonlinear constitutive law and nonlinear equations of systems
Actually struts and ties have limited bearing capacities.
This may in a rst approach described with an uniaxial elastoplastic constitutive law,
see Section 2.5 applied to both the reinforcement and the concrete in the compressive
range.
*
With member strains e
I

xI
given this leads to member stresses
xI
and
member forces s
I
.
*
Stress determination is more elaborated compared to the linear elastic case and
requires housekeeping with an internal state variable, i.e. the actual plastic
strain
Ip
.
Nodal forces from members or internal nodal forces from Eq. (4.10)
f (u) = B
T
s(u) (4.14)
Equilibrium condition in analogy to Eq. (1.63)
r(u) = f (u) p (4.15)
Tangential material stiffness
C
TI
=
_
E
sI
[
xI
[ < f
yI
E
TI
[
xI
[ = f
yI
(4.16)
All these coefcients are collected in a diagonal matrix C
T
.
Tangential system stiffness
2
in analogy to Eq. (4.13)
K
T
= B
T
C
T
B, K
T
R
nn
(4.17)
Solution with Newton-Raphson method according to Eqns. (1.63) (1.65).
2
Distinguish upper T for transposed from lower T for tangential.
Example 4.2 Corbel with elasto-plastic strut and tie model
The corbel problem is discussed in [FS07, Fig. 1b], see Fig. 4.4a.
In the following a slightly modied problem is regarded, which is statically indeterminate.
Corbel geometry, loading and strut-and-tie model are shown in Fig. 4.4. We have m =
14 bar members and n = 12 degrees of freedom, i.e. the system is twofold statically
indeterminate.
Figure 4.4: a) Corbel acc. to [FS07] b) model of Example 4.2
A provisional rst calculation has to be performed to distinguish between tensile and
compressive members. Basing upon this calculation member properties are chosen as
follows:
Tensile members 1 7 are chosen as ties with E
s
= 200 000 MN/m
2
, f
y
= 500 MN/m
2
and a reference cross section area A
s0
= 10 cm
2
= 10 10
4
m
2
. Cross section fac-
tor of each tie is shown in Fig. 4.4b. Compressive members 8 14 are chosen as
struts with E
c
= 30 000 MN/m
2
and a yield limit
3
f
yc
= 40 MN/m
2
. Cross section
of all struts is assumed with A
c
= 4 50 cm
2
= 0.02 m
2
. Finally, a slight hardening
is assumed for reinforcement steel and concrete, see Section 2.5.
System buildung and solution method
The system equations may be built according to Eqns. (4.14) - (4.17).
Alternatively, the formal nite element approach with 2D-bar elements according
to Page 3, material law according to Section 2.5 and system building according to
Section 3.4 may be applied. This leads to the same result.
Loading is applied with prescription of vertical displacement of the loaded node.
The solution is gained incrementally iterative with the Newton-Raphson approach,
see Page 15.
Results
The computed load displacement curve is shown in Fig. 4.5a.
3
Concrete is assumed as elasto-plastic in this context. The assumed value is quite high, codes ordinarily restrict
allowed concrete stresses to a high extent.
95
*
Initially linear elastic up to reaching rst yielding of a tie.
*
Decrease of stiffness with rst yielding of tie, but some more loading can be
applied.
*
Ultimate limit state with yielding of other ties in the statically indeterminate
system such that the system becomes "nearly kinematic".
*
The system does not become "really kinematic" with the hardening assump-
tion, but otherwise the determinant of the tangential system stiffness would be-
come zero without the hardening assumption and the Newton-Raphson method
would not work anymore.
Figure 4.5: a) member stresses b) model of Example 4.2
Member stresses in last computed state see Fig. 4.5b.
Yield limit is reached in most ties and slightly exceeded due to hardening.
Struts are not critical in this particular case, but not far away from reaching nominal
yield limit.
end example 4.2
96 4.4 Ideal Plastic Truss Models
4.4 Ideal Plastic Truss Models
Denition of the ideal plastic problem
Kinematic compatibility see Eq. (4.6), m equations.
Equilibrium see Eq. (4.10), n equations.
Material behavior
*
Limit state condition ow condition
[s
I
[ s
uI
, s
uI
= f
yI
A
I
, I = 1 . . . m (4.18)
with unsigned bearing capacities s
uI
and signed forces s
I
of members.
*
Assumption about strains ow rule
[e
I
[
_
= 0 for [s
I
[ < s
uI
> 0 [s
I
[ = s
uI
, I = 1 . . . m. (4.19)
*
Dissipation condition Kuhn-Tucker conditions
s
I
e
I
0, s e 0, s, e R
m
(4.20)
There is no way for elastic deformations in the ideal plastic approach.
Balance of equations vs. unknowns and solution approach
m bar forces s, n displacements u, m bar strains e vs. 2 m + n equations (4.6),
(4.10), (4.19) plus 2 m constraint equations (4.18), (4.20).
The solution of this problem is not straightforward. It is provided by methods of
optimization and linear programming, see e.g. [Lue84].
For sake of simplicity we restrict to a loading type
f = f
0
, > 0 (4.21)
with a unit load f
0
which is xed and a loading factor .
The basic solution idea is regarding a maximization problem
max (4.22)
while fullling equilibrium and the limit state function, i.e. constraints
B
T
s = f
0
and [s
I
[ s
uI
, I = 1 . . . n (4.23)
The quantities B, s
u
, f
0
are known and s, are unknown in this formulation.
97
Furthermore, linear programming theory states a so called dual problem associated
to the maximization problem. This is a minimization problem
s
u
e min (4.24)
where e collects
4
all [e
I
[ and s
u
all s
uI
. The minimization has constraints
B e = u and f
0
u = 1 (4.25)
i.e. essentially kinematic compatibility. The quantities B, s
u
, f
0
are known and
e, u are unknown in this formulation. Eq. (4.24) can be interpreted as minimization
of energy.
Linear programming theory also states that solutions s
exist for the maximiza-

tion and e
, u
for the minimization. Both have the same optimal value
= s
u
e
(4.26)
with
> 0. It may nally be shown that the solutions fulll the conditions
e
I
= 0 if [s
I
[ < s
uI
and s
> 0 (4.27)
Thus, the solutions s
, e
, u
of the associated optimization problems are also

solutions of the ideal plastic problem as has been formulated before.
The ideal plastic problem is also fullled by s
and e.g. e
, u
with an
arbitrary scalar
5
> 0.
*
The absolute values of the deformations are indeterminate. The relation
between the deformation components is determined to some extent.
The solution s
f
0
is also a solution for the corresponding elasto-plastic problem if
case no hardening occurs in the state of yielding.
Youngs moduli of bars do not inuence this limit state for forces and loads.
But the deformations of the ideal elastoplastic problem are inuenced by Youngs
moduli, due to elastic deformations prior to plastic deformations.
Solution methods
Basically, we are interested in loads f
0
and member bar forces s. Thus, we have to
solve the maximization problem Eqns. (4.22), (4.23).
The solution is simple for statically determinate systems with n = m and a square
B. For small statically indeterminate systems it may be found by inspection or trial
and error.
Larger statically problems require systematic methods of linear programming, e.g.
the simplex method, or general methods of optimization.
4
has to be interpreted as operator on .
5
But this may violate the constraint f
0
u = 1 and the maximization problem is not solved anymore.
To determine deformations, the following approach is appropriate instead of solving
the minimization problem.
*
First of all, from analysis of forces and Eq. (4.19) we can determine a set of
bars with strains e
R
,= 0, R = 1 . . . m
e
and all other bars with e
S
= 0, S =
m
e+1
. . . m.
*
We consider kinematic compatibility Eq. (4.6) and reformulate it with a unit
matrix I
m
R
mm
B u = I
m
e, e R
m
, u R
n
, B R
mn
, m n (4.28)
*
As B has n linear independent rows, Eq. (4.28) can be transformed into
_
I
n
0
_
u =
_
P
Q
_
e, I
n
R
nn
, 0 R
(mn)n
, P R
nm
, Q R
(mn)m
(4.29)
by Gauss elimination, see [ZF97], with a unit matrix I
n
, a zero matrix 0 and
generally fully occupied matrices P, Q. Eq. (4.29) leads to
u = P e, Q e = 0. (4.30)
The last equation describes kinematic compatibility in a tighter sense. In case
m = n the matrix P is the inverse of the matrix B.
*
As e
S
= 0, S = m
e+1
. . . m a submatrix Q
e
of Qis given with
Q
e
e
e
= 0, e
e
R
m
e
, Q
e
R
(mn)m
e
(4.31)
where e
e
collects all yielding e
S
. Two cases have to be distinguished.
m
e
m n: Eq. (4.31) can generally only be fullled with e
e
= 0, i.e.
no deformation occurs at all.
m
e
> mn: a number of m
e
(mn) components of e
e
may be chosen
arbitrarily under the condition Eq. (4.20), while the rest is determined with
Eq. (4.31). But this leaves the absolute deformation undetermined.
*
Only the relations of the deformation components are determined to some ex-
tent. This is a characteristic property of ideal plastic systems.
In case of a statically determinate system m = n things simplify. Kinematic com-
patibility in a tighter sense Eq. (4.30)
2
vanishes and strains of yielding bars may be
given arbitrary values, while all other strains are zero.
With given strains e displacements u can be determined directly by Eq. (4.30)
1
.
Limit theorems of plasticity
1. Any equilibriumstate which fullls the limit state condition, see Eqns. (4.22), (4.23),
gives a lower bound for the loading factor.
2. The work of load bearing capacities on kinematically admissable deformations with
a constraint f
0
u = 1, see Eqns. (4.24), (4.25), gives an upper bound for the loading
factor.
These theorems are generally stated as a kind of postulates. Actually, their exact formu-
lation and proof correspond to linear programming theory.
99
Example 4.3 Corbel for ideal-plastic strut and tie model
We refer to Example 4.2. The same system is computed as ideal plastic on basis of the
1st limit theorem of plasticity. This requires the knowledge of the effective cross section
areas of members and their yield stress, which are taken from Example 4.2.
Youngs modulus and hardening behavior after reaching yield limit are not relevant
in this approach.
Basic parameters are number of bars m = 14, 9 nodes, 3 boundary nodes, number
of degrees of freedom n = 12.
All the following bases upon a linear elastic precalculation distinguishing between
compressive and tensile members.
To start with, states fullling 1st limit theorem of plasticity are determined.
Equilibrium is described by Eq. (4.10)
B
T
s = f
0
(4.32)
with the vector of member forces s, a loading factor , a unit load
f
0
=
_
0 0 0 0 0 0 0 0 0 1 0 0
_
T
(4.33)
applied vertically downward on node 7, see Fig. 4.5b, and
B
T
=
_
_
0 0 0 0 0 0 0 0 1 0 0 .707 0 0
1 0 0 0 0 0 0 0 0 0 0 .707 0 0
0 0 1 0 0 0 1 .243 0 0 0 0 .707 0
0 0 0 0 0 0 0 .971 0 0 0 0 .707 0
0 0 0 1 0 0 0 0 0 0 0 .707 0 .243
0 0 0 0 0 0 0 0 0 0 0 .707 0 .971
0 0 0 0 1 1 0 0 0 0 0 0 .707 0
0 1 0 0 0 0 0 0 0 0 0 0 .707 0
0 0 0 0 0 1 0 0 0 0 .781 0 0 .243
0 0 0 0 0 0 0 0 0 0 .625 0 0 .971
0 0 1 1 0 0 0 0 0 .781 .781 0 0 0
1 1 0 0 0 0 0 0 0 .625 .625 0 0 0
_
_
(4.34)
where the boundary nodes 1, 5 and 9 and their degrees of freedom are disregarded.
With prescription of m = m n + 1 = 3 more or less arbitrary member forces,
Eq. (4.32) constitutes a system of n = 12 equations for the remaining n 1 = 11
member forces plus the loading factor , i.e. n equations for n unknowns. This is
principally solvable for a well behaved structural system.
Figure 4.6: Example 4.3 a) stresses in ultimate limit state b) a displacement mode
The members 1, 7, 11, see Fig. 4.5b, are chosen by chance.
*
They are prescribed with their bearing capacities, i.e. s
1
= 0.05 MN, s
7
=
0.1 MN, s
11
= 0.8 MN.
*
Solution of the modied system Eq. (4.32) as described before leads to s
2
=
0, s
3
= 0.1, s
4
= 0.375, s
5
= 0.6372, s
6
= 0.6372, s
8
= 0, s
9
= 0.05, s
10
=
0.88, s
12
= 0.0707, s
13
= 0, s
14
= 0.0515 and a loading factor =
0.5498.
*
The computed member forces lead to stresses
5
= 637 MN/m
2
,
6
= 637,
10
=
44 in the members 5, 6 and 10. This is not admissable, as the stresses exceed
the material strength and the limit state condition Eq. (4.18) is violated for these
three members.
*
Thus, all member forces are reduced by the less of the factors 500/637 = 0.785
and 40/ 44 = 0.909 which is = 0.785. This leaves the members 5 and 6
in a yielding state.
*
To maintain equilibriumaccording to Eq. (4.32) the loading factor also has to be
reduced by the same factor leading to a nal loading factor of = 0.5498 =
0.43. This is a lower bound for the real loading factor, i.e. the real loading is
not less but may be larger.
*
Deformations can be determined as described on Page 98. With two yielding
members m
e
= 2 and m
e
= m n. Thus, the matrix Q has 2 2 dimension
and Eq. (4.31) can only be fullled with zero deformations. Furthermore, this
solution does not fulll Eq. (4.19) of the ideal plastic problem.
Another choice by chance with the members 9, 11, 13 leads to lower bound of
= 0.081 and leaves only member 1 in a yielding state. No deformation can be
found fullling the kinematic compatibility conditions.
Using the binomial coefcient there are
_
m
m
_
=
m!
m!(m m)!
= 364 (4.35)
possibilities for a choice of m = 3 out of m = 12. This trial and error method to
determine the largest value and therewith a solution of the ideal plastic problem is
obviously cumbersome.
101
Application of simplex method.
The theory of the simplex method has to be omitted here for sake of brevity. It
is a standard method to solve linear constrained optimization problems, where the
problem formulation Eqns. (4.22) - (4.23) comes from, and is used in all disciplines
of science, technology and economics.
The optimization methods yields the member stresses shown in Fig. 4.6a. The cor-
responding loading factor is = 0.44.
*
The horizontal tie in half height is fully utilized which leads to some increase
of the limit load.
*
Compare Fig. 4.5 for the results of the elasto plastic calculation, which leads to
the same ultimate limit load and a similar stress distribution.
Displacements can again be determined as described on Page 98. .
*
With the ve yielding members 1, 3, 5, 6 and 7 it is m
e
= 5 > m n = 2,
i.e. three member strains may principally chosen arbitrarily while the remaining
two are determined on base of Eq. 4.31. Achoice e
1
= 0.1, e
3
= 0.03, e
5
= 0.1
leads to
e =
_
0.1 0 .03 0 0.1 0.1225 0.03125 0 0 0 0 0 0 0
_
T
(4.36)
Energy according Eq. (4.24) amounts to s
u
e = 0.122375
*
Displacements can be determined with Eq. (4.30)
1
. Results are shown in Fig. 4.6b.
The vertical displacement of the loaded node 7 is computed with 0.278125. A
displacement scaling factor of 1/0.278125 = 3.5955 has to be chosen to fulll
Eq. (4.25) with f
0
according to Eq. (4.33). This leads to an energy s
u
e = 0.44
and corresponds to the maximum loading factor as stated by the limit theorem
of plasticity.
In contrast to the elasto-plastic model of Example 4.2 determination of absolute
deformations is not possible with the ideal plastic approach due to arbitrary choice
of some member strains. A deformation mode can be determined to some extent.
end example 4.3
102 4.5 Application Aspects
4.5 Application Aspects
Strut limit forces
A nominal value is determined by cross section of struts and uniaxial compression
strength of concrete.
There is some uncertainty in the effective value of cross section of struts.
Effective compression strength may be reduced to lateral tensile effects.
Some caution has to be considered for the determination of limit forces s
u
of
struts.
Ductility demands
Before reaching its limit state a system has to undergo more or less more deforma-
tions.
*
This starts with pure elastic deformations.
*
Additional plastic deformations will follow in one or more yielding bars. This
is accompanied by a force redistribution in statically indeterminate system with
further increasing load before reaching the systems limit load.
Thus, all bars have to deform to some extent without rupture or brittle failure before
reaching the systems limit load.
This imposes ductility requirements on bars ( minimum strains required) and
nodes (minimum relative rotations of adjoining bars required).
Nodes
Nodes connect struts and ties and have to handle an exchange of forces.
Nodes occupy some extension determined by the adjoining bars.
Figure 4.7: a) compressione node b) compression-tension node c) reinforcement redirection
node
This area of extension may considered as a plate with a two dimensional stress state.
Nodes may fail as bearing elements of a strut-and-tie system.
Basically, their bearing capacity has to be checked.
Bibliography
[CF08] CEB-FIP. Practitioners guide to nite element modelling of reinforced concrete struc-
tures, volume Bulletin Nr. 45. International Federation for Structural Concrete p,
Lausanne, 2008.
[FS07] F. Fingerloos and G. Stenzel. Konstruktion und bemessung von details nach din 1045.
In Betonkalender, pages 325374, Berlin, 2007. Ernst & Sohn.
[Lue84] David G. Luenberger. Linear and Nonlinear Programming. Addison-Wesley, Reading,
Massachusetts, 2. Auage edition, 1984.
[ZF97] R. Zurmhl and S. Falk. Matrizen und ihre Anwendungen. Springer-Verlag, Berlin, 7.
Auage edition, 1997.
103
Chapter 5
Multiaxial Concrete Material Behavior
5.1 Scales
Continuum mechanics provides a framework to describe the behaviour of solids. Within this
framework two scales are considered to describe structural concrete, i.e. mesoscale and macroscale.
The mesoscale distinguishes the cement matrix, aggregates and the interface transition zone.
Each of these material phases is regarded as a continuous isotropic solid with its own constitu-
tive law and its own material parameters. A continuous displacement is assumed along contact
surfaces of different phases in case of deformations.
Regarding a concrete specimen composed of phases its internal geometric characteristics
are random as a matter of principle, e.g. size, shape, position and orientation of aggregates
are random. Thus, two samples chosen out of a collection of specimens of same geometry and
with the same material parameters of phases generally will show different reactions under the
same imposed action. The variation of reactions depends on the size of the specimen relative to
the size of the largest aggregates and the type of the action. The variations of reactions tend to
become smaller for larger specimen sizes. If they are considered to be negligible for the relevant
types of action the specimen size constitutes a representative volume element (RVE).
A mixture of phases may be homogenized for a RVE. Regarding concrete, cement matrix,
aggregates and interface transition zone are homogenized within the macroscale with a com-
mon type of constitutive law and a corresponding single set of material parameters. Assuming
material types the material parameters of the homogenized continuum may basically be de-
termined analytically from the material parameters of the phases by mixture theories, see e.g.
(Mori & Tanaka, 1973), (Wriggers & Moftah, 2006). But such approaches are limited regarding
the mesoscale randomness of concrete. As an alternative, numerical multiscale methods may
provide macroscale parameters derived from numerical mesoscale calculations considering par-
ticular actions and phase properties. Finally, parameters of macroscale constitutive laws may
directly be chosen according to the experimental behaviour of specimens with at least RVE-size
using parameter identication and calibration methods.
Considering numerical simulation methods with a discretization of space like Finite-Element-
Methods a single element generally should not be exposed to larger gradients of reactions within
itself. This leads to upper bounds for elements sizes or neness of discretization, respectively.
Thus, due to local inhomogeneities discretizations of a given structure require an order of ten
times more elements for each spatial dimension within the mesoscale compared to discretiza-
tions within the macroscale. As a further consequence numerical calculations of larger structures
are generally performed in a macroscale using homogenized constitutive laws.
104
105
5.2 Some Basics of Continuum Mechanics
In the following space is measured in a 3D cartesian coordinate system (orthogonal,
right-handed) if not otherwise stated. A space point x has a vector of coordinates
1
( x
1
x
2
x
3
)
T
. A body of material occupies an area of space in a conguration, see
Fig. 5.1a. Due to a loading history this conguration changes with time t. A material
point is identied by the space point X it occupies in a reference conguration at a time
t
0
, i.e. X = x for t = t
0
. Displacements
2
are dened with u = x Xand have a vector
of components u = ( u
1
u
3
u
3
)
T
.
Small strain measurement
A notion of small: displacements have a magnitude of millimetres while the body
has dimensions in the magnitude of metres.
Strain components
ij
=
1
2
_
u
i
x
j
+
u
j
x
i
_
=
ji
, i, j = 1 . . . 3 (5.1)
are dened involving two directions, i.e. a displacement direction and a reference
direction.
Strain components form a symmetric tensor of second order
=
_
_
11

12

13
12

22

23
13

23

33
_
_
(5.2)
A second order tensor obeys particular transformation laws based upon transfor-
mation laws for vectors ( rst order tensors) in case of a transformation of the
underlying coordinate system.
Strain tensor in so-called Voigt notation
=
_
_
_
_
_
_
_
_
11
22
33
23
13
12
_
_
_
_
_
_
_
_
=
_
_
_
_
_
_
_
_
11
22
33
2
23
2
13
2
12
_
_
_
_
_
_
_
_
(5.3)
utilizing symmetry of small strain.
Incremental changes of strain during loading history with progressing time t
= lim
t0
t
(5.4)
Stress
Denition of Cauchy stress with reference to an innitesimal tetrahedron at a
position x with a tetrahedron base area dA, base normal n = ( n
x
n
y
n
z
)
T
exposing a force df and a stress t = df /dA.
1
Change in notation compared to Section 1.3.
2
Per denition displacements become zero in the reference conguration.
106 5.2 Some Basics of Continuum Mechanics
Figure 5.1: a) Continuous body in reference and deformed conguration b) Innitesimal stress
tetrahedron
*
Force df is equilibrated by three exposed forces on each coordinate plane, see
Fig. 5.1b, each force with an exposed stress, each stress with three components.
*
Thus, nine stress components
ij
related to the coordinate planes with i, j =
1, 2, 3 are given. The rst index denotes the plane normal, the second index the
global direction component.
*
Considering an ininitesimal cube it can be shown that
ij
=
ji
due to equi-
librium considerations.
*
Finally, Cauchy stress is given as a symmetric tensor of second order. Its com-
ponents may be written as a matrix
=
_
_
11

12

13
12

22

23
13

23

33
_
_
(5.5)
Relation between exposed stress t and tensorial stress
t = n (5.6)
using matrix algebra with vectors t, n and the matrix arrangement of .
Stress written as vector in so called Voigt notation
=
_
_
_
_
_
_
_
_
11
22
33
23
13
12
_
_
_
_
_
_
_
_
(5.7)
utilizing symmetry of Cauchy stress.
Incremental changes of stress during loading history with progressing time t
= lim
t0
t
(5.8)
Voigt notation is used in the following if not otherwise stated.
General description of material behavior
107
The following general form is assumed
3
= C
T
(5.9)
where the tangential material stiffness C
T
may depend on stress , strain and
internal state variables .
Internal state variables comprise a loading history. They are necessary as an ac-
tual state , for a given may lead to different reponses for different loading
histories.
*
Internal state variables require laws describing their evolution depending on the
evolution of stress and strain.
If the components of C
T
are constant, i.e. C
T
= C , Eq. (5.9) may be integrated in
time t to give
= C (5.10)
i.e. a linear material law.
Change of coordinate system
Another cartesian coordinate system is regarded, with the same origin but different
orientation or different directions of axes, respectively. The relation between these
systems is ruled by three rotation angles, see Fig. 5.2a, and a transformation matrix
Qdepending on the rotation angles. The matrix Qis orthogonal, i.e. Q
1
= Q
T
.
Values of the components of stress and strain differ in the two systems. Given a
strain in the original system there is a strain
= Q (5.11)
in the transformed system
4
. A stress in the initial system is related by
= Q
T

(5.12)
to the stress
in the transformed system. With
= C
the material stiffness

transforms according to
C
= Q C Q
T
, C = Q
T
C
Q (5.13)
Regarding a point of a material body the states of stress and strain essentially remain
the same, but are measured in different coordinate systems
5
.
3
Voigt notation allows to write C
T
as a matrix, otherwise tensor notation is required.
4
This and the following is slightly more complicated as actually written as the transformations for stress and
strain differ slightly when using the Voigt notation. But the essentials remain unchanged.
5
Look a vehicle moving in space: it has a velocity vector, which may be measured in different inertial systems
and has different components. But the velocity essentially remains the same.
108 5.2 Some Basics of Continuum Mechanics
Figure 5.2: a) Rotation of coordinate system b) Principal stress space
Principal values, principal orientations
A strain or stress tensor has principal orientations depending on its tensorial com-
ponents, i.e. a coordinate transformation with vanishing shear components, i.e.
12
=
13
=
23
= 0 or
12
=
13
=
23
= 0.
It remain principal values
1
,
2
,
3
or
1
,
2
,
3
as components in three orthogonal
directions which are the principal directions.
*
Sign convention for the following: compression negative, tension positive.
Principal values and directions are by denition independent fromcoordinate system
orientation.
*
Tensors measured in different coordinate systems having the same principal val-
ues and having rotation angles leading to the same principal directions describe
the same state.
Thus, regarding a material point its state of stress and strain is preferably character-
ized by the corresponding principal values and directions.
Invariants are often used as alternative formulation for principal values, i.e. in case
of stress
6
I
1
=
1
+
2
+
3
J
2
=
1
6
_
(
1

2
)
2
+ (
2

3
)
2
+ (
3

1
)
2
=
1
2
_
(
11

m
)
2
+ (
22

m
)
2
+ (
33

m
)
2
+
23
32
+
13
31
+
12
21
J
3
= (
1

m
)(
2

m
)(
3

m
)
(5.14)
with
m
= (
1
+
2
+
3
)/3.
Principal directions of stress and strain coincide only under restrictive assumptions,
see the following Section 5.3.
Deviatoric and pressure parts
Regarding stress pressure is dened as
p =
1
3
(
11
+
22
+
33
) =
1
3
I
1
(5.15)
6
Exactly spoken: I
1
1st invariant of stress, J
2,3
2nd/3rd invariant of stress deviator.
109
Furthermore, deviatoric part
of stress is derived as
=
_
_
_
_
_
_
_
_
11
22
33
23
13
12
_
_
_
_
_
_
_
_
=
_
_
_
_
_
_
_
_
11
+p
22
+p
33
+p
23
13
12
_
_
_
_
_
_
_
_
=
_
_
_
_
_
_
_
_
2
3

11

1
3

22

1
3

33
2
3

22

1
3

11

1
3

33
2
3

33

1
3

11

1
3

22
23
13
12
_
_
_
_
_
_
_
_
(5.16)
This may also be written as
= I
dev
, I
dev
=
_
_
2
3

