You are on page 1of 24

Paper: ASAT-13-ST-11

13
th
International Conference on
AEROSPACE SCIENCES & AVIATION TECHNOLOGY,
ASAT- 13, May 26 28, 2009, E-Mail: asat@mtc.edu.eg
Military Technical College, Kobry Elkobbah, Cairo, Egypt
Tel : +(202) 24025292 24036138, Fax: +(202) 22621908



1/24

Linear and Nonlinear Finite Element Modeling
of Advanced Isotropic and Anisotropic Beams
Part I: Euler Bernoulli Theory

M. K. Abass
*
M. A. Elshafei
*


Abstract: In the present work, a linear and nonlinear finite element modeling for both
isotropic and anisotropic advanced Euler beams is presented. A hermit cubic and linear shape
functions are used to represent the beam deformations in the modeling. The principle of
virtual work is used to formulate the equilibrium equations for both models. A MATLAB
code is developed to analyze both models linear and nonlinear of steel beam and
Carbon/Epoxy laminated composite beam. A ready-made function is used to solve the
nonlinear system of equilibrium equations. The Steel and the composite beams are tested in
the laboratory to measure the deflections, which compared with the proposed finite element
model. For further verification, a COSMOS software program is used. This comparison
highlighted the need of including the shear strain in the geometric nonlinear analysis for the
case of large deformation.

Keywords: Nonlinear Finite Element Analysis, laminated composite structure, Euler beam
theory, static testing of beams.


1. Introduction
Nonlinearity naturally arises in the true. Based on the assumptions of smallness of certain
quantities of the formulation, the problem may be reduced to a linear problem. Linear
solutions may be obtained with considerable ease and less computational cost when compared
to nonlinear solutions. In many instances, assumptions of linearity lead to reasonable
idealization of the behavior of the system. However, in some cases assumption of linearity
may result in an unrealistic approximation of the response. The type of analysis, linear or
nonlinear, depends on the goal of the analysis and errors in the system's response that may be
tolerated. In some cases, nonlinear analysis is the only option left for the analyst as well as the
designer.

There are two common sources of nonlinearity: geometric nonlinearity and material
nonlinearity. The geometric nonlinearity arises purely from geometric consideration (e.g.
nonlinear strain-displacement relations), and the material nonlinearity is due to nonlinear
constitutive behavior of the material of the system. A third type of nonlinearity may arise due
to changing initial or boundary conditions.



* Egyptian Armed Forces

Paper: ASAT-13-ST-11



2/24

A numerical model for layered composite structures based on a geometrical nonlinear shell
theory was developed on 1993 by F. Guttmann, et.al, [1]. In thin-walled open-section beams
made of fiber-reinforced laminates, at which the bending and torsion are coupled, a non-linear
finite-element (NLFE) analysis based on the updated lagrangian formulation is developed to
solve the problem numerically by B. Omidvar A. Ghorbanpoor on 1996, [2]. 0n 2006 R
Murali et.al investigated systematically the discretisation errors that appear uniquely in a non-
linear beam formulation due to the presence of non-linear derivative terms in the membrane
strain term. These appear in the form of degraded performance (locking), [3]. On 2007 P.
Krawczyk, et.al developed a layer-wise beam model for geometric nonlinear finite element
analysis of laminated beams with partial layer interaction. The model is built assuming first
order shear deformation theory (FSDT) at layer level and moderate interlayer slips, [4]. On
2008 Li Jun, et.al developed the exact dynamic stiffness matrix of a uniform laminated
composite beam based on trigonometric shear deformation theory, [5]. Maxwell Blair, et.al
used Euler-Bernoulli beam mechanics to drive a Finite element beam formulations with
geometric non-linear mechanics including geometric bend-twist coupling in the context of
large deformations and follower forces, [6]. On 2009 Yagci, Baris; Filiz Sinan, et.al presented
a spectral-Tchebychev technique for solving linear and nonlinear beam problems. The
technique uses Tchebychev polynomials as spatial basis functions, and applies Galerkin's
method to obtain the spatially discretized equations of motion, [7]. J. N. Reddy deduced a
nonlinear formulation of straight isotropic beam using Euler Bernoulli beam theory and
Timoshenko beam theory to formulate the kinematic behavior of the beam. The princible of
virtual displacement was used to formulate the equlibrium equations, [8].

Concerning the geometric nonlinearity of the classical beam theory, this paper presents a
simple finite element model able to estimate the large deformations, axial, transverse, and
rotation angle due to distributed and concentrated general loading. The model is able to solve
laminated composite straight beams with different fiber orientation angles of arbitrary number
of layers. The modeled beam has variable cross-section and variable material property along
its longitudinal axis. Linear anisotropic model is also deduced. By applying simplified
assumptions on the derived equations, the linear and nonlinear isotropic models are obtained.

Two beams are introduced as an example of isotropic and anisotropic beams. The two beams
are statically loaded and the deflections are measured in the lab. A MATLAB code is
developed to formulate and solve the system of equilibrium equations to get the required
deflections. The numerical results are compared with the measured data. For further
verification, a COSMOS software program is used.


