Beruflich Dokumente
Kultur Dokumente
1 Preliminaries
This document describes the formulation of stiffness and mass matrices for structural elements
such as truss bars, beams, plates, and cables(?). The formulation of each element involves the
determination of gradients of potential and kinetic energy functions with respect to a set of
coordinates defining the displacements at the ends, or nodes, of the elements. The potential
and kinetic energy of the functions are therefore written in terms of these nodal displacements
(i.e., generalized coordinates). To do so, the distribution of strains and velocities within the
element must be written in terms of nodal coordinates as well. Both of these distributions
may be derived from the distribution of internal displacements within the solid element.
1.1 Displacements
∂ui (x, t)
ii (x, t) = ≡ ui,i (x, t) (4)
∂xi
∂ui (x, t) ∂uj (x, t)
γij (x, t) = + ≡ uj,i (x, t) + ui,j (x, t) (5)
∂xj ∂xi
(6)
Displacement gradient
N
∂ui (x) X ∂
= ψin (x1 , x2 , x3 ) ūn (t) (7)
∂xj n=1 ∂xj
N
X
ui,j (x) = ψin,j (x) ūn (t) (8)
n=1
Strain-displacement relations
N
X
ii (x, t) = ψin,i (x) ūn (t) (9)
n=1
XN
γij (x, t) = (ψin,j (x) + ψjn,i (x)) ūn (t) (10)
n=1
Strain vector
T (x, t) = { 11 22 33 γ12 γ23 γ13 } (11)
(x, t) = [ B(x) ]6×N ū(t) (12)
dx uy(x) + u’(x)
y dx
uy(x) u’(x)dx
x
dx
Figure 2. Transverse deformation u0y (x) and geometric deformation u0x (x).
Geometric deformation can be derived from the Pythagorean theorem and is quadratic in
u0x (x) and u0y (x).
L
u4
u2
u1 u3
Lo
Figure 3. A bar with four displacement coordinates in its original and displaced configurations.
(u4-u2)/Lo = 0.050
100
L1 (Pythagorean Thm.)/Lo-1
10-1 L2 (approx)/Lo-1
strain
log(L1/Lo)
10-2
10-3
10-4 10-3 10-2 10-1
10-3
relative error, L2/L1 - 1
10-4
10-5
10-6
10-7
10-4 10-3 10-2 10-1
(u3-u1)/Lo
Following the development from the previous section, we can consider the planar displace-
ment of a short line segment of length dlo to a line segment of length dl. In the original
configuration, the line segment has a squared length of
dlo2 = dx2 + dy 2
u y (x,y)
+ u y,x (x,y) dx
dl + uy,y (x,y) dy
(x+dx , y+dy)
uy(x,y) u x (x,y)
+ u x,x(x,y) dx
dl o + ux,y (x,y) dy
(x , y)
ux(x,y)
Figure 5. Displacement of line segment dlo to line segment dl through planar displacements
ux (x, y) and uy (x, y).
The difference between the squared lengths, dl2 − dlo2 , is quadratic in dx and dy and can be
expressed in a “matrix form.”
#T "
ux,x + 21 u2x,x + 21 u2y,x 1
+ 12 uy,x + 12 ux,y uy,x
" #" #
2 dx u
2 x,y
dx
l − lo2 = 2 1
dy u + 12 uy,x + 12 ux,y uy,x
2 x,y
uy,y + 12 u2x,y + 12 u2y,y dy
#T "
1
" #" #
dx xx γ
2 xy
dx
= 2 1 (14)
dy γ
2 yx
yy dy
This “matrix” is called the Green-Lagrange strain tensor, or the Lagrangian strain tensor, or
the Green-Saint Venant strain tensor for plane strain. This relationship may be confirmed
in two simple cases:
Generalizing this result to three dimensions, the strain-displacement equations for linear
strain and geometric deformation are:
!2 !2 !2
1 1 1 ∂ui 1 ∂ui 1 ∂uj 1 ∂uk
ii = ui,i + u2i,i + u2j,i + u2k,i = + + + (15)
2 2 2 ∂xi 2 ∂xi 2 ∂xi 2 ∂xi
∂ui ∂uj ∂ui ∂uj ∂uj ∂ui
γij = ui,j + uj,i + ui,j uj,j + uj,i ui,i = + + + (16)
∂xj ∂xi ∂xj ∂xj ∂xi ∂xi
In the following models, finite deformation effects are limited to xx ≈ ux,x + (1/2)u2y,x for
the exclusive purpose of deriving geometric stiffness matrices. For metalic materials in which
strains are typically around 0.001, this approximation does not compromise accuracy.
