Professional Documents
Culture Documents
Three-Dimensional Beam Element Based On A Cross-Sectional Coordinate System Approach
Three-Dimensional Beam Element Based On A Cross-Sectional Coordinate System Approach
Three-Dimensional Beam Element Based On A Cross-Sectional Coordinate System Approach
, (11)
where is the Euclidian norm of the vector. In Equation (11), the vector r
x
is dened on the beam
centerline. A unit length vector, b
y
t
, perpendicular to vectors t
y
t
and r
y
, is dened using the vector
product as follows
b
y
t
=
t
y
t
r
y
_
_
t
y
t
r
y
_
_
, (12)
The third vector to complete the tangential triad can be written as follows
n
y
t
= b
y
t
t
y
t
. (13)
Index y indicates the denition by vector r
y
and the triad independence of vector r
z
. Because vector b
y
t
is
derived using vector r
y
, the tangential triad above is always perpendicular to vector r
y
but not necessarily
to vector r
z
. The tangential frame can also be dened using vector r
z
. This frame is independent of
vector r
y
and is dened as follows
t
z
t
= t
y
t
, n
z
t
=
r
z
t
z
t
_
_
r
z
t
z
t
_
_
, b
z
t
= t
z
t
n
z
t
. (14)
The tangential frames are dened on the centerline of the beam and are depicted geometrically in
Figure 3 where vectors r
y
and r
z
are not on the normal plane. Two orthogonal rotation matrices A
y
t
and
Three-Dimensional Beam Element 317
Figure 3. Geometric interpretation of the tangent frames.
A
z
t
can be created from the tangent frames as follows [13]
A
y
t
=
_
t
y
t
n
y
t
b
y
t
_
, (15a)
A
z
t
=
_
t
z
t
n
z
t
b
z
t
_
. (15b)
If r
y
is perpendicular to the r
z
vector, then both tangent frames are equal. This is valid only in the special
case when the cross-section is not distorted.
2.4. SHEAR DEFORMATION
Rotation matrices, R
=
_
I + v sin
+ 2( v)
2
sin
2
2
_
, (16)
where v is a skew symmetric matrix dened using vector v that denes the rotation axis, and
is
the rotation angle. The shear angles,
1, as a result of which
the second-order term in Equation (16) can be neglected. In this case, the rotation matrices for shear
deformations can be written as follows
R
y
=
1 0 sin
y
0 1 0
sin
y
0 1
, R
z
=
1 sin
z
0
sin
z
1 0
0 0 1
. (17)
Twoshear angles are showninFigure 4usingprojections of the beamto XZ and XY planes. The direction
of the positive rotation is claried using arrows around the principal axes Y and Z. Components of the
318 K. E. Dufva et al.
Figure 4. Denition of shear angles.
rotation matrices R
T
EdV
0
, (21)
where is the second PiolaKirchhoff stress vector and E is the Green strain vector including six strain
components, E = [
xx
yy
zz
2
xy
2
xz
2
yz
]. In Equation (21), integration is carried out over
an initial volume V
0
. The Green strain tensor can be dened in terms of the displacement gradient,
D,
as follows
m
=
1
2
_
D
T
+
D +
D
T
D
_
. (22)
When the beam is initially orientated along the global X-axis without initial stress, the displacement
gradient
D can be written as
D =
u
x
, (23)
320 K. E. Dufva et al.
where x is the vector of spatial coordinates, x, y, and z. In the proposed element, rotation matrices A
y,z
t
are dened at the beam centerline where vectors r
y
and r
z
are functions of x only. Since cross-section
rotation is a function of the x coordinate only, the rotation angle is assumed to have a constant value
over the cross-section and the resulting shear strain is constant. For this reason, a shear correction factor,
k, needs to be used to compensate unnatural strain distribution [16].