2. Theoretical Formulation
There are four different theories to model the kinematics behavior of beams; Euler-Bernoulli
beam theory (EBT) that neglects the transverse shear strain, Timoshenko beam theory (TBT),
which accounts for the transverse shear strain in the simplest way, the second-order beam
theory (SOBT), and the third-order beam theory (HOBT), which add additional terms into the
assumed displacement field [9].

The general form of the assumed displacement field is expressed as, [9];

Paper: ASAT-13-ST-11



3/24

3
2 0 0
0 0 1 2 3
0
( , ) ( ) ( ) ( ) ( )
( , ) 0
( , ) ( )
x x x
dw dw z
u x z u x z C C x C z x C x
dx h dx
v x z
w x z w x

( ( | |
= + + + + +
|
( (
\ .
=
=



Where: , , u v w are the displacements along coordinate directions, longitudinal, lateral, and
transverse, (x,y,z) respectively.
0 0
( ), ( ) u x w x denote the displacement of a point (x, y, 0) on
the mid plane of an undeformed beam along the axial (x) and the transverse (z) directions,
respectively. ( )
x
x and ( )
x
x are functions of x. h is the beam thickness along z-axis.

In the First-Order Beam Theory, FOBT, which known as Timoshenko Beam theory, the
constants are;
0
0 c = ,
1
1 c = ,
2
0 c = ,
3
0 c = , by substituting in Eq. (1), the assumed
displacement field is, [9];

0
0
( , ) ( ) ( )
( , ) 0
( , ) ( )
x
u x z u x z x
v x z
w x z w x
= +
=
=


In the Second-Order Beam Theory (SOBT), the constants are taken as;
0
0 c = ,
1
1 c = ,
2
1 c = ,
3
0 c = , from Eq. (1), the assumed displacement field takes the form, [9];

2
0
0
( , ) ( ) ( ) ( )
( , ) 0
( , )
x x
u x z u x z x z x
v x z
w x z w x
= + +
=
=





In the Higher-Order Beam Theory (HOBT) ,third-order, the constants are taken as;
0
0 c = ,
1
1 c = ,
2
0 c = ,
3
4
( )
3
c h = . By substituting in Eq. (1), the assumed displacement field is,
[9];

3 0
0
0
4
( , ) ( ) ( ) ( )
3
( , ) 0
( , ) ( )
x x
dw z
u x z u x z h x
h dx
v x z
w x z w x

(
= + +
(

=
=



Our concern in this study is geometric nonlinearity of the classical beam theory which is
based on the Euler-Bernoulli hypothesis that plane sections perpendicular to the mid-plane of
the beam before deformation remain plane, rigid (not deform), and rotate such that they
remain perpendicular to the (deformed) mid-plane after deformation, as shown in Fig. 1. The
assumptions lead to neglect the Poisson effect and transverse shear strains.
Paper: ASAT-13-ST-11



4/24


By substituting the following constants,
0
1 c = ,
1
0 c = ,
2
0 c = ,
3
0 c = , in the general form
of displacement field, Eq. (1) ,The assumed displacement field, of EBT, takes the form, [9];


Fig. 1. Kinematics of a beam in the Euler Bernoulli Theory

0
0
0
( , ) ( )
( , ) 0
( , ) ( )
dw
u x z u x z
dx
v x z
w x z w x
=
=
=




3. Strain-Displacement Relations
The general form of the strain-displacement relations in contracted notation can be expressed
as follows, [8]:

1 1
. , i, j, m=1,2,3
2 2
j
i m m
ij
j i i j
u
u u u
x x x x

| | | | c
c c c
= + + | |
| |
c c c c
\ . \ .


where the deformations
1 2 3
, , u u u are , , u v w respectively, the directions
1 2 3
, , x x x are , , x y z
respectively, and strain components
11
,
22
,
33
,
13
,
23
,
12
are
xx
,
yy
,
zz
,
xz
,
yz
,
xy
respectively.

For the case of the beam length is too long in comparison with the other two dimensions, both
yy
and
xy
are assumed equal to zero [12]. According to Kirchhoff's hypothesis, the strain
components
zz
,
xz
, and
yz
are vanish. By applying these assumptions on Eq. (6), the only
nonzero strain is the axial strain,
xx
.

2 2 2
1 1 1
2 2 2
xx
u u v w
x x x x

c c c c | | | | | |
= + + +
| | |
c c c c
\ . \ . \ .


0
dw
dx

0
dw
dx

w
0

u
0

z
x
n
m
n'
m'
Paper: ASAT-13-ST-11



5/24

Omitting large strain terms except the square of
0
dw
dx
, the axial strain can be written as:

2
2
0 0 0
2
1
2
xx
du d w dw
z
dx dx dx

| |
= +
|
\ .



where
2
0
1
2
dw
dx
| |
|
\ .
is the nonlinear term. Eq. (8) can be written as;

0 1
xx xx xx
z = +



where
0
xx
, is strain of mid-plane,
1
xx
is the mid-plane curvatures in the x direction.

2
2
0 1 0 0 0
2
1
,
2
xx xx
du dw d w
dx dx dx

(
| |
= + =
(
|
\ .
(




4. Stress-Strain Relations
For a linear elastic behavior of isotropic materials, Hokes' law can be written as, [14]:

xx xx
E =



and for anisotropic materials this relation takes the form, [10]:

i, j = 1,2,3,...6
i ij j
Q =

where
i
and
j
are the stress and strain components,
ij
Q are the components of the lamina
transformed stiffness matrix.
Thus for our case,
xx
is the only non zero stress, [10].