σ11
1−ν ν ν
11
σ22 ν 1−ν ν 22
σ33 E ν ν 1−ν 33
=
1
(17)
τ12
(1 + ν)(1 − 2ν)
2
−ν
γ12
1
τ23 −ν γ23
2
1
τ13 −ν γ13
2
Stress vector
σ T (x, t) = { σ11 σ22 σ33 τ12 τ23 τ13 } (18)
σ = [ Se (E, ν) ]6×6 (19)
1.5 Potential Energy and Stiffness
Consider a system comprising an assemblage of linear springs, with stiffness ki , each with an
individual stretch, di . The total potential energy in the assemblage is
1X 2
U= ki di
2 i
If displacements of the assemblage of springs is denoted by a vector u, not necessarily equal
to the stretches in each spring, then the elastic potential energy may also be written
1 T
U (u) = u Ku
2
n
1X
= ui fi
2 i=1
n n
1X X
= ui Kij uj
2 i=1 j=1
where K is the stiffness matrix with respect to the coordinates u. The stiffness matrix K
relates the elastic forces fi to the collocated displacements, ui .
f1 = K11 u1 + · · · + K1j uj + · · · + K1N uN
fi = Ki1 u1 + · · · + Kij uj + · · · + KiN uN
fN = KN 1 u1 + · · · + KN j uj + · · · + KN N uN
A point force fi acting on an elastic body is the gradient of the elastic potential energy U
with respect to the collocated displacement ui
∂
fi =U
∂ui
The i, j term of the stiffness matrix may therefore be found from the potential energy function
U (u),
∂ ∂
Kij = U (u) (20)
∂ui ∂uj
1Z
U (ū) = σ(x, t)T (x, t) dΩ (21)
2 ZΩ
1
= (x, t)T Se (E, ν) (x, t) dΩ
2 ZΩ
1
= ū(t)T B(x)T Se (E, ν) B(x)ū(t) dΩ
2 Ω
1 Z h i
= ū(t)T B(x)T Se (E, ν) B(x) dΩ ū(t) (22)
2 Ω N ×N
where M is the constant mass matrix with respect to the generalized coordinates u. The
mass matrix M relates the inertial forces fi to the collocated accelerations, üi .
The i, j term of the constant mass matrix may therefore be found from the kinetic energy
function T ,
∂ ∂ ∂ ∂ ∂
Mij = T (u̇) = T (u̇) (24)
∂ üi ∂t ∂ u̇j ∂ u̇i ∂ u̇j
˙ = 1Z
T (ū) ρ |u̇(x, t)|2 dΩ (25)
2 ZΩ
1
= ρ u̇(x, t)T u̇(x, t) dΩ
2 ZΩ
1 ˙ T Ψ(x)T Ψ(x)ū(t) ˙
= ρ ū(t) dΩ
2 Ω
1 ˙ T Z h i
˙
= ū(t) ρ Ψ(x)T Ψ(x) dΩ ū(t) (26)
2 Ω N ×N
x x
ux (x, t) = 1− ū1 (t) + ū3 (t) (30)
L L
= ψx1 (x) ū1 (t) + ψx3 (x) ū3 (t) (31)
x x
uy (x, t) = 1− ū2 (t) + ū4 (t) (32)
L L
= ψy2 (x) ū2 (t) + ψy4 (x) ū4 (t) (33)
x x
" #
1− L
0 L
0
Ψ(x) = x x (34)
0 1− L
0 L
" #
ux (x, t)
= Ψ(x) ū(t) (35)
uy (x, t)
Strain-displacement relation
!2
∂ux 1 ∂uy
xx = + (36)
∂x 2 ∂x
1
= ψx1,x ū1 + ψx3,x ū3 + {ψy2,x ū2 + ψy4,x ū4 }2 (37)
2
2
1 1 1 1 1
= − ū1 + ū3 + − ū2 + ū4 (38)
L L 2 L L
2
1 1 1 1 1
= − 0 0 ū + − ū2 + ū4 (39)
L L 2 L L
2
1 1 1
= B ū + − ū2 + ū4 (40)
2 L L
1 1
B= − 0 0 . (41)
L L
1 Z
U = xx E xx dΩ (42)
2 Ω
Z L h i
K̄e = BT E B A dx (43)
x=0
1/L2 0 −1/L2
0
Z L
0 0 0 0
= EA dx (44)
x=0 −1/L2 0 1/L2 0
0 0 0 0
1 0 −1 0
EA 0 0 0 0
= (45)
L
−1 0 1 0
0 0 0 0
1 ˙T Z h i
˙
T = ū ρ Ψ(x)T Ψ(x) dΩ ū(t) (46)
2 Ω N ×N
Z L h i
M̄ = ρ Ψ(x)T Ψ(x) A dx (47)
x=0
(1 − Lx )2 (1 − Lx )( Lx )
0 0
x 2
(1 − L ) (1 − Lx )( Lx )
Z L
0 0
= ρA dx (48)
x=0 ( Lx )(1 − Lx ) 0 ( Lx )2 0
x x x 2
0 ( L )(1 − L ) 0 (L)
2 0 1 0
1 0 2 0 1
= ρAL (49)
6
1 0 2 0
0 1 0 2
1 ZL
U = xx E xx A dx (50)
2 0
!2 2
EA Z L ∂ux 1 ∂uy
= + dx (51)
2 0 ∂x 2 ∂x
!2 !2 !4
EA Z L ∂ux ∂ux ∂uy 1 ∂uy
= + + dx (52)
2 0 ∂x ∂x ∂x 4 ∂x
Substitute
∂ux 1 1
= − ū1 + ū3 (53)
∂x L L
∂uy 1 1
= − ū2 + ū4 (54)
∂x L L
to obtain
EA 1
U= (ū3 − ū1 )2 + (ū3 − ū1 )(ū4 − ū2 )2 (55)
2L L
So,
1 0 −1 0 ū1 0 0 0 0 ū1
∂U EA 0 0 0 0
ū2
EA(ū3 − ū1 ) 0 1 0 −1
ū2
= + (56)
∂ū L −1 0
1 0 ū3
L2
0 0 0 0 ū3
0 0 0 0 ū4 0 −1 0 1 ū4
N
= K̄e ū + K̄g ū (57)
L
2D prismatic homogeneous isotropic beam element, neglecting shear deformation and rotatory
inertia.
