The Linear Frame Element

Professor Terje Haukaas University of British Columbia, Vancouver www.inrisk.ubc.
ca
The Linear Frame Element

The beam element, called frame element here to emphasize that it can be used to model
columns too, is by far the most popular finite element in structural analysis. A variety of
alternatives exist, from 2D Euler-Bernoulli beams to 3D beams with torsion, shear
deformation, and geometric stiffness contributions. Figure 1 shows four element
configurations related to 2D structural analysis with axial and bending deformation. The
two lower-most drawings show “basic” configurations; they have DOFs that are
sufficient only for describing deformation, not rigid-body motions. What distinguishes
the two lower-most configurations is inclusion of axial deformation. The upper-most
drawings in Figure 1 show the “local” configuration, also with and without axial
deformation. These are capable of describing any deformation pattern, including rigid
body motion. Transformation from the local configuration to a global structural
coordinate system is addressed in the document on the computational stiffness method.
z, w z, w
3 2 5
1
3 6
2 4 1
x, u 4 x, u
z, w z, w
1 2 3
2
x, u 1 x, u
Figure 1: Beam element configurations for 2D structural analysis.
The stiffness matrix for each element configuration in Figure 1 is known from the
classical stiffness method. For reference, they are:
⎡ 4EI 2EI ⎤
⎢ ⎥
L L
Kb = ⎢ ⎥ (1)
⎢ 2EI 4EI ⎥
⎢ ⎥
⎣ L L ⎦
The Linear Frame Element Updated December 24, 2014 Page 1

Professor Terje Haukaas University of British Columbia, Vancouver www.inrisk.ubc.ca
⎡ EA ⎤
⎢ 0 0 ⎥
⎢ L ⎥
⎢ 4EI 2EI ⎥
Kb = ⎢ 0 ⎥ (2)
⎢ L L ⎥
⎢ 2EI 4EI ⎥
⎢ 0 L L ⎥⎦
⎣
⎡ 12EI 6EI 12EI 6EI ⎤
⎢ − − − ⎥
⎢ L3 L2 L3 L2 ⎥
⎢ 6EI 4EI 6EI 2EI ⎥
⎢ − 2 ⎥
L L L2 L
Kl = ⎢ ⎥ (3)
⎢ 12EI 6EI 12EI 6EI ⎥
⎢ − 3 ⎥
L L2 L3 L2
⎢ ⎥
⎢ − 2
6EI 2EI 6EI 4EI ⎥
⎢ L L L2 L ⎥
⎣ ⎦
⎡ EA EA ⎤
⎢ 0 0 − 0 0 ⎥
⎢ L L ⎥
⎢ 12EI 6EI 12EI 6EI ⎥
⎢ 0 − 0 − − 2 ⎥
⎢ L3 L2 L3 L ⎥
⎢ 0 − 2 0 ⎥
L L L2 L
Kl = ⎢ ⎥ (4)
⎢ − EA 0 0
EA
0 0 ⎥
⎢ L L ⎥
⎢ ⎥
⎢ 0 12EI 6EI 12EI 6EI ⎥
− 0
⎢ L3 L2 L3 L2 ⎥
⎢ 0 − 2 0 ⎥
⎢⎣ L L L2 L ⎥⎦
To illustrate the finite element method, consider the 2D beam element in its basic
configuration, without axial deformations. It is selected to employ the weak form of the
BVP, with the principle of virtual displacements as a starting point. To this end, consider
the equality of internal and external virtual work as a starting point:
δWint = δWext (5)
In particular, the principle of virtual displacements reads
L
∫ σ δε dV = ∫ q δ w dx
V 0
z
(6)
Substitution of material law yields

V
∫ Eε δε dV = ∫ qz δ w dx
0
(7)
Substitution of kinematics yields

L
∫E⋅z ⋅ w ''δ w ''dV = ∫ qz δ w dx (8)