11 xx xx
Q =


5. Variational Formulation
The principle of virtual work will be used to formulate the equilibrium equations, [13].

0
I E
W W W = =



Where
I
W is the internal virtual work, i.e. virtual strain energy stored in the beam due to
actual stress, and
E
W is the work done by external applied load, [13].

Paper: ASAT-13-ST-11



6/24

I ij ij
v
W dv =
}



where v is beam volume.

0 0 0 0 0
0 0
( ) ( ) ( ) ( ) ( ) ( ) ( )
L L
E N P M
w q x w x dx f x u x dx N u x P w x M x = + + + +
} }



where ( ) q x is the distributed transverse load, ( ) f x is the distributed axial load, , , N P M
are concentrated axial, transverse and bending moment load respectively, as shown in Fig. 2.
0 0 0
( ), ( ), ( ) u x w x x are the virtual axial, transverse, and rotational displacements
respectively, and
0
0
( )
( )
d w x
x
dx

| |
=
|
\ .
.



Fig. 2. Beam under general loading

By substituting the stress and strain values, Eq (13), Eq. (9), and performing integration of the
internal virtual work, Eq. (15);

0
L
I xx xx
A
W dAdx =
} }



( )
0 1
0
L
I xx xx xx
A
W z dAdx = +
} }



( )
0 1
11
0
L
I xx xx xx
A
W z Q dAdx = +
} }



N
,
0
( )
N
u x

z
h
b
M
,
0
( )
M
x

q(x),
0
( ) w x

x
1q

x
z
P
,
0
( )
P
w x

L
x
P

x
M

x
2q

x
2f

x
1f

x
N

f(x),
0
( ) u x


y
Paper: ASAT-13-ST-11



7/24

( ) ( )
0 1 0 1
11
0
L
I xx xx xx xx
A
W z Q z dAdx = + +
} }



( ) ( ) ( )
0 0 0 1 1 0 2 1 1
11
0
L
I xx xx xx xx xx xx xx xx
A
W Q z z dAdx
(
= + + +

} }



( ) ( ) ( )
0 0 0 1 1 0 1 1
11 11 11
0
L
I xx xx xx xx xx xx xx xx
W A B D dx
(
= + + +

}



where
( )
2
11 11 11 11
( , , ) (1, , )
k
A
A B D Q z z dA =
}

where
ij
A are the components of extensional stiffness matrix,
ij
D are the components of
bending stiffness matrix,
ij
B are the components of coupling stiffness of the laminate, and
( )
ij k
Q are the components of the transformed stiffness matrix of the k
th
lamina, [10].

2 2
11 11 11 1
1
2 2
( ) ( ) ( )
b h
N
k k k K
k h b
A Q dydz b Q z z

=
= =

} }



2 2
2 2
11 11 11 1
1
2 2
1
( ) ( ) ( )
2
b h
N
k k k k
k h b
B Q zdydz b Q z z

=
= =

} }



2 2
2 3 3
11 11 11 1
1
2 2
1
( ) ( ) ( )
3
b h
N
k k k k
k h b
D Q z dydz b Q z z

=
= =

} }



By taking the variation of the strain components, Eq. (10);

2
0 1 0 0 0 0
2
,
xx xx
d u dw d w d w
dx dx dx dx


(
= + =
(



Substituting in the internal virtual work statement, Eq. (22), we can get;

2 2
0 0 0 0 0 0 0
11
2
2 2 2
0 0 0 0 0 0 0 0
11 2 2 2
2 2
0 0
11 2
1 1
2 2
1
2
I
d u du dw d w dw du dw
A
dx dx dx dx dx dx dx
d u d w d w dw d w d w du dw
W B
dx dx dx dx dx dx dx dx
d w d w
D
dx dx

( | | | |
| | | |
+ + + ( | |
| |
| |
\ . \ .
(
\ . \ .
( | |
| | | |
| |
= + + + ( |
| | |
|
\ . ( \ . \ .
\ .
+
0
2
L
dx
(
(
(
(
(
(
(
(
( | |
(
( |
(
\ .
(

}


Paper: ASAT-13-ST-11



8/24

6. Finite Element Formulation
The displacement of a beam subjected to axial stretching is given by, [15]:

0
2
2
=
c
c
x
u



By solving the above governing equation and applying boundary conditions:

| |{ }
2
0 1 2
1
( )
j j
j
u x u u
=
= =




{ } | |
1 2 1 2
1 , ,
T x x
u u u
L L
= = =

The governing equation for a beam under pure bending is, [16]:

4
4
0
w
x
c
=
c



By solving the above equation and imposing the nodal boundary conditions yields to, [17];

| |{ }
4
0 1 2 3 4
1
( )
J J
J
w x
=
= A = A



where { } | | | |
1 2 3 4 1 1 2 2
T
w w A = A A A A = , and the shape functions take the form;

2 3 2 3
1 2 2 3 2
2 3 2 3
3 4 2 3 2
3 2 2
1 ,
3 2
,
x x x x
x
L L L L
x x x x
L L L L


= + = +
= = +



The element nodal displacements vector is { } d .