1
3

1
3
0 0 0
1
3
2
3

1
3
0 0 0
1
3

1
3
2
3
0 0 0
0 0 0 1 0 0
0 0 0 0 1 0
0 0 0 0 0 1
_
_
(5.17)
with the deviatoric unit matrix I
dev
.
Similar relations are valid for strain.
110 5.3 Basic Linear Material Behavior Description
5.3 Basic Linear Material Behavior Description
Isotropy of materials
Notion of action directions, e.g. principal values of imposed strain and their orien-
tation (principal strain directions).
Notion of reaction directions, e.g. principal values of resulting stresses and their
orientation (principal stress directions).
Notion of material directions, i.e. a cartesian coordinate system being spanned by
four material points in the reference conguration.
An isotropic material behaves in the same way in every loading direction:
*
Principal stress directions change in the same way as principal strain directions,
i.e. principal stress directions coincide with principal strain directions
7
.
*
Principal stress values are independent from principal strain directions, i.e.
principal stress values do not change with a change of principal strain direc-
tions relative to material directions but unchanging principal strain values.
A mathematical description
*
We consider a loading with an associated response = C
0
.
*
The loading direction is rotated into another arbitrary direction
with
= Q
according to Eq. (5.11).
*
Isotropy requires, that the rotated associated response
has the the same ma-

terial law, i.e.
= C
0

, while rotating
according to Eq. (5.12). With

= Q
T

= Q
T
C
0

= Q
T
C
0
Q a requirement
C
0
= Q
T
C
0
Q (5.18)
follows for arbitrary rotations Q.
Triaxial isotropy
We assume an abstract
8
material matrix C
0
C
0
=
_
_
C
11
C
12
C
13
C
14
C
15
C
16
C
21
C
22
C
23
C
24
C
25
C
26
C
31
C
32
C
33
C
34
C
35
C
36
C
41
C
42
C
43
C
44
C
45
C
46
C
51
C
52
C
53
C
54
C
55
C
56
C
61
C
62
C
63
C
64
C
65
C
66
_
_
(5.19)
It can be shown that for arbitrary rotations Q the isotropy requirement Eq. (5.18)
can only be fullled with a form
C
0
=
_
_
C
1
C
2
C
2
0 0 0
C
2
C
1
C
2
0 0 0
C
2
C
2
C
1
0 0 0
0 0 0
1
2
(C
1
C
2
) 0 0
0 0 0 0
1
2
(C
1
C
2
) 0
0 0 0 0 0
1
2
(C
1
C
2
)
_
_
(5.20)
7
More precisely they should keep their same relative orientation, but this extension is theoretical.
8
Without a specic material in background.
111
This is equivalent to the triaxal isotropic linear elastic law
C
0
=
_
_
E(1)
(1+)(12)
E
(1+)(12)
E
(1+)(12)
0 0 0
E
(1+)(12)
E(1)
(1+)(12)
E
(1+)(12)
0 0 0
E
(1+)(12)
E
(1+)(12)
E(1)
(1+)(12)
0 0 0
0 0 0
E
2(1+)
0 0
0 0 0 0
E
2(1+)
0
0 0 0 0 0
E
2(1+)
_
_
(5.21)
with Youngs modulus E and Poissons ratio .
Triaxial orthotropy
If Eq. (5.18) is fullled for three rotations Q, with an angle 180 around each coor-
dinate axis, then an orthotropic material is given. Eq. (5.18) leads to
C
0
=
_
_
C
11
C
12
C
13
0 0 0
C
21
C
22
C
23
0 0 0
C
31
C
32
C
33
0 0 0
0 0 0 C
44
0 0
0 0 0 0 C
55
0
0 0 0 0 0 C
66
_
_
(5.22)
with 12 nonzero components, see [Mal69, (6.2.24)]. This form requires that the
materials symmetry directions coincide with the coordinate directions
910
.
Symmetry is assumed due to thermodynamic reasoning, i.e.
C
0
=
_
_
C
11
C
12
C
13
0 0 0
C
12
C
22
C
23
0 0 0
C
13
C
23
C
33
0 0 0
0 0 0 C
44
0 0
0 0 0 0 C
55
0
0 0 0 0 0 C
66
_
_
(5.23)
with 9 nonzero components. A exibility form with some physical evidence ma-
terial coefcents are directly taken from experimental results is given by
D
0
=
_
_
1
E
1
12
E
1
13
E
1
0 0 0
21
E
2
1
E
2
23
E
2
0 0 0
31
E
3
32
E
3
1
E
3
0 0 0
0 0 0
1
G
4
0 0
0 0 0 0
1
G
5
0
0 0 0 0 0
1
G
6
_
_
(5.24)
where
ij
is a measure for the deformation in i-direction caused by a stress in j-
direction. Relations
12
/E
1
=
21
/E
2
, . . . and so on must hold to have symmetry.
9
Otherwise it will be fully occupied, but 12 independent material parameters remain.
10
Stresses and strains will have the same principal directions if they coincide with the the material principal
directions, otherwise not.
112 5.3 Basic Linear Material Behavior Description
Plane strain, plane stress for the isotropic linear elastic law derived from Eq. (5.21)
In both cases
23
=
13
= 0 holds.
Plane strain is given with
33
= 0, i.e.
_
_
11
22
12
_
_
=
E
1 +
_
_
1
12
12
0
12
1
12
0
0 0
1
2
_
_
_
_
11
22
12
_
_
(5.25)
Plane stress is given with
33
= 0. Using this condition with Eqs. (5.10), (5.21)
leads to
_
_
11
22
12
_
_
=
E
1
2
_
_
1 0
1 0
0 0
1
2
_
_
_
_
11
22
12
_
_
(5.26)
Plane stress for the symmetric orthotropic case
It is assumed that the materials symmetry directions coincide with the coordinate
directions with
33
= 0 and
23
=
13
= 0. Thus, the exibility form Eq. (5.24)
leads to
_
_
11
22
12
_
_
=
_
_
1
E
1
12
E
1
0
21
E
2
1
E
2
0
0 0
1
G
6
_
_
_
_
11
22
12
_
_
(5.27)
The inversion yields the stiffness form
_
_
11
22
12
_
_
=
1
1
12
21
_
_
E
1

12
E
2
0
21
E
1
E
2
0
0 0 (1
12
21
)G
_
_
_
_
11
22
12
_
_
(5.28)
with G = G
6
, see also [CS94, (6.110)]. A notation
12
=
1
( deformation in
1-direction caused by a lateral stress in 2-direction) and
21
=
2
( deformation
in 2-direction caused by a lateral stress in 1-direction) is used in the following.
Requiring symmetry
1
E
2
=
2
E
1
and using a modied Poissons ratio
=

2
E
1
E
1
E
2
=

1
E
2
E
1
E
2

2
=

E
1
E
2
E
1
,
1
=

E
1
E
2
E
2
,
1
2
=
2
(5.29)
leads to
_
_
11
22
12
_
_
=
_
_
E
1
1
2

E
1
E
2
1
2
0

E
1
E
2
1
2
E
2
1
2
0
0 0 G
_
_
_
11
22
12
_
_
(5.30)
whereby E
1
, E
2
0 has been assumed.
113
Invariance of shear exibility for the symmetric orthotropic plane stress case
It is assumed that the shear coefcient G does not change for all plane transforma-
tions Q
T
C
0
Q, see also Eq. (5.18), with
Q =
_
_
cos
2
sin
2
cos sin
sin
2
cos
2
cos sin
2 cos sin 2 cos sin cos
2
sin
2
_
_
(5.31)
i.e. a rotation of stresses and strains around the planes normal with an arbitrary
angle . This requirement can be fullled with a form
C
0
=
_
_
E
1
1
1
2
E
1
1
1
2
0
1
E
2
1
1
2
E
2
1
1
2
0
0 0
E
1
+E
2
2
1
E
2
4(1
1
2
)
_
_
=
_
_
E
1
1
2

E
1
E
2
1
2
0
E
1
E
2
1
2
E
2
1
2
0
0 0
E
1
+E
2
2
E
1
E
2
4(1
2
)
_
_
(5.32)
where the coefcient C
33
is invariant
11
. Notice,
1
E
2
=
2
E
1
and C
33
= E/(2(1+
)) for E
1
= E
2
,
1
=
2
, compare Eq. (5.26).
Still the assumption holds, that symmetry and coordinate directions coincide.
Three material parameters remain with the form Eq. (5.32). This is also a major
motivation for the assumption of invariance of shear exibility, which by the way
absolves from the burden of experimental determination of G.
The unixial behavior in 1-direction with
22
= 0 is given by
11
= E
1
11
,
22
=

E
1
E
2
11
(5.33)
and in 2-direction with
11
= 0 by
22
= E
2
22
,
11
=

E
1
E
2
22
(5.34)
This basically allows to derive an orthotropic material parameters from uniaxial
experimental data, which are given by
11
,
11
,
22
and
22
,
22
,
11
. Notice,
22
in
Eq. (5.33) is different to
22
in Eq. (5.34). The same holds for
11
,
11
.
*
The set of four Eqs. (5.33,5.34) is overdetermined with
11
,
11
,
22
and
22
,
22
,
11
given to determine the values of E
1
, E
2
, . Maybe a fest t has
to be found, e.g. with methods of regression.
11
Other coefcients are not necessarily invariant!
114 5.4 Basics of Nonlinear Material Behavior
5.4 Basics of Nonlinear Material Behavior
5.4.1 Tangential Stiffness
Regarding the uniaxial case nonlinear material behavior is characterized by a decreasing
tangential material stiffness. This basic idea is generalized to the multiaxial case.
A formulation of nonlinear material behavior has been given with Eq. (5.9):
= C
T
(5.35)
The tangential stiffness C
T
is subject to change depending on state of stress, strain
and on loading history. We will consider such materials which in some way become
softer during a load history.
Isotropic nonlinear behavior is characterized in the same way as on Page 110:
An isotropic material behaves in the same way in every action direction. Values of
principal stress increments are independent from principal strain directions. For a
given material state directions of principal stress increments coincide with directions
of principal strain increments.
An isotropic tangential material stiffness Eq. (5.9) obeys Eq. (5.18), i.e.
C
T
= Q
T
C
T
Q (5.36)
Hence, the tangential material stiffness C
T
has to follow a form like Eq. (5.20)
which allows only for two independent coefcients.
Starting with the initial unloaded state, all isotropic solid materials including such with
essentially nonlinear behavior may be described initially by Eq. (5.9), whereby the
tangential stiffness is given by Eq. (5.21) with an initial Youngs modulus and an initial
Poissons ratio.
But initially isotropic materials may become anisotropic in higher loading regimes
due to the type of loading.
Thus, forms like Eq. (5.23) or Eq. (5.32) may become appropriate for the tangential
stiffness in case of load induced anisotropy, with the coefcients depending on state
of stress, strain and on loading history.
5.4.2 Stress Limit States
Stress limit states mark the other end compared to initial states. They describe strength of
materials. For initially isotropic materials such strength limit states may be described in
a form
f(
1
,
2
,
3
) = 0 (5.37)
using principal stress values
1
,
2
,
3
. Stress states with f(
1
,
2
,
3
) 0 are admiss-
able, states f(
1
,
2
,
3
) > 0 cannot be sustained. Orientation of principal stress direc-
tions has no inuence in case of isotropic materials.
In a general case the strength limit may be a function of principal stress values and
principal stress directions relative to the material directions.
Stress-strain behavior Eq. (5.35) is often separated from strength limit states Eq. (5.37)
and formally both are treated independently.
5.4.2 Stress Limit States 115
To have a consistent material description the integration of Eq. (5.35) should not
lead to stress states violating the strength limit state Eq. (5.37).
Regarding isotropic strength limit states, they principally allow for anisotropic stress
strain behavior.
Regarding concrete a load induced anisotropy e.g. tensile strength is reached in
one direction while a compressive strength is utilized in orthogonal directions may
be combined with an isotropic strength limit state formulation.
Eq. (5.37) forms a surface in triaxial principal stress space. A stress point on this sur-
face can be described by pressure Eq. (5.39), deviatoric length Eq. (5.40) and deviatoric
angle
12
Eq. (5.42).
Principal stress values span a cartesian coordinate system (principal stress space)
and the corresponding state may be described by a vector. The following elements
are signicant the in principal stress space:
*
Hydrostatic axis: a space diagonal with a direction
13
n
= ( 1 1 1 )
T
/
3.
*
Projection of a stress vector = (
1

2

3
)
T
on the hydrostatic axis:
= ( 1 1 1 )
T
/
3 with length
14
= (
1
+
2
+
3
)/
3.
*
With given, its corresponding deviatoric plane: this plane has as normal,
i.e. ( ) = 0 or
1
+
2
+
3
=
3.
*
Projection of a stress vector = (
1

2

3
)
T
on its deviatoric plane
= =
1
3
_
_
2
1

2

3
1
+ 2
2

3
1

2
+ 2
3
_
_
(5.38)
*
Projection of the particular stress vector = ( 1 0 0 )
T
on the deviatoric
plane:
1
= 2/3( 1
1
2

1
2
)
T
. It has a direction
1
=
_
2/3( 1
1
2

1
2
)
T
called Rendulic direction in the following.
Description of the principal stress state by Haigh-Westergaard coordinates:
*
Length of the hydrostatic projection
=
1
3
(
1
+
2
+
3
) =
I
1
3
I
1
=
3 (5.39)
*
Length of the deviatoric projection (Eq. (5.38))
= [[ =
_
2J
2
J
2
=

2
2
(5.40)
*
Deviatoric angle from Rendulic direction to deviatoric direction
cos =
1

1
(5.41)
A commonly used variation of this formulation is given with
cos 3 = 4 cos
3
3 cos =
3
3
2
J
3
_
J
3
2
(5.42)
with the second and third invariant J
2
, J
3
of the stress deviator, see Eq. (5.14).
Eq. (5.42) yields one solution in the range 0 60. But any interchanging
of
1
,
2
,
3
in invariants leads to the same solution .
12
Also called Lode angle.
13
A direction is a vector of length 1 by denition.
14
This is different to mean stress (
1
+
2
+
3
)/3 or hydrostatic pressure (
1
+
2
+
3
)/3.
Figure 5.3: Deviatoric length and angle in deviatoric plane
*
and span a Rendulic plane.
Regarding isotropic materials principal stress values may interchange their position in
the strength limit condition as the orientation of principal stress directions with respect
to material directions is not relevant: f(
1
,
2
,
3
) = f(
1
,
3
,
2
) = f(
2
,
3
,
1
) =
f(
2
,
1
,
3
) = f(
3
,
1
,
2
) = f(
3
,
2
,
1
). Thus,
1

2

3
(5.43)
(signed!) may be set without loss of generality. From this point of view the deviatoric
angle Eq. (5.42) uniquely characterizes a state of stress.
Strength limit conditions have their own special elements in principal stress space:
Looking at a cylindrical specimen with conning stresses in radial direction and a
distinguished stress in longitudinal direction is useful for the following.
Compressive meridian
1
=
2
>
3
: cylindrical specimen with compression
3
<
0 in the longitudinal direction and circumferential conning pressure
1
=
2
<
0, [
1
[ < [
3
[. From Eq. (5.14) we get J
2
= (
1

3
)
2
/3 and J
3
= 2(
1

3
)
3
/27 and from Eq. (5.42) cos 3 = 1 or = 60.
Tensile meridian
1
>
2
=
3
: cylindrical specimen with circumferential conn-
ing pressure
2
=
3
< 0 and a longitudinal compression
1
< 0, [
1
[ < [
3
[.
From Eq. (5.14) we get J
2
= (
1

3
)
2
/3 and J
3
= 2(
1

3
)
3
/27 and from
Eq. (5.42) cos 3 = 1 or = 0.
Compressive and tensile meridian are determined as the intersection of stress limit
surface with the Rendulic planes with = 60 and = 0.
Strength limit surfaces of concrete form a smoothed, curved tetrahedron, see Fig. 5.4.
Its tip is located in the positive octant (
1
> 0,
2
> 0,
3
> 0) near to the origin
(triaxial tensile strength).
It opens in the negative octant(
1
< 0,
2
< 0,
3
< 0), i.e. generally strength
grows under pressure.
*
More precisely, the admissable deviatoric length increases with the amount of
pressure for a certain range of pressures. This also depends on the deviatoric
angle.
5.4.2 Stress Limit States 117
*
Deviatoric concrete strength under high pressures is not yet really known.
*
From a theoretical point of view there is no strength limit for a pure pressure
15
.
But a material is highly compacted under high pressures with large deforma-
tions.
Its compressive and tensile meridian are slightly curved. The tensile meridian falls
below the the compressive meridian.
Stress-strain behavior within this surface is initially linear elastic and becomes in-
creasingly nonlinear when approaching the limit surface.
Figure 5.4: Stress limit surfaces a) general view direction b) pressure axis view direction
Selection of some popular isotropic strength limit functions for concrete:
Ottosen [Ott77]
f = a
J
2
f
2
c
+
J
2
f
c
+b
I
1
f
c
1 = 0 (5.44)
where f
c
is compression strength (unsigned), a, b are constants, and is a function
of 3
= k
1
cos
_
1
3
arccos(k
2
cos 3)
, for cos 3 0
= k
1
cos
_
3

1
3
arccos(k
2
cos 3)
, for cos 3 0
(5.45)
in which k
1
, k
2
are constants. The four parameters a, b, k
1
, k
2
are determined from
tensile strength f
ct
, biaxial strength, and points on the compressive meridian.
Hsieh/Ting/Chen [HTC82]
f = a
J
2
f
2
c
+b
J
2
f
c
+c

1
f
c
+d
I
1
f
c
1 = 0 (5.46)
with constants a, b, c, d and largest principal stress
1
. This may be written as
f = a
_

f
c
_
2
+
_
b cos + c
_

f
c
+

d

f
c
1 = 0 (5.47)
with hydrostatic length Eq. (5.39), deviatoric length Eq. (5.40) and deviatoric
angle Eq. (5.42).
15
Practically, pure pressure is not reachable in experimental setups. Small deviatoric parts cannot be avoided.
Willam / Warnke [WW75], [CS94, Section 5.5].
=
2
c
(
2
c

2
t
) cos +
c
(2
t

c
)
_
4(
2
c

2
t
) cos
2
+ 5
2
t
4
t
c
4(
2
c

2
t
) cos
2
+ (
c
2
t
)
2
(5.48)
with
= a
0
+a
1
t
+a
2
2
t
,

= b
0
+b
1
c
+b
2
2
c
(5.49)
and

= /f
c
, = /f
c
. The parameters
t
describe the normalized tensile merid-
ian, i.e. = 0 and
c
the normalized compressive meridian, i.e. = 60. The
parameters a
0
, a
1
, a
2
, b
0
, b
1
, b
2
are material constants. As compressive and tensile
meridian should meet at the same point on the -axis a
0
= b
0
.
With given the values of
t
,
c
are determined from Eq (5.49). This may be used
to determine with Eq. (5.48) depending on .
These approaches in a rst view provide the same shapes of stress limit state surfaces, see
Fig. 5.4. Differences are given with details, e.g.
Number of material constants
Exact course of compressive and tensile meridian
Simplicity of formulation
Occurence of sharp edges, i.e. lines with non-unique normals
*
Hsieh has a sharp compressive meridian, Willam/Warnke has no sharp edges.
Biaxial strength as a special case
Biaxial strength is combined with a plane stress state, i.e. one principal stress com-
ponent is zero.
Figure 5.5: a) Biaxial stress limit state b) stress paths
Stress limit states under this condition are given as a special case of the triaxial
limit surface, i.e. the intersection of the triaxial limit surface with any of the planes
1
= 0 or
2
= 0 or
3
= 0. An example is given with Fig. 5.5. We have a closed
line in stress plane instead of a surface in stress space.
For biaxial experimental results see e.g. [KH69].
Uniaxial strength as a further special case
5.4.3 Phenomenological Approach for Biaxial Anisotropic Stress-Strain Relation 119
Uniaxial strength is combined with a uniaxial stress state, i.e. two principal stress
components are zero.
Stress limit states under this condition are given as a special case of the biaxial limit
curve, i.e. the intersection of the biaxial limit curve with the stress coordinate axis
cutting the uniaxial compressive and tensile strength.
Remarks involving stress strain behavior of concrete
Concrete may be regarded as isotropic in its unloaded initial state.
As far as it is known stress limit states are independent from the stress path, i.e.
different loading histories with different ways aiming at the same point of the stress
limit surface actually reach it as nal point.
But the stress strain behavior along these ways may be different.
*
Regarding the stress paths D1 and D2 of Fig. 5.5b path D1 will lead to a load
induced anisotropy due to approaching the tensile strength and cracking rst,
while some amount of compressive strength and stiffness remains in the orthog-
onal direction.
*
On the other hand path D2 will have a more or less isotropic stress strain be-
havior.
*
A simple formulation for load induced anisotropy will be given in the following
Section 6.2.
Approaching the stress limit states stress strain behavior becomes increasingly non-
linear. This may be be described by the formats of damage and plasticity.
5.4.3 Phenomenological Approach for Biaxial Anisotropic Stress-Strain Relation
Load induced anisotropy, which is characteristic for concrete, shall be covered for the
plane stress case.
Basic approach
Consider a material point during its loading history with its principal strains.
These principal strain directions span a coordinate system which will be used in the
following. Each of the two coordinate direction aligns to a distinguished material
direction called principal material direction in the following.
Each principal material direction is regarded as uniaxial. The uniaxial behavior is de-
scribed by a generalized form of Eq. (2.1), see [CS94, 6.8.2]
=
a
1 +
_
a
p
p
2
_

p
+
_

p
_
2
(5.50)
where , are stress and strain in a principal direction,
p
,
p
experimentally determined
values of maximum stress and corresponding strain in a principal direction under biax-
ial conditions and an experimentally determined value a representing the initial tangent
modulus.
The effective initial modulus is assumed with
a =
E
0
1
(5.51)
with the initial
16
Youngs modulus E
0
under uniaxial loading, the corresponding
initial Poissons ratio and the ratio of the principal stress in the orthogonal
direction to principal stress in the direction considered. This approach makes a
material stiffer under biaxial compression.
The parameter a is not known a priori. It inuences the values of
p
,
p
. Its value
has to be estimated and to be corrected if necessary.
The tangential stiffness corresponding to Eq. (5.50) is given by
E
T
=
a
_
1
_

p
_
2
_
_
_
1

p
_
2
+
a
p
_
2
(5.52)
whereby the parameters a,
p
,
p
are assumed as constant. The tangential material
stiffness in a principal directions depends on the strain in that direction. A value
= 0 yields E
T
= a and =
p
yields E
T
= 0.
A combination of Eqs. (5.35), (5.32), and (5.52) is proposed for anisotropic stress-strain
behavior of concrete under plane stress conditions, see [CS94, 6.8.2], [MK96, 3.4.2.3],
and the coefcients of Eq. (5.32) are replaced by values acc. to Eq. (5.52).
This includes a stress limit condition.
Regarding
1
,
2
some margin of discretion remains. An approach is given in the
following Example 5.1
Example 5.1 Modeling of biaxial stress strain behavior with orthotropic hypoelasticity
A stress ratio value
=

2
1
= 0.5 (5.53)
is assumed, which basically rules the nonlinear material behavior.
Corresponding experimental results of [KH69] provide
p1
= 36 MN/m
2
,
p1
= 0.003
and
p2
= 18 MN/m
2
,
p2
= 0.001. Furthermore, E
0
= 30 000 MN/m
2
, = 0.2
are assumed. Firstly, this leads to a = 30 000/(1 0.2 0.5) = 33 333 MN/m
2
with
Eq. (5.51).
Using these values Eq. (5.52) yields for the compressive range
E
T1
=
33 333
_
1
_

1
0.003
_
2
_
_
_
1

1
0.003
_
2
33 333
36

1
_
2
, E
T2
=
33 333
_
1
_

2
0.001
_
2
_
_
_
1

1
0.001
_
2
33 333
18

2
_
2
(5.54)
where strains
1
,
2
have to be considered with their signs. The computed values depend-
ing on the absolute value of
1
,
2
are shown in Fig. 5.6a.
16
The initial material is assumed as isotropic.
121
Figure 5.6: Example 5.1 a) b)
Regarding the values of the Poissons ratio, see Eq. (5.29) with E
1
, E
2
replacing E
T1
, E
T2
,
1
= is assumed leading to
2
=
E
T2
E
T1
(5.55)
This may yield special effects in case of e.g. E
T1
0, but shall not be examined with all
details.
With values determined for E
T1
, E
T2
,
1
,
2
depending on
1
,
2
, a tangential material
stiffness for plane stress conditions may be determined according to Eq. (5.32).
end example 5.1
Limitations
The stress ratio parameter , which has been assumed as constant in deriving the
coefcients of the tangential material stiffness, generally is not constant in applica-
tions. Thus, the values a,
p
,
p
will change during a load history.
Principal directions of stress and strain generally will not coincide due to variations
in . Thus, shear stresses will arise in the principal directions of strains, which are
assmued as natural coordinate system for load induced anisotropy.
It has to be concluded that the extension of phenomenological approaches, which work
well for uniaxial behavior, may become awkward for biaxial and especially triaxial behav-
ior. This motivates for the application of frameworks like damage and plasticity, which
generalize important aspects of mechanical behavior of solids.
5.5 Isotropic Damage
Basic approach for isotropic damage
= (1 D) C
0
(5.56)
This can be applied for the triaxal case with C
0
according to Eq. (5.21), or for plane strain
according to Eq. (5.25) or plane stress according to Eq. (5.26).
Eq. (5.56) introduces a state parameter D. This scalar damage variable by denition has
a range
0 D 1 (5.57)
122 5.5 Isotropic Damage
D = 0 denotes a fully undamaged material, D = 1 a fully damaged material leading to
= 0 for every . The value of D is not allowed to decrease. It may retain its value
during a loading process or increase, i.e.