Consider the geometry of a deformed beam. The functions ux (x) and uy (x) describe the
translation of points along the neutral axis of the beam as a function of the location along
the un-stretched neutral axis.
We will describe the deformation of the beam as a function of the end displacements (ū1 , ū2 , ū4 , ū5 )
and the end rotations (ū3 , ū6 ). In a dynamic context, these end displacements will change
with time.
6
X
ux (x, t) = ψxn (x) ūn (t)
n=1
X6
uy (x, t) = ψyn (x) ūn (t)
n=1
The functions ψxn (x) and ψyn (x) satisfy the boundary conditions at the end of the beam and
the differential equation describing bending of a Bernoulli-Euler beam loaded statically at
the nodal coordinates. In such beams the effects of shear deformation and rotatory inertia
are neglected. For extension of the neutral axis,
x
ψx1 (x) = 1 −
L
x
ψx4 (x) =
L
CC BY-NC-ND December 13, 2017, H.P. Gavin
Structural Element Stiffness, Mass, and Damping Matrices 13
and ψx2 = ψx3 = ψx5 = ψx6 = 0 along the neutral axis. For bending of the neutral axis,
2
x 3 x
ψy2 (x) = 1 − 3 +2
L L
2 3 !
x x x
ψy3 (x) = −2 + L
L L L
2 3
x x
ψy5 (x) = 3 −2
L L
2 3 !
x x
ψy6 (x) = − + L
L L
and ψy1 = ψy4 = 0.
x
1− L 0 0 x 0 0
L
Ψ(x) = x
2 x
3 x x
2 x
3 x
2 x
3
x
2 x
3 (58)
0 1−3 L
+2 L L
−2 L
+ L
L 0 3 L
−2 L
− L
+ L
L
" #
ux (x, t)
= Ψ(x) ū(t) (59)
uy (x, t)
These expressions are analytical solutions for the displacements of Bernoulli-Euler beams
loaded only with concentrated point loads and concentrated point moments at their ends.
Internal bending moments are linear within beams loaded only at their ends, and the beam
displacements may be expressed with cubic polynomials.
In extension, the elastic potential energy in a beam is the strain energy related to the uni-
form extensional strain, xx . If the strain is small, then the extensional strain within the
cross section is equal to an extension of the neutral axis, (∂ux /∂x), plus the bending strain,
−(∂ 2 uy /∂x2 )y.
∂ux ∂ 2 uy
xx = − y
∂x ∂x2
6 6
X ∂ X ∂2
= ψxn (x) ūn − ψ (x) y ūn
2 yn
(60)
n=1 ∂x n=1 ∂x
6 6
0 00
X X
= ψxn (x) ūn − ψyn (x) y ūn
n=1 n=1
X6
= Bn (x, y) ūn
n=1
= B(x, y) ū (61)
where
1 6y 12xy 4y 6xy 1 −6y 12xy 2y 6xy
B(x, y) = − , − 3 , − 3, , + 3 , − 2 . (62)
L L2 L L L L L2 L L L
The elastic stiffness matrix can be found directly from the strain energy of axial strains xx .
1 Z
U = xx E xx dΩ (63)
2 Ω
Z L Z h i
K̄e = B(x, y)T E B(x, y) dA dx. (64)
x=0 A
Note that this integral involves terms such as A y 2 dA and A ydA in which the origin of the
R R
coordinate axis is placed at the centroid of the section. The integral A y 2 dA is the bending
R
R
moment of inertia for the cross section, I, and the integral A ydA is zero.
It is also important to recognize that the elastic strain energy may be evaluated separately
for extension effects and bending effects. For extension, the elastic strain energy is
1Z L
U = EA (xx )2 dx
2 x=0
6
!2
1Z L X
0
= EA ψxn (x) ūn dx
2 x=0 n=1
In bending, the elastic potential energy in a Bernoulli-Euler beam is the strain energy related
to the curvature, κz .
6 6
∂ 2 uy X ∂2 X
00
κz = = ψyn (x) ūn = ψyn (x) ūn
∂x2 n=1 ∂x 2
n=1
The kinetic energy of a particle within a beam is half the mass of the particle, ρAdx, times
its velocity, u̇, squared. For velocities along the direction of the neutral axis,
6
ψxn (x) ū˙ n ,
X
u̇x (x) =
n=1
The kinetic energy function and the mass matrix may be by substituting equation (58) into
equations (26) and (29).
