EN234: Three-dimentional Timoshenko beam element undergoing

axial, torsional and bending deformations

Wenqiang Fang

December 16, 2015

1 Introduction
Timoshenko beam theory is applied to discribe the behaviour of short beams when the cross-sectional dimensions of
the beam are not small compared to its length. Here in this project, we develop the theoretical formulation for three-
dimentional Timoshenko beam element undergoing axial, torsional and bending deformations. Then we implement it
into EN234FEA.

2 The kinemation relations

Consider a typical two-node beam element of length l, where each node has six degrees of freedom. In local reference
system, the beam is along with x-axis. The elastic deformation vector of the beam element is d = [U V W ]T that can
be expressed as:
∂v ∂w
U =u−y −z ,
∂x ∂x
V = −zθ x + v, (1)

W = yθ x + w.
where u is the axial stretch displacment, v, w consist of contributions vb , wb and vs , w s due to bending and transverse
shear, that is
v = vb + v s , w = w b + w s . (2)

The relationships between total slope, bending rotation and transverse shear are

∂v ∂vb ∂vs
= + = θ z + γx y ,
∂x ∂x ∂x (3)
∂w ∂wb ∂w s
= + = −θ y + γ xz ,
∂x ∂x ∂x
where γ x y and γ x z are shear strains in the (xy)- and (xz)-planes, respectively. The two rotations θ y , θ z are related to
the bending deformations vb , wb by the expressions

∂vb ∂wb
θz = , θy = − (4)
∂x ∂x

1
 The strain components are

dθ z dθ y dv dw
ε x x = ux − y +z , γx y = − θ z , γ xz = + θy (5)
dx dx dx dx

3 Derivation of shape functions

Shape function matrices for axial and torsional deformation are given by[1]

[Nu ] = [Nθ x ] = [(1 − ξ) ξ], (6)

where ξ = x/l is the dimensionless axial coordinate. Shape functions for bending deformation in the (xy)-plane are
derived as follows:
v(x) = a0 + a1 x + a2 x 2 + a3 x 3 (7)

The shear strain is assumed to be constant along the finite element γ x y = γ0 .[2] Thus the slope due to bending can be
obtained:
θ z = a1 + 2a2 x + 3a3 x 2 − γ0 (8)

and the moment-curvature relationship is

∂θ z
Mz = −EIz z = −EIz z (2a2 + 6a3 x) (9)
∂x
the shear force is related to the transverse shear strain by

Q y = κG Aγ x y (10)

where κ is the shear correction factor that accounts for the non-uniform distribution of the shear stress over the cross-
section A; E is the modulus of elasticity, and G is the shear modulus, Iz z is the second moment of area about the z-axis.
The bending moment Mz and the shearing force Q y are related by

dMz
− Qy = 0 (11)
dx
Combining which equation (9) and (10), we get the expression of γ0

EIz z EIz z
γ0 = −6 a3 = −6Λz a3 , Λz = (12)
κG A κG A
The following boundary conditions must be satisfied:

v(0) = v1 , v(l) = v2 , θ(0) = θ z1 , θ(l) = θ z2 (13)

This can be written in matrix form


 v1 

1 0 0 0   a0 
     

θ z1
    
 = 0 1 0 6Λz a1

 
  

   



 (14)



 v2 



1 l l2 l 3  
 

 a2 



θ z2 (3l + 6Λz )  
2
 
   
 0 1 2l a3
   
 

2
Or in more compact form
d = Aa (15)

Sloving for a and substituting the values of a j into the expression of v(ξ = x/l) and simplifying, we obtains

v(ξ) = Nv1 v1 + Nv2 θ z1 + Nv3 v2 + Nv4 θ z2 (16)

where  
Nv1 = Φz 1 − 3ξ 2 + 2ξ 3 + Φz (1 − ξ) ,
 
Nv2 = lΦz ξ − 2ξ 2 + ξ 3 + Φz (ξ − ξ 2 )/2 ,
  (17)
Nv3 = Φz 3ξ 2 − 2ξ 3 + Φz ξ ,
 
Nv4 = lΦz −ξ 2 + ξ 3 + Φz (−ξ + ξ 2 )/2 .
Hence θ z (ξ) can be written in the form

θ z (ξ) = Nθ z1 v1 + Nθ z2 θ z1 + Nθ z3 v2 + Nθ z4 θ z2 (18)

where
Φz  
Nθ z1 = 6 −ξ + ξ 2 ,
l 
Nθ z2 = Φz 1 − 4ξ + 3ξ 2 + Φz (1 − ξ) ,
(19)
Φz  
Nθ z3 = −6 −ξ + ξ 2 ,
l 
Nθ z4 = Φz −2ξ + 3ξ 2 + Φz ξ .
where
1 12Λz 12EIz z
Φz = , Φz = 2 = (20)
1 + Φz l κG Al 2

Shape functions for bending in the (xz)-plane are obtained in a similar manner

1 12EIy y
Φy = , Φy = (21)
1 + Φy κG Al 2
 
Nw1 = Φy 1 − 3ξ 2 + 2ξ 3 + Φy (1 − ξ) ,
 
Nw2 = −lΦy ξ − 2ξ 2 + ξ 3 + Φy (ξ − ξ 2 )/2 ,
  (22)
Nw3 = Φy 3ξ 2 − 2ξ 3 + Φy ξ ,
 
Nw4 = −lΦy −ξ 2 + ξ 3 + Φy (−ξ + ξ 2 )/2 .