D 0.
The damage variable D needs an evolution law describing its development from 0 to 1
during a loading history.
Stress based damage approach D depends on stress history
Strain based damage approach D depends on strain history
In the following some simple forms of evolution laws are given for D. A lot of alternative
forms exist, se e.g. [???]. The conventions
1

2

3
or
1

2

3
with signed
values have to be followed.
Strain based damage
Figure 5.7: Graph of Eq. (5.58)
Damage is connected to an equivalent strain , e.g.
D() =
_
0 e
0
1 e
e
0
e
d
g
d
> e
0
, (5.58)
see Fig. 5.7, with constant material parameters e
0
, e
d
, g
d
. This form guarantees the
condition 0 D 1 for arbitrary values 0.
Equivalent strain is connected to strain with a so called damage function, e.g.
*
Approach 1
F =
_

1

1
> 0
0 else
(5.59)
with the larged principal strain
1
and a material constant .
*
Approach 2
F = c
1
J
2,
+
_
c
2
_
J
2,
+c
3
1
+c
4
I
1,
_
2
(5.60)
with the largest principal strain
1
, the 1st strain invariant I
1,
of and the 2nd
invariant J
2,
of the deviator of , see Eq. (5.14) where strain principal values
are used instead of stress principal values. The coefcients c
1
. . . c
4
are further
constant material parameters.
123
In case of uniaxial tension is
2
=
3
=
1
leading to
J
2,
=
1
3
(1 +)
2
2
1
, I
1,
= (1 2)
1
F = c
1
(1+)
2
3

2
1
+
_
c
2
1+
3
+c
3
+c
4
(1 2)
_
1

2
(5.61)
In case of uniaxial compression is
1
=
2
=
3
leading to
J
2,
=
1
3
(1 +)
2
2
3
, I
1,
= (1 2)
3
F = c
1
(1+)
2
2
3
3
+
_
c
2
1+
3
c
3
+c
4
(1 2)
_
3

2
(5.62)
*
A general form is given with F = F(, ) whereby the dependence on has to
be isotropic, i.e. only strain invariants are involved. The equivalent strain is
connected to damage D by, e.g., Eq. (5.58).
Such approaches may be used for triaxial behavior and include biaxial and uniaxial
behavior as a special cases.
So called Kuhn-Tucker constraints relate a change of the damage variable to the
damage function and distinguish loading from unloading:
F 0,

D 0, F

D = 0 (5.63)
*
In case F < 0 is

D = 0, i.e. unloading occurs and damage will not change.
*
In case F = 0 is

D 0, i.e. loading occurs and damage may increase. This
implies a consistency condition
F =
F
+
F
= 0 =
1
F
(5.64)
Stress based damage
Denition of an equivalent stress . Possible approaches:
*
Determination from known yielding functions, e.g. Mises, Drucker-Prager
*
Denitions derived from thermodynamic considerations, e.g. as the comple-
mentary energy norm of the stresses
=
_
2
0
() (5.65)
with the elastic complementary energy function
0
of the virgin material.
Equivalent stress is connected to stress with a damage function, e.g.
F = r (5.66)
with the equivalent damage threshold r. The threshold r is given by e.g.
r = max [r
0
; max
s
] (5.67)
with an initial threshold r
0
and the maximum value max
s
of the equivalent damage
threshold reached during loading history so far.
The connection of r to the damage variable D is given by relatively complex rela-
tions, in most cases not explicitely.
Stress based damage has also to be completed by Kuhn-Tucker constraints, see
Eq. (5.63).
124 5.5 Isotropic Damage
E
0
[MN/m
2
] 30 0000 e
0
[] 2.836 10
4
c
1
[] 1.738
[] 0.2 e
d
[] 1.910 10
3
c
2
[] 0.608
g
d
[] 2 c
3
[] 7.700
c
4
[] 2.993
Table 5.1: Material parameters of Example 5.2
Example 5.2 Modeling of uniaxial stress strain behavior with strain based damage approach 2
The material parameters given in Table 5.1 including the initial value E
0
of Youngs
modulus.
Considering continuously increasing compression with
3
< 0,

D > 0 according to
Kuhn-Tucker constraints Eq. (5.62) yields F =
3
= 0 and =
3
. Eqns. (5.56),
(5.58) in case
3
e
0
lead to
3
= (1 D) E
0
3
= e
3
e
0
e
d
g
d
E
0
3
(5.68)
Considering continuously increasing tension with
1
> 0,

D > 0 Eq. (5.61) yields F =
1
= 0 and =
1
with depending on , c
1
. . . c
4
. The values given in Table 5.1
result in = 10. Eqns. (5.56), (5.58) in case
1
e
0
/ lead to
1
= (1 D) E
0
1
= e
1
e
0
e
d
g
d
E
0
1
(5.69)
Figure 5.8: Example 5.2 a) Uniaxial stress strain curve loading only b) loading, unloading,
reloading
The stress-strain curve resulting from Eqns. (5.68), (5.69) is shown in Fig. (5.8).
The computed initial value of Youngs modulus ratio of stress and strain in the
uniaxial case exactly reproduces the prescribed value of Table 5.1.
125
Uniaxial compressive strength is computed with f
c
= 30 MN/m
2
at a strain
c1
=
0.0015 = 1.5, uniaxial tensile strength with f
ct
= 3 MN/m
2
.
With given values of E
0
, the values of f
c
,
c1
, f
ct
are determined by choosing
e
0
, e
d
, . There is systematic method to determine the values of e
0
, e
d
, with cho-
sen values of f
c
,
c1
, f
ct
[???].
Loading, unloading and reloading
Loading in a range 0
3
1.4 10
3
with
3
< 0,

D > 0, F = 0.
Unloading in a range 1.4 10
3

3
0 with
3
> 0,

D = 0, F =
3
< 0,
i.e.
3
>
with
= 1.4 10
3
.
Change of index in stress and strain from 3 to 1 due to change from compressive
into tensile regime. Actually the physical direction is not changed.
Elastic reloading ranging 0
1

/ with
1
> 0,

D = 0, F =
1

< 0.
Resumed loading ranging
/
1
0.5 10
3
with
1
> 0,

D > 0, F = 0.
end example 5.2
Tangential material stiffness for isotropic strain based approach
From Eq. (5.56)
= (1 D) C
0
C
0

D = (1 D) C
0

0

D (5.70)
On the other hand damage D is a function of the equivalent damage strain , see
e.g. Eq. (5.58), leading to
D =
dD
d
=
dD
d
F
(5.71)
with Eq. (5.64). Finally is
= C
T
(5.72)
with
17
C
T
=
_
(1 D) C
0
+
dD
d
F
0
F
for loading
(1 D) C
0
unloading
(5.73)
The quantities
F
,
F
,
dD
d
have to be computed from the forms for F, D.
17
The form
0
F
is an outer product of two vectors. An outer product ab yields a matrix c with components
c
ij
= a
i
b
j
. This is not necessarily symmetric.
126 5.6 Isotropic Plasticity
5.6 Isotropic Plasticity
General format for stress based plasticity
General triaxial approach for stresses
= C
0
(
p
) (5.74)
with the isotropic linear elastic material matrix C
0
according to Eq.(), total strains
18
and permanent strains
p
. This leads to zero stresses = 0 in case of total strains
equalizing permanent plastic strains =
p
. The rate form
= C
0
(
p
) (5.75)
has to be used for general purposes.
Plastic strains are variable. They are derived with a ow rule
19

p
=

(5.76)
using a ow potential G(,
p
), a plastic multiplier and a state variable
p
com-
prising the load history.
Plastic ow occurs in the case yielding or loading, respectively. Yielding is ruled by
a yield function
F(,
p
) = 0 (5.77)
Loading is distinguished fromunloading by Kuhn-Tucker conditions similar to Eq. (5.63)
F 0,

0, F

= 0 (5.78)
*
In case F < 0 is

= 0, i.e. elastic loading / reloading or unloading occurs and
permanent strains will not change.
*
In case F = 0 is

0, i.e. plastic loading may occur and permanent strains
may chance. This implies a consistency condition similar to Eq. (5.64)
F =
F
+
F
p

p
= 0 (5.79)
The state variable
p
has an initial value
p0
for the unloaded virgin material.
Finally, the formalism has to be completed with an evolution law for the internal
state variable. This is assumed with

p
=

H(,
p
) (5.80)
The functions F(,
p
), G(,
p
), H(,
p
) are material functions specic for a
particular material. They have to be known and to be dened in advance. Further-
more, with an actual state of the material described by a given stress and a given
state variable
d
the loading condition may be controlled.
*
In case of unloading F < 0 the stress increment is given by
= C
0
(5.81)
18
Actually measured with a strain gauge or so.
19
This has shear components
23
,
13
,
12
instead of
23
,
13
,
12
, see Eq. (5.3).
127
*
In case of loading, i.e. F = 0, Eq. (5.79) is combined with Eqs. (5.75), (5.80)
to yield
F
[C
0
(
p
)] +
F
H = 0 (5.82)
Using Eq. (5.76) this can solved for

leading to
=
1
A
F
C
0
, A =
F
p
H +
F
C
0

G
(5.83)
Combining Eq. (5.83), (5.76) and (5.75) leads to an incremental material ac-
cording to Eq. (5.35)
= C
T
(5.84)
with a tangential material stiffness
20
C
T
= C
0

1
A
C
0

G
C
0
(5.85)
with C
0
*
A correct evaluation of the tangential material stiffness is essential for nonlinear
equation solution, see Section 1.7 and Eq. (1.40). Regarding incremental ap-
proaches or discretization in time algorithms like radial return and algorithmic
material modulus are appropriate, [BLM00, 5.9]. Regarding Mises plasticity,
see Example 5.3 these aspects are discussed in Appendix D.
Associated plasticity with identity of ow potential and yield condition G = F is
often assumed. This simplies the whole formalism.
Example 5.3 Modeling of uniaxial stress strain behavior with Mises plasticity
Mises plasticity is a simple approach to describe plastic behavior of metals. It is stress
based and has an associated ow rule with a yield function limiting the deviatoric length,
see Eq. (5.40)
21
F = G =
_
3
2

p
=
_
3J
2
p
,
F
=
G
=
1
2
_
3
J
2
p
= 1 (5.86)
with the 2nd invariant J
2
of the stress deviator, see Eq. (5.14)
2
, and the stress deviator
,
see Eq. (5.16).
Regarding the tangential stiffness according to Eq. (5.85) it may be shown that
F
C
0
= G
_
3
J
2
, G =
E
2(1 +)
(5.87)
with the shear modulus G, and
22
A = H + 3G, C
T
= C
0

G
1 +
H
3G
1
J
2
(5.88)
20
The form
G
is again an outer product of two vectors. An outer product ab yields a matrix c with compo-
nents c
ij
= a
i
b
j
. This is not necessarily symmetric.
21
The tensor components
23
,
32
and so on have to be distinguished while building F/.
22
Full tensor notation for stress and elasticity tensor is required to derive Eq. (5.88)
1
.
128 5.6 Isotropic Plasticity
In case of a uniaxial stress state with
22
=
33
=
23
=
13
=
12
= 0 is
J
2
=
1
3

2
11
, F =
11

p
(5.89)
with the initial Youngs modulus E and Poissons ratio , see Eq. (5.21). For this case the
tangential stiffness, see Eq. (5.85), after some rearrangement is given with
C
T
=
E
1 +
_
_
1
3
(3+13+)
(12)(1+)
1
3
(3++1)
(12)(1+)
1
3
(3++1)
(12)(1+)
0 0 0
1
3
(3++1)
(12)(1+)
1
6
(6+564)
(12)(1+)
1
6
(6+81)
(12)(1+)
0 0 0
1
3
(3++1)
(12)(1+)
1
6
(
6+81)
(12)(1+)
1
6
(6+564)
(12)(1+)
0 0 0
0 0 0
1
2
0 0
0 0 0 0
1
2
0
0 0 0 0 0
1
2
_
_
(5.90)
with = (1 +)
2
3
H
E
.
In case of uniaxial stress also is
22
=
33
=
23
=
13
=
12
= 0. Thus, regarding
Eqs. (5.84), (5.90) with the Voigt-notation according to Eqs. (5.3), (5.7) leads to
23
=
13
=
12
= 0 and

22
=
33
=
3 + 1 +
3 + 2 + 2

11
(5.91)
This in turn may be used to determine

11
= C
T,11

11
+C
T,12

22
+C
T,13

33
=
E
+
2
3
(1 +)

11
=
H
1 +
H
E

11
(5.92)
in case of loading. The elastic part of longitudinal strain may be determined with
el,11
=

11
/E yielding a plastic strain

p,11
=
11

el,11
=
1
H

11

11
= H
p,11
(5.93)
The material function H is assumed as constant for Mises plasticity and plays a role
as hardening modulus. Regarding Eq. (5.89)
2
with F = 0 in case of loading leads to
11
=
p
. But the longitudinal stress
11
also corresponds to a current uniaxial yield
stress f
y
. Thus
p
= f
y
and f
y
replaces
p
for Mises plasticity.
Pure shear with
11
=
22
=
33
=
23
=
13
= 0 and
12
,= 0 may be treated in an
analogous way leading to

12
=
H
3 +
H
G

12
, G =
E
2(1 +)
,
p,12
=
12

el,12
=
3
H

12
(5.94)
Mises plasticity is characterized by four material constants: initial Youngs modulus E,
Poissions ratio , hardening modulus H and initial uniaxial yield stress f
y0
. The value
of f
y0
may be directly taken from a given unixial bilinear stress-strain relation, see e.g.
Fig. 2.10a. The value of H may be indirectly determined from such a relation by trans-
forming Eq. (5.92) into
H =
11
11
1
1
E
11
11
(5.95)
The current yield stress f
y
starts with the initial value f
y0
and changes in case of plastic
loading as is ruled by the hardening modulus. This applies in the same way to tension and
compression.
129
The values for H, f
y0
derived from a uniaxial bilinear stress strain relation may also be
used for the general triaxial case whereby the current yield stress is determined by means
of Eq. (5.86)
1
, i.e. f
y
=
3J
2
.
Ideal Mises plasticity is given with H = 0 and f
y
= f
y0
= const. as a special case.
end example 5.3
For isotropic materials the material functions can be described as functions of stress in-
variants or principal stress values
F = F(
1
,
2
,
3
,
p
), G = G(
1
,
2
,
3
,
p
), H = H(
1
,
2
,
3
,
p
) (5.96)
Orientation of principal stress directions has no inuence for isotropic materials. This
allows the representation of yield criteria as surfaces in principal stress space.
Yield criteria are strongly related to stress limit states, see Section 5.4.2, as the latter are
xed and not depending upon a state variable and furthermore form a boundary for yield
criteria, i.e. F f.
Yield criteria for concrete
Figure 5.9: Surfaces of Mohr-Coulomb and Drucker-Prager yield criteria in principal stress
space
Mises plasticity is not appropriate for concrete as it is independent from pressure
and treats compression and tension in the same way.
Extending the mises plasticity the the Drucker-Prager criterion assumes
F =
p
a I
1
+
_
3J
2

p
(5.97)
For stress invariants I
1
, J
2
see Eq. (5.14). The parameter a is assumed as material
constant and
p
as state variable. Using Haigh-Westergaard coordinates this may be
reformulated as
F =
p
a
3 +
_
3/2
p
(5.98)
With F = 0 this forms a circular cone in principal stress space with a radius
=
_
2/3
p
in the deviatoric plane = 0 and an apex ( = 0) located at
= 1/(
3 a). With a > 0 the yield cone opens in the compressive octant. The
parameter a is measure for the allowed tensile stress.
130 5.8 Localization and Regularization
The Mohr-Coulomb criterion based on Coulomb friction makes the bearable shear
stress in a plane dependent on the planes normal stress
=
p
(c tan ) , 0
p
tan 1 (5.99)
with a nominal cohesion c and a nominal angle of internal friction
23
. Regard-
ing triaxial stress states with Eq. (5.43) the maximum shear stress and its attached
normal stress are given by
m
= (
1

3
)/2,
m
(
1
+
3
)/2, see [Mal69, 3.4].
They are related to a pair , with the largest ratio [/[, see Mohrs circle, by
=
m
+
m
sin , =
m
cos . This leads Eq. (5.99) to
F =
1
2
(
1

3
) +

p
sin
2
(
1
+
3
)
p
c cos (5.100)
F = 0 spans a plane between the compressive meridian F = 0,
2
=
1
>
3
(signed!) and the tensile meridian F = 0,
2
=
3
<
1
(signed!). Regarding
compressive and tensile meridians see Page 116. Cyclic interchanging of principal
stresses leads to totally six planes forming a hexangular cone with an apex at
1
=
2
=
3
= c cot , see Fig. 5.9.
*
The Mohr-Coulomb criterion has the Rankine criterion as a special case with
p
tan = 1. This restricts tensile sustainable tensile stresses but allows un-
bounded compressive stresses.
Both former approaches have characteristic drawbacks regarding concrete strength,
see Page 116: Drucker-Prager has identical compressive and tensile meridians, Mohr-
Coloumb has sharp edges along the meridians with undened yield surface gradi-
ents. A yield criterion to cover these drawbacks may base on the the stress limit
function of Willam / Warnke, see Page 118.
Flow rules for concrete
5.7 Microplane
5.8 Localization and Regularization
Softening
Mesh dependence of energy
Overview regularization methods
Crack band approach
Nonlocal damage
Gradient damage
23
Notice: friction angle here is positive in clockwise direction. A nominal value is considered as material
constant. To cover a hardening the variable multiplier
p
is introduced which serves as an internal state variable.
131
5.9 Long Term Behavior
Creep
Shrinkage
5.10 Short Term Behavior
Strain rate effect
Bibliography
[BLM00] T. Belytschko, W.K. Liu, and B. Moran. Nonlinear Finite Elements for Continua and
Structures. John Wiley & Sons, Chichester, 2000.
[CS94] W.F. Chen and A.F. Saleeb. Constitutive Equations for Engineering Materials, Vol-
ume 1: Elasticity and Modeling. Elsevier Science B.V., Amsterdam, 2. Auage
edition, 1994.
[HTC82] S.S. Hsieh, E. Ting, and W.F. Chen. A plasticity fracture-model for concrete. Int. J.
Solids Structures, 18:181197, 1982.
[KH69] H. Kupfer and H.K. Hilsdorf. Behavior of concrete under biaxial stresses. ACI
Journal, 66:656666, 1969.
[Mal69] L. E. Malvern. Introduction to the Mechanics of a Continuous Medium. Prentice-
Hall, Englewood Cliffs, New Jersey, 1. auage edition, 1969.
[MK96] G. Mehlhorn and J. Kollegger. Anwendung der nite elemente methode im stahlbe-
tonbau. In G. Mehlhorn, editor, Der Ingenieurbau - Rechnerorientierte Baumechanik,
pages 293425. Ernst & Sohn, 1996.
[Ott77] N.S. Ottosen. A failure criterion for concrete. Journal of Engineering Mechanics,
103:527535, 1977.
[WW75] K.J. Willam and E. Warnke. Constitutive model for the triaxial behavior of concrete.
In IABSE Proceedings Vol. 19. International Association for Bridge and Structural
Engineering, 1975.
132
Chapter 6
Deep Beams
6.1 Limit Analysis
Basic ideas
A deep beam with plane stress conditions is given with geometry, boundary condi-
tions, material properties, a unit loading
1
.
*
Material properties in a rst approach are described by Youngs modulus E and
Poissons ratio .
*
Material properties are complemented with values for the uniaxial strength of
concrete and reinforcement.
*
A multiplier ruling load intensity (loading factor) is used as variable.
A state of equilibrium can be determined with a linear elastic analysis. The corre-
sponding state of stress can be linearily scaled with the loading factor.
In a further step, the loading factor is determined such that the internal stress state
does not exceed the strength of the materials ( limit state condition, compare
Eq. (4.18)) all around the deep beam.
*
Thus, the actual stress state in any plate point has to be compared to the mate-
rial strength. This includes strength of concrete as well as strength of reinforce-
ment.
*
The crucial point is to combine resistance of concrete and reinforcement while
regarding the restricted tensile strength of concrete.
The determined loading factor corresponds to a load lower or equal to the limit load
of the system regarding the assumed unit loading type.
*
This procedure is justied by the rst limit theorem of plasticity. It has been
demonstrated for truss models on page 98, but it is principally valid for every
type of structural element.
The procedure conforms to a proof of the limit load for a given system.
A variant is given with the design procedure for a given load, i.e. compressive strength
of concrete or amount of reinforcement are adjusted such, that the limit state condition is
not violated.
The shape and dimension of the concrete body are generally assumed as given in
this context.
1
Load or largest load in a load combination has a value 1.
133
134 6.1 Limit Analysis
Linear elastic analysis
Closed form analytical solutions are available for simple cases, see e.g. [Gir74].
Finite element solutions are appropriate for more complex situations, an example
has been demonstrated on Page 88. A suitable element type is given with the 4-node
quadrilateral, see page 4. Youngs modulus and Poissons ratio, see Eq. (1.21), can
be chosen according to the initial values of concrete.
The calculation yields a stress state
x
,
y
,
xy
for every point of a plate or for every
integration point of every nite element, respectively.
Plane principal stresses
A given plane stress state
x
,
y
,
xy
has principal stresses
1
,
2
with an angle
measured positive counterclockwise from the x-axis
cos 2 =
y
2
_
(
y
2
)
2
+
2
xy
(6.1)
This has one solution for in the range 0 . . . /2, whereupon multiplied by
sign
xy
indicates the direction of
1
. The direction of
2
is perpendicular. The
principal stress values are given by
1
=

x
+
y
2
+
_
_
y
2
_
2
+
2
xy
,
2
=

x
+
y
2

_
_
y
2
_
2
+
2
xy
(6.2)
This has a well known representation by Mohr cycles, see Fig. 6.1a.
An alternative and more general representation of stresses is given with
1
=
x
cos
2
+
y
sin
2
+
xy
2 cos sin
2
=
x
sin
2
+
y
cos
2

xy
2 cos sin
(6.3)
whereby indicates the direction of
1
and +/2 the direction of
2
. This form
allows to control the stress values depending on the direction .
Solving Eq. (6.3)
2
for
x
,
y
,
xy
results in
x
=
1
cos
2
+
2
sin
2
y
= sin
2
1
+ cos
2
xy
= sin cos (
1

2
)
(6.4)
Please note: equations in the sets Eq. (6.1), (6.2) cannot be mixed with or comple-
mented to equations in the set Eq. (6.4).
2
Requiring additionally (
y

x
) cos sin +
xy
(cos
2
sin
2
) = 0 as condition for vanishing of shear
stresses.
135
Figure 6.1: Mohr circles a) common view b) collecting applied, concrete and reinforcement
stress
Concrete stress and reinforcement stresses at a point have to be derived from a given stress
state
x
,
y
,
xy
.
It is assumed that concrete can sustain a compression in a direction
c
, but no
stresses are transferred in the orthogonal direction. This implies cracking in the
orthogonal direction and a principal stress state with
c,1
< 0,
c,2
00. This can
be assumed without loss of generality with an appropriate choice of
c
. Therefore,
using Eq. (6.4) concrete contribution in the global directions is given by
c,x
=
c,1
cos
2
c
,
c,y
=
c,1
sin
2
c
,
c,xy
=
c,1
sin
c
cos
c
(6.5)
A notation
c,1
=
c
is used in the following.
Furthermore, two sets of (distributed, smeared) reinforcement are considered with
prescribed directions
s1
,
s2
and geometric reinforcement ratios
s1
,
s2
. Each
reinforcement set can sustain stresses in its own direction. Eq. (6.4) yields also
reinforcement contributions in the global directions
s1,x
= cos
2
s1
s1,1
,
s1,y
= sin
2
s1
s1,1
,
s1,xy
= sin
s1
cos
s1
s1,1
s2,x
= cos
2
s2
s2,1
,
s2,y
= sin
2
s2
s2,1
,
s2,xy
= sin
s2
cos
s2
s2,1
(6.6)
with
s1,2
=
s2,2
= 0.
Summed contributions are in equilibrium with the given stress state
x
,
y
,
xy
s1
s1,x
+
s2
s2,x
+
c,x
=
x
s1
s1,y
+
s2
s2,y
+
c,y
=
y
s1
s1,xy
+
s2
s2,xy
+
c,xy
=
xy
(6.7)
Inserting Eqns. (6.5), (6.6) into Eq. (6.7) yields three equations for the reinforcement
stresses
s1,1
,
s2,1
, the concrete stress direction
c
and the concrete stress
c
. Thus,
one parameter out of these may be prescribed, and the others result from Eqns. (6.7).
Eq. (6.7) is illustrated by Mohr cycles in Fig. 6.1b.
A remark concerning a stress state parameter open for prescription: reality has a unique
solution. We might miss this solution in the current approach as deformation behavior is
not taken into account. By choosing an appropriate value for, e.g., the concrete compres-
sion direction
c
we hope to yield a good approximation for the real solution.
A common special case
It is assumed that reinforcement directions are aligned to global coordinate axes, i.e.
s1
= 0,
s2
= /2. Regarding Eq. (6.6) this leads to
s1,x
=
s1,1
,
s2,y
=
s2,1
,
s1,y
=
s1,xy
=
s2,x
=
s2,xy
= 0 (6.8)
A notation
s1,x
=
sx
,
s2,y
=
sy
,
s1
=
x
,
s2
=
y
is used in the following.
From Eqns. (6.5)
3
, (6.7)
3
sin
c
cos
c
=

xy
c
(6.9)
With given values
3
of
c
and
xy
this provides up to two solutions for
c
in the
range /2
c
/2 under the condition [
xy
/
c
[ 0.5, see Fig. 6.2.
Figure 6.2: Course of sin
c
cos
c
indicating solution range of Eq. (6.9)
From Eqns. (6.7)
1,2
, (6.8)
sx
=
x

c
cos
2
c
,
y

sy
=
y

c
sin
2
c
(6.10)
Eqns. (6.9), (6.10) are the basis for the following.
A further special case
A cross section, e.g. of a beam, of height z is considered with the conditions
s2
=
0,
y
= 0.
Eqs. (6.7), (6.5), (6.6) yield with omitted index for reinforcement
s
cos
2
s
+
c
cos
2
c
=
x
s
sin
2
s
+
c
sin
2
c
= 0
s
sin
s
cos
s
s
+
c
sin
c
cos
c
=
xy
(6.11)
Let us prescribe
xy
,
s
,
c
and
s
. The remaining values
s
,
c
,
x
may be deter-
mined from the equations. This will be performed for
s
= /2,
xy
= V/z,
s
=
f
y
and leads to
c
=
V
z sin
c
cos
c
,
s
=
V
f
y
z cot
c
, N = V cot
c
(6.12)
with the assumption 0 <
c
< /2, V < 0 and
x
= N/z. Eq. (6.12)
2
cor-
responds to the design rules for stirrups. From Eq. (6.12)
3
the shifting distance
a for the longitudinal reinforcement derives with V = (zN/2)/a leading to
a = z cot
c
/2.
3
Please note:
c
< 0 by denition and
xy
has to be taken with the correct sign.
137
Design strategy for the common special case
With the concrete stress
c
limited by concrete compressive strength f
c
(unsigned)
Eq. (6.9) leads to a constraint [
xy
[ 0.5f
c
. The case [
xy
[ = 0.5f
c
is connected
with [
c
[ = /4. Larger deviations from this direction make the constraint more
restrictive.
To begin with,
c
may freely be chosen within the admissable range not violating
the constraint. This leads to an admissable concrete stress and with Eq. (6.10) to
values
x
sx
and
y
sy
.
*
We assume
x
sx
0,
y
sy
0 and desire to choose
c
such that the total
amount of reinforcement determined by adding Eqns. (6.10)
tot
=
x
+
y
=
1
f
y
_
x

c
cos
2
c
+
y

c
sin
2
c
_
=
1
f
y
(
x
+
y

c
)
(6.13)
is minimized with a choice
sx
=
sy
= f
y
. The contribution
c
is positive
by denition, i.e.
c
=
xy
/sin
c
cos
c
has to be minimized.
*
This is reached with
c
= /4, see Fig. 6.2, depending on the sign of
xy
and leading to
tot,min
=
1
f
y
(
x
+
y
+ 2[
xy
[) (6.14)
Please note: this result is valid only for orthogonal reinforcement meshes with
coordinate directions aligned to reinforcement directions.
Further constraints arise with
x
sx
0 and
y
sy
0. We consider a case that
the foregoing procedure leads to a result of, e.g.
x
sx
< 0.
*
Setting
x
sx
= 0 leaves
y

sy
and
c
and
c
as unknown values to be deter-
mined from three Eqs. (6.9), (6.10).
*
This remaining set of equations is nonlinear and cannot be solved directly. A
numerical method like the Newton-Raphson method may be used instead, see
Eq. (1.65). Collecting Eqns. (6.10), (6.9) by
u =
_
_
y