{ } | |
1 1 2 2 3 4
T
d u u = A A A A




6.1. Nonlinear stiffness matrix of anisotropic materials

By taking the variation of Eq. (29) and Eq. (32), then substituting the shape functions into Eq.
(27), we can get the following;

Paper: ASAT-13-ST-11



9/24

2 4
0
1 1
11
2 4
0 0
1 1
2 2 4 4
0
2 2
1 1
11
2
1
.
2
1
. . .
2
.
j
J i
i j J
j J
j
J I
I j J
j J
J J i I
i J I J
J J
I
I
I
d
dw d d
u u
dx dx dx dx
A
d
dw dw d d
u
dx dx dx dx dx
d dw d d d
u
dx dx dx dx dx
W B
d
dx

= =
= =
= =
(
+ A
(

`
(

+ A + A
(

)
( (
A + A A
( (

=
+ A



2 4
0
0
2
1 1
2 2 4
11 2 2
1
1
. .
2
.
L
j
J
j J
j J
J I
I J
J
dx
d
dw d
u
dx dx dx
d d
D
dx dx

= =
=












``
(

+ A
(

)



+ A A
`
)



)
}




where i,j=1,2 and I,J=1,2,3,4 from now on.

2
11
1
0
2 4
0
11 11 2
1
0 0
2 2
0
11 11 2
1
0 0
2
0
11
. .
1
. . . .
2
. . . .
1
2
L
j
i
j
j
i
L L
J J i i
J
J
L L
j j
I I
I
j
j
i
d
d
A dx u
dx dx
u
d dw d d d
A dx B dx
dx dx dx dx dx
d d
dw d d
W
A dx B dx u
dx dx dx dx dx
dw
A
dx

=
=
=
(
(

`
(

+ A
(

)
(
=

(

| |
A
|
\ .
+

} }

} }
2
11 2
4
0 0
2 2 2
1
0
11 11 2 2 2
0 0
. . .
1
. . . .
2
L L
J J I I
J
L L
J
J J I I
d d d d
dx B dx
dx dx dx dx
dw d d d d
B dx D dx
dx dx dx dx dx

=










`


(


` (

(

A
(

(
+

(
) )
} }

} }



The internal virtual work can be expressed in matrix form as:

{ } { } | |
{ }
{ }
2
6 6
T T
I
X
u
W u K
1
41


(
= A
`

A

)


Where the stiffness matrix
| |
11 12
2 2 2 4
21 22
4 2 4 4
K K
K
K K


( ( (

(
=
(
( (





and its coefficients are,

Paper: ASAT-13-ST-11



10/24

11
11
0
2
12 0
11 11 2
0 0
2
21 0
11 11 2
0 0
2
2
22 0
11 11
0
. .
1
. . . .
2
. . . .
1
. .
2
L
j
i
ij
L L
J J i i
iJ
L L
j j
I I
Ij
L
J I I
IJ
d
d
K A dx
dx dx
dw d d d d
K A dx B dx
dx dx dx dx dx
d d
dw d d
K A dx B dx
dx dx dx dx dx
dw d d d d
K A dx B
dx dx dx dx





=
| |
=
|
\ .
| |
=
|
\ .
| |
=
|
\ .
}
} }
} }
} 2
0
2 2 2
0
11 11 2 2 2
0 0
.
1
. . . .
2
L
J
L L
J J I I
dx
dx
dw d d d d
B dx D dx
dx dx dx dx dx

+
}
} }



Its clear that if we substitute the value of
0
( ) w x as a function of the shape functions, it will
arise a square term in
j
u and
j
A , which leads to form a nonlinear, second order, system of
algebraic equations. Thus to solve this system we should make iteration by assuming an initial
value of
0
( ) w x , then calculate the corresponding stiffness matrix | | K , Eq.(39). Then, get
the corrected value of
0
( ) w x from the calculated element nodal displacements { } d , Eq.(32).
Repeat the iteration process and substitute the new value of
0
( ) w x in the stiffness matrix and
calculate the nodal displacement until the system of equations converge and reach the
required accuracy.

Note that the stiffness matrix is asymmetric matrix;
12 21
T
K K ( ( =

. As marked above,
12
K (

contain the factor
1
2
, where
21
K (

does not, Eq.(39). To make it symmetric
i.e.
12 21
T
K K ( ( =

, split the linear strain
0
du
dx
, in Eq.(27), into two equal parts, as marked in
the following, [8];

2 2
0 0 0 0 0 0 0 0
11
2
2 2 2
0 0 0 0 0 0 0 0
11 2 2 2
2
11
1 1 1 1
2 2 2 2
1
2
I
d u du dw d w dw du du dw
A
dx dx dx dx dx dx dx dx
d u d w d w dw d w d w du dw
W B
dx dx dx dx dx dx dx dx
d w
D

( | | | |
| | | |
+ + + + ( | |
| |
| |
\ . \ .
(
\ . \ .
( | |
| | | |
| |
= + + + ( |
| | |
|
\ .
( \ . \ .
\ .
+
0
2
0 0
2 2
L
dx
d w
dx dx
(
(
(
(
(
(
(
(
( | |
(
( |
(
\ .
(