Φy  
Nθ y1 = 6 −ξ + ξ 2 ,
l  
Nθ y2 = −Φy 1 − 4ξ + 3ξ 2 + Φy (1 − ξ) ,
(23)
Φy  
Nθ y3 = −6 −ξ + ξ 2 ,
l 
Nθ y4 = −Φy −2ξ + 3ξ 2 + Φy ξ .

3
By the above shape functions, the kinematic relation can be expressed as

{d}6×1 = [U V W Θx Θy Θz ]T = [N]6×12 {e}12×1 (24)

4 Element stiffness and residual force

The total potential energy
Z l Z l
dθ y 2
Z l
1 dθ z 2 1 1
ΠP = EIy y ( ) dx + EIz z ( ) dx + G Aκ(γ x2 y + γ xz
2
)dx
0 2 dx 0 2 dx 0 2
Z l Z l Z l Z l
1 du 2 1 dθ x 2 (25)
+ E Aκ( ) dx + G Jρ ( ) dx − qv vdx − qw wdx
0 2 dx 0 2 dx 0 0
− Q∗v v l −Q∗w w l −My∗ θ y l −Mz∗ θ z l −Tx∗ θ x l −Nx∗ u l

Do finite element discretization. We get the local element stiffness matrix and the residual force vetcor formed as
Z 1   Z 1  
K= EIy y Bby T Bby + EIz z Bbz T Bbz + G Aκ(Bsy T Bsy + Bsz T Bsz ) dξ + E ABs T Bs + G Jρ Br T Br dξ (26)
0 0

Z 1
1
R= qNT dξ , F = NT f 0 (27)
0

The degrees of freedom are found by solving the linear equation

Ku = R + F (28)

Where K can be sloved out analytically[3]

 E A 
 l 
 0 12Φz 3E Iz z 
l
 12Φ y E I y y 
 0 0 l3 
GJ ρ
 0 0 0 l symmetr y 
 
6Φ E I (4+Φ y )Φ y E I y y
 0 0 − yl 2 y y 0 l

 6Φ z E I z z (4+Φ z )Φ z E I z z 
 0 l2
0 0 0 l 
 − E A 0 0 0 0 0 EA 
 l 12Φ z E I z z
l
− 6ΦzlE2 Iz z 12Φ z E I z z

 0 − l 3 0 0 0 0 l3

 12Φ y E I y y 6Φ y E I y y 12Φ y E I y y 
 0 0 − l3
0 l2
0 0 0 l3 
GJ ρ GJ ρ
 0 0 0 − l 0 0 0 0 0 l

6Φ y E I y y (2−Φ y )Φ y E I y y 6Φ y E I y y (4+Φ y )Φ y E I y y
 
 0 0 − l2 0 l 0 0 0 l2
0 l 
6Φ z E I z z (2−Φ z )Φ z E I z z
 0 0 0 0 0 − 6ΦzlE2 Iz z 0 0 0 (4+Φ z )Φ z E I z z
 
l2 l l 
(29)
For uniformly distributed load qv0 , qw0 , the residual vector yields

l
R= [0 6qv0 6qw0 0 − lqw0 lqv0 0 6qv0 6qw0 0 lqw0 − lqv0 ]T (30)
12

4
5 The transformation between local reference system and global reference
system
The above formulation is only valid in local reference system. For every element, we need to do coordinate transforma-
tion from global reference system to local reference system. The degree of freedom of one element in global reference
system is
u = [u1 v1 w1 θ x1 θ y1 θ z1 u2 v2 w2 θ x2 θ y2 θ z2 ] (31)

The degree of freedom of one element in local reference system is

u0 = [u10 v10 w10 θ x1

0
θ y1
0
θ z1
0
u20 v20 w20 θ x2
0
θ y2
0
θ z2
0
] (32)

The tranformation matirx is

T 0
cos(x x) cos(x 0 y) cos(x 0 z) 
*. +/  
T
Q = ... // , T = cos(y 0 x) cos(y 0 y) cos(y 0 z) 
 (33)
T //
0
 cos(z x) cos(z 0 y) cos(z 0 z) 
.  
, T -12×12

where T satisfies



 ui0 






 ui 






 θ xi
0 


  θ xi


 



 v 0  = [T]  vi  ,  θ 0
      = [T] 
  θ yi  , i = 1, 2

(34)
 i     yi   
 w0 

 
 
 wi 
  
  θ0
 

  θ zi

 


 i     zi   
Thus, the transformation from global reference system to local reference system of displacment, force and element
stiffness matrix can be expressed as
u0 = Qu , F0 = QF , K0 = QKQT (35)