sy
c
_
_
, f (u) =
_
_
x

c
cos
2
y

c
sin
2
c

y

sy
xy
c
sin
c
cos
c
_
_
= 0 (6.15)
the application of Newton-Raphson, see Eq. (1.65), leads to
u
(+1)
= u
()
_
f
1
u
1
f
1
u
2
f
1
u
3
f
2
u
1
f
2
u
2
f
2
u
3
f
3
u
1
f
3
u
2
f
3
u
3
_
_
1
u=u
()
f (u
()
) (6.16)
With an appropriate start value u
(0)
this generally converges leading to a solu-
tion
c
,
y
sy
,
c
.
*
An analogous method can be used in case with prescribed
y
sy
= 0 and un-
known
x
sx
,
c
,
c
.
The design strategy has to be performed for all relevant points of a deep beam with a
loading
x
,
y
,
xy
given for each point.
Example 6.1 Limit design of reinforcement for deep beam
We refer to the example of Fig. 4.1 with the same system and loading. A state of stress has
been calculated for each integration point of each element with a linear elastic calculation.
Uniaxial concrete compressive strength (unsigned) is assumed with f
c
= 15 MN/m
2
,
reinforcement yield strength with f
y
= 500 MN/m
2
. The width of the deep beam
is t = 0.6 m. Safety factors are not regarded.
Figure 6.3: Example 6.1 a) characteristic stress points b) upper right reinforcement
Four points with different characteristic types of stress state are considered.
Biaxial principal compression
Element 61 at lower right integration point with
x
= 0.86,
y
= 3.66,
xy
=
0.87. This leads to prinicpal compressive stresses with values 3.91, 0.61. A
reinforcement is theoretically not necessary.
Biaxial principal tension
Element 118 at upper right integration point with
x
= 3.90,
y
= 2.27,
xy
=
2.76. This leads to principal tensile stresses with values 5.96, 0.21.
The direction of concrete compressive stress is chosen with
c
= /4. Using
Eq. (6.9) this leads to
c
= 2
xy
= 5.52 MN/m
2
. Setting
sx
=
sy
= f
y
yields from Eq. (6.10)
x
=
1
f
y
_
x

c
cos
2
c
_
= 0.0133
y
=
1
f
y
_
y

c
sin
2
c
_
= 0.0101
(6.17)
This corresponds to reinforcement cross sections with t in cm units a
sx
= t
x
=
79.8 cm
2
/m and a
sy
= t
y
= 60.4 cm
2
/m. These related values are required
locally only, not necessarily over a cross sectional length of 1 m.
To actually reach this state with both reinforcement directions under tensile yielding
and the diagonal concrete direction under compression may require a considerable
cracking. Crack directions should be aligned to concrete compression direction.
139
Mixed principal stresses
Element 91 at upper left integration point with
x
= 0.61,
y
= 0.30,
xy
=
2.24. This leads to principal stresses 2.70, 1.78.
Concrete compressive direction is again chosen with
c
= /4 leading to
c
=
4.48 MN/m
2
. This yields
x
= 0.0033,
y
= 0.0039 in the same way like
Eq. (6.17).
Mixed principal stresses with constraint violation
Element 48 at upper left integration point with
x
= 4.01,
y
= 0.11,
xy
= 0.02.
This leads to principal stress values 4.01, 0.12.
Concrete compressive direction
c
= /4 leads to
y

sy
< 0. Thus,
y

sy
= 0
is prescribed and an iteration is performed according to Eq. (6.16). This yields
c
= 0.10 MN/m
2
,
c
= 0.43 and
x
= 0.0080 with
sx
= f
y
.
Reinforcement design
A minimum constructive reinforcement ratio is assumed with
x,min
=
y,min
=
min
= 0.0015 = 0.15 %, see [din08, 13.6]. This value is not explicitely taken
into account for the computation, but the shown results are modied insofar, as only
those reinforcement ratios are shown in Fig. 6.3b, which exceed
min
.
Additional mesh / bar reinforcement: has to be completed
end example 6.1
Analogy
The basic approach perform a linear elastic calculation for internal forces and
perform a reinforcement design and concrete proof with methods of limit analysis
has strong analogies with the common practice for design and proof of reinforced
concrete beams with respect to bending, shear and torsion.
Concrete strength
The concrete model used within this scope is uniaxial. Actually, there may be ten-
sion in the direction orthogonal to the principal compression direction. Lateral ten-
sion may lead to a decrease in concrete strength. This is regarded by reduction
factors applied on f
c
, see e.g. [CF08, Section 2.2.7] to provide a "safe" value,
which is used as concrete strength in the limit analysis.
There may be more reasons to reduce concrete strength for such calculations.
Ductility requirements
In order to reach stress limit states a larger redistribution of internal forces may be
necessary. This may require larger deformations or a sufcient ductility, respec-
tively.
Construction rules to ensure ductility demands remain to be added.
Limitations
Limit analyis does not regard deformations, i.e. serviceability is not proven.
Examples
See [CF08, Section 2.2.8]
140 6.2 2D Crack Modeling
6.2 2D Crack Modeling
Cracking has already been discussed in Section 2.1, Page 21 and in Section 2.4. These
basic concepts are adapted to the 2D-case in the following.
Rankine cracking criterion
A crack initiates when the largest principal stress reaches the tensile strength.
The crack direction is given with the direction perpendicular to the direction of the
largest principal stress.
Determination of crack length depends on a scan method.
Scan method
A scan method denes how points for testing the cracking criterion are selected.
A mesh based method scans a whole domain in discrete points constituting a regular
or irregular mesh.
*
Regarding 2D nite elements, such a testing mesh is naturally given by the
element integration points.
*
If the testing mesh is sufciently dense, crack propagation is described by an
increasing number of cracked points, e.g. points with a stress state fullling the
Rankine criterion.
*
As a consequence, the crack length question is not explicitely addressed in the
mesh based scan method.
A boundary based method distinguishes crack initiation and crack propagation.
*
A crack in most cases initiates from a boundary. Thus, boundary points are
scanned for testing the cracking criterion. If a scanned point fullls the criterion
for crack initiation, it becomes a crack tip.
*
A crack may propagate with the movement of its crack tip. Thus, a criterion for
the event of crack tip movement has to be dened
4
.
*
Upon fullling such a criterion, a rule to determine the direction of crack tip
movement has to be dened. An additional criterion has to be dened to deter-
mine the length of the crack tip movement.
*
Altogether, this requires quite sophisticated models to describe crack propaga-
tion.
The mesh based method will be used in the following. This method has proven its practi-
cability and avoids dealing with the crack length problem. A drawback may be considered
insofar, as a crack geometry is not precisely captured, but cracking of concrete is a diffuse
matter anyway.
4
The Rankine criterion is not suitable in all cases for this particular purpose.
141
Rotating crack vs. xed crack
Testing of a particular point with a known stress state is performed with, e.g., the
Rankine criterion. This decides whether this point has to be regarded as cracked, and
if so, to determine its orientation. Cracking requires a reformulation of the material
stiffness, which in the 2D case depends on the crack orientation.
Regarding a load history, principal stress orientations may change. This leads to two
alternative concepts regarding stress orientations, xed crack and rotating crack.
*
Fixed crack: crack orientation is xed with that value occurring in crack initi-
ation. This corresponds to clearly separated opposite crack surfaces or macro
cracks, see Fig. 2.2b.
*
Rotating crack: crack orientation follows the direction of principal stresses, i.e.
it may change during a load history. This corresponds to crack bands and micro
cracking, where the orientation of a bunch of micro cracks may change upon
changing of principal tensile stress direction, see Fig. 2.2b.
Although the xed crack concept seems to be more realistic from a phenomenolog-
ical point of view it shows some disadvantages.
*
In practical applications it yields a too stiff behavior of the cracked material in
many cases.
*
Furthermore, it introduces a relative sliding of crack surfaces resulting in fric-
tion forces. This requires the formulation of a friction-slip law which is difcult
to formulate.
Thus, in the following we will start with the rotating crack concept which is easier to
handle.
Cohesive crack
An uncracked point with a computed stress state is considered. A local coordinate
system is chosen aligned to principal stress directions with the x-axis in principal
tensile direction.
Figure 6.4: a) Bilinear cohesive crack model b) smeared crack concept
A crack is assumed to initiate if the computed principal tensile stress exceeds the
tensile strength in a certain point. Principal tensile stress is orthogonal to crack sur-
faces by denition in the rotating crack concept. Therefore, the immediate cracking
situation can be regarded as uniaxial and direction indices are dropped temporarily.
A process zone arises on crack initiation. It is characterized by high strains within a
band width b
w
, see Page 22. The distance b
w
accounts roughly to a small multiple
of the largest aggregate diameter and is kept constant.
A crack width w is dened as the distance change between left and right border of
the process zone. This distance starts with a value b
w
. Assuming a uniform strain
cb
within the process zone the crack width is given by
w = b
w

cb
(6.18)
see also Eq. (2.4). With a developing macro crack the assumption of a
cb
is not
strictly realistic anymore, but the denition of a crack width still holds.
Uniaxial cracking behavior of concrete is characterized by the cohesive model with
crack traction depending on crack width, see Section 2.4 and Fig. 2.8.
The crack traction-crack width relation will be simplied as bilinear in the follow-
ing, see Fig. 6.4a. This leads to a relation
=
_
f
ct
_
1
wb
w
ct
w
cr
_
b
w
ct
< w w
cr
0 w > w
cr
(6.19)
with the concrete strain
ct
= f
ct
/E upon cracking, the crack traction and the
crack width w. The gray shaded area corresponds to the crack energy G
f
, see
Eq. (2.6), which is assumed as material constant. The smallest crack width with
vanishing crack tractions, the so called critical crack width w
cr
is determined by
w
cr
= 2G
f
/f
ct
+b
w
ct
as a material constant.
Single Smeared crack
We introduce the notion of a black box embracing of piece of material with an
orientation in the crack normal direction. The box is chosen such, that one crack
arises in it, but the exact point of cracking is not localized.
Furthermore, the concept of a crack band is incorporated, see Page 22, with a crack
band width b
w
within the box.
Assigning the box with a length L
c
b
w
leads to a box strain in crack normal
direction
=
1
L
c
_
(L
c
b
w
)
m
+b
w

cb
_
= (1 )
m
+
cb
, =
b
w
L
c
(6.20)
see Eq. (2.28) and Fig. 6.4b, with the strain
m
of the uncracked bulk material and
the strain
cb
within the crack band.
Essentially, the smeared crack approach allows to use conventional continuum 2D
nite elements with a continuous displacement approach as described on Page 4 to
model cracks, which are are more or less discontinuous in nature.
143
Stress strain relation for single smeared crack with monotonuous loading
Direction indices are reused again.
Because of equilibrium considerations the stress component in the bulk material in
crack normal direction has to be the same as the crack traction
1
from Eq. (6.19).
Bulk material stress in crack tangential direction is
2
. Thus,
m
1
=
m
1
(
1
,
2
) in a
2D case. For the sake of brevity a linear elastic plane stress behavior according to
Eq. (1.21) or (5.26) is assumed in the following
5
, i.e.
m
1
=
1
E
(
1

2
) ,
2
=
1
E
(
2

1
) . (6.21)
with the bulk material strain
2
in the crack tangential direction and its Youngs
modulus E and Poissons ratio .
Furthermore, from Eq. (6.20)
1
=
1
L
c
_
(L
c
b
w
)
m
1
+b
w

cb
1
_
= (1 )
m
1
+
cb
1
, =
b
w
L
c
(6.22)
Strains
1
,
2
are generally given from a superordinated calculation. With given con-
stant values E, , b
w
, L
c
, f
ct
, w
cr
there remain ve unknowns
1
,
2
, w
1
,
m
1
,
cb
1
and ve equations (6.18), (6.19), (6.21), (6.22).
Furthermore, b
w
L
c
or 1, i.e. element size is large compared to crack band
width, and
ct
b
w
w
cr
, i.e. elastic energy until cracking is small to crack energy,
will be assumed to simplify the relations.
Regarding a range
1
>
ct
, w
1
< w
cr
with
ct
= f
ct
/E, = L
c
/w
cr
a solution
for stresses is given by
1
= f
ct
1
1

2
1
ct
(1
2
)
,
2
= E
c
(1
ct
)
2

ct

1
+
ct
1
ct
(1
2
)
(6.23)
and for the crack width
w =

1
+
2

ct
(1
2
)
1
ct
(1
2
)
L
c
(6.24)
Eq. 6.23 leads to a tangential material stiffness
C
T
=
E
1
ct
(1
2
)
_

ct

ct
ct
1
ct
_
(6.25)
The parameter
ct
somehow plays a keyrole. It should be
ct
< 1, which leads to
restrictions to L
c
for a given material.
The relations Eq. (6.21), (6.25) are derived in a local coordinate system aligned to
the principal axes of stress and strain
6
. This is based on a strain state
x
,
y
,
xy
leading to principal strains
1
,
2
and a principal strain orientation
7
. Stresses and
tangential material stiffness have to be transformed to the global system for usage
with, e.g., Eqns. (1.33)
1
, (1.40) according to Eqns. (5.12), (5.13).
5
Plane strain, nonlinear or anisotropic approaches essentially follow the same procedure.
6
Both coincide within this frame due to the rotating crack approach.
7
This is calculated by Eq. (6.1) with
x
,
y
,
xy
replaced by
x
,
y
,
xy
/2 as
xy
dened by Eq. (1.2).
Extension to multiple cracks
Remains to be completed.
Un- and reloading
Unloading of a crack band, i.e. a reduction of crack width, leads to decreasing
strains and decreasing stresses within the crack band. A simple model is shown in
Fig. 6.4a.
Reloading of a crack band approximately goes along the unloading path until the lat-
est maximum crack width is reached. Then it follows the loading path, see Fig. 6.4a.
Eqns. (6.23) - (6.25) have to be modied for these cases.
Softening, regularization and characteristic length of elements
The cohesive crack approach, e.g. Eq. (6.19), implies softening, i.e. increascing
deformations with decreasing stresses. This leads to a localization in nature, see
Fig. (2.3), and also in numerical models with nite elements.
Localization is generally given with relatively narrow bands of high strains com-
pared to the adjacent areas.
If no special measures are taken a localization band in a nite element mesh spreads
along one element in its normal direction. This holds in the same way for small
element sizes and for large element sizes.
A key point for numerical models arises with the proper description of the crack
energy, see Page 22. It must be the same for small nite elements and large nite
elements.
*
Regarding the smeard crack concept this is reached with Eqns. (6.19) - (6.18)
and the incorporation of L
c
, w
cr
, f
ct
. The latter both yield a prescribed crack
energy G
f
.
Discrete Crack modeling
The smeared crack crack approach allows to model discontinuous cracks with con-
tinuous displacement elds. Cracks are in a black box, which can be seen as a whole
only.
For an explicit modeling of cracks or displacement discontinuities nite element
extensions like enhancements with internal degrees of freedom or extended nite
elements (XFEM) or other interpolation techniques like elementfree Galerkin meth-
ods (EFG) are necessary.
145
6.3 2D Modeling of Reinforcement and Bond
To start with a single reinforcement bar may be modeled with 2Dbar elements, see Page 3.
Material behavior of reinforcement is uniaxial and may be described by the uniaxial
elasto-plastic law, see Section 2.5.
Thus, modeling of a reinforcement bar may be considered as a special case of a
elasto-plastic truss, which has already been discussed in Section 4.3
But this truss is embedded in concrete and interacts with it via bond.
Bond
For bond mechanisms see Section 2.6.
Figure 6.5: Bond a) rigid b) exible
Rigid bond model
*
It is assumed that slip between concrete and reinforcement can be disregarded.
*
As a consequence, nite elements for reinforcement on one hand and concrete
on the other hand share the same nodes, see Fig. 6.5a.
*
This enforces the same displacements of concrete and reinforcement in nodes.
There might be minor displacement differences along a rebar axis between
nodes, but this is not signicant if the mesh is not too coarse.
Flexible bond model
*
Slip between concrete and reinforcement in the longitudinal direction is re-
garded while both have the same displacement in the lateral direction basically.
*
As a consequence nite elements for reinforcement on one hand and concrete
on the other hand have to have their own nodes, see Fig. 6.5b.
*
Concrete nodes and reinforcement nodes have to be connected by a special type
of spring elements, see Page 4.
*
This type of element is a so called bond element, which is constrained or very
stiff in the lateral rebar direction and has a bond law, see Fig. 2.11b, in the
longitudinal direction.
For sake of brevity and simplicity a rigid bond will be assumed in the following, if
not otherwise stated.
Reinforcement mesh
146 6.3 2D Modeling of Reinforcement and Bond
To model each bar of a reinforcement mesh with a nite element bar is very elab-
orate. A so called smeared model is used instead, i. e. reinforcement bars of one
direction are modelled as a sheet. The cross section of a reinforcement mesh is
charcterized by single bar cross section A
s
and bar distance s. This leads to a sheet
thickness
t =
A
s
s
(6.26)
A more common notation is abbreviated with a
s
which denotes the total reinforce-
ment cross section in cm
2
related to a width of 1 m, i.e. a
s
= 100 t if t is measured
in m.
A reinforcement sheet is regarded as a plate with an anisotropic behavior.
*
It is assumed that a reinforcement sheet has a direction
s
given by the direc-
tion
8
of its bars. A rotated cartesian coordinate system is assigned with the
x
-axis in the bar direction.

*
A given global strain state =
_

x

y

xy
_
T
can be transformed to the
rotated system using Eq. (5.11)
= Q (6.27)
with
Q =
_
_
cos
2
s
sin
2
s
cos
s
sin
s
sin
2
s
cos
2
s
cos
s
sin
s
2 cos
s
sin
s
2 cos
s
sin
s
cos
2
s
sin
2
s
_
_
(6.28)
This leads to a strain
x
which can be used to determine a stress
x
according to
the uniaxial material law Eqs. (2.30), (2.31) appropriate for the reinforcement.
The rotated stress state is given with
=
_

x
0 0
_
T
. This is transformed
back to the global system with Eq. (5.12)
= Q
T

(6.29)
*
The tangential material stiffness in the rotated system is given by
C
T
=
_
_
C
T
0 0
0 0 0
0 0 0
_
_
(6.30)
with C
T
= E
s
in case of loading belowyield limit and unloading and C
T
= E
T
at the yield limit, see Fig. 2.10.
*
This is transformed to the global system with Eq. (5.13)
C
T
= Q
T
C
T
Q = C
T
_
_
cos
4
s
cos
2
s
sin
2
s
cos
3
s
sin
s
cos
2
s
sin
2
s
sin
4
s
sin
3
s
cos
s
cos
3
s
sin
s
sin
3
s
cos
s
cos
2
s
sin
2
s
_
_
(6.31)
leading to a symmetric but fully occupied matrix C
T
.
With Eqs. (6.27)-(6.31) the necessary components are given for modeling a rein-
forcement mesh with 2D continuum elements, e.g. 4-node quadrilateral elements as
described on Page 4.
8
Measured positive counterclockwise starting from the global x-direction.
147
Overlay of elements
Figure 6.6: Overlay of elements
Each reinforcement direction may be modeled separately with its own nite element
mesh as has been described before.
In the same way the concrete part may be modeled separately.
An appropriate discretization approach is given by an element overlay, i.e. nite
elements for several reinforcement directions and nite elements for concrete share
the same geometry and assuming a rigid bond the same nodes, see Fig. 6.6.
Finally bar elements for single reinforcement bars may be superposed.
Example 6.2 Nonlinear simulation for deep beam
We refer to the deep beam example given with Fig. 4.1. In contrast to Example 6.1 the
nonlinear stress-strain behavior of concrete and reinforcement will be considered to some
extent.
Thickness of deep beam is t = 0.6 m.
Concrete material model according to Page 143ff. The following material properties
are assumed: Youngs modulus E
c
= 31 900 MN/m
2
, Poissons ratio = 0.2,
tensile strength f
ct
= 1.0 MN/m
2
, crack energy G
f
= 50 Nm/m
2
.
Reinforcement material model according to Section 2.5. The following properties
are assumed: Youngs modulus E
s
= 200 000 MN/m
2
, yield limit f
y
= 500 MN/m
2
,
ultimate strength f
t
= 550 MN/m
2
, strain at ultimate strength
u
= 0.05. The latter
values lead to a hardening modulus E
T
= 1053 MN/m
2
.
Smeared reinforcement mesh model according to Page 145. An orthogonal mesh
is chosen with
s1
= 0 , t
s1
= 0.006 m and
s2
= 90 , t
s2
= 0.006 m which
corresponds to a reinforcement ratio of
s1
=
s2
= 1 %.
Plane stress conditions are assumed throughout the example.
Loading
Loading is applied by displacement control of a node, where a concentrated load
acts upon, see Fig. 4.1a. Distributed loads are neglected. Actual loading is given by
the reaction force of the displaced node.
*
In contrast to load control the control of displacements allows to model limit
states of a structure, i.e. limited loading with increasing displacements. The
term loading is maintained also for control of displacements.
The target displacement of the loaded node is prescribed with 0.036 m.
Discretization and computation method
Discretization is shown in Fig. 6.7a. It consists of 209 nodes and 3 176 + 1 four-
node quadrilateral elements. Thereof, 176 elements are used for the concrete layer
and 176 elements each for the two reinforcement directions.
Figure 6.7: Example 6.2: a) discretization b) load deection curve
Integration order for 4-node quadrilateral elements, see Eqs. (1.28), (1.33) and (1.62),
is generally chosen with 2 2.
Special care has to be given for the lower left hand support. A nodal support of
the RC deep beam itself is not appropriate as this leads to concentrated high tensile
stresses. A more realistic approach is given with a support by an extra element,
whereby this element is assumed as linear elastic with E
c
= 31 900 M/m
2
, = 0.2
and with a vertical nodal support of its lower two nodes.
As the problemis physically nonlinear with a non-smooth, rough characteristic
9
spe-
cial care has to be given to the iteration matrix in an incrementally-iterative solution
approach, see Page 15. The BFGS method in combination with line search is used
instead of the Newton-Raphson method, as the latter will not lead to convergence
during the equilibrium iterations.
In case of a new crack detection an equilibrium iteration sequence is performed
without applying a load increase. This procedure is separately performed for each
new crack. New loading is only applied on a cracked system in equilibrium.
Proper selection of load or pseudo time step increments is a matter of experience
and sometimes intuition is required in such problems. A value t = 0.01 is chosen
here with a target value t = 1.0. Thus, 100 iteration sequences arise caused by
load increments. But there are roughly 500 more iteration sequences caused by
cracking. Each iteration sequence may have a large number of single iterations
before it reaches convergence or equilibrium, respectively. Taking all together the
computation time is quite high.
Results of numerical computation
9
This is mainly caused by the multi-linear post cracking behavior, see Fig. 6.4a, leading to derivatives with jumps.
149
The computed curve for the load depending on the deection of the loaded node
is shown in Fig. 6.7a. Four stages can be recognized as is typical for reinforced
concrete:
*
Uncracked state I.
*
Ongoing cracking state IIa with elastic reinforcement behavior.
*
Finally cracked state IIb with elastic reinforcement behavior.
*
Finally cracked state III with yielding reinforcement.
*
The transitions are gradual in the current example, as a number of elements are
included in a state change.
The computed loading for a displacement of 4 cm is computed with P = 14.3 MN.
This should be near to the ultimate limit load due to the horizontal characteristic of
the load-deection curve.
Concrete stresses in the nally computed step are shown in Fig. 6.8a.
Figure 6.8: Example 6.2: a) concrete principal stresses b) reinforcement principal stresses
*
Low tensile stresses remain according to the prescribed tensile strength.
*
Minimumconcrete stress (maximumcompression) is computed with 96 MN/m
2
near the load application point. Remember, a linear elastic behavior is assumed
for compressive concrete in the current model.
This computed value is beyond the compressive strength of concrete.
On the other hand, a point load as assumed is not realistic.
Furthermore, a biaxial compressive state is given, which leads to a strength
increase of concrete.
Reinforcement stresses in the nally computed step are shown in Fig. 6.8b.
*
Horizontal lines belong to the stresses of the horizontal reinforcement direction,
vertical lines to the stresses of the vertical stress direction.
*
Horizontal stresses reach the yield strength in the lower midspan area and the
upper right side support area.
*
Small reinforcement areas have compression due to bond with concrete.
Control and comparative calculations
External loading is computed with P = 14.3 MN in the nal state and vertical
support reaction on the left hand lower support with A = 7.46 MN.
A comparative calculation with a simple beam model span L = 10.5 m, loaded at
L
p
= 3.5 m, hinge support on left side, fully restrained on right side leads to a left
hand support reaction of A = 7.41 MN and a span moment M
f
= 22.2 MNm.
Figure 6.9: Simple beam model of Example 6.2
An internal lever arm of d = 3 m between resulting compression and tension forces
is assumed regarding the cross section below the loaded node. This leads to a
reinforcement force of F
s
= 22.3/3 = 7.4 MN leading to a reinforcement of
A
s
150 cm
2
. With a reinforcement ratio = 1 % and a deep beam thickness
t = 0.6 m this requires a reinforcement area of 150/0.6 2.5 m in vertical di-
rection beginning from the bottom. This roughly matches to the area of computed
reinforcement yielding, see Fig. 6.8b.
Summary
The model of this example combines equilibrium (in an integral form), nonlinear
material behavior with limited material strength and, in contrast to the companion
Example 6.1, kinematic compatibility.
An issue remains with the actually limited compressive strength of concrete, which
has not been considered.
end example 6.2
151
6.4 Biaxial Concrete Material Laws
Concrete under biaxial compression plane strain
Biaxial orthotropic extension of Saenz proposal, see Eq. (2.1), according to [LNS72],
se also [CS94, 6.8].
*
Classication: Fit of Measurements (FoM) approach.
*
Monotonic loading without un- and reloading.
*
The coordinate system is coinciding with the principal strain directions.
*
We assume stresses and a stress ratio
22

11
< 0, =

11
22
1 (6.32)
*
The stress ratio can be related to a pair of strength values f
c2
, f
c1
(unsigned),
compare Fig. 5.5, and corresponding strain values
c2
,
c1
(signed).
The transfer of Eq. (2.1) for the x-direction of the biaxial case is given with
22
=
E
22
1 +
_

E
E
c2
2
_

22
c2
+
_
22
c2
_
2
,

E =
E
c0
1
, E
c2
=
f
c2
c2
(6.33)
with the initial Youngs modulus E
c0
and Poissons ratio , which is assumed as
constant. This leads to a corresponding tangential stiffness, see Eq. (2.2)
E
t2
=
d
22
d
22
=
E
_
1

2
22
2
c2
_
_
1 +
_

E
E
c2
2
_

22
c2
+

2
22
2
c2
_
2
(6.34)
where the stress ratio is assumed as constant with respect to differentiation
10
.
In the same way we get a relation between
11
,
11
, where index 2 is replaced by
index 1 in the Eqs. (6.33), (6.34). This leads to different behavior in the 1- and 2-
coordinate directions depending on the loading ratio (load induced anisotropy).
Special cases
*
Uniaxial loading = 0: this leads to