1 ˙T Z h i
˙
T = ū ρ Ψ(x)T Ψ(x) dΩ ū(t) (67)
2 Ω N ×N
Z L h i
M̄ = ρ Ψ(x)T Ψ(x) A dx (68)
x=0
It is important to recognize that kinetic energy and mass associated with extensional velocities
may be determined separately from those associated with transverse velocities. The kinetic
energy for extension of the neutral axis is
1Z L
T = ρA (u̇x )2 dx
2 x=0
6
!2
1Z L X
˙
= ρA ψxn (x) ūn dx
2 x=0 n=1
1 ZL 1Z L Z L
U = EA u2x,x dx + EI u2y,xx dx + EA ux,x u2y,x dx . (75)
2 0 2 0 0
Substitute
6
0
X
uy,x = ψyn (x) ūn (76)
n=1
6
00
X
uy,xx = ψyn (x) ūn (77)
n=1
6
X
0 N
ux,x = ψxn ūn = (78)
n=1 EA
so that,
N
K̄ = K̄e + K̄g (80)
L
For prismatic homogeneous isotropic beams, substituting the expressions for the functions
ψxn and ψyn into equations (65) - (70), or substituting equation (62) into equation (64) and
(58) to equation (68) results in element stiffness matrices K̄e , M̄, and K̄g .
EA
L
0 0 − EA
L
0 0
12EI 6EI
0 − 12EI 6EI
L3 L2 L3 L2
4EI
− 6EI 2EI
L
0 L2 L
K̄e =
(81)
EA
L
0 0
sym
12EI
− 6EI
L3 L 2
4EI
L
140 0 0 70 0 0
156 22L 0 54 −13L
4L2 2
−3L
0 13L
ρAL
M̄ = (82)
420
140 0 0
sym
156 −22L
4L2
0 0 0 0 0 0
6 L L
0 − 56
5 10 10
2L2 L 2
0 − 10 − L30
15
N
K̄g = (83)
L
0 0 0
sym
6 L
− 10
5
2
− 2L
15
Consider again the geometry of a deformed beam. When shear deformations are included
sections that are originally perpendicular to the neutral axis may not be perpendicular to
the neutral axis after deformation. The functions ux (x) and uy (x) describe the translation of
points along the neutral axis of the beam as a function of the location along the un-stretched
neutral axis. If the beam is not slender (length/depth < 5), then shear strains will contribute
significantly to the strain energy within the beam. The deformed shape of slender beams is
different from the deformed shape of stocky beams.
The beam carries a bending moment M (x) related to axial strain xx and a shear force, S
related to shear strain γxy . The potential energy has a bending strain component and a shear
strain component.
1Z T
U = σ dΩ
2 ZΩ
1 1Z
= σxx xx dΩ + τxy γxy dΩ
2 Ω 2 Ω
1 Z L Z M (x)y M (x)y 1 Z L Z SQ(y) SQ(y)
= dA dx + dA dx
2 0 A I EI 2 0 A Ib(y) GIb(y)
1 Z L M (x)2 Z 2 1 Z L S 2 Z Q(y)2
= y dA dx + dA dx
2 0 EI 2 A 2 0 GI 2 A b(y)2
1 Z L M (x)2 1 Z L S2
= dx + dx (84)
2 0 EI 2 0 G(A/α)
where the shear area coefficient α reduces the cross section area to account for the non-uniform
distribution of shear stresses in the cross section,
A Z Q(y)2
α= 2 dA .
I A b(y)2
For solid rectangular sections α = 6/5 and for solid circular sections α = 10/9 [2, 3, 4, 5, 8].
The transverse deformation of a beam with shear and bending strains may be separated into
a portion related to shear deformation and a portion related to bending deformation,
where
x x
EIu00(b)y (x) = M (x) = −M1 1− + M2 (86)
L L
1
G(A/α)u0(s)y (x) = S(x) = (M1 + M2 ) (87)
L
It can be shown that the following shape functions satisfy the Timoshenko beam equations
(equations (85), (86) and (87)) for transverse displacements.
" #
1 x 2 x 3 x
ψy2 (x) = 1−3 +2 + 1− Φ
1+Φ L L L
" 2 3 2 ! #
L x x x 1 x x
ψy3 (x) = −2 + + − Φ
1+Φ L L L 2 L L
" #
x 2
3
1 x x
ψy5 (x) = 3 −2 + Φ
1+Φ L L L
" 2 ! #
L x 2 x 3 1 x x
ψy6 (x) = − + − − Φ
1+Φ L L 2 L L
The term Φ gives the relative importance of the shear deformations to the bending deforma-
tions, 2
12EI r
Φ= 2
= 24α(1 + ν) , (88)
G(A/α)L L
q
where r is the “radius of gyration” of the cross section, r = I/A, ν is Poisson’s ratio. Shear
deformation effects are significant for beams which have a length-to-depth ratio less than 5.