2
V 0
Separation of the left-hand side integral into a cross-section integral and a longitudinal
integral yields
L L
∫ EI ⋅ w ''δ w ''dx = ∫ qz δ w dx
0 0
(9)
This is the weak form of the BVP for beam bending. The all-important discretization of
this problem by means of shape functions follows. The element has two basic DOFs
related to bending. Imagine the element in a horizontal position with the coordinate x
running from 0 at the left end to L at the right end. Let the clockwise rotation of the left
end be denoted u1 and let the clockwise rotation at the right end be denoted u2. Third-
order polynomial shape functions are possible, given two rotational DOFs and zero
displacement at the element ends. Consequently, the shape functions are:
⎧⎪ u1 ⎫⎪
w(x) = Nu = ⎡ N1 (x) N 2 (x) ⎤ ⎨ (10)
⎣ ⎦ u2 ⎬
⎪⎩ ⎪⎭
where
1 3 2 2
N1 (x) = − x + x −x
L2 L (11)
1 1
N 2 (x) = − 2 x 3 + x 2
L L
Substitution into the weak form yields
( )( ) dx − ∫ qz ( Nδ u ) dx = 0
L L
∫0
EI ⋅ N ''u ⋅ N ''δ u
0
(12)
where δu is the virtual nodal deformations because the virtual displacements are
discretized by the same shape functions as the actual displacements. Rearranging yields
(
δ u ⎡⎢ ∫ EI ⋅ N ''T N '' dx ⎤⎥ u − ∫ qz NT dx = 0
⎣ 0
L
⎦ 0
L
) (13)
Furthermore, because the virtual displacements are arbitrary the parenthesis must be zero
for this equation to be generally valid. Consequently, it is rewritten
⎡L ⎤ L
⎢ ∫ EI ⋅ N '' N ''dx ⎥ u = ∫ qz N dx
T
(14)
⎣0   ⎦ 
0
  
Stiffness matrix, K Load vector, F

where the stiffness matrix and load vector are identified. Again it is noted that the finite
element method yields integral expressions for the stiffness matrix and the load vector.
Substitution of Eq. (11) into Eq. (14) and assuming that the distributed element load is
uniform yields
⎡ 4EI 2EI ⎤ ⎧ q L2 ⎫
⎢ ⎥⎧ u ⎫ ⎪ − ⎪
z
⎢ L L ⎥ ⎪⎨ 1 ⎪⎬ = ⎪⎨ 12 ⎪
⎬ (15)
⎢ 2EI 4EI ⎥ ⎪ u2 ⎪ ⎪ qz L2 ⎪
⎢ ⎥⎩ ⎭
⎪ 12 ⎪
⎣ L L ⎦ ⎩ ⎭
This stiffness matrix is equal to that of the classical stiffness method because the third-
order polynomial shape functions match the solution of the differential equation for beam
bending. Under such circumstances the finite element method is exact. Notice that the
right-hand side load vector has the correct sign for upward-acting qz and clockwise
rotation DOFs.
Shear Deformation
The inclusion of shear deformation in the stiffness matrix for beam elements is possible
by employing the unit virtual load method to establish the flexibility matrix, followed by
inversion to obtain the stiffness matrix. This takes place in the basic element
configuration. As a starting point, recall that the a stiffness coefficient kij is “force along
degree of freedom number i due to a unit displacement/rotation along degree of freedom
number j.” However, the unit virtual load method yields deformations, not forces. For
that reason it is convenient to first establish the flexibility coefficients fij instead of
stiffness coefficients kij. Subsequently, the flexibility matrix is inverted to obtain the
stiffness matrix. Figure 2 shows the virtual and real section force diagrams used to obtain
f11, f12, f21, and f22, where fij is the displacement or rotation along degree of freedom
number i due to a unit force along degree of freedom number j. Each flexibility
coefficient is computed by evaluating the virtual work integral
M ⋅δ M V ⋅δV
L L
fij = ∫ dx + ∫ dx (16)
0
EI 0
GAv
By employing the “quick integration” formulas from the unit virtual work method, the
following results are obtained:
1 1 ⎛ 1⎞ ⎛ 1⎞ L 1
f11 = ⋅1⋅1⋅ L + ⋅⎜ − ⎟ ⋅⎜ − ⎟ ⋅ L = + (17)
3EI GAv ⎝ L ⎠ ⎝ L ⎠ 3EI GAv L
1 1 ⎛ 1⎞ ⎛ 1⎞ L 1
f21 = − ⋅1⋅1⋅ L + ⋅⎜ − ⎟ ⋅⎜ − ⎟ ⋅ L = − + (18)
1 1 ⎛ 1⎞ ⎛ 1⎞ L 1
f12 = − ⋅1⋅1⋅ L + ⋅⎜ − ⎟ ⋅⎜ − ⎟ ⋅ L = − + (19)