E = E
c0
and Eqs. (2.1), (6.33) are the
same.
*
Biaxial isotropic loading = 1 and
11
=
22
and
11
/
c1
1:
i
= E
c0
/(1
)
i
, i = 1, 2. This corresponds to Eq. (5.26).
Both tangential stiffnesses E
t2
, E
t1
are used to construct in incremental material law
like Eq. (5.9), with an orthotropic tangential material stiffness formally according to
Eq. (5.32)
= C
T
, C
T
=
_
_
E
t1
E
t2
0
0
0 0
E
t1
E
t2
E
t1
+E
t2
+2E
t2
_
_
, =
E
t1
E
t1
E
t2
2
(6.35)
which regards also a shear strain increment
12
.
The tangential material stiffness with respect to directions of principal strain / or-
thortopy has to be transformed into the global coordinate system using Q
T
C
T
Q
(???) with Qaccording to Eq. (5.31).
10
This should at least be approximately true for proportional loading, i.e.
22
=
20
,
11
=
10
with xed
values for
20
,
10
.
152 6.5 Further Aspects and Application Case Studies
6.5 Further Aspects and Application Case Studies
Loading by external load control vs. displacement control
Arc length method
Benchmarks
Comparison of different approaches for concrete material laws
Complex loading paths
Extension of 2D panels to 3D folded plate
thin walled girders
Bibliography
[CF08] CEB-FIP. Practitioners guide to nite element modelling of reinforced concrete
structures, volume Bulletin Nr. 45. International Federation for Structural Concrete
p, Lausanne, 2008.
[CS94] W.F. Chen and A.F. Saleeb. Constitutive Equations for Engineering Materials, Volume
1: Elasticity and Modeling. Elsevier Science B.V., Amsterdam, 2. Auage edition,
1994.
[Gir74] K. Girkmann. Flchentragwerke. Springer-Verlag, Wien, 6. Auage edition, 1974.
[LNS72] T.C.Y. Liu, A.H. Nilson, and F.O. Slate. Biaxial stress-strain relations for concrete.
Journal of the Structural Division, 98:1025, 1034 1972.
153
Chapter 7
Slabs
7.1 Cross Sectional Behavior
7.1.1 Kinematic Basics
Reference plane
Coordinates x, y in the reference plane directions and z in the transverse direction.
The reference plane is placed in the midst of a slab. Bottom and top coordinates are
given by z
1
= h/2, z
2
= h/2 with the slab thickness h.
Bernoulli-hypothesis (cross sections remain plane during deformation)
Displacements
w(x, y, z) = w(x, y, 0) = w(x, y)
u(x, y, z) = u(x, y) z
_
w(x,y)
x

x
(x, y)
_
v(x, y, z) = v(x, y) z
_
w(x,y)
y

y
(x, y)
_
(7.1)
with the displacements u(x, y), v(x, y) of the reference plane. This inlcudes plates as a
special case with the inclusion of displacement components u, v.
Strains, compare Eqns. (1.1)-(1.3)
x
=
u
x
=
u
x
z
_
2
w(x,y)
x
2

x
(x,y)
x
_
y
=
v
y
=
v
y
z
_
2
w(x,y)
y
2

y
(x,y)
y
_
xy
=
u
y
+
v
x
=
u
y
+
v
x
z
_
2

2
w
xy

x
y

y
x
_
xz
=
u
z
+
w
x
=
w
x
+
x
+
w
x
=
x
yz
=
v
z
+
w
y
=
w
y
+
y
+
w
y
=
y
(7.2)
The strains of the reference plane are given by

x
=
u
x
,
y
=
v
y
,
xy
=
u
y
+
v
x
(7.3)
154
7.1.1 Kinematic Basics 155
With the inclusion of shear deformations curvature is advantageously dened as
1
x
=

2
w
x
2

x
x
y
=

2
w
y
2

y
y
xy
= 2

2
w
xy

x
y

y
x
(7.4)
With Eq. (3.2) this leads to strains
x
=
x
z
x
,
y
=
y
z
y
,
xy
=
xy
z
xy
(7.5)
with
x
,
y
,
xy
linearily varying along the beam height with extremal values at the top
and bottom of the cross section and constant
xz
=
y
,
yz
=
y
along the beam height.
Nevertheless, all these strains are varying with the reference plane coordinates x, y.
At a given location x, y these strains may be transformed to other directions leading
to values
x
,
y
,
xy
. Direction angle is measured by counterclockwise against the
x-axis. The transformation rule is given by
_
_
xy
_
_
=
_
_
cos
2
sin
2
cos sin
sin
2
cos
2
cos sin
2
sin
2
_
_
_
_
xy
_
_
(7.6)
This transformation is connected with a principal direction
with
xy
= 0 and
principal strains
1
,
2
. For a given x, y the principal values
1
,
2
,
may vary
along the height z.
For a given x, y the reference axis strain
x
,
y
,
xy
and curvature
x
,
y
,
xy
have
the same transformation rule Eq. (7.6). Thus, they also have principal values and
directions
1
,
2
,

and
1
,
2
,
, respectively.
The orientation of the principal systems generally must not be the same, i.e.
,

and
generally may have different values. But
=

in case of
x
=
y
=
xy
= 0 and
in case of
x
=
y
=
xy
= 0 due to Eq. (7.5).
Regarding kinematics there are displacement variables u, v, w and deformation vari-
ables
x
,
y
,
xy
,
x
,
y
,
xy
,
x
,
y
. Corresponding are internal forces n
x
, n
y
, n
xy
,
m
x
, m
y
, m
xy
, q
x
, q
y
, see Fig. (7.1).
At a slab location x, y these are given by
n
x
=
_
h/2
h/2
x
dz, n
y
=
_
h/2
h/2
y
dz, n
xy
=
_
h/2
h/2
xy
dz
m
x
=
_
h/2
h/2
x
z dz, m
y
=
_
h/2
h/2
y
z dz, m
xy
=
_
h/2
h/2
xy
z dz
q
x
=
_
h/2
h/2
xz
dz, q
y
=
_
h/2
h/2
yz
dz
(7.7)
where
x
,
y
,
xy
,
xz
,
yz
depend on
x
,
y
,
xy
,
xz
,
yz
. For comparisons with beams
see Eq. (3.8).
1
x
,
x
,
xy
and
x
,
y
,
xy
each form the component of a second order tensor which are each dened by a force
direction and the direction of a reference plane.
156 7.1 Cross Sectional Behavior
7.1.2 Linear elastic behavior
We consider plane stress Eq. (1.21) and with Eq. (7.5) gain as normal stresses
x
=
E
1
2
(
x
+
y
) =
E
1
2
[(
x
+
y
) z (
x
+
y
)]
y
=
E
1
2
[(
y
+
x
) z (
y
+
x
)]
(7.8)
and as a shear stress in planes parallel to the reference plane with G = E/2(1 +).
xy
= G
xy
= G(
xy
z
xy
) (7.9)
The both remaining shear stress components are given similarily
xz
= G
xz
= G
x
,
yz
= G
yz
= G
y
(7.10)
Internal forces (
_
h/2
h/2
zdz = 0!)
n
x
=
E
1
2
(
x
+
y
)
_
h/2
h/2
dz = K
n
(
x
+
y
)
n
y
= K
n
(
y
+
x
)
n
xy
=
E
2(1 +)

xy
_
h/2
h/2
dz = (1 ) K
n

xy
2
m
x
=
E
1
2
(
x
+
y
)
_
h/2
h/2
z
2
dz = K (
x
+
y
)
m
y
= K (
y
+
x
)
m
xy
=
E
2(1 +)

xy
_
h/2
h/2
z
2
dz = (1 ) K

xy
2
= (1 ) K

2
w
xy
q
x
= G
x
_
h/2
h/2
dz = Gh
x
q
y
= Gh
y
(7.11)
with the slab stiffnesses K
n
=
E h
1
2
, K =
E h
3
12(1
2
)
, see also [Gir74, Section 189.], and
the shear reduction factor , see Page 44.
Balance of equations
2
Variables: 3 displacement variables u, v, w, 8 deformation variables
x
,
y
,
xy
,
x
,
y
,
xy
,
x
,
y
,
8 force variables n
x
, n
y
, n
xy
, m
x
, m
y
, m
xy
, q
x
, q
y
, which makes together 19 vari-
ables.
Equations: 8 force-displacement relations Eq. (7.11), 6 kinematic relations Eqs. (7.3),
(7.4), which makes together 14 equations. This is complemented with 5 equilibrium
conditions, see Eqs. (7.12)-(7.14).
2
The slopes w/x, w/y need not to be adressed in this particular balance. In case the referred variables
are directly determined from the referred equations the slopes result from the derivatives of the deection surface
w(x, y).
7.1.3 Reinforced cracked sections 157
7.1.3 Reinforced cracked sections
The following considerations regard a slab point x, y, i.g. an integration point.
Layer model
Subdivision of slab thickness into layers.
Strain of each layer is determined by the kinematic assumptions Eq. (7.2)
Stress of each layer is computed from strains with an appropriate material law. This
has some aspects.
*
Separation of in-plane reaction stresses from
x
,
y
,
xy
from vertical shear
reaction stresses from
xz
,
yz
.
*
In-plane reaction is biaxial. Thus, a biaxial material law can be used.
*
Basically, each layer may have its own material law which allows to distinguish
between concrete layers and reinforcement layers
3
.
*
Regarding concrete, a rst approach to model cracks in a biaxial setting has
been described in Section 6.2.
This includes limited tensile strength, the Rankine crack criterion, smeared
cracks, crack orientation, crack width, crack tractions and softening.
Generally, cracking states may vary along the slab thickness. This concerns
the cracking event itself, but furthermore actual values for crack orientation
and crack width in case a crack occurs.
*
Regarding reinforcement, smeared layers are appropriate as have been described
in Section 6.3, Page 145.
Numerical integration of stresses along thickness to determine resulting internal
forces according to Eq. (7.7)
The layer model is computationally expensive.
More issues
*
Treatment of shear forces
*
Tangential stiffness
Decoupled model
This is based on principal deformations for reference plan strains
x
,
y
,
xy
and
curvature
x
,
y
,
xy
. It is assumed that the corresponding principal directions are
at least approximately the same. Thus, a transformation to a common principle
system is possible.
Each principal direction is treated like a beam cross section with unit width, see
Section 3.1.3.
*
Lateral expansion effects seem to be disregarded.
Integration of stresses over slab thickness in each principal direction leads to prin-
cipal moments m
1
, m
2
, which are transformed back to the global system to yield
moments m
x
, m
y
m
xy
.
The same procedure may be basically applied to vertical shear deformations and
shear forces.
3
This implies rigid bond due to the kinematic approach.
158 7.2 Equilibrium of slabs
7.2 Equilibrium of slabs
u, v are used as symbols for reference plane displacements u, v in the following.
Figure 7.1: Slab equilibrium
Strong differential equilibrium
We regard a place x, y of the reference plane with a distributed loading p
x
, p
y
, p
z
.
Equilibrium of normal forces
n
x
x
+
n
xy
y
+p
x
= 0,
n
y
y
+
n
xy
x
+p
y
= 0 (7.12)
Equilibrium of shear forces
q
x
x
+
q
y
y
+p
z
= 0 (7.13)
Equilibrium of moments
m
x
x
+
m
xy
y
+q
x
= 0,
m
y
y
+
m
xy
x
+q
y
= 0 (7.14)
The linear elastic Kirchhoff case
4
Shear deformations are neglected. Then Eq. (7.4) leads to
x
=

2
w
x
2
,
y
=

2
w
y
2
,
xy
= 2

2
w
xy
(7.15)
From Eqns. (7.14), (7.13)
2
m
x
x
2
+ 2
2
m
xy
xy
+

2
m
y
y
2
= p
z
(7.16)
From Eq. (7.11)
46
2
m
x
x
2
= K
_
x
x
2
+

2
y
x
2
_
= K
_
4
w
x
4
+

4
w
x
2
y
2
_
2
m
y
y
2
= K
_
y
y
2
+

2
x
y
2
_
= K
_
4
w
y
4
+

4
w
y
2
x
2
_
2
m
xy
xy
= (1 )
K
2
xy
xy
= (1 ) K

4
w
x
2
y
2
(7.17)
4
Other derivations use an opposite sign convention for z, w compared to Fig. 7.1. This reverses the sign in the
moment-curvature relations.
159
4
w
x
4
+

4
w
x
2
y
2
+ 2(1 )

4
w
x
2
y
2
+

4
w
y
4
+

4
w
y
2
x
2
=
p
z
K
(7.18)
or
4
w
x
4
+ 2

4
w
x
2
y
2
+

4
w
y
4
=
p
z
K
(7.19)
Boundary conditions are given with appropriate combinations of, e.g. w, w/x,
2
w/x
2
,
3
w/x
3
. The physical meaning of the latter two is given by e.g. m
x
=
K(
2
w/x
2
+
2
w/y
2
) and m
x
/x = q
x
m
xy
/y, respectively.
Weak integral equilibrium
Extending Eq. (7.4) the following kinematic variables are dened
x
=
w
x

x

x
=
w
x

x
y
=
w
y

y

y
=
w
y

y
x
=

2
w
x
2

x
x
=

x
x
y
=

2
w
y
2

y
y
=

y
y
xy
= =

x
y
+

y
x
(7.20)
introducing two further displacement variables
x
,
y
and two further kinematic
relations.
An equivalent to strong differential equilibrium formulation Eqs. (7.12)(7.14) is
given by a weak integral equilibrium formulation for a slab of Area A with poten-
tially coupling normal forces and and moments
_
A
u
_
n
x
x
+
n
xy
y
+p
x
_
da +
_
A
v
_
n
y
y
+
n
xy
x
+p
y
_
da +
_
A
w
_
q
x
x
+
q
y
y
+p
z
_
da
+
_
A
x
_
m
x
x
+
m
xy
y
+q
x
_
da +
_
A
y
_
m
y
y
+
m
xy
x
+q
y
_
da = 0
(7.21)
Those terms with derivatives of internal forces are partially integrated:
_
A
u
_
n
x
x
+
n
xy
y
_
da =
_
A
_
u
x
n
x
+
u
y
n
xy
_
da +
_
A
u(n
x
e
x
+n
xy
e
y
)ds
_
A
v
_
n
y
y
+
n
xy
x
_
da =
_
A
_
v
y
n
y
+
v
x
n
xy
_
da +
_
A
v(n
xy
e
x
+n
y
e
y
)ds
_
A
w
_
q
x
x
+
q
y
y
_
da =
_
A
_
w
x
q
x
+
w
y
q
y
_
da +
_
A
w(q
x
e
x
+q
y
e
y
)ds
_
A
x
_
m
x
x
+
m
xy
y
+q
x
_
da =
_
A
_
x
x
m
x
+

x
y
m
xy

x
q
x
_
da
+
_
A
x
(m
x
e
x
+m
xy
e
y
)ds
_
A
y
_
m
y
y
+
m
xy
x
+q
y
_
da =
_
A
_
y
y
m
y
+

y
x
m
xy

y
q
y
_
da
+
_
A
y
(m
xy
e
x
+m
y
e
y
)ds
(7.22)
with a unit normal vector e with components e
x
, e
y
along the boundary.
160 7.2 Equilibrium of slabs
Combining Eqs. (7.22), (7.21) yields
_
A
(
x
n
x
+
y
n
y
+
xy
n
xy
+
x
m
x
+
y
m
y
+
xy
m
xy
+
x
q
x
+
y
q
y
) da
=
_
A
(up
x
+v p
y
+wp
z
) da
+
_
A
_
u(n
x
e
x
+n
xy
e
y
) +v(n
xy
e
x
+n
y
e
y
) +w(q
x
e
x
+q
y
e
y
)
+
x
(m
x
e
x
+m
xy
e
y
) +
y
(m
xy
e
x
+m
y
e
y
)
_
ds
(7.23)
with virtual deformations
x
= u/x,
y
= v/y,
xy
= u/y+v/x
and
x
=
x
/x,
y
=
y
/y,
xy
=
x
/y +
y
/x and
x
=
w/x
x
,
y
= w/y
y
.
A generalizing matrix notation of Eq. (7.23) is given with
_
A
T
da =
_
A
u
T
pda +
_
A
U
T
t ds (7.24)
with
=
_

x

y

xy

x

y

xy

x

y
_
T
=
_
n
x
n
y
n
xy
m
x
m
y
m
xy
q
x
q
y
_
T
u =
_
u v w
x

y
_
T
p =
_
p
x
p
y
p
z
0 0
_
T
U =
_
u
s
v
s
w
s

sx

sy
_
T
t =
_
n
sx
n
sy
q
s
m
sx
m
sy
_
T
(7.25)
with the displacement boundary values u
s
, v
s
, w
s
,
sx
,
sy
and
n
sx
= n
x
e
x
+n
xy
e
y
n
sy
= n
xy
e
x
+n
y
e
y
q
s
= q
x
e
x
+q
y
e
y
m
sx
= m
x
e
x
+m
xy
e
y
m
sy
= m
xy
e
x
+m
y
e
y
(7.26)
A coupling of normal forces and moments as may arise with reinforced cracked
concrete is realized with a dependence of normal forces on strains and additionally
curvatures and also of moments on curvature and strains.
Decoupling of normal forces and moments
In case normal forces do not depend on curvature and moments do not depend on
strains e.g. for a linear elastic material behavior weak equilibrium can be formu-
lated independently for Eq. (7.12) and for Eqs. (7.13), (7.14).
In case of normal forces this leads to
=
_

x

y

xy
_
T
=
_
n
x
n
y
n
xy
_
T
u =
_
u v
_
T
p =
_
p
x
p
y
_
T
U =
_
u
s
v
s
_
T
t =
_
n
sx
n
sy
_
T
(7.27)
while Eq. (7.24) is still valid. Basically this corresponds to the biaxial plane strain
or plane stress case.
161
In case of bending moments the relations
=
_

x

y

xy

x

y
_
T
=
_
m
x
m
y
m
xy
q
x
q
y
_
T
u =
_
w
x

y
_
p =
_
p
z
0 0
_
U =
_
w
s

sx

sy
_
T
t =
_
q
s
m
sx
m
sy
_
T
(7.28)
hold with Eq. (7.24) unchanged. These relations correspond to the Reissner-Mindlin
slab.
w is decoupled from
x
,
y
by the shear deformation
x
,
y
. Thus, w,
x
,
y
may
be interpolated as independent displacement variables within the Finite Element ap-
proach. As a further consequence, C
0
-continuity is sufcient for the interpolation
of w,
x
,
y
, see also Section 1.6.
Kirchhoff-slab
Bending moments are considered only.
w
x
=
x

x
= 0,
w
y
=
y

y
= 0.
Eq. (7.28) is modied with
=
_

x

y

xy
_
T
=
_
m
x
m
y
m
xy
_
T
u =
_
w
x

y
_
p =
_
p
z
0 0
_
U =
_
w
s

sx

sy
_
T
t =
_
q
s
m
sx
m
sy
_
T
(7.29)
with the same Eq. (7.24).
According to Eq. (7.4) the relations between strains and deformations are given by
x
=

2
w
x
2
,
y
=

2
w
y
2
,
xy
= 2

2
w
xy
(7.30)
As a consequence, C
1
-continuity is required for test and trial functions for w in order
to ensure compatibility and integrability of Eq. (7.24), see also Section 1.6. This
requirement is not trivial to fulll within the two dimensional setting. A number of
ways to relax it have been investigated.
With a linear elastic material behavior according to Eqs. (7.11)4 6 the relation
between generalized strains and stresses can be formulated as
= C , C = K
_
_
1 0
1 0
0 0
1
2
_
_
, K =
E h
3
12(1
2
)
(7.31)
with Youngs modulus E, Poissons ratio and the slab height h.
162 7.3 Structural Slab Elements
7.3 Structural Slab Elements
7.3.1 Area coordinates
Basic approach
x = L
1
x
1
+L
2
x
2
+L
3
x
3
y = L
1
y
1
+L
2
y
2
+L
3
y
3
L
1
+L
2
+L
3
= 1
(7.32)
This leads to
L
1
= a
1
+b
1
x +c
1
y, L
2
= a
2
+b
2
x +c
2
y, L
3
= a
3
+b
3
x +c
3
y (7.33)
with
a
1
=
y
3
x
2
y
2
x
3
2
, b
1
=
y
2
y
3
2
, c
1
=
x
3
x
2
2
a
2
=
y
1
x
3
y
3
x
1
2
, b
2
=
y
3
y
1
2
, c
2
=
x
1
x
3
2
a
3
=
y
2
x
1
y
1
x
2
2
, b
3
=
y
1
y
2
2
, c
3
=
x
2
x
1
2
(7.34)
and
=
1
2
det
1 x
1
y
1
1 x
2
y
3
1 x
2
y
3
= area123 =
1
2
(y
1
x
3
+y
3
x
2
+y
2
x
1
y
2
x
3
y
3
x
1
y
1
x
2
)
(7.35)
Nodal values of the area coordinates
x = x
1
, y = y
1
L
1
= 1, L
2
= 0, L
3
= 0
x = x
2
, y = y
2
L
1
= 0, L
2
= 1, L
3
= 0
x = x
3
, y = y
3
L
1
= 0, L
2
= 0, L
3
= 1
(7.36)
Derivatives of area coordinates with b
i
, c
i
L
i
x
= b
i
,
L
i
y
= c
i
(7.37)
7.3.2 A triangular Kirchhoff element
Triangular element, three nodes, 9 DOF, see [ZT91, 1.5] with a basic approach
w =
1
L
1
+
2
L
2
+
3
L
3
+
4
L
1
L
2
+
5
L
2
L
3
+
6
L
3
L
1
+
7
L
2
1
L
2
+
8
L
2
2
L
3
+
9
L
2
3
L
1
w
x
=
w
L
1
L
1
x
+
w
L
2
L
2
x
+
w
L
3
L
3
x
=
_
1
+
4
L
2
+
6
L
3
+ 2
7
L
1
L
2
+
9
L
2
3
_
b
1
+
_
2
+
4
L
1
+
5
L
3
+
7
L
2
1
+ 2
8
L
2
L
3
_
b
2
+
_
3
+
5
L
2
+
6
L
1
+
8
L
2
2
+ 2
9
L
1
L
3
_
b
3
w
y
=
w
L
1
L
1
y
+
w
L
2
L
2
y
+
w
L
3
L
3
y
=
_
1
+
4
L
2
+
6
L
3
+ 2
7
L
1
L
2
+
9
L
2
3
_
c
1
+
_
2
+
4
L
1
+
5
L
3
+
7
L
2
1
+ 2
8
L
2
L
3
_
c
2
+
_
3
+
5
L
2
+
6
L
1
+
8
L
2
2
+ 2
9
L
1
L
3
_
c
3
(7.38)
This yields for node 1 with L
1
= 1, L
2
= L
3
= 0
w
1
=
1
1x
=
1
b
1
+ (
2
+
4
+
7
) b
2
+ (
3
+
6
) b
3
1y
=
1
c
1
+ (
2
+
4
+
7
) c
2
+ (
3
+
6
) c
3
(7.39)
7.3.2 A triangular Kirchhoff element 163
and for node 2 with L
2
= 1, L
2
= L
3
= 0
w
2
=
2
2x
= (
1
+
4
) b
1
+
2
b
2
+ (
3
+
5
+
8
) b
3
2y
= (
1
+
4
) c
1
+
2
c
2
+ (
3
+
5
+
8
) c
3
(7.40)
and for node 3 with L
3
= 1, L
1
= L
2
= 0
w
3
=
3
3x
= (
1
+
6
+
9
) b
1
+ (
2
+
5
) b
2
+
3
b
3
3y
= (
1
+
6
+
9
) c
1
+ (
2
+
5
) c
2
+
3
c
3
(7.41)
The solution is
1
= w
1
2
= w
2
3
= w
3
4
= c
3
2x
+w
2
w
1
+b
3
2y
5
=
3x
c
1
w
2
+w
3
+b
1
3y
6
= c
2
1x
+w
1
w
3
+b
2
1y
7
= b
3
2y
+c
3
2x
b
3
1y
+c
3
1x
2w
2
+ 2w
1
8
= b
1
2y
+c
1
2x
b
1
3y
+
3x
c
1
+ 2w
2
2w
3
9
= b
2
1y
+c
2
1x
b
2
3y
+c
2
3x
2w
1
+ 2w
3
(7.42)
leading to
w = L
1
(1 +L
2
2L
1
L
2
L
3
+ 2L
2
3
)w
1
+L
1
(c
3
L
1
L
2
+L
2
3
c
2
L
3
c
2
)
1x
+
L
1
(b
2
L
2
3
+b
3
L
1
L
2
b
2
L
3
)
1y
+L
2
(2L
2
1
L
3
+L
1
+ 1 + 2L
2
L
3
)w
2
+
L
2
(c
3
L
1
+c
3
L
2
1
+c
1
L
2
L
3
)
2x
+L
2
(b
3
L
1
+b
3
L
2
1
+b
1
L
2
L
3
)
2y
+
L
3
(L
2
+ 2L
3
L
1
2L
2
2
+ 1 L
1
)w
3
+L
3
(c
1
L
2
2
+c
2
L
3
L
1
c
1
L
2
)
3x
+
L
3
(b
1
L
2
2
+b
2
L
3
L
1
b
1
L
2
)
3y
(7.43)
Higher derivatives, see also (7.4) with
x
=
y
= 0
2
w
x
2
=