To neglect shear deformation, set Φ = 0. These displacement functions are exact for frame
elements with constant shear forces S and linearly varying bending moment distributions,
M (x), in which the strain energy has both a shear stress component and a normal stress
component,
6
!2 6
!2
1Z L X
00 1Z L X
0
U= EI ψ(b)yn (x)ūn dx + G(A/α) ψ(s)yn (x)ūn dx (89)
2 0 n=1 2 0 n=1
where the bending and shear components of the shape functions, ψ(b)yn (x) and ψ(s)yn (x) are:
" #
1 x 2 x 3
ψ(b)y2 (x) = 1−3 +2
1+Φ L L
Φ x
ψ(s)y2 (x) = 1−
1+Φ" L
2 3 2 ! #
L x x x 1 x x
ψ(b)y3 (x) = −2 + + 2 − Φ
1+Φ L L L 2 L L
LΦ 1 x
ψ(s)y3 (x) = −
1 + Φ"2 L
2 3 #
1 x x
ψ(b)y5 (x) = 3 −2
1+Φ L L
Φ x
ψ(s)y5 (x) =
1 + Φ "L
2 3 2 ! #
L x x 1 x
ψ(b)y6 (x) = − + + Φ
1+Φ L L 2 L
L 1x
ψ(s)y6 (x) = − Φ
1+Φ 2L
The geometric stiffness matrix for a Timoshenko beam element may be derived as was done
with the Bernoulli-Euler beam element from the potential energy of linear and geometric
strain,
Z L
0 0
K̄ij = EA ψxi (x)ψxj (x) dx
0
Z L
00 00
+ EI ψ(b)yi (x)ψ(b)yj (x) dx
0
Z L
0 0
+ G(A/α) ψ(s)yi (x)ψ(s)yj (x) dx
0
Z L
0 0
+ N ψyi (x)ψyj (x) dx (90)
0
where the displacement shape functions ψ(x) are provided in section 4.1.
For prismatic homogeneous isotropic beams, substituting the previous expressions for the
functions ψxn (x) and ψ(b)yn (x), and ψ(s)yn (x) into equation (90) and (68), results in the
Timoshenko element elastic stiffness matrices K̄e , mass matrix M̄, and geometric stiffness
matrix K̄g EA
− EA
L
0 0 L
0 0
12 EI 6 EI 12 EI 6 EI
1+Φ L3 1+Φ L2
0 − 1+Φ L3 1+Φ L2
4+Φ EI 6 EI 2−Φ EI
− 1+Φ
1+Φ L
0 L2 1+Φ L
K̄e = (91)
EA
L
0 0
sym
12 EI 6 EI
− 1+Φ
1+Φ L3 2
L
4+Φ EI
1+Φ L
280 0 0 140 0 0
312 + 588Φ + 280Φ2 (44 + 77Φ + 35Φ2 )L 0 108 + 252Φ + 175Φ2 −(26 + 63Φ + 35Φ2 )L
(8 + 14Φ + 7Φ2 )L2 0 (26 + 63Φ + 35Φ2 )L 2
−(6 + 14Φ + 7Φ )L 2
ρAL
M̄ =
(92)
840
280 0 0
sym
312 + 588Φ + 280Φ2 −(44 + 77Φ + 35Φ2 )L
(8 + 14Φ + 7Φ2 )L2
0 0 0 0 0 0
6/5+2Φ+Φ2 L/10 −6/5−2Φ−Φ2 L/10
(1+Φ)2 (1+Φ)2
0 (1+Φ)2 (1+Φ)2
2L2 /15+L2 Φ/6+L2 Φ2 /12 −L/10 −L2 /30−L2 Φ/6−L2 Φ2 /12
0
(1+Φ)2 (1+Φ)2 (1+Φ)2
N
K̄g = (93)
L
0 0 0
sym
6/5+2Φ+Φ2 −L/10
(1+Φ)2 2
(1+Φ)
2 2 2 2
2L /15+L Φ/6+L Φ /12
(1+Φ)2
Consider again the geometry of a deformed beam with linearly-varying axial beam displace-
ments outside of the neutral axis. The functions ux (x, y) and uy (x, y) now describe the
translation of points anywhere within the beam, as a function of the location within the
beam. We will again describe these displacements in terms of a set of shape functions,
ψxn (x, y) and ψyn (x), and the end displacements ū1 , · · · , ū6 .
6
X
ux (x, y, t) = ψxn (x, y) ūn (t)
n=1
X6
uy (x, t) = ψyn (x) ūn (t)
n=1
The shape functions for transverse displacements ψyn (x) are the same as the shape functions
ψyn (x) used previously. The shape functions for axial displacements along the neutral axis,
ψx1 (x, y) and ψx4 (x, y) are also the same as the shape functions ψx1 (x) and ψx4 (x) used
previously. To account for axial displacements outside of the neutral axis, four new shape
functions are derived from the assumption that plane sections remain plane, ux (x, y) =
−u0(b)y (x)y.
!
x x 2 y
0
ψx2 (x, y) = −ψy2 y= 6 −
L L L
2 !