Vectors r
y
and r
z
are not necessarily perpendicular since distortion of the cross-section is kinemati-
cally admissible. Note that these vectors may also change their length. For this reason, strain components
yz
,
yy
, and
zz
cannot be neglected. For the sake of simplicity these strain components are approx-
imated using linear interpolation without considerable loss of accuracy. Strain component
yy
at the
nodal point can be approximated as follows [8]
n
yy
=
1
2
_
r
n
T
y
r
n
y
1
_
. (24)
where n is the node number. It is important to note that when linear interpolation is used, the contribu-
tion of these strain components to strain energy is small. For this reason, different approximation for
transverse strain components can be used as follows [10]
n
yy
=
_
_
r
n
y
_
_
1. (25)
It is noteworthy that all numerical examples introduced in this paper were tested with both approxima-
tions, Equations (24) and (25), and no differences in displacements were perceived. The strain
n
yy
can
now be interpolated over the element length as follows
yy
=
_
1
x
l
_
1
yy
+
x
l
2
yy
= N
1
1
yy
+ N
2
2
yy
. (26)
Similarly the strain components
zz
and
yz
can be dened as follows
zz
= N
1
1
zz
+ N
2
2
zz
, (27)
yz
= N
1
_
r
1
T
y
r
1
z
_
+ N
2
_
r
2
T
y
r
2
z
_
= N
1
1
yz
+ N
2
2
yz
. (28)
It is important to note that using approximations in Equations (26)(28) does not affect the ob-
jectivity of the strains since they are calculated at the nodal points according to the Green strain
denition.
The second PiolaKirchhoff stress tensor can be written as = CEwhere Cis the constitutive matrix.
As pointed out in [5], coupling between the axial and normal stress will lead to residual transverse normal
stresses in bending as a result of which overly stiff behavior of the element occurs. In order to avoid this
phenomenon, Poissons effect is neglected using a diagonal constitutive matrix [1]. However, the cross-
section can still deform under an appropriate loading. For an isotropic material, the Young modulus
E and shear modulus G are the same for all directions. This leads to a similar form of strain energy
function, U, as used by e.g., Danielson and Hodges [17] and Sharf [18] as follows
U =
1
2
_
V
0
_
E
_
2
xx
+
2
yy
+
2
zz
_
+ 4kG
_
2
xy
+
2
xz
+
2
yz
__
dV
0
. (29)
Three-Dimensional Beam Element 321
The beam elastic forces can be obtained as follows [7]
Q =
U
e
. (30)
4. Beam Inertia
In a continuum mechanics approach, the beam cross-section has been dened as a deformable eld
while no coordinate system has been attached to dene the cross-section. As pointed out by Yakoub
and Shabana [8], when a cross-section coordinate system is used, the velocity vector can be written as
r = S
0
e
0
+ [S
s
e
s
], (31)
where vector is the cross-section angular velocity vector and S
0
is the shape function matrix that
contains shape functions dened at the element centerline as being functions of the x coordinate only, e
0
is a vector that contains velocities of the nodal coordinates related to the shape function matrix S
0
. These
nodal coordinates are global positions and gradients with respect to the x coordinate of the nodes. In
Equation (31), matrix S
s
includes those shape functions where y and z coordinates are also included, and
vector e
s
is the nodal coordinate vector related to matrix S
s
. Using the continuum mechanics approach,
the velocity vector can be dened as follows [8]
r = S
0
e
0
+ S
s
e
s
. (32)
Comparing Equations (31) and (32) it can be seen that if the cross-section coordinate system is used,
the velocity vector is more complex. This will lead to a more complex expression of the kinetic energy
and nonlinear mass matrix [8]. In practice, Equations (31) and (32) lead to similar results and in the
proposed element the velocity description is dened as follows
r = S e. (33)
When substituting this expression to the expression of kinetic energy, one can obtain the constant mass
matrix as follows [8]
M =
_
V
0
S
T
SdV
0
. (34)
5. Numerical Examples
Element formulation and equations of motion are implemented in the mathematical software Matlab.
Numerical integration is used to obtain the vector of elastic forces and due to the complex expres-
sion of elastic forces the possibility to use C language is utilized in element implementation in
Matlab.
322 K. E. Dufva et al.
5.1. FLEXIBLE PENDULUM
In the rst example, the dynamics of a simple beam structure is studied. In this structure, a beam with a
rectangular cross-section is hinged with a pin joint. The height and width of the beam are both assumed
to be 0.05 m. The beam has a length of 1.2 m, the second moment of the area of 5.2083 10
7
m
4
,
Youngs modulus E of 0.710
6
N/m
2
, the shear modulus G of 2.692310
5
N/m
2
, and a mass density of
5540 kg/m
3
. Initially, the beamis orientedalongthe global X-axis assumingthe initial velocitytobe zero.
The gravity constant is 9.81 m/s
2
. A freefalling exible pendulum under an evenly distributed gravity
force is a conservative system in which the total energy must remain constant during the simulation.
The energy sum for the whole system is as follows
n
i
(T
i
+U
i
+ V
i
) = const. (35)
where T
i
is the element kinetic energy, U
i
is the strain energy, V
i
is the potential energy of the element,
and n is the number of elements [19]. Figure 5 depicts the energy distribution between the different
components as a function of time. It can be seen that there is no energy loss during time integration.