2
w
xL
1
L
1
x
+

2
w
xL
2
L
2
x
+
w
2
xL
3
L
3
x
=
x
2
w
y
2
=

2
w
yL
1
L
1
y
+

2
w
yL
2
L
2
y
+
w
2
yL
3
L
3
y
=
y
2
w
xy
=

2
w
xL
1
L
1
y
+

2
w
xL
2
L
2
y
+
w
2
xL
3
L
3
y
=

xy
2
2
w
yx
=

2
w
xy
(7.44)
or
x
= B
11
w
1
+B
12
1x
+B
13
1y
+B
14
w
2
+B
15
2x
+B
16
2y
+B
17
w
3
+B
18
3x
+B
19
3y
y
= B
21
w
1
+B
22
1x
+B
23
1y
+B
24
w
2
+B
25
2x
+B
26
2y
+B
27
w
3
+B
28
3x
+B
29
3y
xy
2
= B
31
w
1
+B
32
1x
+B
33
1y
+B
34
w
2
+B
35
2x
+B
36
2y
+B
37
w
3
+B
38
3x
+B
39
3y
(7.45)
with
B
11
= 8b
3
b
1
L
3
+ 4b
2
1
L
2
+ 2(4b
2
b
1
2b
2
3
)L
1
+ 2(b
2
b
1
+b
3
b
1
)
B
12
= 4b
1
c
2
b
3
L
3
+ 2c
3
b
2
1
L
2
+ 2(2c
3
b
1
b
2
+c
2
b
2
3
)L
1
2b
1
c
2
b
3
. . .
(7.46)
These derivations demonstrate the basic procedure. But this particular approach has draw-
backs, e.g. the patch test is not fullled for arbitrary element shapes. Improved forms are
proposed by [Spe88] using forth order terms instead of cubic terms in Eq. (7.38)
1
.
Details are omitted here and have to be taken from [Spe88], [ZT91, 1.5].
7.4 System Building and Solution Methods
System building and solution methods basically follow the outlines as have already been
described with their general aspects in Section 1.7 and with applications for beams in
Section 3.4.
Strains are approximated with Finite Element approximations
= B u
I
(7.47)
see Eq. (1.6), with u
I
3
, (7.28)
3
or (7.29)
3
applied to all nodes
of an element and according to Eq. (7.25)
1
, (7.28)
1
or (7.29)
1
for the coupled
problem, the Reissner-Mindlin slab or the Kirchhoff slab, respectively.
*
In case of, e.g., a Kirchhoff slab the matrix B is determined by Eq. (7.45).
The weak equilibrium condition Eq. (7.24) has to be integrated element by element.
As has been demonstrated before, see Section 1.5, integration results in
r(u) = f (u) p = 0 (7.48)
with nodal internal forces f , nodal displacements u of all elements and external
nodal loads p, see also Eq. (1.63).
In case of linear material behavior internal nodal forces may be written as
f (u) = K u K u = p (7.49)
with a constant stiffness matrix Kassembled from element stiffness matrices
K
I
=
_
V
I
B
T
C BdV (7.50)
see also Eqs. (1.36), (1.37), with Caccording to, e.g., Eq. (7.31) in case of a Kirch-
hoff slab. This yields an immediate solution for the displacement with a given load
p
u = K
1
p (7.51)
Numerical integration schemes for triangular 3-node elements
Triangular elements often use area coordinates to mark a spatial point, see Sec-
tion 7.3.1.
The approach to integrate for internal nodal forces, external loads and stiffness ma-
trices evaluates an integrand at a number of sampling points within an element,
multiplies these sampled values with weighting factors and sums up the weighted
values.
*
This resembles the procedure of Page 14.
Sampling points and weighting factors are given in Table 7.1 up to integration order
2.
Boundary conditions
First of all, kinematic boundary conditions have to provide a stable support and to
prevent rigid body motions.
Boundary forces are basically given by Eq. (7.26) with a unit normal vector e with
components e
x
, e
y
along the boundary.
165
n
i
L
1
L
2
L
3

i
0 1/3 1/3 1/3 1
1 1/ 2 1/2 0 1/3
0 1/2 1/2 1/3
1/ 2 0 1/2 1/3
2 1/3 1/3 1/3 - 27/48
0.6 0.2 0.2 25/48
0.2 0.6 0.2 25/48
0.2 0.2 0.6 25/48
Table 7.1: Sampling points and weights for triangular numerical integration
In order to simplify the formulations a boundary edge parallel to the global y-axis
with x = const., e
x
= 1, e
y
= 0 is considered examplarily, i.e.
q
s
= q
x
, m
sx
= m
x
, m
sy
= m
xy
(7.52)
along the boundary edge.
*
Free edge
Values q
s
, m
sx
, m
sy
are prescribed in most cases with 0 and go directly
into the boundary force vector t, see Eqs. (7.28)
6
, (7.29)
6
.
*
Simply supported edge
Regarding Eqs. (7.28)
5
, (7.29)
5
kinematic boundary conditions w = w
s
=
0 and
y
=
sy
= 0 are given. As
x
is not prescribed the corresponding
force m
sx
= m
x
has to be prescribed, generally with a value 0. On the
other hand, as
y
is prescribed the corresponding boundary force m
sy
=
m
xy
has to result from the computation and is transmitted to the slabs
support
5
.
*
Clamped edge
Regarding Eqs. (7.28)
5
, (7.29)
5
kinematic boundary conditions w = w
s
=
0 and
x
=
sx
= 0,
y
=
sy
= 0 are given. All corresponding boundary
forces are determined from the computation, whereby m
sy
= m
xy
= 0 as
x
/y = 0.
A boundary edge parallel to the global x-axis with y = const. is treated in the same
way with indices exchanged. A straight skew or curved boundary edge with, e.g.,
simply support leads to a prescribed coupling of
x
,
y
and m
x
, m
y
, m
xy
along the
boundary. These may be regarded as additional constraint conditions.
Kinematic boundary conditions are applied to nodes and directly implemented upon
assembling the system, e.g. by the modication of the systems stiffness matrix and
load vector, see Page 63 for the basic approach.
Boundary force reactions are automatically computed as internal nodal forces for
those boundary degrees of freedom which have kinematic boundary conditions pre-
scribed. These particular internal nodal forces are not equilibrated by external nodal
loads.
5
It is connected to a compensatory shear force which is combined with the ordinary shear force q
x
.
Example 7.1 A linear elastic rectangular plate with large opening and free edges
Geometry and boundary conditions are shown in Fig. 7.2a. A single support is given in
the lower left corner. The left and the lower edge are not supported. The upper and right
edge are sinply supported (hinged). The slab has an opening with dimensions given in
Fig. 7.2a.
The material properties are given with E = 31 900 MN/m
2
, = 0.2. The slab thickness
is h = 0.3 m A uniform loading is assumed with p = 10 kN/m
2
. It does not act in the
opening area.
Kirchhoff theory is assumed with a neglection of shear deformations. The element type
is chosen according to Section 7.3.2. The discretization is performed with 132 elements,
which are indicated in Fig. 7.2b.
Figure 7.2: Example 7.1 a) system b) discretization and principal moments
Results
Principal moments are shown in Figure 7.2b.
*
Principal moments are aligned to free edges
*
Uniaxial behavior along free edges
*
Twisting moments in the upper right corner
*
Reversed twisting near lower left single support
The computed deections are shown in Fig. 7.3. They obviously conform to pre-
scribed boundary conditions. The maximum deection is in the range of 3 cm, i.e.
w
max
/L 1/230. Thus, thickness and stiffness are too low for serviceability.
Reaction forces
*
The computed reaction forces in the boundary nodes are given in Fig. 7.3b.
Number belongs to the lower left single support, the numbers 2-6 to the right
edge, the number 7 to the right upper corner and the numbers 8-14 to the upper
edge.
*
The sum of all reaction forces equals 330 kN and thus equals the total loading.
*
The by far largest reaction force is given the simple support.
*
The upper right reaction force corresponds to an uplift. This conforms to the
theory of slabs with drilling resistance.
167
Figure 7.3: Example 7.1 a) deections b) boundary suppoer reactions
end example 7.1
168 7.5 Reinforcement Design with linear elastic internal forces
7.5 Reinforcement Design with linear elastic internal forces
Moment transformation and principal moments
Moments m
x
, m
y
, m
xy
as they are dened by Eq. (7.7) similar to stresses behave
like second order tensors.
Transformation of moments into a coordinate system rotated by an angle (positive
counterclockwise, see Eq. (B.3))
m
= T m (7.53)
with
m
=
_
_
m
x
m
y
m
xy
_
_
, T =
_
_
cos
2
sin
2
2 cos sin
sin
2
cos
2
2 cos sin
cos sin cos sin cos
2
sin
2
_
_
, m=
_
_
m
x
m
y
m
xy
_
_
(7.54)
The principal direction is dened with the condition
m
xy
= 0
_
1 cos
2
2
m
y
m
x
2
+ cos 2 m
xy
= 0 (7.55)
Within a range 0 /2 a unique solution is determined by
cos 2 =
m
x
m
y
2
_
(
m
x
m
y
2
)
2
+m
2
xy
(7.56)
for m
xy
,= 0. Principal moments are already given by m
x
, m
y
for m
xy
= 0.
A solution multiplied by sign m
xy
the sign of m
xy
indicates the direction of a
principal moment m
1
. A second principal moment m
2
is perpendicular. They have
the values
m
1
=
m
x
+m
y
2
+
_
m
x
m
y
2
_
2
+m
2
xy
, m
2
=
m
x
+m
y
2

_
m
x
m
y
2
_
2
+m
2
xy
(7.57)
Corresponding forces
According to the denition of Eq. (7.7) and Fig. 7.1 a moment m
x
leads to stresses
x
, m
y
to
y
and m
xy
to
xy
. Assigning an internal lever armz yields corresponding
couple force resultants
t
x
=
m
x
z
, t
y
=
m
y
z
, t
xy
=
m
xy
z
(7.58)
Each of the two components may be attached to a lower and upper layer of the slab.
All considerations regarding reinforcement design of deep beams, see Section 6.1 Pages 134-
137, may be applied to the layer forces or upper or lower reinforcement, respectively. This
is summarized in the following, which has to be applied to the upper and lower layer in
the same way.
169
Reinforcement design basics
According to Eq. (6.4) forces are connected to principal forces by
6
t
x
= t
1
cos
2
+t
2
sin
2
t
y
= t
1
sin
2
+t
2
cos
2
t
xy
= (t
1
t
2
) sin cos
(7.59)
This form is alternative to the form of Eqs. (7.56,7.57) with the angle measured
from the x-axis to the 1-axis (counterclockwise positive, see Eq. (B.4)).
Two reinforcement directions
s1
,
s2
and a concrete direction
c
are considered.
These directions describe the principal 1-directions for each part. The corresponding
principal stress values are transformed to the global directions applying Eq. (7.59)
to each part, whereby the stresses in the principal 2-directions vanish due to the
uniaxial behavior of each part. This leads to
t
c,x
= t
c,1
cos
2
c
, t
c,y
= t
c,1
sin
2
c
, t
c,xy
= t
c,1
sin
c
cos
c
t
s1,x
= t
s1,1
cos
2
s1
, t
s1,y
= t
s1,1
sin
2
s1
, t
s1,xy
= t
s1,1
sin
s1
cos
s1
t
s2,x
= t
s2,1
cos
2
s2
, t
s2,y
= t
s2,1
sin
2
s2
, t
s2,xy
= t
s2,1
sin
s2
cos
s2
(7.60)
see also Eqs. (6.5,6.6). A notation t
c
= t
c,1
, t
s1
= t
s1,1
, t
s2
= t
s2,1
will be used
in the following. Reinforcement forces are connected to reinforcement stresses via
reinforcement cross sections a
s1
, a
s2
, i.e.
t
s1
= a
s1
s1
, t
s2
= a
s2
s2
(7.61)
The parts contribute to total forces according to Eq.(6.7)
t
c,x
+t
s1,x
+t
s2,x
= t
x
t
c,y
+t
s1,y
+t
s2,y
= t
y
t
c,xy
+t
s1,xy
+t
s2,xy
= t
xy
(7.62)
Usage of of Eqs. (7.60), (7.61) in Eq. (7.62) rst of all yields three equations for the
8 design parameters t
c
,
c
,
s1
, a
s1
,
s1
,
s2
, a
s2
,
s2
.
Three further Eqs. (7.58) connect t
x
, t
y
, t
xy
to moments m
x
, m
y
, m
xy
by the in-
ternal lever arm z. Thereby m
x
, m
y
, t
xy
are given, e.g., from a linear elastic FE-
calculation which has been performed in advance.
It is reasonable to prescribe the four values
s1
,
s1
,
s2
,
s2
. Thus, the parameters
t
c
,
c
, a
s1
, a
s2
and z remain open for design.
6
May also be derived according to Eq. (7.54) with a sign reversal of and a zero mixed component.
A common special case and its design strategy basics
It is assumed that reinforcement directions are aligned to global coordinate axes, i.e.
s1
= 0,
s2
= /2 and thus sin
s1
= 0, cos
s1
= 1, sin
s2
= 1, cos
s2
= 0.
This leads to
t
s1,x
= t
sx
= a
sx
s1
, t
s1,y
= 0, t
s1,xy
= 0
t
s2,x
= 0, t
s2,y
= t
sy
= a
sy

s2
, t
s2,xy
= 0
(7.63)
Insertion into Eq. (7.62) together with Eq. (7.60)
1
yields
t
c
cos
2
c
+a
sx
s1
= t
x
t
c
sin
2
c
+a
sy

s2
= t
y
t
c
sin
c
cos
c
= t
xy
(7.64)
To exploit reinforcement load carrying capacity stresses
s1
=
s2
= f
y
are used in
the following with the reinforcement yield limit f
y
.
With Eq. (7.64)
3
the concrete force negative by denition is given by
t
c
=
t
xy
sin
c
cos
c
(7.65)
The bearing capacity of the compression zone can be estimated with appropriate
assumptions about the values and the distribution of concrete stresses.
*
A constant concrete stress distribution is assumed with a value
c
= f
c
, a
zero line of bending x measured from the compressed side and the compression
stress zone height k x. Values = 0.95, k = 0.8 can be used, see DIN 1045-1
[din08, 9.1.6]. This leads to t
c
= f
c
k x. Furthermore, z = d k x/2 with
the structural height d. Assuming t
c
= m
c
/z yields
x =
d
k
_
1
1 +
2km
c
f
c
d
2
_
, z =
d
2
_
1 +
1 +
2km
c
f
c
d
2
_
(7.66)
with

f
c
= f
c
k and m
c
< 0 by denition.
A reinforcement can be determined with Eq. (7.64)
a
sx
=
t
x
t
c
cos
2
c
f
y
=
m
x
m
c
cos
2
c
z f
y
a
sy
=
t
y
t
c
sin
2
c
f
y
=
m
y
m
c
sin
2
c
z f
y
(7.67)
whereby the +sign is used for the lower slab side and the sign for the upper side,
see Fig. 7.1, and derived from Eqs. (7.65), (7.58), t
c
= m
c
/z with
m
c
sin
c
cos
c
= m
xy
(7.68)
Summarizing, four equations
7
(7.68), (7.67), (7.66)
2
hold for the ve design param-
eters a
sx
, a
sy
, m
c
, z,
c
.
Regarding
c
, a rst choice
c
=
_

4
for m
xy
0
4
m
xy
< 0
(7.69)
is appropriate as has been shown on Page 137. This leads to t
c
= 2[t
xy
[, m
c
=
2[m
xy
[ and a
sx
= m
x
+[m
xy
[/(z f
y
), a
sy
= m
y
+[m
xy
[/(z f
y
). The in-
ternal lever arm z can directly be determined from Eq. (7.66)
2
and therefore also
a
sx
, a
sy
.
7
Setting z = 1, m = and omitting Eq. (7.66)
2
recovers the deep beam case, see Page 135.
171
Figure 7.4: Slab reinforcement
A value a
si
< 0 indicates reorientation of the concrete stress direction, i.e. the angle
c
has to be changed to gain a
si
= 0.
*
This case is covered by the nonlinear system of four equations (7.66)
2
, (7.67)
with a
si
= 0 and (7.68) for four unknowns
c
, m
c
, z and a
sj
with i ,= j. This
system has to be solved iteratively in a similar way as has been demonstrated
for deep beams, see Page 137.
The zero line x of bending measured should not exceed d/2. Acondition x d/2
leads to (k = 0.8).
[m
c
[
1
_
1
k
2
_
2
2
f
c
d
2
= 0.32

f
c
d
2
(7.70)
Example 7.2 Reinforcement design for slab with linear elastic internal forces
We refer to Example 7.1 with the same system and loading. Moments have been calcu-
lated for each element integration point with a linear elastic calculation.
Uniaxial concrete compressive strength (unsigned) is assumed with f
c
= 17.0 MN/m
2
(concrete grade C30/37,

f
cd
= 12.92), reinforcement strength with f
y
= 435 MN/m
2
.
The structural height of the slab is d = 0.25 m. Safety factors are not regarded for the
loading.
Figure 7.5: Example 7.2 a) reinforcement of section b) selected points
Selected points with different characteristics of bending are considered.
Point x = 0.17, y = 2.5 near midspan left free edge.
Computed moments are m
x
= 0.001 MNm/m, my = 0.051, m
xy
= 0.020 leading
to principal moments m
1
= 0.058, m
2
= 0.006 with an orientation of
1
= 71.
Lower reinforcement: Assuming = 45 yields a
sx
= 2.02 cm
2
/m, a
sy
= 6.87
with z/d = 0.95 and x/d = 0.10.
Upper reinforcement:
c
= 45 leads to a
sy
< 0. Thus, an iteration has to be
performed leading to a
sx
= 0.67 cm
2
/m, a
sy
= 0 with z/d = 0.92, x/d = 0.16
and
c
= 69 computed.
Point x = 6.5, y = 4.17 near upper right corner of simple support.
x
= 0.004 MNm/m, my = 0.004, m
xy
= 0.028 lead-
ing to m
1
= 0.032, m
2
= 0.024 with
1
= 45.
Curvature is negative in the corners diagonal direction and positive lateral to it.
This indicates a load transfer over edge supporting a load transfer in the corners
diagonal and corresponds to the well known drilling effect.
Lower reinforcement: = 45 yields a
sx
= 3.21 cm
2
/m, a
sy
= 3.14 with z/d =
0.0.93, x/d = 0.14.
Upper reinforcement: = 45 yields a
sx
= 2.32 cm
2
/m, a
sy
= 2.39 with
z/d = 0.93, x/d = 0.14 (must here be the same as the upper values).
Point x = 3.5, y = 0.17 near midspan lower free edge.
x
= 0.060 MNm/m, my = 0.007, m
xy
= 0.008 leading
to m
1
= 0.061, m
2
= 0.006 with
1
= 8.
Lower reinforcement: = 45 yields a
sx
= 6.33 cm
2
/m, a
sy
= 1.39 with z/d =
0.98, x/d = 0.04.
An upper reinforcement is not necessary as both principal stresses are positive and
lead to a upper biaxial compressive stress state.
Point x = 0.5, y = 0.17 near lower left single support.
x
= 0.02 MNm/m, my = 0.011, m
xy
= 0.046 leading
to m
1
= 0.061, m
2
= 0.030 with
1
= 42.
Curvature is positive in the corners diagonal direction and negative lateral to it.
This corresponds to the major load transfer along the free edges compared to the
diagonal.
Lower reinforcement:
c
= 45 yields a
sx
= 6.94 cm
2
/m, a
sy
= 5.89 with
z/d = 0.88, x/d = 0.24.
Upper reinforcement:
c
= 45 yields a
sx
= 2.63 cm
2
/m, a
sy
= 3.68 with z/d =
0.88, x/d = 0.24 (must here be the same as the upper values).
As compressive heights are computed with x/d < 0.5 for all points load bearing capacity
of concrete is provided.
end example 7.2
Basically the same comments about high beams regarding concrete strength, ductility
requirements and serviceability, see Page 139, are also valid for slabs.
Shear forces
173
Shear forces have to be calculated as derivatives of bending moments in case of
Kirchhoff slabs
q
x
=
m
x
x

m
xy
y
, q
y
=
m
y
y

m
xy
x
(7.71)
see Eq. (7.14).
*
Bending moments are determined from curvature with Eq. (7.31) in case of lin-
ear elastic material behavior. Furthermore, curvature is determined from Finite
Element interpolation with Eq. (7.47). To do this analytically is very elaborate,
e.g. to compute third derivatives of displacements in order to determine rst
derivatives of moments. Furthermore, the accuracy decreases with computa-
tion of higher displacement derivatives.
*
A separate interpolation or approximation of moments within elements should
be less expansive and yield reliable shear force values. This may base on a
linear approach
m = a x +b y +c,
m
x
= a,
m
y
= b (7.72)
where m stands for m
x
, m
y
, m
xy
. The coefcients a, b, c need at least three
computed values, i.e. three integration points, within an element and may be
determined using a linear regression analysis, see Appendix C.0.1.
Transformation rule for shear forces
*
Across section with normal vector e = ( cos sin )
T
is regarded, whereby
denotes the angle with the global x-axis. The shear force in this cross section
is calculated from q
x
, q
y
, see Fig. 7.1, by
8
q
= q
x
cos +q
y
sin (7.73)
Shear forces generally become relevant near supported edges. A shear design cal-
culation can be performed as follows:
*
Shear basically arises from change of longitudinal forces resulting from change
of moments. This change of forces in adjacent cross sections leads to forces in
horizontal sections and adjoined shear forces in cross sections. In case of rein-
forced concrete these forces are realized by concrete struts and reinforcement
ties. The mechanisms are basically the same for beams and slabs.
*
Principal longitudinal forces t
1
, t
2
computed from t
x
, t
y
, t
xy
are used in case
of slabs. The force t
1
acts at a cross section orientated with a normal direction
angle , the force t
2
at a cross section perpendicular to it.
*
Each of these forces is completed by the opposite longitudinal force both
forming a couple for moments m
1
, m
2
and a shear force computed with
Eq. (7.73).
*
Each cross section has a corresponding longitudinal section which can be con-
sidered as a section of a beam in a rst approach. The strut and tie approach for
beams shear should be applicable along this section.
*
A corresponding reinforcement force is given by all transformed reinforcement
contributions. With t
sx
= a
sx
f
y
, t
sx
= a
sx
f
y
, see Eq. (7.63), the transforma-
8
This corresponds to Eq. (7.26)
3
.
tion of reinforcement forces to the -direction is basically determined accord-
ing to Eq. (7.54) and leads to
9
t
s
= t
sx
cos
2
+t
sy
sin
2
a
s
= a
sx
cos
2
+a
sy
sin
2
(7.74)
with the effective reinforcement a
s
in - direction.
Shear design itself can be performed as for beams within this setting. It has to be
performed for both principal directions, if necessary.
Example 7.3 Computation of shear forces and shear design
We refer to Examples 7.1, 7.2 with the same system and loading. With four integration
points given per element, see Table 7.1 and Fig. 7.2b, four values have been computed for
each of m
x
, m
y
, m
xy
.
Alinear approximation of moments is performed within each element according to Eq. (7.72)
with details given in Appendix C.0.1. Derivatives of moments are determined according
to Eqs. C.6, C.7, which are constant within an element. Thus, shear forces q
x
, q
y
com-
puted with Eq. (7.71) are also constant within an element but may differ from element to
element. The results are shown in Fig. 7.6a.
Figure 7.6: Example 7.3 a) computed shear forces b) critical shear points
Two computed shear force values within elements adjacent to the right and upper
boundary edge are specied. Regarding an element base length of 1 m these values
basically match to the corresponding support reactions, see Fig. 7.3b. Their positive
sign, see Fig. 7.1 for sign conventions, corresponds to the negative (downward)
external loading.
Larger negative shear forces predominate around the left lower single support. This
also corresponds to the negative external loading.
Shear forces distribution looks somehow confuse around the central opening. A
ner discretization may presumably help to have more evidence regarding this area.
A shear design calculation can be performed according to Page 173.
9
This transformation differs from Eq. (7.73), as t
sx
, t
sy
are in plane and subject to a transformation of both
direction and reference length, while q
x
, q
y
are orthogonal to plane and subject to transformation of reference length
only.
175
A basic design value is given by the shear force admittable in case without shear
reinforcement, see [din08, 10.3.3]
v
Rd,ct
= 0.10 (100 f
ck
)
1/3
d (7.75)
with = 1.89, f
ck
= 30 MN/m
2
, d = 0.25 m in the current example. The param-
eter denotes the reinforcement ratio, i.e. = a
s
/d in the current context. A
typical value is given by a
s
= 5 cm
2
/m and = 5 10
4
/0.25 = 0.002. This leads
to v
Rd,ct
= 0.086 MN/m
2
.
A value q
according to Eq. (7.73) is computed throughout the slab and compared to