0 x x
ψx3 (x, y) = −ψy3 y= −1 + 4 − 3 y
L L
2 !
0 x x y
ψx5 (x, y) = −ψy5 y= 6 − +
L L L
2 !
0 x x
ψx6 (x, y) = −ψy6 y= 2 −3 y
L L
Because ψyn , ψx1 and ψx4 are unchanged, the stiffness matrix is also unchanged. The kinetic
energy of the beam, including axial and transverse effects is now,
6
!2 6
!2
1 Z L Z h/2 1Z L
ψxn (x, y) ū˙ n ψyn (x) ū˙ n
X X
T = ρb(y) dy dx + ρA dx
2 x=0 y=−h/2 n=1 2 x=0 n=1
Evaluating equation (24) using the new shape functions ψx2 , ψx3 , ψx5 , and ψx6 , results in a
mass matrix incorporating rotatory inertia.
1 1
0 0 0 0
3 6
13 6 r2 11 1 r2 9 6 r2 13 2
1 r
+ L + 0 − − 420 L+
35 5 L2 210 10 L 70 5 L2 10 L
1 2 2 13 1 r2
L2
105
+ 15
r 0 420
L + 10 L
0
M̄ = ρAL
(94)
1
sym 3
0 0
13 6 r2 11 2
1 r
+ − 210 L+
35 5 L2 10 L
1 2 2
105
L2 + 15
r
Beam element mass matrices including the effects of shear deformation on rotatory inertia
are more complicated. Refer to p 295 of Theory of Matrix Structural Analysis, by J.S.
Przemieniecki (Dover Pub., 1985).
EI 2EI 6EI
L
θ L
θ L2
∆
←− ←− ←− ←−
3EI 3EI 6EI 12EI
L3
∆ L2
θ L2
θ L3
∆
EA
N1
L
0 0 − EA
L 0 0
ū1
12EI 6EI
− 12EI 6EI
V1 0 ū2
L3 L2 L3 L2
4EI
− 6EI 2EI
M1 0 ū3
L L2 L
=
EA
N2 0 0 ū4
L
sym
12EI
− 6EI
V2 ū5
L3 L2
4EI
M2 ū6
L
f̄ = K̄ ū
EA 2 EA
− EA − EA
2
L c L cs L c L cs
12EI 2 12EI
+ L3 s − L3 cs − L2 s − L3 s + L3 cs − 6EI
6EI 12EI 2 12EI
L2 s
EA 2
− EA − EA
2
L s L cs L s
+ 12EI
L3 c
2 6EI
L2 c + 12EI 12EI 2
L3 cs − L3 c
6EI
L2 c
4EI 6EI
− 6EI 2EI
L2 s L2 c
L L
K=
EA 2 EA
L c L cs
+ 12EI 2
− 12EI 6EI
L3 s L3 cs L2 s
sym
EA 2
L s
12EI 2
− 6EI
+ L3 c L2 c
4EI
L
f =Ku
N1
140 0 0 70 0 0
ū¨1
ū¨2
V1 156 22L 0 54 −13L
4L2 13L −3L2 ū¨3
M1 0
ρAL
=
420
ū¨4
N2 140 0 0
sym
ū¨5
V2 156 −22L
ū¨6
4L2
M2
¨
f̄ = M̄ ū
where
c s 0 x2 − x1
c = cos θ =
−s c 0 0
L
0 0 1
T=
c s 0
0 −s c 0
y2 − y1
s = sin θ =
L
0 0 1
140c2 −16cs −22sL 70c2 16cs 13sL
+15s2 +54s2
140s2 70s2
22cL 16cs −13cL
+156c2 +54c2
4L2 −13sL 13cL 2
−3L
ρAL
M=
420
140c2 −16cs 22sL
+156s2
sym
140s2 −22cL
+156c2
2
4L
f =Mu
Approximate element stiffness and mass matrices based on assumed distribution of internal
displacements.
Consider the geometry of a rectangle with edges aligned with a Cartesian coordinate system.
(0 ≤ x ≤ a, 0 ≤ y ≤ b) The functions ux (x, y, t) and uy (x, y, t) describe the in-plane
displacements as a function of the location within the element.
The eight coefficients c1 , · · · , c8 may be found uniquely from matching the displacement
coordinates at the corners.