}


Paper: ASAT-13-ST-11



11/24

2 2
0 0 0 0 0 0 0 0 0 0
11
2
2 2 2
0 0 0 0 0 0 0 0
11 2 2 2
1 1 1
2 2 2
1
2
I
d u du dw d w dw du d w dw du dw
A
dx dx dx dx dx dx dx dx dx dx
d u d w d w dw d w d w du dw
W B
dx dx dx dx dx dx dx dx

( | | | |
| | | |
+ + + + ( | |
| |
| |
\ . \ .
(
\ . \ .
( | |
| | | |
| |
= + + + ( |
| | |
|
\ . \ . \ .
\ .
0
2 2
0 0
11 2 2
L
dx
d w d w
D
dx dx

(
(
(
(
(
(
( (
(
( | |
(
+
( |
(
\ .
(

}



By substituting the shape functions in the internal virtual work:

2 4 2
0 0
1 1 1
11
2
4
0 0
1
2 4 4
0
2
1 1
11
1 1
. .
2 2
1
.
2
.
j j
J i I
i j J I j
j J j
J I
I J
J
J i I
i J I J
J J
I
d d
dw d dw d d
u u u
dx dx dx dx dx dx dx
A
d du dw d
dx dx dx dx
d dw d d
u
dx dx dx dx
W B



= = =
=
= =
(
+ A + A
(

`
(
| |
+ A A +
(
|

\ .
(
)
(
A + A A
(

=

2
2
2 2 4
0
0
2
1 1
2 2 4
11 2 2
1
1
. .
2
.
J
L
j
J I
I j J
j J
J I
I J
J
d
dx
dx
d
dw d d
u
dx dx dx dx
d d
D
dx dx

= =
=








(

(


` `
(

+ A + A
(

)



+ A A
`
)




)




The internal virtual work can be expressed in matrix form as:

{ } { }
{ }
{ }
2
6 6
I
X
u
W u K
T T
1
41


(
( = A
`


A

)



and the stiffness matrix
11 12
2 2 2 4
21 22
4 2 4 4
K K
K
K K


( ( (

(
( =

(
( (


, which is symmetric.
The symmetric stiffness matrix coefficients are;

Paper: ASAT-13-ST-11



12/24

11 11
11
0
2
12 12 0
11 11 2
0 0
2
21 0
11 11 2
0 0
2
22 0 0
11
. .
1
. . . .
2
1
. . . .
2
1
2
L
j
i
ij ij
L L
J J i i
iJ iJ
L L
j j
I I
Ij
IJ
d
d
K K A dx
dx dx
dw d d d d
K K A dx B dx
dx dx dx dx dx
d d
dw d d
K A dx B dx
dx dx dx dx dx
du dw d
K A
dx dx




= =
| |
= =
|
\ .
| |
=
|
\ .
(
| |
= +
(
|
\ .
(

}
} }
} }
2
11 2
0 0
2 2 2
0
11 11 2 2 2
0 0
. . .
1
. . . .
2
L L
J J I I
L L
J J I I
d d d
dx B dx
dx dx dx dx
dw d d d d
B dx D dx
dx dx dx dx dx

+
} }
} }



It is clear that the term
1
2
exist in the two submatrices
12
K (

and
21
K (

, which give
symmetric stiffness matrix.

In the symmetric case
0 0
( ), ( ) u x w x should be known from a previous iteration to calculate
the stiffness matrix. Thus we can solve the equilibrium equations for the nodal displacements
{ } d , and repeat the iteration process.


6.2. Linear stiffness matrix of anisotropic materials
Omit the nonlinear term
2
0
dw
dx
| |
|
\ .
from the longitudinal strain expression, Eq.(8). So, the linear
strain will be;

2
0 0
2 xx
du d w
z
dx dx
=



This results to omit all nonlinear terms in the stiffness matrix coefficients as shown:

11 11
11
0
2
12 12
11 2
0
2
21
11 2
0
2 2 2
22
11 11 2 2 2
0 0
. .
.
.
. . .
L
j
i
ij ij
L
J i
iJ iJ
L
j
I
Ij
L L
J J I I
IJ
d
d
K K A dx
dx dx
d d
K K B dx
dx dx
d
d
K B dx
dx dx
d d d d
K B dx D dx
dx dx dx dx


= =
= =
=
= +
}
}
}
} }



Paper: ASAT-13-ST-11



13/24

6.3. Nonlinear stiffness matrix of isotropic materials
For isotropic materials,
11
( ) Q E x = . If the geometric x -axis is along the principle centroidal
axis,
11
0 B = ,
11
( ) ( ) A E x A x = , and
11
( ) ( )
y
D E x I x = , where ( ) A x is the beam cross-
section area, ( ) E x is the beam modulus of elasticity, and ( )
y
I x is the beam cross-section
second moment of area about y-axis.
Thus, Euler nonlinear symmetric stiffness matrix coefficients for isotropic materials are as
follows;

11 11
0
12 12 0
0
21 0
0
2
22 0 0
0
( ). ( ). . .
1
( ). ( ). . . .
2
1
( ). ( ). . . .
2
1
( ). ( ). . .
2
L
j
i
ij ij
L
J i
iJ iJ
L
j
I
Ij
L
J I
IJ
d
d
K K E x A x dx
dx dx
dw d d
K K E x A x dx
dx dx dx
d
dw d
K E x A x dx
dx dx dx
du dw d d
K E x A x dx
dx dx dx dx