the admittable shear force according to Eq. (7.75) with the effective reinforcement
Eq. (7.74).
*
The points computed with [q
[ > v
Rd,ct
are shown in Fig. 7.6b. Most of them
are insofar not critical as the computed necessary bending reinforcement is low,
i.e. the necessary shear bearing capacity may be reached with larger bending
reinforcement, which is built in anyway to have minimum uniform bending
reinforcement.
end example 7.3
Punching
Punching is a particular occurence of shear forces, i.e. shear forces are regarded
with respect to single supports.
A design value for punching results from a computed support reaction force.
Design itself obeys the corresponding methods of ordinary reinforced concrete.
Bibliography
[Spe88] B. Specht. Modied shape functions for the three-node plate bending element passing
the patch test. Int. J. Num. Meth. Eng., 26:705715, 1988.
[ZT91] O. C. Zienkiewicz and R. L. Taylor. The Finite Element Method, Volume 2. McGraw
Hill, London, 4.Auage edition, 1991.
176
Chapter 8
Shells
8.1 Preliminary Remarks
Shell kinematics is quite complex, see [???]. Thus, deviating from the standard way for struc-
tural elements up to now kinematics, generalized material behavior, equilibrium formulated in
generalized forces, appropriate element types, system building a relatively short track coupled
to a simple standard nite shell element is discussed in the following. The chosen element is the
continuum based 4-node shell element as has been discussed, e.g., by [Bat96, 5.4.2].
Shells include plates and slabs as special cases. In particular, they can model slabs exposed
to the combined action of lateral and in-plane actions. This effect has already been discussed
for cracked reinforced concrete beams in a simplier setup, see Examples 3.2, 3.4.
8.2 Approximation of Geometry and Displacements
Shells are an extension of slabs where the reference plane becomes a simply or doubly
curved reference surface. The geometry of a surface in space is described in a global
cartesian system with base vectors e
1
, e
2
, e
3
by its coordinates
x
1
= x
1
(r, s), x
2
= x
2
(r, s), x
3
= x
3
(r, s) (8.1)
whereby isoparametric coordinates r, s serve as independent variables. Thus, a pair r, s
identies a point of the reference surface or a shell point. Every shell point has thickness
a. Reference surface and thickness occupy a shell body.
Furthermore, every shell point has a so called director. This is a unit vector V
n
describing
a direction of a cross section on which the Bernoulli-hypothesis is applied. Basically the
shell director is chosen independently from the geometriy denition of Eq. (8.1), but more
or less it coincides with the normal of the reference surface in case of smooth forms.
A local cartesian coordinate system is created for a shell point using, e.g. the unit vector
e
y
of the global coordinate system with the vector cross product leading to a unit vector
V
1
V
1
=
e
y
V
n
[e
y
V
n
[
, e
y
V
n
=
_
_
V
nz
0
V
nx
_
_
(8.2)
and another unit vector V
2
V
2
= V
n
V
1
=
_
_
V
1z
V
ny
V
1x
V
nz
V
1z
V
nx
V
1x
V
ny
_
_
(8.3)
177
178 8.2 Approximation of Geometry and Displacements
V
1
, V
2
, V
n
in this sequence form an orthogonal, normalized, right-handed coordinate
system which more or less snuggles against the reference surface.
Shell geometry is approximated by an isoparametric Finite-Element interpolation. Nodes
are placed in the reference surface spanning a mesh of quadrilaterial elements whereby
each element has four nodes. The body occupied by the undeformed shell is interpolated
by
x
i
(r, s, t) =
4
K=1
N
K
(r, s) x
iK
+
t
2
4
K=1
a
K
N
K
(r, s) V
niK
, i = 1 . . . 3 (8.4)
with
x
i
i-th coordinate of shell body
N
K
(r, s) =
1
4
(1 +r
K
r)(1 +s
K
s) acc. to Eqs. (1.4,1.15)
r
K
, s
K
local isoparametric coordinates of node K
x
iK
i-th coordinate of node K
a
K
shell thickness at node K
V
niK
i-th component of director at node K
r, s local isoparametric coordinates within the reference surface
t local isoparametric coordinate lateral to the reference surface
x
i
, x
iK
, a
K
have dimensions of [length] while 1 r, s, t 1 and r
K
, s
K
= 1 are
dimensionless. Eq. (8.4) leads to the Jacobian similar to Eq. (1.16)
J =
_
_
J
11
J
12
J
13
J
21
J
22
J
23
J
31
J
32
J
33
_
_
=
_
_
x
1
r
x
1
s
x
1
t
x
2
r
x
2
s
x
2
t
x
3
r
x
3
s
x
3
t
_
_
(8.5)
connecting global coordinates with local isoparametric coordinates. Its components are
given by
x
i
r
=
4
K=1
b
rK
X
iK
+
t
2
4
K=1
b
rK
a
K
V
niK
x
i
s
=
4
K=1
b
sK
X
iK
+
t
2
4
K=1
b
sK
a
K
V
niK
x
i
t
=
1
2
4
k=1
N
K
a
K
V
niK
i = 1 . . . 3 (8.6)
with
b
rK
=
N
K
r
=
1
4
r
K
(1 +s
K
s), b
sK
=
N
K
s
=
1
4
s
K
(1 +r
K
r) (8.7)
Shell displacements have to be approximated in the next step. For this purpose we intro-
duc a small rotation around the vector V
1
, see Eq. (8.2), and a small rotation around
the vector V
2
, see Eq. (8.6). This leads to a vector v
v = V
2
+ V
1
(8.8)
lying in the plane spanned by V
1
, V
2
. This vector is in particular given at nodes with
v
K
= V
2K
+ V
1K
and V
2K
, V
1K
determined from Eqs. (8.2,8.6) using the par-
ticular director V
nk
. The vector v
K
are used to deform the directors V
nK
leading to an
interpolation of displacements
u
i
(r, s, t) =
4
K=1
N
K
(r, s) u
iK
+
t
2
4
K=1
a
K
N
K
(r, s) v
iK
, i = 1 . . . 3 (8.9)
with
179
u
i
i-th component of displacement
u
iK
i-th component of displacement of node K
v
iK
=
K
V
2iK
+
K
V
1iK
i-te component of director change at node K
This approach realizes Bernoullis hypothesis of plane cross sections whereby the orien-
tation of a cross section is dened by the direction of the corresponding director.
8.3 Approximation of Deformations
Shell deformations are derived from shell displacements by their derivatives with respect
to spatial coordinates. We start with the local isoparametric coordinates
_
_
u
i
r
u
i
s
u
i
t
_
_
=
4
K=1
_
_
b
rK
t g
1iK
b
rK
t g
2iK
b
rK
b
sK
t g
1iK
b
sK
t g
2iK
b
sK
0 g
1iK
N
K
g
2iK
N
K
_
_
_
_
u
iK
K
_
_
, i = 1 . . . 3 (8.10)
with scaled rotation axes
g
1K
=
1
2
a
K
V
2K
,
_
_
g
1xK
g
1yK
g
1zK
_
_
=
1
2
a
K
_
_
V
2xK
V
2yK
V
2zK
_
_
g
2K
=
1
2
a
K
V
1K
,
_
_
g
2xK
g
2yK
g
2zK
_
_
=
1
2
a
K
_
_
V
1xK
V
1yK
V
1zK
_
_
(8.11)
This is transformed into derivatives with respect to global coordinates with the inverse of
the Jacobian
_
_
_
u
i
x
1
u
i
x
2
u
i
x
3
_
_
_
= J
1
_
_
u
i
r
u
i
s
u
i
t
_
_
i = 1 . . . 3 (8.12)
with
J
1
=
_
_
J
1
11
J
1
12
J
1
13
J
1
21
J
1
22
J
1
23
J
1
31
J
1
32
J
1
33
_
_
=
_
_
r
x
1
r
x
2
r
x
3
s
x
1
s
x
2
s
x
3
t
x
1
t
x
2
t
x
3
_
_
(8.13)
Eqs. (8.10,8.12,8.13) may be used to obtain the interpolation of the small strain tensor
components
ij
=
1
2
_
u
i
x
j
+
u
j
x
i
_
, i, j = 1 . . . 3 (8.14)
This is identical to the strains of a three dimensional body. The difference arises from the
constrained freedom to deform according to Eq. (8.9) including Bernoullis hypothesis.
The second order strain tensor as a whole is given by
E =
3
i=1
3
j=1
ij
e
i
e
j
(8.15)
whereby e
i
e
j
is the tensor product of the global system unit vectors e
i
and e
j
.
180 8.3 Approximation of Deformations
Deformations according to Eq. (8.14) are measured in the global cartesian system and are
inconvenient for thin curved shell bodies. A covariant or so called natural coordinate
system is more suitable
1
. Its base vectors are formed by the Jacobian J as
G
1
=
_
_
(
11
(
12
(
13
_
_
=
_
_
J
11
J
21
J
31
_
_
, G
2
=
_
_
(
21
(
22
(
23
_
_
=
_
_
J
12
J
22
J
32
_
_
, G
3
=
_
_
(
31
(
32
(
33
_
_
=
_
_
J
13
J
23
J
33
_
_
(8.16)
with j-th component (
ij
of base vector i measured in the absolute cartesian system. G
1
is the tangential vector along the space curve given with varying r while s, t are hold
constant, G
2
the tangential vector along the curve with varying s and r, t constant and
G
3
with varying t and r, s constant. The covariant system generally is skew and not
normalized, i.e. G
i
G
j
,= 0 for i ,= j and G
i
G
j
,= 1 for i = j.
Thus, a contravariant coordinate system is introduced with base vectors
G
1
=
_
_
(
11
(
12
(
13
_
_
=
_
_
J
1
11
J
1
12
J
1
13
_
_
, G
2
=
_
_
(
21
(
22
(
23
_
_
=
_
_
J
1
21
J
1
22
J
1
23
_
_
, G
3
=
_
_
(
31
(
32
(
33
_
_
=
_
_
J
1
31
J
1
32
J
1
33
_
_
(8.17)
utilizing the inverse J
1
of the Jacobian. Due to the denitions of G
i
, G
j
the properties
G
i
G
j
= 0 hold for i ,= j and G
i
G
j
= 1 for i = j. Contravariant and covariant systems
may also formally be derived for cartesian coordinate systems, but then they coincide due
to normalization and orthogonality.
Following the approach in [Bat96, 2.4, 6.5.2], [DB84] the strain as a whole is described
as
E =
3
i=1
3
j=1

ij
G
i
G
j
(8.18)
with so called covariant strain components
2

ij
or natural strains. Natural strain com-
ponents have a dimension of [length]
2
as G
i
, G
j
each have a dimension of [length]
1
.
Identity of Eqs. (8.15,8.18) together with Eq. (8.14) leads to
3

ij
=
3
r=1
3
s=1
(
ir
(
js
rs
=
1
2
3
k=1
_
(
ik
u
k
j
+(
jk
u
k
i
_
, i, j = 1 . . . 3 (8.19)
with
1
= r,
2
= s,
3
= t.
A nal remark: The components (
ij
of the covariant base form a second order tensor. But
it is not symmetric, i.e. (
ij
,= (
ji
. The same holds for the contravariant base: (
ij
,= (
ji
.
Taking Eqs. (8.19,8.10) together the interpolation of contravariant strain components is
given by
=
4
K=1
B
K
u
K
(8.20)
1
This should improve the elements behavior in case of mesh distortions [Bat96, p. 425].
2
Writing indices of
ij
as subscripts and thus making these quantities contravariant is pure convention but quite
convenient in the context of tensor calculus.
3
May be shown using (1) G
i
G
j
= 0 for i = j, (2) e
j
G
i
= G
i
e
j
= G
ij
and (3) G
ij
= J
ji
= x
j
/
i
.
181
with
=
_

11

22

33
2
23
2
13
2
12
_
T
B
K
=
_
_
(
11
b
rK
(
12
b
rK
(
13
b
rK
t b
rK
1
11K
t b
rK
1
12K
(
21
b
sK
(
22
b
sK
(
23
b
sK
t b
sK
1
21K
t b
sK
1
22K
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
_
_
u
K
=
_
u
1K
u
2K
u
3K

K

K
_
T
(8.21)
and
1
11K
= (
11
g
1xK
+(
12
g
1yK
+(
13
g
1zK
, 1
12K
= (
11
g
2xK
+(
12
g
2yK
+(
13
g
2zK
1
21K
= (
21
g
1xK
+(
22
g
1yK
+(
23
g
1zK
, 1
22K
= (
21
g
2xK
+(
22
g
2yK
+(
23
g
2zK
.
.
.
.
.
.
.
.
.
.
.
.
(8.22)
with (
ij
according to Eqs. (8.16,8.5), b
rK
, b
sK
according to Eq. (8.7), both depending
on r, s, and g
ixK
, g
iyK
according to Eq. (8.11) whereby K indicates the elements nodes.
The approach Eq. (8.21) still includes a non-zero strain component
33
normal to the
shells reference surface. This results from the continuum based approach. Its absolute
value should be considerably smaller compared to the other components in practical ap-
plications.
Regarding Eq. (8.21), the discretized strain state of every point in the shell body is ruled
by ve degrees of freedom per node. Thus, a so called ve-parameter shell model is given.
8.4 Shell Stresses and Material Laws
The concept of Cauchy stresses has already been discussed in Section 5.2, Page 105.
Stress components have been introduced with respect to the global cartesian system. The
stress tensor as whole in analogy to Eq. (8.15) is given by
S =
3
i=1
3
j=1
ij
e
i
e
j
(8.23)
Within the context of shells it is appropriate to use the covariant system Eq. (8.16) as a
base for stress components
S =
3
i=1
3
j=1

ij
G
i
G
j
(8.24)
with so called contravariant stress components
ij
. Identity of Eqs. (8.23,8.24) leads to
4

ij
=
3
r=1
3
s=1
(
ir
(
js
rs
(8.25)
Motivation of introducing contravariant stress components is given by formulating the
rate of internal specic strain energy which is dened in the global cartesian system as
e =
3
i=1
3
j=1
ij

ij
(8.26)
4
May be shown using (1) G
i
G
j
= 0 for i = j, (2) e
j
G
i
= G
i
e
j
= G
ij
.
182 8.4 Shell Stresses and Material Laws
This particular formulation of the strain energy establishes, e.g., the formulation of the
principle of virtual displacements, see Eqs. (1.28,1.29), which on the other hand is a
basis for the Finite-Element-Method. Using the transformation rules Eq. (8.19,8.25) and
regarding G
i
G
j
= 0 for i ,= j and G
i
G
j
= 1 for i = j it can be shown that
3
i=1
3
j=1
ij

ij
=
3
i=1
3
j=1

ij

ij
(8.27)
i.e. contravariant stress components are complementary to covariant strain components
from an energetically point of view.
Thus, while using covariant or natural strain components to describe shell deformation
it is mandatory to use contravariant stress components while utilizing weak equilibrium
conditions like the principle of virtual displacements.
To give a comprehensible description of the material behavior of shells it is appropriate
to use the local systemV
1
, V
2
, V
n
as has been introduced with the shell director V
n
and
Eqs. (8.2,8.3). The local system is an orthogonal, normalized and right-handed or carte-
sian coordinate system, respectively. On one hand it leans against the shells reference
surface, thus it generally changes with every point of the reference surface
5
. On the other
hand it is appropriate for the description of material behavior due to its normalization and
orthogonality.
To facilitate the notation V
3
= V
n
is used in the following. In analogy to Eqs. (8.15,8.23)
local strain and stress components
ij
,
ij
are given by
6
E =
3
i=1
3
j=1

ij
V
i
V
j
, S =
3
i=1
3
j=1

ij
V
i
V
j
(8.28)
To determine local material behavior natural strains as they are derived using Eq. (8.20)
have to be transformed into local strains . The identity of Eqs. (8.28
1
,8.18) leads to

ij
=
3
r=1
3
s=1
T
ir
T
sj

rs
(8.29)
with
T
ij
= V
i
G
j
, T
ij
,= T
ji
(8.30)
This may be written as a matrix operation
= T (8.31)
with , ordered according to Eq. (5.3) and the components of T derived by Eq. (8.30).
To simplify a linear elastic material behavior is considered. The shell body differs from
the three-dimensional continuum ruled by the linear elastic law Eq. (5.21) insofar as its
normal stress in a plane normal to the reference surface should be small compared to all
other stress components. Thus, we use
=

C (8.32)
5
Also called co-rotational.
6
As they are referenced in a cartesian system no distinction between contravariant and covariant is necessary,
183
with ordered as in Eq. (5.7) and
C =
_
_
E(1)
(1+)(12)
E
(1+)(12)
0 0 0 0
E
(1+)(12)
E(1)
(1+)(12)
0 0 0 0
0 0 0 0 0 0
0 0 0
E
2(1+)
0 0
0 0 0 0
E
2(1+)
0
0 0 0 0 0
E
2(1+)
_
_
(8.33)
according to Eq. (5.21) whereby the inuence of the colateral components
33
,
33
has
been neglected due to the assumption of thin shells. The relation Eq. (8.32) may be
generalized as incremental form

=

C
T

, see also Eq. (5.9), with the tangential
material stiffness

C
T
covering nonlinear behavior as desired.
Local stress components
ij
have to be transformed into contravariant components
ij
,
see following Section 8.5. It can be shown that
= T
T
(8.34)
with ordered as in Eq. (5.7) and the transposed of the transformation matrix T from
Eq. (8.31). This bases on the identity of Eqs. (8.28)
2
,(8.25).
Finally, combination of Eqs. (8.34,8.32,8.31) leads to
= T
T

C T =

C (8.35)
which yields a transformation law
C = T
T

C T (8.36)
for the material matrix. This transformation law is also valid for a tangential and nonlinear
material matrix. Material matrices are required to determine the tangential stiffness of
discretized systems, see Page 9.
8.5 System Building
The theory under consideration treats the shell body as a continuum with constraints re-
garding deformations. Thus, in a rst approach the general form Eq. (1.28) is used to
describe weak equilibrium but by the product
T
replaced by
T
with accord-
ing to Eq. (8.21) and according to Eq. (8.34)
_
V
T
dV +
_
V
u
T
udV =
_
V
u
T
pdV +
_
A
t
u
T
t dA (8.37)
For the evaluation of integrals see Eqs. (1.33)-(1.35). Integration is performed by numer-
ical methods, the basic approach has been described in Section 1.7 on Page 14. It will be
extended to the case of continuum based shells.
Integration of internal nodal forces is performed with
f
I
=
_
V
I
B
T
(r, s, t) (r, s, t) dV =
+1
_
1
+1
_
1
+1
_
1
B
T
(r, s, t) (r, s, t) J(r, s, t) dtdrds
(8.38)
184 8.5 System Building
see Eq. (1.33), with B assembled by the B
K
s from Eq.(8.21), from Eq. (8.34), lo-
cal isoparametric coordinates r, s, t and the determinant J = detJ of the Jacobian, see
Eq. (8.5). It is evaluated numerically, i.e.
f
I
=
n
u
i=0
n
u
j=0
n
v
k=0
k
B
T
(r
i
, s
j
, t
k
) (r
i
, s
j
, t
k
) J(r
i
, s
j
, t
k
) (8.39)
with integration orders n
u
, n
v
, sampling points r
i
, s
j
, t
k
and weighting factors
i,j,k
. Let
us assmune that a Gaussian integration is used. Then it may be appropriate to use different
integration orders n
u
, n
v
along the reference surface with local coordinates r, s and along
the collateral direction with local coordinate t.
The collateral direction may need a different treatment compared to the in-surface direc-
tions due to shell bending with transverse linear strains. In case of linear elastic material
behavior a choice u = v = 1, n
u
= n
v
= 2, see Table 1.1, is appropriate as stresses
also arise linearily. In case of nonlinear material behavior stresses may vary nonlinearily
with kinks and jumps. This generally requires a higher integration order for the collateral
directions, e.g. u = 1, n
u
= 2 and v = 4, n
v
= 5. It may also be appropriate to choose
a different integration scheme for the collateral directions, e.g. a Lobatto-scheme which
yields a higher accuracy in cases with extremal integrand values on the boundary.
As every node has ve kinematic degrees of freedom Eq. (8.38) leads to ve components
for internal forces f
I
at every node I which are three force components with respect to the
global coordinate system and two bending moment components with respect to the local
directions V
1
, V
2
, see Eqs. (8.2,8.3).
Prescribed distributed loads p, see Eq. (1.28), are given as forces per volume and pre-
scribed surface tractions

t as forces per area, each with directions related to the global
coordinate system. Corresponding nodal forces, see Eq. (1.33), again have ve compo-
nents for each node. Element stiffness and mass matrices have 20 20 entries with the
four node element. Assembling of element contributions is performed in the standard
way, see Page 8.
Due to the continuum based approach structural response is described by strains and
stresses varying with the position in the reference surface and the collateral direction dis-
tance. Regarding shells and slabs a more familiar approach is given with, e.g., moments
and shear forces. It is appropriate to refer them to the local system, see Page 182. In an
analogous way as for slabs, see Eq. (7.7), resulting local internal forces are derived from
local stresses , see Eq. (8.32), by
n
1
=
a
2
_
1
1

11
dt, n
2
=
a
2
_
1
1

22
dt, n
12
=
a
2
_
1
1

12
dt
m
1
=
a
2
4
_
1
1

11
tdt, m
2
=
a
2
4
_
1
1

22
tdt, m
12
=
a
2
4
_
1
1

12
tdt
q
1
=
a
2
_
1
1

13
dt, q
2
=
a
2
_
1
1

23
dt
(8.40)
with the local shell thickness a and the isoparametric local coordinate 1 t 1. In
practice the integration is again performed numerically. Thus, nonlinear materials are au-
tomatically covered. For integration method and order see the remarks above concerning
integration of system integrals.
185
8.6 Slabs and Beams as a Special Case
A rectangular slab element of constant thickness a is considered as a special case of
the general shell element, see Fig. 8.1. Directors are given by V
1
=
_
1 0 0
_
T
,
Figure 8.1: Slab element as a special case of a shell
V
2
=
_
0 1 0
_
T
and V
n
=
_
0 0 1
_
T
. Thus, after some calculations the matrix
B of interpolation functions Eq. (8.20) is written as
B
K
=
_
_
J
11
b
rK
0 0 0 t
a
2
J
11
b
rK
0 J
22
b
sK
0 t
a
2
J
22
b
sK
0
0 0 0 0 0
0 0 J
33
b
sK

a
2
J
22
N
K
0
0 0 J
33
b
rK
0
a
2
J
11
N
K
J
11
b
sK
J
22
b
rK
0 t
a
2
J
22
b
rK
t
a
2
J
11
b
sK
_
_
(8.41)
with the components J
ij
of the Jacobian according to Eq. (8.5), b
sK
, b
rK
according to
Eq. (8.7) and for N
K
see Eq. (8.4).
Adisplacement in the x
1
x
3
-plane is applied with u
1
= u
4
=
_
u
1
0 w
1
0
1
_
T
,
u
2
= u
3
=
_
u
2
0 w
2
0
2
_
T
leading to
_
_
_
_
_
_
_
_

11

22

33
2
23
2
13
2
12
_
_
_
_
_
_
_
_
=
_
_
_
_
_
_
_
_
1
2
J
11
u
1
t
a
4
J
11
1
+
1
2
J
11
u
2
+t
a
4
J
11
2
0
0
0
1
2
J
33
w
1
+
a
4
J
11
(1 r)
1
+
1
2
J
33
w
2
+
a
4
J
11
(1 +r)
2
0
_
_
_
_
_
_
_
_
(8.42)
see Eqs. (8.20,8.4-8.7). Furthermore, we have J
11
= L
1
/2, J
33
= a/2, see Eq. (8.5).
For rectangular slabs it is convienient to transform natural strain components
ij
back into
the global cartesian coordinate system. Transformation rules may again be derived by the
identity of Eqs. (8.15,8.18). For the special case under consideration the transformation
is given by
11
=
_
(
11
_
2

11
=
4
L
2
1

11
,
13
= (
11
(
33

13
=
4
L
1
a

13
(8.43)
186 8.7 Locking
This nally leads to
_

11
2
13
_
=
1
L
1
_
1 0 t
a
2
1 0 t
a
2
0 1
L
1
2
(1 r) 0 1
L
1
2
(1 +r)
_
_
_
_
_
_
_
_
_
u
1
w
1
1
u
2
w
2
2
_
_
_
_
_
_
_
_
(8.44)
The interpolation of strains according to Eq. (8.44) corresponds to the interpolation of
strains of the 2D Timoshenko beam element, see Eq. (3.78). This becomes obvious with
(1) setting 2
13
= , (2) regarding the reversed orientation of rotations, see Eq. (8.8),
and (3) adopting Eq. (3.4) ruling beam kinematics to the current case with
11
= +
t
a
2
, 1 t 1.
8.7 Locking
As a special case a state of uniform bending in the longitudinal x
1
-direction is applied to
Eq. (8.44) with
2
=
1
= /2 and u
1
= w
1
= u
2
= w
2
= 0. Eq. (8.44) yields
11
= t
a
2
L
1
,
13
=
1
2
r, 1 r, t 1 (8.45)
The term z = t
a
2
with the thickness a describes the distance from the reference plane
and

L
1
corresponds to a curvature. To simplify a linear elastic behavior is assumed with
= 0, see Eqs. (5.3,5.7,5.10,5.21),
11
= E
11
= E t
a
2
L
1
= E z

L
1
,
13
= G
13
=
1
2
G r (8.46)
with a normal stress
11
in the longitudinal x
1
-directions, a shear stress
13
in the vertical
x
3
-direction, Youngs modulus E and G =
E
2
. This corresponds to stresses in beams with
=

L
1
, see Eqs. (3.6,3.7) and leads to a resulting moment and shear force per unit width
m = E
a
3
12
L
1
, q =
1
2
Ga r (8.47)
Thus, the applied deformation results in a constant bending moment and a linearily vary-
ing shear force along the element. This obviously violates equilibrium conditions locally,
as a zero shear force is required throughout the element in case of constant bending mo-
ment. A spurious transverse shear force arises with this type of element. The local error
shall be measured by
q
m
= 3
L
1
a
2
r, 1 r 1 (8.48)
It becomes larger with more slender elements, i.e. with decreasing thickness or increasing
element length.
Local or strong equilibrium is not enforced within the Finite-Element-Method, but weak
or integral equilibrium, see Section 1.5. As a consequence, nodal forces resulting from
187
integration of internal forces have to be in equilibrium. For the special case under consid-
eration nodal forces according to Eq. (8.38) are given by
r =
1
_
1
1
_
1
1
L
1
_
_
1 0
0 1
t
a
2
L
1
2
(1 r)
1 0
0 1
t
a
2
L
1
2
(1 +r)
_
_
E t
a
2
L
1
1
4
E r
_
a
2
dt
L
1
2
dr = E
_
_
_
_
_
_
_
_
_
0
0
a
3
12
L
1
a
24
L
1
0
0
a
3
12
L
1
+
a
24
L
1
_
_
_
_
_
_
_
_
_
(8.49)
The non-zero nodal forces are conjugate to the applied rotation angle /2 on the left
hand side r = 1 and /2 on the right hand side r = 1. Insofar moments are given
from a mechanical point of view forming an equilibrated system. As has already been
mentioned /L
1
is a curvature. Furthermore, Ea
3
/12 is the bending stiffness per unit
width. Thus, the rst term in each entry corresponds to a reasonable mechanical behavior.
The second part EaL
1
/24 leads to an additional spurious moment resulting from the
spurious shear force. This spurious moment corresponds to an additional spurious stiff-
ness of this element. The inuence of this spurious effects increase with decreasing thick-
ness a with a constant element length L
1
, i.e. with increasing element slenderness. The
spurious effects reduce with ner discretizations, i.e. decreasing L
1
for a constant a.
Convergence may principally be reached, but it is reached slowly with a large parameter
c, see Section 1.6 Eq. (1.59). This type of element under consideration yields much too
stiff models in practical applications due to spurious transverse shear forces. So called
transverse shear locking occurs. A number of locking phenomena are known, e.g.
transverse shear locking, i.e. spurious transverse shear forces in case of transverse
bending
in-plane shear locking, i.e. spurious in-plane shear forces in case of in-plane bending
membrane locking, i.e. spurious membrane forces in case of transverse bending
and others [Bis99, 6.4].
Transverse shear locking is a major cause for deciences of slab and shell elements. It has
the property that spurious shear stresses disappear in distinguished points of an element,
e.g in the point r = 0 for the case under consideration in the previous Section. On the
other hand, such shear forces that are reasonable from a mechanical point of view yield a
value in these distinguished points.
This motivates popular approaches to avoid locking:
Reduced integration of system integrals, see Eqs. (1.33,1.36,1.40,1.61,1.62).
This corresponds to an integration at r = 0, s = 0 for the element under consider-
ation. Reduced integration does not affect integration order along the local t-axis.
Albeit, a numerical instability of results, so called hour-glassing, may occur with
reduced integration. The occurence of hour-glassing depends on the discretized ge-
ometry and applied boundary and loading conditions
Transverse shear strains are approximated with their own elds applying a mixed
interpolation
These elds are connected to the elds given by Eq. (8.20) through the values of

13
,
23
in those distinguished points with vanishing spurious transverse shear forces.
188 8.7 Locking
These points are given by the coordinates A : r = 0, s = 1, B : r = 1, s = 0,
C : r = 0, s = 1 and D : r = 1, s = 0 for the element under consideration. The
particular strains determined by Eq. (8.20) are given by
A
13
,
B
13
,
C
13
,
D
13
and
213
correspondingly. Anchored by these values the elds for transverse shear strains are
assumed with

13
(r, s) =
1
2
(1 +s)
A
13
+
1
2
(1 s)
C
13

23
(r, s) =
1
2
(1 +r)
D
23
+
1
2
(1 r)
B
23
(8.50)
The approach is called Assumed-Natural-Strain-Method (ANS) [DB84] and leads to
a modication of the rows 4 and 5 of the matrix B
K
, see Eq. (8.21). This modica-
tion is straightforward with evaluating Eq. (8.20) in points A, B, C, D and combin-
ing it with Eq. (8.50).
Assuming strains partially independent from displacements or applying mixed inter-
polations in a rst instance is not covered by the principle of virtual displacements
Eqs. (1.28,1.29). Thus, an extended weak form like the principle of Hu-Washizu
[Bat96, 4.4.2] is required. This involves elds for stresses and strains as indepen-
dent solution variables.
Its applicability requires an additive split of the matrix B, see Eq. (8.20). The cur-
rent approach of Eq. (8.50) leading to an extension of Eq. (8.21) as indicated above
allows such a split. Independent parts of the stress eld may be eliminated in ad-
vance applying reasonable assumptions. Thus, the mixed interpolation may nally
be applied in the framework of the principle of virtual displacements.
Example 8.1 Convergence study for linear simple slab
We consider a quadratic linear elastic slab with a span L = L
1
= L
2
= 8.0 m, a thick-
ness a = 0.25 m and material parameters E = 33 000 MN/m
2
, = 0.2 (0.25) in ac-
cordance with common concrete grades. The slab is simply supported along its edges,
i.e. hinged without vertical displacements. A constant vertical loading is assumed with
q = 16 kN/m
2
downward.
Assuming the Kirchhoff theory, see Page 7.2, neglection shear an exact solution for this
problem is described in [Gir74, 7.8 b)]. The maximum deection in the center point is
given by an innite double sum
w
e
max
=
16q L
4
K
6
,
n
sin
n
2
sin
m
2
mn(m
2
+n
2
)
2
, K =
1
1
2
E a
3
12
(8.51)
With the parameters given this yields a converged value w
e
max
= 5.95 10
3
m.
A small convergence study is performed with meshes of 1 element up to 16 elements
whereby quarter symmetry is used. Fig. 8.2 shows the meshes. Boundary conditions
of nodes along symmetry axes are given by prescribing appropriate zero rotations. The
maximum deection w
n
max
arises at right upper corner node. Computed values are
Discretization 1 1 2 2 3 3 4 4
w
max
[m] 5.39 10
3
5.87 10
3
5.95 10
3
5.98 10
3
189
Figure 8.2: Example 8.1 a) b)
whereby the Assumed-Natural-Strain-Method, see Page 188 and Eq. (8.50), has been ap-
plied to avoid transverse shear locking
7
. Inclusion of shear deformations leads to slightly
larger converged displacement compared to Kirchhoff theory.
end example 8.1
8.8 Reinforced Concrete Shells
Basically the approach Eq. (8.32)
(r, s, t) =