ux (x, y, t) = x̂ŷ ū1 (t) + (1 − x̂)ŷ ū3 (t) + (1 − x̂)(1 − ŷ) ū5 (t) + x̂(1 − ŷ) ū7 (t) (95)
uy (x, y, t) = x̂ŷ ū2 (t) + (1 − x̂)ŷ ū4 (t) + (1 − x̂)(1 − ŷ) ū6 (t) + x̂(1 − ŷ) ū8 (t) (96)
Strain-displacement relations
∂ux 1 ∂ux
xx = =
∂x a ∂ x̂
∂uy 1 ∂uy
yy = =
∂y b ∂ ŷ
∂ux ∂uy 1 ∂ux 1 ∂uy
γxy = + = +
∂y ∂x b ∂ ŷ a ∂ x̂
so that
ū1
ū2
ū3
−ŷ/a −(1 − ŷ)/a (1 − ŷ)/a
ŷ/a 0 0 0 0
xx
ū4
yy = 0 x̂/b 0 (1 − x̂)/b 0 −(1 − x̂)/b 0 −x̂/b
ū5
γxy x̂/b ŷ/a (1 − x̂)/b −ŷ/a −(1 − x̂)/b −(1 − ŷ)/a −x̂/b (1 − ŷ)/a
ū6
ū7
ū8
or
(x, y, t) = B(x, y) ū(t)
6.2.1 Plane-Stress
6.2.2 Plane-Strain
1Z
V = σ(x, y, t)T (x, y, t) h dx dy (103)
2 A
1 T
Z h i
= ū(t) B(x, y)T Se (E, ν) B(x, y) h dx dy ū(t) (104)
2 A 8×8
˙ = 1Z
T (ū) ρ |u̇(x, y, t)|2 h dx dy (106)
2 A
1 ˙ TZ h i
˙
= ū(t) ρ Ψ(x, y)T Ψ(x, y) h dx dy ū(t) (107)
2 A 8×8
6.5 2D Rectangular Plane-Stress and Plane-Strain Element Stiffness and Mass Matrices
Eh
K̄e = 12(1−ν 2 )
·
4c + kA kB −4c + kA /2 −kC −2c − kA /2 −kB 2c − kA kC
kB 4/c + kD kC 2/c − kD −kB −2/c − kD /2 −kC −4/c + kD /2
−4c + kA /2 kC 4c + kA −kB 2c − kA −kC −2c − kA /2 kB
−kC 2/c − kD −kB 4/c + kD kC −4/c + kD /2 kB −2/c − kD /2
−2c − kA /2 −kB 2c − kA kC 4c + kA kB −4c + kA /2 −kC
−kB −2/c − kD /2 −kC −4/c + kD /2 kB 4/c + kD kC 2/c − kD
2c − kA −kC −2c − kA /2 kB −4c + kA /2 kC 4c + kA −kB
kC −4/c + kD /2 kB −2/c − kD /2 −kC 2/c − kD −kB 4/c + kD
Eh
K̄e = 12(1+ν)(1−2ν) ·
The element mass matrix for the plane-stress and plane-strain elements is the same.
4 0 2 0 1 0 2 0
0 4 0 2 0 1 0 2
2 0 4 0 2 0 1 0
ρabh 0 2 0 4 0 2 0 1
M̄ =
(109)
36
1 0 2 0 4 0 2 0
0 1 0 2 0 4 0 2
2 0 1 0 2 0 4 0
0 2 0 1 0 2 0 4
Note, again, that these element stiffness matrices are approximations based on an assumed
distribution of internal displacements.
If the structure is in a fluid (liquid or gas), the motion of the structure is resisted by the
fluid viscosity. At low speeds (low Reynolds numbers), this damping effect can be taken
to be linear in the velocity, and the damping forces are proportional to the total rate of
displacement (not the rate of deformation). If the fluid is flowing past the structure at high
flow rates (high Reynolds numbers), the motion of the structure can interact with the flowing
medium. This interaction affects the dynamics (natural frequencies and damping ratios) of
the coupled structure-fluid system. Potentially, at certain flow speeds, the motion of the
structure can increase the transfer of energy from the flow into the structure, giving rise to
an aero-elastic instabililty.
Damping can also arise within structural systems from friction forces internal to the structure
(the micro-slip within joints and connections) inherent material viscoelasticity, and inelas-
tic material behavior. In many structural systems, a type of damping in which damping
stresses are proportional to strain and in-phase with strain-rate are assumed. Such so-called
“complex-stiffness damping” or “structural damping” is commonly used to model the damp-
ing in soils. Fundamentally, this kind of damping is neither elastic nor viscous. The force-
displacement behavior does not follow the same path in loading and unloading, behavior
but instead follows a “butterfly” shaped path. Nevertheless, this type of damping is com-
monly linearized as linear viscous damping, in which forces are proportional to the rate of
deformation.
In materials in which stress depends on strain and strain rate, a Voigt viscoelasticity model
may be assumed, in which stress is proportional to both strain and strain-rate ,˙
σ = [ Se (E, ν) ] + [ Sv (η) ] ˙
The internal virtual work of real viscous stresses Sv ˙ moving through virtual strains δ is
Z
˙ =
δW (ū) σ(x, t)T δ(x, t) dΩ (110)
ZΩ
= ˙ t)T Sv (η) δ(x, t) dΩ
(x,
ZΩ
= ˙ T B(x)T Sv (η) B(x)δū(t) dΩ
ū(t)
Ω
Z h i
˙ T
= ū(t) B(x)T Sv (η) B(x) dΩ δū(t) (111)
Ω N ×N
Given a material viscous damping matrix, Sv , a structural element damping matrix can be
determined for any type of structural element, through the integral in equation (111), as has
been done for stiffness and mass element matrices earlier in this document. In doing so, it
may be assumed that the internal element displacements ui (x, t) (and the matrices [Ψ] and
[B]) are unaffected by the presence of damping, though this is not strictly true. Further,
the parameters in Sv (η) are often dependent of the frequency of the strain and the strain
amplitude. Damping behavior that is amplitude-dependent is outside the domain of linear
analysis.