= =
| |
= =
|
\ .
| |
=
|
\ .
(
| |
= +
(
|
\ .
(

}
}
}
}
2 2
2 2
0
( ) ( ) . .
L
J I
y
d d
E x I x dx
dx dx

+
}




6.4. Linear stiffness matrix of isotropic materials
The stiffness matrix coefficients are as follow;

11
0
12
21
2 2
22
2 2
0
. .
0
0
. .
L
j
i
ij
iJ
Ij
L
J I
IJ y
d
d
K EA dx
dx dx
K
K
d d
K EI dx
dx dx


=
=
=
=
}
}


By substituting the shape functions into external virtual work
E
W , Eq.(16), we can get the
following;

2
1 1 2 2 1 1 2 3
1
0
4
1 2 2 4
1
0
( ) ( )
( ) ( )
L
E i i
i
L
I I
I
W N u N u f x x u dx P P
q x x dx M M


=
=
= + + + A + A
+ A + A + A

}


{ } { }
T
E
W d F =

where the element nodal load vector is { } { } { }
c b
F F F = +
Paper: ASAT-13-ST-11



14/24


where { }
c
F is the concentrated nodal load vector;

{ } | |
1 1 1 2 2 2
T
c
F N P M N P M =



and { }
b
F is the equivalent body load vector;

{ } | |
1 1 2 2 3 4
0
( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) .
L
T
b
F f x x q x x q x x f x x q x x q x x dx =
}



Thus the equilibrium equation is;

{ } { }

. K d F
(
=



where

K
(

is the rearranged symmetric stiffness matrix, according to the displacement
vector { } d , Eq.(34), [8]. The linear symmetric stiffness matrices of isotropic and anisotropic
models are presented in appendix (A).

11 12 12 11 12 12
11 11 12 12 13 14
21 22 22 21 22 22
11 11 12 12 13 14
21 22 22 21 22 22
21 21 22 22 23 24
11 12 12 11 12 12
21 21 22 22 23 24
21 22 22 21 12 22
31 31 32 32 33 34
11 22 22 21 22 22
41 41 42 42 43 44

K K K K K K
K K K K K K
K K K K K K
K
K K K K K K
K K K K K K
K K K K K K
(
(

(
=


(
(
(
(
(
(
(



7. Solution of Nonlinear System of Equations
After assembling the elements stiffness matrices and force vectors, a new system of nonlinear
algebraic equations will be achieved.

{ } ( ) { } { }
g g g g
K d d F
(
=



Where
{ } ( ) g g
K d
(

is the global stiffness matrix, which is function of the, unknown, global
nodal displacement vector
{ }
g
d , and
{ }
g
F is the global force vector.
This nonlinear system should be linearized to be solved and to get the nodal displacements
{ }
g
d . Two linearization methods were used in this study; the direct iteration and Newton-
Raphson iterative methods.

In the direct iteration procedure, the solution at the (r
th
) iteration is determined from the
following assembled set of equations, [8];
Paper: ASAT-13-ST-11



15/24


{ }
( )
{ }
( )
( )
{ }
1
1
r
r
g g g
d K d F

(
=
(



where the global stiffness matrix is determined using the nodal displacement vector from the
previous iteration, { }
( ) 1 r
d

.

In the Newton Raphson procedure, the linearized element equation is of the form, [8, 11];

{ }
( )
{ }
( )
{ }
( )
{ }
( ) { }
1
1 ( 1) ( 1) r r r r
g g g g
d d T d R d


(
=
(



where the residual
{ }
( ) { }
{ }
( )
{ }
( 1) ( 1) r r
g g g g
R d K d F

=

and the tangent stiffness matrix
{ }
( )
( 1) r
g
T d

(
(

element is calculated using the definition,
[8, 11];

{ }
( )
{ } ( ) { }
{ }
( ) 1
( 1)
r
g r
g
g
R d
T d
d

| |
c
( |
=
( |

c
|
\ .


The tangent stiffness matrix associated with the Euler Bernoulli beam is as follows,[8];

11 11
12 12
21 21
2
22 22 0 0
11
0
ij ij
iJ iJ
Ij Ij
L
J I
IJ IJ
T K
T K
T K
du dw d d
T K A dx
dx dx dx dx

( ( =

( ( =

( ( =

| |
| |
( ( = + + |
|

|
\ .
\ .
}




In both methods, direct and Newton Raphson, the first iteration can be calculated using linear
stiffness matrix, i.e. assume
{ }
( ) 1
0
r
g
d

= , and calculate
{ }
( ) r
g
d using Eq.(56) or Eq.(57).
Then calculate the residual, and repeat iteration process till reach a sufficient residual. At the
exact solution, the residual equals zero. The iteration process is explained by a flow chart in
the appendix (B).

During solution, direct iteration and Newton Raphson iteration methods, sometimes are not
able to converge to a certain solution. A MATLAB ready-made function "fsolve" is used for
iteration process to obtain the proposed results.