1 1 ⎛ 1⎞ ⎛ 1⎞ L 1
f22 = ⋅1⋅1⋅ L + ⋅⎜ − ⎟ ⋅⎜ − ⎟ ⋅ L = + (20)
1 Virtual: 1 Virtual:
1 δM 1 δM
–1/L δV –1/L δV
f11 f12
1 Real: Real: 1
1 M
1M
–1/L V –1/L V
Virtual: 1 Virtual: 1
1 δM 1 δM
–1/L δV –1/L δV
f21 f22
1 Real: Real: 1
1 M
1M
–1/L V –1/L V
Figure 2: Beam cases for virtual work computations.
The results are summarized by the equation

⎡ θ1 ⎤ ⎡ f11 f12 ⎤ ⎡ M 1 ⎤
⎢ ⎥=⎢ ⎥⎢ ⎥ (21)
⎢⎣ θ 2 ⎥⎦ ⎢⎣ f21 f22 ⎥ ⎢ M 2
⎦⎣ ⎥⎦
By defining the auxiliary coefficient
12EI
α= (22)
GAv L2
and summarizing the above results, the flexibility matrix in the basic element
configuration is:

⎡ ⎛ α⎞ ⎛ α⎞ ⎤
⎡ θ1 ⎤ ⎢ ⎜2+ ⎟ − ⎜ 1− ⎟ ⎥ ⎡
L ⎢ ⎝ 2⎠ ⎝ 2 ⎠ ⎥ M1 ⎤
⎢ ⎥= ⋅⎢ ⎢ ⎥ (23)
⎢⎣ θ 2 ⎥⎦ 6EI ⎛ α⎞ ⎛ α ⎞ ⎥ ⎢⎣ M 2 ⎥⎦
⎢ − ⎜ 1− ⎟ ⎜⎝ 2 + ⎟⎠ ⎥
⎢⎣ ⎝ 2⎠ 2 ⎥⎦
In passing, it is noted that the expression for a for a rectangular cross-section with zero
Poisson’s ratio is
b ⋅ h3
12 ⋅ E ⋅ 12 ⎛ h ⎞
2
α= 12 = ⋅⎜ ⎟ (24)
E 5 ⎝ ⎠
⋅ ⋅b ⋅ h ⋅ L2 5 L
2 6
which reasonably suggests that beams with higher h/L ratio exhibits more shear
deformation. Next, inversion of the flexibility matrix in Eq. (23) yields the stiffness
matrix in the basic configuration, modified with shear deformation:
EI ⎡ ( 4 + α ) (2 − α ) ⎤
Kb = ⋅⎢ ⎥ (25)
(1 + α )L ⎢ ( 2 − α ) ( 4 + α ) ⎥
⎣ ⎦
Transformation to the local coordinate system with the transformation matrix that is
shown in the document on the computational stiffness method yields:
⎡ 12EI 6EI 12EI 6EI ⎤
⎢ − − − ⎥
⎢ L3 L2 L3 L2 ⎥
⎢ 6EI (4 + α ) ⋅ EI 6EI (2 − α ) ⋅ EI ⎥
⎢ − 2 ⎥
1 L L L2 L
Kl = ⋅⎢ ⎥ (26)
(1 + α ) ⎢ 12EI 6EI 12EI 6EI ⎥
⎢ − L3 L2 L3 L2 ⎥
⎢ ⎥
⎢ − 6EI (2 − α ) ⋅ EI 6EI (4 + α ) ⋅ EI ⎥
⎢ L2 L L2 L ⎥
⎣ ⎦
To ease the extraction of values from Eq. (26) in hand calculations, the stiffness
coefficients in Eq. (26) are provided for two fundamental beam cases in Figure 3.