Cs = αMs + βKs
ωn 21
2ζ1 ωn1 1
R̄T Cs R̄ =
...
= α
...
+β ...
(112)
2ζN ωnN 1 ωn 2N
where ωn 2j is an eigen-value (squared natural frequency) and the columns of R̄ are mass-
normalized eigen-vectors (modal vectors) of the generalized eigen-problem
From equations (112) it can be seen that the damping ratios satisfy
α 1 β
ζj = + ωnj
2 ωnj 2
and the Rayleigh damping coefficients (α and β) can be determined so that the damping
ratios ζj have desired values at two frequencies. The damping ratios modeled by Rayleigh
damping can get very large for low and high frequencies. Rayleigh damping grows to ∞ as
ω → 0 and increases linearly with ω for large values of ω. Note that the Rayleigh damping
matrix has the same banded form as the mass and stiffness matrices. In other words, with
Rayleigh damping, internal damping forces are applied only between coordinates that are
connected by structural elements.
The Caughey damping matrix is a generalization of the Rayleigh damping matrix. Caughey
damping matrices can involve more than two parameters and can therefore be used to provide
a desired amount of damping over a range of frequencies. The Caughey damping matrix for
an assembled model for a structural system is
j=n
X2
Cs = Ms αj (M−1
s Ks )
j
j=n1
where the index range limits n1 and n2 can be positive or negative, as long as n1 < n2 . As
with the Rayleigh damping matrix, the Caughey damping matrix may also be diagonalized
by the real eigen-vector matrix R̄. The coefficients αj are related to the damping ratios, ζk ,
by
1 1 j=n
X2
ζk = αj ωk2j
2 ωk j=n1
The coefficients αj may be selected so that a set of specified damping ratios ζk are obtained
at a corresponding set of frequencies ωk . If n1 = 0 and n2 = 1, then the Caughey damp-
ing matrix is the same as the Rayleigh damping matrix. For other values of n1 and n2 the
Caughey damping matrix loses the banded structure of the Rayleigh damping matrix, imply-
ing the presence of damping forces between coordinates that are not connected by structural
elements.
Structural systems with classical damping have real-valued modes r̄j that depend only on
the system’s mass and stiffness matrices (equation (113)), and can be analyzed as a sys-
tem of uncoupled second-order ordinary differential equations. The responses of the system
coordinates can be approximated via a modal expansion of a select subset of modes. The
convenience of the application of modal-superpostion to the transient response analysis of
structures is the primary motivation
An element Rayleigh damping matrix may be easily computed from the element’s mass and
stiffness matrix C = αM+βK and assembled into a damping matrix for the structural system
Cs . The element damping is presumed to increases linearly with the mass and the stiffness of
the element; larger elements will have greater mass, stiffness, and damping. System damping
matrices assembled from such element damping matrices will have the same banding as the
mass and stiffness matrices; internal damping forces will occur only between coordinates
connected by a structural element. However, such an assembled damping matrix will not be
diagonizeable by the real eigenvectors of the structural system mass matrix Ms and stiffness
matrix Ks .
The damping matrix for a linear viscous damper connecting a node at (x1 , y1 ) to a node at
(x2 , y2 ) is found from the element coordinate transformation,
" #T " #" #
c s 0 0 0 0 cd −cd c s 0 0 0 0
C6×6 =
0 0 0 c s 0 −cd cd 0 0 0 c s 0
where c = (x2 − x1 )/L and s = (y2 − y1 )/L. Structural systems with supplemental damping
components generally have non-classical system damping matrices.
References
[1] Clough, Ray W., and Penzien, Joseph, Dynamics of Structures, 2nd ed. (revised), Com-
puters and Structures, 2003.
[2] Cowper, G.R., “Shear Coefficient in Timoshenko Beam Theory,” J. Appl. Mech.,
33(2)(1966):335-346
[3] Dong, S.B, Alpdogan, C., and Taciroglu, E. “Much ado about shear correction factors in
Timoshenko beam theory,” Int. J. Solids & Structures, 47(2010):1651-1655.
[4] Gruttman, F., and Wagner, W., “Shear correction factors in Timoshenko’s beam theory
for arbitrary shaped cross-sections,” Comp. Mech. 27(2001):199-207.
[5] Kaneko, T., “An experimental study of the Timoshenko’s shear coefficient for flexurally
vibrating beams,” J. Phys. D: Appl. Phys. 11 (1978): 1979-1988;
[6] Paz, Mario, Structural Dynamics Theory and Computation, Chapman & Hall, 2000.
[8] Rosinger, H.E., and Ritchie, I.G., “On Timoshenko’s correction for shear in vibrating
isotropic beams,” J. Phys. D: Appl. Phys., 10 (1977): 1461–1466.