C (r, s, t) (8.52)
with local isoparametric coordinates
8
r, s, t or its incremental form
(r, s, t) =

C
T

(r, s, t) (8.53)
see also Section 5.2 Page 106, allow for an arbitrary material behavior with variable ma-
terial stiffness

C or tangential material stiffness

C
T
. Regarding a shell point r, s in the
reference surface it is appropriate to introduce the notion of layers.
A layer is a plane through the point r, s, t, i.e. thickness t is regarded in contrast to
the related shell point. This plane has the shell points director V
n
(r, s) as normal.
The behavior with respect to a layer is characterized by the components
11
,
22
,
12
and

11
,
22
,
12
, respectively. Furthermore, for reinforced cracked concrete it is appropriate to
decouple transverse shear characterized by
13
,
23
,
13
,
23
from layer behavior. These
assumptions motivate a generalization
C =
_
_
C
11
C
12
0 0 0 C
16
C
21
C
22
0 0 0 C
26
0 0 0 0 0 0
0 0 0 c
44
0 0
0 0 0 0 c
55
0
C
61
C
62
0 0 0 C
66
_
_
(8.54)
of Eq. (8.33) with component ordering according to Eqs. (5.3,5.7). Uppercase coefcients
mark layer behavior while lower case coefcients mark transverse shear behavior.
7
If it is not applied, i.e. Eq. (8.21) is used as is without modications regarding the entries for
23
,
13
, the
computed maximum deection would be wmax = 1.44 10
3
m for the 4 4-discretization. That is an error of
roughly 80%.
8
Local isoparametric coordinates tranlate into physical coordinates with Eq. (8.4).
190 8.8 Reinforced Concrete Shells
Transverse shear stiffness has already been discussed Section 3.5.5 in the context of struc-
tural beams. Structural beams and shells both rely on the assumption of plane deformed
cross sections. Regarding transverse shear stiffness of shells a substancially better ap-
proach than that used for beams is not yet available. This leads to a proposal
c
44
= c
55
= G, G =
E
2(1 +)
(8.55)
with a reduction factor and the initial values of Youngs modulus E, Poissons ratio ,
shear modulus G. According to Section 3.5.5 the reduction factor may be chosen with
= 0.5.
Layer behavior is described by a general form
_
_

11

22

12
_
_
=
_
_
C
11
C
12
C
16
C
21
C
22
C
26
C
61
C
62
C
66
_
_
_
_

11

22

12
_
_
(8.56)
or the corresponding incremental form. A layer has a biaxial plane stress state.
Thus, local layer behavior corresponds to local behavior of deep beams.
The thickness of a counterpart deep beam corresponds to the height of a layer. The
height of layer is a matter of t-integration of stresses into nodal forces, see Eq. (8.38),
or resulting internal forces, see Eq. (8.40). Such a height is implicitely included in a
numerical integration process.
The modeling of reinforced concrete deep beams as already been in dicussed in Chapter 6.
The approach may be directly transferred to layers of reinforced concrete shells.
Modeling of cracks due to limited tensile strength of concrete is treated in Sec-
tion 6.2. This leads to a stress-strain relation in a principal local coordinate system
9
,
see Eq. (6.25). Its transformation to the local system leads to a form covered by
Eq. (8.56). This may be transferred to the form Eq. (8.54) and nally be used in
Eq. (8.32).
This material relation is based on principal strains
1
,
2
derived from
11
,
22
,
12
.
Prinicpal strain directions also rule crack directions. Such calculations are per-
formed for each concrete layer t. Basically, principal strain direction may change
with t for a given shell point r, s, i.e. crack faces may become curved surfaces in
space.
Reinforcement and bond are treated in Section (6.3). Rigid bond is assumed in the
following in order to simplify. Regarding thin reinforcement meshes basically the
same procedure as described for deep beams, see Page 145, may be applied for
reinforced shell layers.
A difference is given as shell reinforcement layers are not implemented as sepa-
rate elements but subject to integration along the collateral t-direction within the
frame given by Eqs. (8.38,8.40). A reinforcement layer of sheet thickness a
R
, see
Eq. (6.26), is regarded with a collateral coordinate t
R
. The contribution of this layer
9
This involves the coordinate system transformations natural system local system principal system and
backward. This looks elaborate but is justiable to cover geometrical and physical complexity. This note basically
also concerns the treatment of reinforcement meshes.
191
to, e.g. internal forces, see Eq. (8.40), is given by
n
R1
= a
R

R11
, n
R2
= a
R

R22
, n
R12
= a
R

R12
m
R1
= a
R

R11
a
2
t
R
, m
R2
= a
R

R22
a
2
t
R
, m
R12
= a
R

R12
a
2
t
R
(8.57)
with local reinforcement stresses
R11
,
R22
,
R12
determined in analogy to Eq. (6.29)
and its related equations. Thin reinforcement layers generally do not contribute to
transverse shear forces. The described approach may be applied to multiple rein-
forcement layers while adding up their contributions.
Alternative methods to deal with reinforced concrete shells.
Larger reinforced concrete shells were built in a large number for a wide span of applica-
tions during the twenties up to the sixties of the 20th century. An excellent documentation
is given by [Joe62]. Due to expensive formwork and other upcoming restrictions their ap-
plication mainly reduced to cooling towers of large power plants nowadays. A future
perspective may arise with, e.g., upwind solar chimneys.
The relevance of the shell approach for reinforced concrete arises with the combination
of bending with membrane forces or slabs with deep beams, respectively. As has already
been demonstrated for structural beams bending and and membrane forces may interact in
case of cracked reinforced concrete sections. This will be demonstrated with the following
Example 8.2.
Example 8.2 Nonlinear calculation for simple slab
We consider the system of Example 8.1 with the same dimensions. A concrete grade
C30/37 according to EC2 [EN 04, 3.1] is chosen. This gives a Youngs modulus E
cm
=
33 000 MN/m
2
and a characteristic concrete strength of f
ck
= 30 MN/m
2
.
Reinforcing steel properties adhere to EC2 [EN 04, 3.2] with Youngs modulus E
s
=
200 000 MN/m
2
, yield strength f
yk
= 500 MN/m
2
and tensile strength f
t
= 525 MN/m
2
at a strain of
uk
= 25 10
3
. The cover to reinforcement is chosen with c = 0.02 m and
the effective depth of cross section with d 0.22 m.
Self weight is given by g = 0.25 25 = 6.25 kN/m
2
and a variable service load of
q = 5.0 kN/m
2
is assumed. These are characteristic values and have to be multiplied by
safety factors to have a design load of p = 1.35 6.25 + 1.50 5.0 = 16 kN/m
2
.
To determine a reinforcement a design is performed according to [EN 04, 6.1]. Internal
forces are determined based on Eqs. (7.11)
46
whereby deections and curvatures are de-
termined with an extension of Eq. (8.51) taking variable coordinates into account. Values
of bending and twisting moment are given by
m
x
= m
y
= 45.3 kNm/m, m
xy
= 38.0 kNm/m (8.58)
Reinforcement forces are determined by f
s
= m/z with the internal lever arm z. A good
estimation is given with z = 0.8 d. The design value for the reinforcement strength is
chosen with f
yd
= 500/1.15 = 435 MN/m
2
. Finally, the required reinforcement cross
section area a
s
obeys f
yd
a
s
= m/(0.8 d). This yields
10
a
sx
= a
sy
=
1
435
45.310
3
0.80.22
5.9 cm
2
/m
a
sxy
=
1
435
38.010
3
0.80.22
5.0 cm
2
/m
(8.59)
The bending moments m
x
, m
y
occur in the slab midpoint while the twisting moments
occur in the corner points. The bending moments require a bottem reinforcement while
the twisting moment also requires an upper reinforcement. A bottom reinforcement of
a
sx,lower
= a
sy,lower
= 6.0 cm
2
/m and an upper reinforcement a
sx,upper
= a
sy,upper
=
5.0 cm
2
/m are chosen throughout the slab in the following.
A more detailed design approach shows that the failure of this slab occurs due reinforce-
ment failure while concrete compression remains approximately in the elastic regime. But
the concrete will crack due to tensile stresses. This activates the reinforcement and leads
to a different behavior compared to the elastic case.
The approach of Section 8.8 with cracked concrete behavior as described in Section 6.2
will be used in the following. The tension stress of concrete is restricted to a value of f
ct
=
1 MN/m
2
. A reinforcement is regarded as has been determined before. The following
model parameters are chosen:
sheet thickness a
R
[m] sheet height coordinate t
R
[m]
lower x-direction 0.6 10
3
-0.10
lower y-direction 0.5 10
3
-0.10
upper x-direction 0.5 10
3
0.10
upper y-direction 0.5 10
3
0.10
Sheet height coordinates t
R
are assumed to be the same in x- and y-directions to preserve
symmetry. The tension stiffening effect, see Section 2.8, is regarded.
Discretization is chosen with 44 elements for the quarter slab. Due to the physical non-
linearities the loading is applied in 10 steps. An equilibrium iteration has to be performed
within each loading step leading to an incrementally iterative scheme, see Section 1.7,
Page 15. A Gaussian quadrature is used for integration of system matrices and vectors.
Integration orders, see Section 8.5, Eq. (8.38), are chosen with n
u
= 2 in the reference
surface directions and with n
v
= 4 in the collateral direction. Collateral Integration of
reinforcement contributions is performed separately while considering their discrete po-
sitions t
R
.
Results of numerical computation
The relation between the load factor and the mid point deection is shown in Fig. 8.3a.
Again thress stages can be seen: (I) stage I with basically linear behavior and only
sporadic cracking. (IIa) Stage IIa with progressing cracking leading to a nearly -
nal state of cracking. The load level is hold constant during ths stage. Deections
increase to a multiple of the value at the end of stage I, which may happen rapidly
due to constant loading. (IIb) Stage IIb of basically nal cracking state with a much
lower stiffness compared to stage I. The nal deection of w
max
= 5.15 10
2
m is
10
Twisting moments may be seen as skew principal moments which are directed along the diagonals of the slabs
corners. This leads to, e.g., upper diagonal tension which is hold by reinforcement in x- and y-direction. Transfor-
mation of stresses and of their reference length leads to the relation for a
sxy
.
193
Figure 8.3: Example 8.2 a) load deection curve b) principal strains in upper and lower concrete
layer
roughly 8 times higher compared to the linear elastic case of Example 8.1 and leads
to deection-span ratio of 1/150
11
.
A smeared crack concept is basically used as described in Section 6.2, Page 143.
It is extended to two directions. Smeared strains in cracked integration points are
shown in Fig. 8.3b for the case under consideration. The top and bottom layer are
chosen out of the four concrete layers. The lateral thickness direction is not to scale
with the ground directions.
Length and orientation of red bars indicate size and orientation of principal values of
smeared strains. Cracks are not shown explicitely but arise orthogonal to the strain
directions. Top and and bottom show a different behavior due to bending.
*
The bottom layer has two orthogonal cracks in the central region due to bending
moments and single cracks in diagonal direction in corner regions due to twist-
ing moments. Corresponding principal strains are across the diagonal direction
for the latter.
*
The top layer has no cracks in the central region as it is under compression.
But it has single cracks across the diagonal direction due to twisting moments.
Corresponding principal strains are in diagonal direction.
Computed prinicipal moments are shown in Fig. 8.4a. Bar directions indicate the
directions of the corresponding stresses. A green color indicates a positive moment,
a red color a negative one. A positive moment has tension on the lower surface and
compression on the upper surfacce.
Skew principal moment directions with opposite signs in the corner region corre-
spond to twisting moments. The negative moment in the diagonal direction has diag-
onal tension on the upper side which is compensated by upper x- and y-reinforcement.
The positive moment across the diagonal direction has corresponding tension on the
lower side. It is compensated by lower x- and y-reinforcement.
Central areas also have skew principal moments but with nearly the same sign and
size. Thus, there will be only minor twisting moments. Lower surface tension is
compensated by lower x- and y-reinforcement.
Anewphenomenon compared to Example 8.1 is given with membrane forces, which
are determined in analogy to the internal forces n
1
, n
2
, n
12
of Eq. (8.40)
13
. This
11
This ratio should be smaller than 1/250 in ordinary cases according to codes like Eurocode 2. But such a value
has to be maintained with load safety factors of 1.
Figure 8.4: Example 8.2 a) principal moments b) principal membrane forces
is caused by the elongation effect of cracked reinforced cross sections which has
already been discussed for beams in Example 3.2, Page 65. In contrast to beams
an enlongation may have a different direction for every point of the slabs reference
surface as the crack directions may be different.
This leads to eigenstresses, i.e. self-equilibrating internal forces without reaction
forces on supports. Due to this eigenstresses kinematic compatibility is preserved
even in case local elongations or also contractions. The corresponding principal
membrane forces are shown in Fig. 8.4b. In case that horizontal movements of the
slab are prevented on the supporting edges horizontal reaction force would arise and
the whole slab will come under compression.
end example 8.2
Bibliography
[Bat96] K.J. Bathe. Finite Element Procedures. Prentice Hall, Englewood Cliffs, New Jersey,
1996.
[Bis99] M. Bischoff. Theorie und numerik einer dreidimensionalen schalenformulierung.
Technical report, Bericht Nr. 30, Institut fr Baustatik, Universitt Stuttgart, 1999.
[Cze99] F. Czerny. Tafeln fr rechteckplatten. In Betonkalender 1999, Band I, chapter 277-339,
pages 1145. Verlag Ernst u. Sohn, Berlin, 1999.
[DB84] E.N. Dvorkin and K.J. Bathe. A continuum mechanics based four-node shell element
for general nonlinear analysis. Eng. Comput., 1:7788, 1984.
[EN 04] EN 1992-1-1. Eurocode 2: Design of concrete structures - Part 1-1: General rules
and rules for buildings, December 2004.
[Joe62] J. Joedicke. Schalenbau. Dokumente der Modernen Architektur 2. Karl Krmer Ver-
lag, Stuttgart, 1962.
195
Appendix A
ConFem
Coordinate system
Figure A.1: Coordinate System
196
Appendix B
Transformations of coordinate systems
In the following rectangular cartesian coordinates are considered.
Figure B.1: Coordinate transformations
Transformation of 2D vectors
A vector
a =
_
a
x
a
y
_
(B.1)
is initially given within a coordinate system x, y. Another coordinate system x, y
is rotated with an angle (positive in the counterclockwise direction). The compo-
nents of a in the system are
_
a
x
a
y
_
T
. Then the following relations hold
a = T a, a = T
T
a, T =
_
cos sin
sin cos
_
(B.2)
Transformation of 2D stresses
_
_
_

11

22

12
_
_
_
=
_
_
cos
2
sin
2
2 cos sin
sin
2
cos
2
2 cos sin
cos sin cos sin cos
2
sin
2
_
_
_
_
_
11
22
12
_
_
_
(B.3)
Inversion
_
_
_
11
22
12
_
_
_
=
_
_
cos
2
sin
2
2 cos sin
sin
2
cos
2
2 cos sin
cos sin cos sin cos
2
sin
2
_
_
_
_
_

11

22

12
_
_
_
(B.4)
197
198 APPENDIX B. TRANSFORMATIONS OF COORDINATE SYSTEMS
Transformation of 2D strains
_
_
_
D
11
D
22
2

D
12
_
_
_
=
_
_
cos
2
sin
2
cos sin
sin
2
cos
2
cos sin
2
sin
2
_
_
_
_
_
D
11
D
22
2D
12
_
_
_
(B.5)
Inversion
_
_
_
D
11
D
22
2D
12
_
_
_
=
_
_
cos
2
sin
2
cos sin
sin
2
cos
2
cos sin
2
sin
2
_
_
_
_
_
D
11
D
22
2

D
12
_
_
_
(B.6)
Appendix C
Linear regression analysis applications
For basics of linear regression analysis see [BSMM00, 4.4.3.1].
C.0.1 Determination of moments in triangular slab elements
Area coordinates, see Section 7.3.1, are used in the following. It is assumed, that four
sampling points according to Table 7.1 are used for numerical integration.
An linear approximation approach
m = a L
1
+b L
2
+c L
3
(C.1)
with L
1
, L
2
, L
3
according to Eq. (7.33) is chosen for the eld m. Regarding the sampling
points of numerical integration the approximation yields
m= X a, m=
_
_
_
_
m
(1)
m
(2)
m
(3)
m
(4)
_
_
_
_
, X =
_
_
1/3 1/3 1/3
0.6 0.2 0.2
0.2 0.6 0.2
0.2 0.2 0.6
_
_
, a =
_
_
a
b
c
_
_
(C.2)
The target values of the m (direct results of FEM-calculation) are collected in
m=
_
_
_
_
m
(1)
m
(2)
m
(3)
m
(4)
_
_
_
_
(C.3)
The linear regression equation to determine the coefcients a is given by
Aa = X
T
m, A = X
T
X =
1
225
_
_
124 88 88
88 124 88
88 88 124
_
_
, A
1
=
1
6
_
_
53
2
11 11
11
53
2
11
11 11
53
2
_
_
(C.4)
and
a = A
1
X
T
m (C.5)
Nodal values of m may be determined by applying Eq. (C.1) with nodal area coordinates.
Node 1: L
1
= 1, L
2
= 0, L
3
= 0 m = a
Node 2: L
1
= 0, L
2
= 1, L
3
= 0 m = b
199
200 APPENDIX C. LINEAR REGRESSION ANALYSIS APPLICATIONS
Node 3: L
1
= 0, L
2
= 0, L
3
= 1 m = c
With known coefcients a the derivatives with respect to global coordinate axes are de-
termined by
m
x
=
m
L
1
L
1
x
+
m
L
2
L
2
x
+
m
L
3
L
3
x
= a b
1
+b b
2
+c b
3
(C.6)
and
m
x
= a c
1
+b c
2
+c c
3
(C.7)
with b
i
, c
i
Appendix D
Numerical Integration of Elastoplastic
Material Laws
We refer to Section 5.6. The material behavior is described by a set of rst order differ-
ential equations Eqs. (5.75), (5.76), (5.79), (5.80)
C
0
(
p
) = 0

p

= 0

p

H(,
p
) = 0
F(,
p
) = 0
(D.1)
in case of loading. This is a set of 14 equations driven by total strain (t) to determine
14 unknowns (t),
p
(t), (t),
p
(t). This is discretized in time for t = t
n+1
with the
implicit Euler-scheme, i.e.

n+1
=
1
t
(
n+1

n
) =

t
n+1
=
1
t
(
n+1

n
) =

t

p,n+1
=
1
t
(
p,n+1

p,n
) =

p
t

n+1
=
1
t
(
n+1

n
) =

t
(D.2)
and
F
n+1
= F
n
+
F
+
F
p
= 0 (D.3)
leading to a set of algebraic equations
a = C
1
0
+
G
= 0
b = H
p
= 0
f =
F
+
F
p
+F
n
= 0
(D.4)
Herein, is given from a superordinated calculation, the values
n
,
p,n
and F
n
are
assumed to be known
1
and , ,
p
remain to be determined. Depending upon
the characteristics of
G
,
F
,
F
p
, H this set of equations may be nonlinear. Associated
plasticity F = G with F/ = G/ = r is assumed in the following to simplify
the notation. The Newton-Raphson method, see Eq. (1.65), may be used to determine
solutions. It is applied with an iteration rule
_
_
C
1
0
+
r
p
r
H
p
1 H
r +
r
+
r
2
F
2
p
p
+
F
p
0
_
_
i
_
_
_
_
=
_
_
a
b
f
_
_
i
(D.5)
1
A value F
n
< 0 may occur due to a nite discrecization in time.
201
202 APPENDIX D. NUMERICAL INTEGRATION OF ELASTOPLASTIC MATERIAL LAWS
and
_
_
_
_
i+1
=
_
_
_
_
i
+
_
_
_
_
(D.6)
Mises plasticity, see Example 5.3, is considered as an exemplary case. The material
functions are given by
r =
_
3
4J
2
_
_
_
_
_
_
_
_
11
22
33
23
13
12
_
_
_
_
_
_
_
_
,
r
p
= 0,
F
p
= 1,

2
F
2
p
= 0, H = const.,
H
=
H
p
= 0
(D.7)
see Eqs. (5.86), (5.95). Thus, the residuals and the iteration matrix of Eq. (D.5) take a
form
a
i
=
_
_
a
b
f
_
_
i
=
_
_
C
1
0
+r
H
p
r
p
+F
n
_
_
i
, A
i
=
_
_
C
1
0
+
r
0 r
0 1 H
r +
r
1 0
_
_
i
(D.8)
Finally we have C
1
0
derived from Eq. (5.21), and
r
=
3
2
3J
2
_
I
dev

2
3
r r
_
=
3
2
3J
2
I (D.9)
with the deviatoric unit matrix I
dev
according to Eq. (5.17). The matrix
I has the following

properties
I
2
=
I,

Ir = 0,

II = 0,

II
dev
=
I and

I = 0
r
= 0 (D.10)
see [BLM00, (5.9.21) and following clause], with the unit matrix I.
The foregoing relations for the Mises plasticity may be simplied to a large extent, see
[BLM00, 5.9.3]. Eqs. (D.4), (D.8) are reformulated as
_
U
11
U
12
U
21
0
_
i
_
v
_
=
_
w
f
_
i
(D.11)
with
U
11
=
_
C
1
0
+
r
0
0 1
_
, U
12
=
_
r
H
_
, U
21
=
_
r 1
(D.12)
using Eq. (D.11)
6
and
v =
_

p
_
, w =
_
a
b
_
=
_
C
1
0
+r
H
p
_
(D.13)
Eq. (D.11) yields a solution for the plastic multiplier increment
=
U
21
U
1
11
w+f
U
21
U
1
11
U
12
i
(D.14)
203
A rst key point is that U
1
11
in case of Mises plasticity is given by
U
1
11
=
_
C
0
2c
I 0
0 1
_
, c =
2a
1 + 2a
, a =
3
2
p
, =
E
2(1 +)
(D.15)
leading to
U
21
U
1
11
=
_
r C
0
1
, U
21
U
1
11
U
12
= r C
0
r +H = 3 +H (D.16)
using Eqs. (D.10). A second key point is that the stress increment may be generally
determined with
= C
0
(
p
) = C
0
( r) = C
1
+r (D.17)
based on Eqs. (D.1)
1
, (5.76) instead of using the linearized system Eqs. (D.5), (D.6). This
yields w =
_
0 H
p
_
T
and in case of Mises plasticity
U
21
U
1
11
w+f = f +
p
H = F
n
+r H (D.18)
using Eqs. (D.4)
3
, (D.7)
3
and completes the iteration rule Eq. (D.14).
A corresponding iteration algorithm is given as follows. It is performed on an integration
point level, see Page 14.
1. i=0: Given are from a superordinated calculation and
n
, F
n
,
p,n
from the pre-
vious time step. The values of E, , H and shear modulus are material constants.
Initial iteration values are assumed with
0
= 0,
p,0
= 0.
2. Stress is predicted with
i
= C
0
(
p,i
) ,
n+1,i
=
n
+
i
(D.19)
This also yields a value for r
i
.
3. If i = 0 and F
n+1,i
< 0 from Eq. (5.86) an elastic state is given. Upate F
n
with
F
n+1,i
. Finish the iteration.
4. A Newton increment for the plastic multiplier is computed according to Eqs. (D.14),
(D.18), (D.16)
=
F
n
+r
i

i
H
i
3 +H
(D.20)
5. If [[ < Tol a convergent state has been reached within a prescribed tolerance Tol.
Finish the iteration and update internal state variable by going to 8.
6. Update the plastic multiplier increment and plastic strains
i+1
=
i
+,
p,i+1
=
i+1
_
r
i,1
r
i,2
r
i,3
2r
i,4
2r
i,5
2r
i,6
_
T
(D.21)
7. i := i + 1. Continue with step 2.
8. Update internal state variable
p,n+1
=
_
3J
2,n+1
according to Eq. (5.86) using a
converged stress
n+1
. Update F
n
with 0. Compute tangential material stiffness
with Eq. (5.88) for a consistent systems tangential stiffness.
9. Proceed with next equilibrium iteration step of superordinated calculation.
Appendix E
ACCESS lectures in summer term 2013
Preliminary overview
1. Tue 09.04.2013 Lecture: Reinforced Concrete Tension Bars
2. Mon 15.04.2013 Exercise:
3. Tue 16.04.2013 Lecture: Finite Element ingredients for RC tension bars
4. Tue 23.04.2013 Lecture:
6. Tue 30.04.2013 Lecture:
7. Tue 07.05.2013 Lecture:
9. Tue 14.05.2013 Lecture:
10. Mon 27.05.2013 Exercise:
11. Tue 28.05.2013 Lecture:
12. Tue 04.06.2013 Lecture:
13. Mon 10.06.2013 Exercise:
14. Tue 11.06.2013 Lecture:
15. Tue 18.06.2013 Lecture:
16. Mon 24.06.2013 Exercise:
17. Tue 25.06.2013 Lecture:
18. Tue 02.07.2013 Lecture:
19. Mon 08.07.2013 Exercise:
20. Tue 09.07.2013 Lecture:
21. Tue 16.07.2013 Lecture:
204
Bibliography
[BLM00] T. Belytschko, W.K. Liu, and B. Moran. Nonlinear Finite Elements for Continua
and Structures. John Wiley & Sons, Chichester, 2000.
[BSMM00] Bronstein, Semendjajew, Musiol, and Muehlig. Taschenbuch der Mathematik. Ver-
lag Harri Deutsch, Frankfurt/Main, 5. auage edition, 2000.
205

Computational Methods in Reinforced Concrete

Enviado por

Dados do documento

Título original

Direitos autorais

Formatos disponíveis

Compartilhar este documento

Compartilhar ou incorporar documento

Opções de compartilhamento

Você considera este documento útil?

Este conteúdo é inapropriado?

Direitos autorais:

Formatos disponíveis

Computational Methods in Reinforced Concrete

Enviado por

Direitos autorais:

Formatos disponíveis

Computational Engineering for Reinforced Concrete

t and arbitrary virtual displacements u.

) according to Fig. 2.2a.

and material functions E

(). The parameter

= 1. Using Eqns. (2.8),

(0) = 0. Eq. (2.12) describes a Kelvin-Voigt element with a

, , may be determined based on experiments. This results in

= 100 [d] for = 0.5, i.e. half of

for any selection of L

(s), see Fig. 2.11b.

is used for abbreviation.

according to Eqs. (3.28,3.29). To simplify A

. The same holds for the derivatives. Thus, we have

. These test functions have to be kinematically

. Combining Eq. (3.53) leads to

. The same holds for the derivatives and it is set

, i.e. deformation results from

and and w are decoupled by the shear

remain as variables. The latter

= 100 [d] for

from beam deformation,

, and for the states following the xing of tendons,

, using Eqs. (3.133), (3.135) with the appropriate deformations or displace-

exist for the maximiza-

for the minimization. Both have the same optimal value

of the associated optimization problems are also

in the transformed system. With

the material stiffness

has the the same ma-

according to Eq. (5.12). With

-axis in the bar direction.

generally may have different values. But

according to Eq. (7.73) is computed throughout the slab and compared to

I has the following

Você também pode gostar