6EI 6EI
−
(1+ α )⋅ L2 (1+ α )⋅ L2
(4 + α )⋅ EI
(1+ α )⋅ L (2 − α )⋅ EI
(1+ α )⋅ L
θ = 1.0
12EI
(1+ α )⋅ L3 12EI
−
6EI (1+ α )⋅ L3
−
(1+ α )⋅ L2
Δ = 1.0 6EI
−
(1+ α )⋅ L2
Figure 3: Amendment of fundamental beam cases with shear deformation terms.
Geometric Stiffness
Significant axial compressive force in a beam-column member reduces its lateral
stiffness. Conversely, axial tension force increases the lateral stiffness. This change in
stiffness due to axial force is called geometric stiffness, and is often referred to a “P-delta
effects.” In this section, the finite element approach is utilized to amend the stiffness
matrix for frame members. The resulting geometric stiffness matrix is approximate
because the shape functions differ from the solution to the differential equation. The
expression for the exact geometric stiffness matrix is more complicated, and is presented
under the theory for beam members with axial force. In the following, the principle of
virtual displacements is first employed with polynomial shape functions. Start by
considering the differential equation for beam bending, amended with the P-delta effect.
In the absence of distributed load q, the weighted and integrated version reads
L
∫ ( EI ⋅ w ''''+ P ⋅ w '')δ w dx = 0
0
(27)
Integration by parts and cancelling boundary terms yield

L L
∫ EI ⋅ w ''δ w ''dx − ∫ P ⋅ w 'δ w 'dx = 0

0 0
(28)
Discretization of the real and virtual displacement fields by shape functions, i.e.,
w(x)=Nu and δw(x)=Nδu, where the vector u collects the displacements along the
degrees of freedom, yields
⎛ ⎞
⎜L L
⎟
⎜ ∫ EI ⋅ N '' ⋅ N ''dx − P ⋅ ∫ N ' ⋅ N 'dx ⎟ ⋅ u = 0
T T
(29)
⎜0 
0
 ⎟
⎝ kG ⎠

where the geometric stiffness matrix, kG, is defined. Two paths can now be followed to
obtain the geometric stiffness matrix in the local element configuration. Because the
basic configuration neglects rigid-body movement of the element, using the shape
functions for that configuration will miss the “P/L terms.” In other words, the P-delta
effects associated with rigid-body movement of the element will be missed if the integral
in Eq. (29) is conducted in the basic element configuration. For that reason, the following
third-order polynomial shape functions for the beam element in the local configuration
are employed:
2x 3 3x 2
N1 (x) = 3 − 2 + 1
L L
x 2x 2
3
N 2 (x) = − 2 + −x
L L
(30)
2x 3 3x 2
N 3 (x) = − 3 + 2
L L
3 2
x x
N 4 (x) = − 2 +
L L
As a result, the following total stiffness matrix is obtained:
⎡ 12EI 6EI 12EI 6EI ⎤ ⎡ 6 1 6 1 ⎤
⎢ − − − ⎥ ⎢ − − − ⎥
⎢ L3 L2 L3 L2 ⎥ ⎢ 5L 10 5L 10 ⎥
⎢ 6EI 4EI 6EI 2EI ⎥ ⎢ 1 2L 1 L ⎥
⎢ − 2 ⎥ ⎢ − 10 −
L L L2 L 15 10 30 ⎥
k=⎢ ⎥− P⋅⎢ ⎥ (31)
⎢ 12EI 6EI 12EI 6EI ⎥ ⎢ − 6 1 6 1 ⎥
− 3
⎢ L L2 L3 L2 ⎥ ⎢ 5L 10 5L 10 ⎥
⎢ 6EI 2EI 6EI 4EI ⎥ ⎢ 1 L 1 2L ⎥
⎢ − 2 ⎥ ⎢ − − ⎥
⎢⎣ L L L2 L ⎥⎦ ⎢⎣ 10 30 10 15 ⎥⎦
!######"###### $
kG
If the integration had been conducted in the basic configuration, as mentioned above, it
would be necessary to add the “P/L terms” to the result:
⎡ 1 1 1 1 ⎤
⎢ − − − ⎥ ⎡
5L 10 5L 10 ⎤
⎢ ⎥ ⎢ P 0 −
P
0 ⎥
⎢ 1 2L 1 L ⎥ ⎢ L L ⎥
⎢ − − ⎥ ⎢ 0
10 15 10 30 0 0 0 ⎥
kG = P ⋅ ⎢ ⎥+⎢ ⎥ (32)
⎢ 1 1 1 1 ⎥ ⎢ −P P
− 0 0 ⎥
⎢ 5L 10 5L 10 ⎥ ⎢ L L ⎥
⎢ 1 L 1 2L ⎥ ⎢ 0 0 0 0 ⎥⎦
⎢ − − ⎥ ⎣
⎢⎣ 10 30 10 15 ⎥⎦
To ease the extraction of values from Eq. (31) in hand calculations, the stiffness
coefficients in Eq. (31) are provided for two fundamental beam cases in Figure 4.