The method to determine T is discussed as below. Firstly, we need to slove the components of axial vector i in the
global reference system. Given the global coordinates of the beam (x 1 , y1 , z1 , x 2 , y2 , z2 ), the first row of T is
x2 − x1 y2 − y1 z2 − z1
T11 = , T12 = , T13 = (36)
l l l
where q
l= (x 2 − x 1 ) 2 + (y2 − y1 ) 2 + (z2 − z1 ) 2 (37)

Then given the cross-section direction vector in global reference system k0 = (k10 , k 20 , k30 ) (This is define in the input
file along with the second moment of area Iy y , Iz z ), the local (xz)-plane is formed by vectors i and k0. we can slove
vector j = −i × k0, thus the second row of T is

T12 k30 − T13 k20 T13 k 10 − T11 k 30 T11 k20 − T12 k10
T21 = − , T22 = − , T23 = − (38)
A A A
where q
A= (T12 k30 − T13 k20 ) 2 + (T13 k 10 − T11 k 30 ) 2 + (T11 k20 − T12 k10 ) 2 (39)

Now the normal vector k of local (xy)-plane is given by k = i × j. Therefore the third row of T is given by

T12T23 − T13T22 T13T21 − T11T23 T11T22 − T12T21

T31 = , T32 = − , T33 = − (40)
B B B

5
where q
B= (T12T23 − T13T22 ) 2 + (T13T21 − T11T23 ) 2 + (T11T22 − T12T21 ) 2 (41)

6 EN234FEA implement
The twelve properties we difined for one beam element is

• E, G Young’s modulus and shear modulus

• A The area of cross section

• k 0 The position vector of local z-direction

• Iz z , Iy y , Jρ The second moments of area about the z-axis and the y-axis, polar moment of area

• κ Shear correction factor depending on the shape of the cross section, typically 5/6 for rectangular section

• l y , hz The maxima distance from the edge to the center of the cross section in y-direction and z-direction

• qv0 , qw0 Uniformly distributed load on y-direction and z-direction

In our code, we plan to calculate the following state variables and output them by user_print subroutine:

• S1 Normal stress due to stretch

• T1 Maximum shear stress due to torsion

• S2 Maximum Normal stress due to bending in y-direction

• T2 Shear stress due to bending in y-direction

• S3 Maximum Normal stress due to bending in z-direction

• T3 Shear stress due to bending in z-direction

They can be calculated from the following formulas,

E(u2 − u1 )
S1 =
l
G(θ x2 − θ x1 )
T1 = max(l y , hz )
l
El y (θ z2 − θ z1 )
S2 =
l
(42)
−GΦz Φz (2v1 + θ z1 l − 2v2 + θ z2 l)
T2 = Gγ0y =
2l
Ehz (θ y2 − θ y1 )
S3 =
l
−GΦy Φy (2w1 + θ y1 l − 2w2 + θ y2 l)
T3 = Gγ0z =
2l

6
7 The test problems
To verify the code, I’ll do several tests with axial, torsional and bending deformations. The bending loads contain
distributed force, focus shear force and force moment.

(a) Axial stretch test (b) Tortion test

(c) bending test with point load (d) bending test with moment

(e) bending test with distributed load (f) Test with all the above load

(g) Test with all the above load in 3d space (h) Complex Test in 3d space

Figure 1: Test models

In the first step, different kinds of forces are applied seperatelly. The analytical solution can be derived easily and
the numerial solutions agree with them very well.

In the second step, these loads are applied on the elements at the same time. The deformation and stresses also

7
superpose linearly since it is under small deformation.

In the third step, I make a rotation to let the beam not along with x-axis. By appling the same loads, I verify that
the stresses in local coordinate system of the beam do not change.

In the fourth step, I build a complex model without analytical solution. By plotting the deformation configures in
Tecplot, I infer that the result is reasonable qualitatively.

The input files and result data files of all the tests are available in the beam file directory.

(a) Initial configure (b) Second configure

(c) Third configure (d) Final configure

Figure 2: Tecplot visualization of the solution of the last test

8
References
[1] A Bazoune, YA Khulief, and NG Stephen. Shape functions of three-dimensional timoshenko beam element.
Journal of Sound and Vibration, 259(2):473–480, 2003.

[2] R Davis, RD Henshell, and GB Warburton. A timoshenko beam element. Journal of Sound and Vibration,
22(4):475–487, 1972.

[3] Janusz S Przemieniecki. Theory of matrix structural analysis. Courier Corporation, 1985.