Paper: ASAT-13-ST-11



16/24

8. Validation and Discussion
Two examples of cantilever beams, steel and carbon fiber, are loaded, as shown in Fig.3, and
their deflections are measured in the laboratory. MATLAB code is used to analyze the
existing beams, for both linear and nonlinear models. After solving the obtained systems of
equations, the calculated deflections are compared with the measured ones. A relation
between nondimensional maximum deflections and the applied load is drawn for each beam.


Example (1)
A cantilever beam made of steel has the following data:

Table 1. Steel cantilever beam dimensions and Properties

Length (L) 800 mm
Width (b) 20 mm
Height (h) 6 mm
Modulus of elasticity (E) 2.1e11 Pa
Shear modulus (G) 79.3 GPa
Poisson's ratio( ) 0.3
Load (P)
Transverse concentrated load
Vary from 0.58 N to 60.58 N
Load location(x
P
) 600 mm from fixed end
Density 7850 kg/m
3










Fig. 3. Cantilever beam dimensions


On the lab, the test rig shown in Fig. 4 was used for measurements. The beam was fixed on
the vertical stand, and at a specific distance, different weights were hanged. The maximum
traverse displacement of each weight was measured by displacement gage, for small
displacements, and a ruler, for larger ones. The different results were noted.

To get the relation between the maximum traverse displacement and the applied load
numerically, A MATLAB code is developed to formulate and solve the system of equilibrium
equations previously formulated in the theoretical part of the present study.


x
P

L
P
b
h
Paper: ASAT-13-ST-11



17/24


Fig. 4. Deflection measurement of steel cantilever beam



Fig. 5. Convergence of linear and nonlinear solutions of steel cantilever
at load = 8.08 N

Studying the convergence of the model, different number of elements is used for both models;
linear and nonlinear, once at a light load (8.08 N), and another at a heavy load (76.58 N) Fig.
5, 6. The results show that, at the light load a convergence occurs for the linear model at 12
elements, and for the nonlinear model at 20 elements, Fig.5. Accordingly this specific
numbers of elements is selected for the numerical calculations. However, for the relatively
heavy load linear solution is converging away from the measured data, and nonlinear solution
needs larger number of elements to converge to the solution. This takes a lot of time, and it is
some times the iteration process proves fruitless, Fig.6.

Paper: ASAT-13-ST-11



18/24


Fig. 6. Convergence of linear and nonlinear solutions of steel cantilever
at load = 76.58 N

For further verification, a COSMOS software program is used. The beam is solved using the
Nonlinear 2D Beam Element (BEAM2D), which is 2-node uniaxial beam element for two
dimensional nonlinear structural models. The element has three degrees of freedom, two
translations and one rotation, per node. The element include the shear strain effect.

In this step, the numerical results are compared with the measured data, Fig.7.

Fig. 7. Nondimensional maximum deflection for steel cantilever beam

Paper: ASAT-13-ST-11



19/24

Example (2)
The previous procedures are re-applied on a carbon fiber beam. The laminated Carbon
Fabric/Epoxy Prepreg cantilever beam has the following data:

Table 2. Laminated Carbon /Epoxy cantilever beam dimensions and Properties

Length (L) 500 mm
Width (b) 20 mm
Height (h) 2.35 mm
E
1
39.8 GPa
E
2
37.9 GPa
G
12
1.9027
12

0.14
Number of plies 8 plies
Orientation angles [0]
8

Load (P)
Transverse concentrated load
Vary from 0.58 N to 8.58 N
Load location(x
P
) 400 mm from fixed end
Density 1227.565 kg/m
3




Fig. 8. Deflection measurement of Carbon /Epoxy cantilever beam

Studying the convergence of the model, different number of elements is used for both models;
linear and nonlinear, once at a light load (1.08 N), and another at a heavy load (7.08 N) Fig. 9,
10. The results show that, at the light load a convergence occurs for the linear model at 15
elements, and for the nonlinear model at 20 elements, Fig.9. Accordingly this specific
numbers of elements is selected for the numerical calculations. However, for the relatively
heavy load linear solution is converging away from the measured data, and nonlinear solution
needs larger number of elements to converge to the solution. This takes a lot of time, and it is
some times the iteration process proves fruitless, Fig.10.

Paper: ASAT-13-ST-11



20/24


Fig. 9. Convergence of linear and nonlinear solutions of Carbon /Epoxy cantilever
at load = 1.08 N




Fig. 10. Convergence of linear and nonlinear solutions of Carbon /Epoxy cantilever
at load = 7.08 N
Paper: ASAT-13-ST-11



21/24




Fig. 11. Nondimensional maximum deflections for Carbon /Epoxy cantilever beam


In this step, the numerical results are compared with the measured data, Fig.7, 11. It is clear
that both linear and nonlinear solutions converge to the measured data and COSMOS
solutions at relative deflection less than
max
0.1
w
L
= . For relative deflection range of (0.1 - 0.3),
the linear model solutions are quite close to COSMOS linear and nonlinear solutions and the
measured data also. For large transverse displacements, the Euler linear model solutions and
COSMOS linear solutions are quite identical, while COSMOS nonlinear solutions are closer
to measured data. However, the transverse deflections of the Euler nonlinear model go far
from the measured data for the relative deflection more than 0.1.


9. Conclusion and Future Work
This paper is investigating the geometric nonlinearity of advanced Euler Bernoulli beam to
estimate the large deformations due to multi applied loads. When applied loads on the beam
are large, the linear load-deflection relationship is not valid, because the beam develops
internal forces that resist deformation, and the magnitude of internal forces increases with
loading as well as the deformation.