6EI P 6EI P
! + !
L2 10 L2 10
4EI 2PL
! 2EI PL
L 15 +
L 30
! = 1.0
12EI 6P
!
L3 5L 12EI 6P
6EI P ! +
! 2 + L3 5L
L 10
! = 1.0 !
6EI P
+
L2 10
Figure 4: Amendment of fundamental beam cases with geometric stiffness terms.
Exact Geometric Stiffness Matrix

The geometric stiffness matrix in Eq. (31) is approximate because polynomial shape
functions are employed while the solution above contains trigonometric functions. As an
alternative, column by column of the exact stiffness matrix can be established by solving
the differential equation for this problem, which is provided in the document on beam
members with axial force. According to the classical stiffness method, the columns of the
stiffness matrix are established by setting the degrees of freedom shown in Figure 5 equal
to one, one at a time.
u1 u3
u2 u4
Figure 5: Degrees of freedom for beam element.
The first column of the stiffness matrix is obtained by setting u1=1 and u2=u3=u4=0 and
computing the force along each degree of freedom:
F1 = V (0) = EI ⋅ w '''(0)
F2 = M (0) = EI ⋅ w ''(0)
(33)
F3 = −V (L) = −EI ⋅ w '''(L)
F4 = −M (L) = −EI ⋅ w ''(L)
To simplify expressions, R.K. Livesley introduced the following auxiliary functions in
his book entitled “Matrix Methods of Structural Analysis” published by Pergamon Press
in Oxford, UK in 1975:

β
φ1 =
tan(β )
1 β2
φ2 = ⋅
3 1 − φ1
1 3
φ3 = ⋅ φ1 + ⋅ φ2 (34)
4 4
1 3
φ4 = − ⋅ φ1 + ⋅ φ2
2 2
φ5 = φ1 ⋅ φ2
where
L P
β= ⋅ (35)
2 EI
which leads to the following exact stiffness matrix, including both elastic stiffness and
geometric stiffness:
⎡ 6φ5 −3Lφ2 −6φ5 −3Lφ2 ⎤
⎢ ⎥
2EI ⎢ −3Lφ2 2L2φ 3 3Lφ2 L2φ 4 ⎥
K = 3 ⋅⎢ ⎥ (36)
L ⎢ −6φ5 3Lφ2 6φ5 3Lφ2 ⎥
⎢ −3Lφ L2φ 4 −3Lφ2 2L2φ 3 ⎥
⎣ 2
⎦
Torsion
While other documents describe the theory of St. Venant torsion and warping torsion, the
stiffness matrix for a beam element with torsional DOFs is presented here. First it is
noted that elements that only carry torque by St. Venant torsion have a rather simple
stiffness matrix. In fact, the solution to the differential equation for St. Venant torsion
yields
T
φ= ⋅L (37)
GJ
for an element with length L subjected to constant torque, T. This immediately produces
the following stiffness matrix for St. Venant torsion:
⎡ GJ GJ ⎤
⎢ − ⎥
L L
k St .V . = ⎢ ⎥ (38)
⎢ GJ GJ ⎥
⎢ − L L ⎥
⎣ ⎦
The more general element carries torque with both shear and axial stresses, i.e., both St.
Venant torsion and warping torsion. For this situation, consider the full set of DOFs
shown in Figure 6. Two of the DOFs represent ordinary rotation around the x-axis, while