Energy balance is studied using a four-element model.
The vertical displacement of the end tip of the beam is studied using different numbers of elements.
Figure 6 shows the deviation of displacements using 2-, 4-, and 8-element models compared to a model
with 16 elements. It can be concluded fromFigure 6 that the proposed element shows a good convergence
rate in this example. The displacement obtained when 2 and 16 elements are used is depicted in
Figure 7.
Figure 5. The energy balance of the pendulum. , T is the kinetic energy; - - -, U the strain energy; and , V the potential
energy of the system. Sum of the energy is marked as solid line .
Three-Dimensional Beam Element 323
Figure 6. Deviation of displacements at vertical direction of, , 2-element; - - -, 4-element; and , 8-element models from
16-element model.
Figure 7. The vertical displacement of the end tip of the pendulum. , 2 elements, and , 16 elements.
5.2. STATIC ANALYSES
In order to ensure element behavior in linear and nonlinear static problems, various analyses with a
cantilever beam are carried out. The material model of the beam is considered to remain within a linear
324 K. E. Dufva et al.
Table 2. Parameters for analyses of the cantilever beam.
Width, w [m] Height, h [m] Length, l [m] S
L
Tip load [N]
Model I 0.1 0.5 2 74 20 10
6
h
3
Model II 0.01 0.01 2 1.8 10
5
1.5 EI /l
2
Table 3. The deformed position of the beam end tip, Model I.
Number of elements Tip position (X, Y) [m]
1 1.999 0.0324
4 1.999 0.0324
Analytical 2.000 0.0324
elastic range during deformation. The centerline of the beam can rotate during the deformation, and the
position vector gradient r/x in the longitudinal direction is set to be free. In these examples, Youngs
modulus is 2.07 10
11
N/m
2
and the shear modulus is 7.9615 10
10
N/m
2
. The shear correction factor
k for the rectangular cross-section is 5/6. The geometrical data, slenderness ratio S
L
, and loads for
analyses are given in Table 2. The ratio S
L
is calculated as GAl
2
/EI. If the slenderness value is high
(S
L
> 1000), the EulerBernoulli beam theory is assumed valid since shear deformation does not play
a signicant role.
5.2.1. Cantilever with Tip Load
In the rst case, the cantilever beam is subjected to a vertical tip load. The results of Model I are
presented in Table 3 where the positions of the end tip of the beam are shown for a different number of
elements. As can be seen in Table 3, the results match the analytical results when only one element is
used.
In the second case of the cantilever beam, large deections are considered. The results are compared
to an analytical solution from Gere and Timoshenko [16] as well as to a previously proposed absolute
nodal coordinate beam element [8]. The analytical solution is based on the differential equation for an
exact deection curve. In this example, the applied force is obtained from the ratio Fl
2
/EI that is set to
1.5. The results of the analysis are shown in Table 1. Using the proposed element, the analytical result is
achieved in this example when 16 elements are used. In this example, the effect of linear interpolation of
shear deformation is also studied. In Table 1 the model where linearization is not used is marked with an
asterisk. It can be concluded from Table 1 that improved convergence is obtained using a linearization
of the shear deformation.
5.2.2. Semicircle Bending Test
In order to ensure behavior of the element in highly nonlinear cases, the moment is applied to the end of
the cantilever beam. The applied moment is capable of bending the beam into a semicircle. The results
are compared to solutions obtained from the element proposed by Yakoub and Shabana [8]. The beam
is modeled using three and six elements. The length l of the cantilever structure is 1 m, the height and
the width of the cross-section are 0.035 m. Youngs modulus is 2.0 10
8
N/m
2
and Poissons ratio is 0.
The applied external moment at the end of the beam is EI /l. Deformations of different models are
Three-Dimensional Beam Element 325
Figure 8. Deformation of the cantilever subjected to end tip moment load.
shown in Figure 8 after one load step, = 1. Using the proposed element, the semicircle is obtained
with three elements. The formulation proposed by Yakoub and Shabana does not reach a semicircle
even if six elements are used, which implies overly stiff element responses under pure bending. The use
of 16 elements gives approximately a full semicircle when the formulation proposed by Yakoub and
Shabana is used.
5.3. 180
TWISTED BEAM
In this test the cantilever beam is subjected to pure torque in order to study element capability to
capture torsional behavior. The structure is twisted 180