Although the nonlinear model is used, Euler Bernoulli theory falls short to estimate the larger
deflections, because it neglects the Poisson effect and transverse shear strains. So it is
prescribed to include shear strain in the geometric nonlinear analysis. Timoshenko beam and
higher order beam models provide a better hypothesis to enhance the estimation of large
deflections. For future work, part II provides a complementary study that focuses on finding
better solutions with fewer limitations, saving time and achieving much satisfying results.
Paper: ASAT-13-ST-11



22/24

References
[1] F. Guttmann, W. Wagner, L. Meyer, and P. Wriggers, "A nonlinear composite shell
element with continuous interlaminar shear stresses," Computational Mechanics, Vol.
13, pp. 175-188, 1993.
[2] B. Omidvar A. Ghorbanpoor, " Nonlinear FE Solution for Thin-Walled Open-Section
Composite Beams," Journal of Structural Engineering, Vol. 122, November 1996, pp.
1369-1378, 1996.
[3] R Murali and G Prathap, Field-consistency Aspects of Locking in a Geometrically Non-
linear Beam Formulation, CSIR Centre for Mathematical Modelling and Computer
Simulation, RR CM 0601, 2006.
[4] P. Krawczyk, F. Frey, and A.P. Zielinski, " Large deflections of laminated beams with
interlayer slips: Part 1: model development," Engineering Computations, Vol. 24, pp. 17
32, 2007.
[5] 2008 Li Jun, and Hua Hongxing, " dynamic stiffness analysis of a laminated composite
beams using trigonometric shear deformation theory," Composite Structures, Vol. 89,
Issue 3, pp. 433-442, July 2009.
[6] 2008,Maxwell Blair, and Alfred G. Striz, "Finite Element Beam Assemblies with
Geometric Bend-Twist Coupling," AIAA, 2008.
[7] Yagci, Baris; Filiz, Sinan; Romero, Louis L.; Ozdoganlar, O. BurakThis, "A spectral-
Tchebychev technique for solving linear and nonlinear beam equations," Journal of
Sound and Vibration, Vol. 321, Issue 1-2, pp. 375-404, 2009.
[8] J. N. Reddy, An Introduction to Nonlinear Finite Element Analysis, Oxford University
Press, Oxford.
[9] A.A. Khdeir and J. N. Reddy, An exact Solution for the Bending of Thin and Thick
Cross-ply Laminated Beams, Composite Structure, Vol. 37, pp. 195-203, 1997.
[10] Ronald F. Gibson, Principles of Composite Material Mechanics, McGraw Hill Inc.,
New York, 1994.
[11] William H., Saul A. Teukolsky, William T. Vetterling, and Brain P. Flannery,
Numerical Recipes in Fortran, 2nd ed., Cambridge University Press, Cambridge, 1992.
[12] Logan, D.L., A First Course in the Finite Element Method, PWS-KENT, 1992.
[13] David H. Allen, and Walter E. Haisler, Introduction to Aerospace Structural Analysis,
John Wiley & Sons, New York, 1985, pp. 250-328
[14] Ferdlanand P. Beer, and E. Russell Johnston, Jr., Mechanics of Materials, McGraw- Hill
International Book Company, Auckland, 1981.
[15] Yildirm V., Sancaktar, E. and Kiral, E., Comparison Of The In-Plane Natural
Frequencies of Symmetric Cross-Ply Laminated Beams Based on The Bernoulli-Euler
and Timoshenko Beam Theories, Journal of Applied Mechanics, Vol. 66, pp. 410-417
(1999).
[16] Robbins, D. H., and Reddy, J. N., Analysis of Piezoelectrically Actuated Beams Using
a Layer-Wise Displacement Theory, Computers & Structures, Vol. 41, No. 2, pp. 265-
279, 1991.
[17] Bendary L. M., El Shafei M. A. and Riad A. M., Finite Element Model of a Smart
Isotropic Beam with Distributed Piezoelectric Actuators, Proceeding of the 13th
International AMME Conference,MTC, 27-29 May, 2008.
Paper: ASAT-13-ST-11



23/24

Appendix (A)

Linear stiffness matrix of anisotropic materials:-

11 11 11 11
11 11 11 11
3 2 3 2
11 11 11 11
2
11 11
11 11
3 2
11
0 0
12 6 0 12 6
4 6 2

0
. 0 12 6
4
A B A B
L L L L
D D D D
L L L L
D B D D
L L L L
K
A B
L L
D D
sym
L L
D
L
(

(
(
(

(
(
(

(
(
=
(
(
(
(

(
(
(
(




Linear stiffness matrix of isotropic materials:-

3 2 3 2
2
3 2
0 0 0 0
12 6 0 12 6
4 0 6 2

0 0
. 12 6
4
EA EA
L L
EI EI EI EI
L L L L
EI EI EI
L L L
K
EA
L
EI EI
sym
L L
EI
L
(

(
(
(
(
(
(

(
(
=
(
(
(
(

(
(
(
(

Paper: ASAT-13-ST-11



24/24

Appendix (B)
Flow chart of solution of nonlinear system of equations by iteration

You might also like