the other two represent warping deformation. In particular, the DOFs written in terms of
deformation are:
⎧ u1 ⎫ ⎧ φ x=0 ⎫
⎪ ⎪ ⎪ ⎪
⎪ u2 ⎪ ⎪ φ ' x=0 ⎪
u=⎨ ⎬= ⎨ ⎬ (39)
⎪ u3 ⎪ ⎪ φ x=L ⎪
⎪ u4 ⎪ ⎪ φ ' x=L ⎪
⎩ ⎭ ⎩ ⎭
and the DOFs written in terms of forces are:
⎧ F1 ⎫ ⎧ Tx=0 ⎫
⎪ ⎪ ⎪ ⎪
⎪ F2 ⎪ ⎪ Bx=0 ⎪
F=⎨ ⎬= ⎨ ⎬ (40)
⎪ F3 ⎪ ⎪ Tx=L ⎪
⎪ F4 ⎪ ⎪ Bx=L ⎪
⎩ ⎭ ⎩ ⎭
i.e., the two warping-related DOFs correspond to the bi-moment defined in the theory of
warping torsion.
2 4
1 3
x
Figure 6: DOFs for torsion.
The starting point for the derivation of the stiffness matrix is the complete differential
equation for torsion, including the equilibrium equation mx=–dT/dx, on residual form:
−GJ ⋅ φ ''+ ECw ⋅ φ ''''− mx = 0 (41)
Weighting and integrating yields:
L
∫ ( −GJ ⋅φ ''+ EC
0
w ⋅ φ ''''− mx ) δφ dx = 0 (42)
Integration by parts and cancellation of boundary terms yield:

L L L
∫ GJ ⋅φ 'δφ 'dx + ∫ EC
0 0
w ⋅ φ ''δφ ''dx − ∫ mx δφ dx = 0
0
(43)
Next, the real and virtual rotation fields are discretized with the same third-order
polynomial shape functions:
φ (x) = Nu
(44)
δφ (x) = Nδ u

where u is the vector of DOFs in Eq. (39) and the vector N consists of the functions:
2x 3 3x 2
N1 (x) = 3 − 2 + 1
L L
x 2x 2
3
N 2 (x) = 2 − +x
L L
(45)
2x 3 3x 2
N 3 (x) = − 3 + 2
L L
3 2
x x
N 4 (x) = 2 −
L L
The only difference between these shape functions and the ones in Eq. (30) is that the
sign of N2 and N4 are flipped. This is done because those two functions here represent the
amount of φ-rotation along the member due to a unit value of dφ/dx at the end, rather than
clockwise bending rotation at the member ends. For example, N4 represents the amount
of rotation along the member when dφ/dx=1 (positive unit slope) at the right-most end.
Substitution of Eq. (44) into Eq. (43) and the observation that the virtual DOFs are
arbitrary yields:
L L L
∫0 GJ ⋅ N ' N 'dx ⋅ u + ∫0 ECw ⋅ N '' N ''dx ⋅ u = ∫0 mx N dx ⇒ ku = F (46)

T T T
  

k St .V . k warping F
where:
⎡ 6GJ GJ 6GJ GJ ⎤ ⎡ 12ECw 6ECw 12ECw 6ECw ⎤
⎢ − ⎥ ⎢ − ⎥
⎢ 5L 10 5L 10 ⎥ ⎢ L3 L2 L3 L2 ⎥
⎢ 2GJL GJ GJL ⎥ ⎢ 4ECw 6EC 2ECw ⎥
⎢ − − ⎥ ⎢ − 2w ⎥ (47)
15 10 30 L L L
k=⎢ ⎥+⎢ ⎥
⎢ 6GJ GJ ⎥ ⎢ 12ECw 6ECw ⎥
− − 2
⎢ 5L 10 ⎥ ⎢ L3 L ⎥
⎢ 2GJL ⎥ ⎢ 4ECw ⎥
⎢ ⎥ ⎢ ⎥
⎢⎣ 15 ⎥⎦ ⎢⎣ L ⎥⎦
To compute axial and shear stresses from torsion it is necessary to compute section forces
beyond those directly available in F in Eq. (40). Specifically, the shear stress from St.
Venant torsion requires TSt.V. and the shear stress from warping torsion requires B’, i.e.,
Twarp. Eq. (47) shows that these two contributions to the total torque, T, are readily
obtained in matrix analysis simply by evaluating the two contributions to the forces
separately, as F=(kSt.V.+kwarp)u, with the two stiffness matrices given in Eq. (47).
Exact Stiffness Matrix for Torsion
The stiffness matrix in Eq. (47) is an approximation, because the solution to the
differential equation for combined St. Venant torsion and warping torsion is more
complex than the simple third-order polynomials that were employed as shape functions
here. The exact stiffness matrix based on the solution to the differential equation is:

⎡ 2 ⋅α 2 ⋅α ⎤
⎢ β − β ⎥
⎢ L L ⎥
⎢ L ⎛ 1 1⎞ L ⎛ 1 1⎞ ⎥
⎢ ⋅⎜ β + − ⎟ −β − ⋅ ⎜ −β + − ⎟ ⎥
GJ ⎢ 2 ⎝ β α⎠ 2 ⎝ β α⎠ ⎥ (48)
k= ⋅
2(α − β ) ⎢ 2 ⋅α ⎥
⎢ −β ⎥
⎢ L ⎥
⎢ L ⎛ 1 1⎞ ⎥
⎢ ⋅ β+ − ⎟ ⎥
⎢⎣ 2 ⎜⎝ β α⎠ ⎥
⎦
where
L GJ
α= ⋅ (49)
2 ECw
and
eα − e−α
β = tanh(α ) = (50)
eα + e−α
Shear Walls
An element that is employed in “shear wall analysis” is shown in Figure 7. The objective
of shear wall analysis is to determine the forces on lateral force resisting systems
consisting of columns and shear walls. In particular, the objective of the type of analysis
considered here is to determine the forces along the three DOFs shown in Figure 7. While
the transformation matrices for the assembly of the stiffness matrix for the entire floor is
provided in the document on matrix structural analysis, the element stiffness matrix is
discussed here.
u3,wall
u2,wall
u1,wall
Figure 7: Shear wall element.

The stiffness matrix for the element in Figure 7 is 3-by-3, but the content depends on the
boundary conditions of the shear wall. If the shear wall is restrained against warping
deformation as well as rotations at top at bottom then the stiffness matrix is
⎡ 12EI 2 6P ⎤
⎢ − 0 0 ⎥
⎢ (1+ α 1 )H
3
5H ⎥
⎢ 12EI1 6P ⎥
k=⎢ 0 − 0 ⎥ (51)
⎢ (1+ α 2 )H 3
5H ⎥
⎢ 6GJ 12ECw ⎥
⎢ 0 0 + ⎥
⎢⎣ 5H H3 ⎥⎦
where shear deformation and P-delta effects are included. If warping torsion, shear
deformation, and P-delta effects are neglected, and the shear is free to rotate at the top
because the bending stiffness of the slab is small, then the stiffness matrix for the element
in Figure 7 simplifies to
⎡ 3EI 2 ⎤
⎢ 3
0 0 ⎥
⎢ H ⎥
⎢ 3EI1 ⎥ (52)
k=⎢ 0 0 ⎥
H3
⎢ ⎥
⎢ 0 0
GJ ⎥
⎢ H ⎥⎦
⎣
Other cases are not listed here, but it is noted that the previously presented element
matrices are employed to determine the stiffness associated with the DOFs of the element
in Figure 7. It is reiterated that the removal of DOFs from an element without restraining
them is achieved by static condensation, which is described in the document on matrix
structural analysis.
Generalized DOFs
It is straightforward to establish the shape functions in Eq. (11) for this beam element.
However, for pedagogical purposes another approach is explored. It is referred to as
Hermitian interpolation. It takes as a starting point the terms of a general polynomial,
namely 1, x, x2, x3, x4, and so forth. These are collected in the vector Nq. Each term is
multiplied by a factor, qi, which is called a generalized DOF or generalized displacement.
The generalized DOFs are collected in the vector q. Rather than considering the beam
element in its basic configuration, the local element configuration, still without axial
deformations, is adopted. The reason is that even in the basic configuration addressed
earlier the two displacement DOFs were implicitly included. When establishing the shape
functions the displacement DOFs were set to zero to focus only the DOFs that are
necessary to describe the element deformation. Now the local element configuration is
considered, in which all the DOFs needed to describe deformation and rigid-body motion
are included. Given four element DOFs, four generalized DOFs are selected:

⎧ q1 ⎫
⎪ ⎪
{1 } ⎪ q2 ⎪
w(x) = N q q = x x2 x3 ⎨ ⎬ (53)
⎪ q3 ⎪
⎪ q4 ⎪
⎩ ⎭
Next, the generalized DOFs are related to the actual DOFs by the equation
u = Aq (54)
In turn, the sought shape functions N in Eq. (14) are obtained by combining Eqs. (53) and
(54):
w(x) = N q q = N q A −1 u (55)

N
The matrix A is established by writing each actual DOF, ui, in terms of the element
displacement, w(x):
u1 = w(0) = q1
u2 = − w′(0) = q2
(56)
u3 = w(L) = q1 + q2 L + q3 L2 + q4 L3
u4 = − w′(L) = q2 + q3 L + q4 L2
which is summarized in the A-matrix:
⎧ u1 ⎫ ⎡ 1 0 0 0 ⎤ ⎧ q1 ⎫
⎪ ⎪ ⎪ ⎪
⎪ u2 ⎪ ⎢ 0 1 0 0 ⎥ ⎪ q2 ⎪
⎨ ⎬= ⎢ ⎥⎨ ⎬ (57)
⎪ u3 ⎪ ⎢ 1 L L2 L3 ⎥ ⎪ q3 ⎪
⎪ ⎢ 2 ⎥
u4 ⎪ ⎣ 0 1 L L ⎦ ⎪ q4 ⎪
⎩ ⎭  ⎩ ⎭
A
Eq. (55) yields the sought shape functions, N = N q A −1 , which substituted into Eq. (14)
produces the stiffness matrix in Eq. (3). This is the same result as any other approach
based on polynomial functions for Euler-Bernoulli beams, because for this element the
solution to the differential equation is indeed polynomial.
Complete Stiffness Matrix

Figure 8 shows the complete 3D element with 7 DOFs per node, including the warping
DOFs, number 7 and 14. This is the DOF numbering that is utilized in the structural
analysis program St.

z, w
y, v
6 13
5 12
3 10
2 1 4 9 x, u
8 11
7 14
Figure 8: Complete 3D beam element.

The Linear Frame Element

Caricato da

Informazioni sul documento

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

The Linear Frame Element

Caricato da

Copyright:

Formati disponibili

Professor Terje Haukaas University of British Columbia, Vancouver www.inrisk.ubc.

The Linear Frame Element

Figure 1: Beam element configurations for 2D structural analysis.

The Linear Frame Element Updated December 24, 2014 Page 1

Substitution of material law yields

The Linear Frame Element Updated December 24, 2014 Page 2

Substitution of kinematics yields

∫E⋅z ⋅ w ''δ w ''dV = ∫ qz δ w dx (8)

The Linear Frame Element Updated December 24, 2014 Page 3

The Linear Frame Element Updated December 24, 2014 Page 4

Figure 2: Beam cases for virtual work computations.

The results are summarized by the equation

The Linear Frame Element Updated December 24, 2014 Page 5

The Linear Frame Element Updated December 24, 2014 Page 6

Integration by parts and cancelling boundary terms yield

∫ EI ⋅ w ''δ w ''dx − ∫ P ⋅ w 'δ w 'dx = 0

The Linear Frame Element Updated December 24, 2014 Page 7

The Linear Frame Element Updated December 24, 2014 Page 8

Exact Geometric Stiffness Matrix

Figure 5: Degrees of freedom for beam element.

The Linear Frame Element Updated December 24, 2014 Page 9

The Linear Frame Element Updated December 24, 2014 Page 10

Integration by parts and cancellation of boundary terms yield:

The Linear Frame Element Updated December 24, 2014 Page 11

∫0 GJ ⋅ N ' N 'dx ⋅ u + ∫0 ECw ⋅ N '' N ''dx ⋅ u = ∫0 mx N dx ⇒ ku = F (46)

  

The Linear Frame Element Updated December 24, 2014 Page 12

Figure 7: Shear wall element.

The Linear Frame Element Updated December 24, 2014 Page 13

The Linear Frame Element Updated December 24, 2014 Page 14

Complete Stiffness Matrix

The Linear Frame Element Updated December 24, 2014 Page 15

Figure 8: Complete 3D beam element.

The Linear Frame Element Updated December 24, 2014 Page 16

Potrebbero piacerti anche