Coordinate Transformation
Transformation Matrix of Axial Deformation and Bending of
U5
U6
x
U4
U2
j
U3
q
X
i
U1
u5 u6 u2 u4
y x
u3
u1
Y
y
U2
i
x
u1
u2
U1
(14.23a)
(14.23b)
In matrix form
{u} = [ R]{U }
where
2
(14.24)
{u}T = {u1
u2
u3 U3
u4 U4
u5
u6 } U5 U6}
{U }T = {U 1 U 2
U1 U 2 U 3 U 4 U 5 U 6 . . . C S . - S C . . . . . . 1 . . . [ R] = . . . C S . . . . - S C . . . . . . 1 66
L2 = ( X j - X i ) 2 + (Y j - Yi ) 2
(14.28)
cosq =
X j - Xi L
sin q =
Y j - Yi L
{ P } = [ R]T { p}
C = cosq = S = sin q =
X j - Xi L Y j - Yi L
[ K e ]6 6 = [Te ]T [k e ][Te ]
[ M e ]66 = [ Re ]T [m e ][ Re ]
{ Pe }61 = [ R]T { pe }
Note: If q = 0 , then [ R] = [ I ] [ K ] = [k ]
[ M ] = [m ]
{ P } = { p}
Axial Deformation and Bending of Beam on Plane 1 [ K ( e ) ]nn = [ Le ]T [ K e ]66 [ Le ]6n Ve = {U }T [ K ( e ) ]{U } 2 1 [ M ( e ) ]nn = [ Le ]T [ M e ][ Le ] Te = {U }T [ M ( e ) ]{U } 2
dW nc ,e = {dU }T { P ( e ) }
n : system DOF
{ P ( e ) }n1 = [ Le ]T { Pe }
T = Te
e =1
NE
[ M ] = [ M (e ) ]
e =1
V = Ve
e =1
1 NE {U }T [ K ( e ) ]{U } 2 e =1 1 = {U }T [ K ]{U } 2 =
NE
NE
[ K ] = [ K (e ) ]
e =1
dW nc = dW nc ,e + {dU }T { Pext }
e =1
NE ( e ) = {dU }T { P } + { P } ext e =1
= {dU } { P }
T
NE
{P} =
{ P ( e ) } + { Pext }
e =1
d T T V + = {P} & } {U } {U } dt {U
Lagrange Equations
s x ( x , t ) = Ee x ( x , t )
= [ D]11 [ B ]14 {v ]41
V= 1 T ex s x dV V 2 1 = {v }T ( V [ B ]T [ D][ B ]dV ){v } 2 1 = {v }T [k ]{v } 2
[ D] = E
For a Beam
[k ] = V [ B ]T [ D][ B ]dV
T=
( 4 ) V = Ve
e =1
NE
[ K e ] = [ Re ]T [k ][ Re ] [ K ( e ) ] = [ Le ]T [ K e ][ Le ]
NE
[ K ] = [ K (e ) ]
e =1
NE
dW nc = dW nc ,e + {dU } { Pext }
e =1
dW nc = {dU } { P ]
{ P } = { P ( e ) } + { Pext }
e =1
(5)
d T T V + = {P} & } {U } {U } dt {U
&& } + [ K ]{U ] = { P } [ M ]{U
[ K e ] = [Te ]T [k ][Te ] [ K ( e ) ] = [ Le ]T [ K e ][ Le ]
NE
[ K ] = [ K (e ) ]
e =1
(5) dW nc - dV + dW inertia = 0
14.8 Finite Element Solutions for Natural Frequencies and Modes a. A hand-crank 2-DOF solution based on the finite element model b. The effect of increasing the number of degrees-of-freedom in a finite element model based on a consistent mass matrix. c. The effect of reducing out freedoms by using the Guyan reduction method. d. The comparison of lumped mass models with consistent mass models Example14.7 Accuracy of 2-DOF Model
v( x , t )
U2 U1
rAL 156
420 - 22 L
- 22 L U 1 EI 12 - 6 L U 1 0 + 3 = 2 U U L 4 L2 6 L 4 L 2 2 0
b. Natural Frequencies
&&} + [ K ]{u} = 0 [ M ]{u
U ( t ) = f cos w t
(2) (3)
[ K - mM ]{f } = {0}
m=
w 2 rAL4
240 EI
(4)
det[ K - mM ] = 0
12 - 6 L 156 - 22 L f1 0 det m - 22 L 4 L2 f = 0 2 6 L 4 L 2
(5) (6)
m 1 = 2.97147 10 -2 m 2 = 2.88457
w 12 = 12.4802 w 22 = 1211.52
EI 4 rAL EI 4 rAL
1/ 2
EI w 1 = 3.533 rAL4
w1
1/ 2
exact
EI = 3.516 rAL4
1/ 2
w 2 = 34.81
EI 4 r AL
w2
exact
EI = 22.03 rAL4
1/ 2
(7)
FE Model Comment:
(w 1 ) FEM > (w 1 ) exact
Continuum Model
Why? The FEM model is a constrained model. Therefore it is stiffer than a real structure.
9
Example 14.10 Comparison of Consistent Mass Finite Element Models of a Uniform Cantilever Beam with Continuum Model
3
1 2
5
4
Element coordinates
2 1
E = r = A= L= I =1
System coordinates
(1)
( N e - 1)
(Ne )
Table 14.1 Comparison of Consistent Mass Finite Element Models of a Uniform Cantilever Beam with Continuum Model
Ne Mode No. 1 2 3 4 5 6 7 8 9 10 1 3.53273 34.8069 2 3.51772 22.2215 75.1571 218.138 3 3.51637 22.1069 62.4659 140.671 264.743 527.796 4 3.51613 22.0602 62.1749 122.657 228.137 366.390 580.849 953.051 5 3.51606 22.0455 61.9188 122.320 203.020 337.273 493.264 715.341 1016.20 1494.88 Exact (Reference 14.1) 3.51602 22.0345 61.6972 120.903 199.860 298.566 416.991 555.165 713.079 890.732
10
5 5 5
L1 = 0.5
4 2
3 5
L2 = 0.5
Table 14.2 Frequencies Obtained by Using Guyan Reduction to Reduce out Rotations of Uniform Cantilever Beams
Ne Mode No. 1 2 3 4 5 1 3.56753 2 3.52198 22.2790 3 3.51699 22.2362 62.6685 4 3.51628 22.0946 62.9703 123.545 5 3.51611 22.0573 62.2180 124.725 205.277 Exact (Reference 14.1) 3.51602 22.0345 61.6972 120.903 199.860
Comment - l FEM > lexact - as the number of elements increases, the error becomes smaller - the higher modes have larger errors
11
Lumped Mass Models Table 14.3 Frequencies of a Uniform Cantilever Beam Based on Lumped Mass Models\
Ne Mode No. 1 2 3 4 5 1 2.44949 2 3.15623 16.2580 3 3.34568 18.8859 47.0284 4 3.41804 20.0904 53.2017 92.7302 5 3.45266 20.7335 55.9529 104.436 153.017 Exact (Reference 14.1) 3.51602 22.0345 61.6972 120.903 199.860
- as the number of elements increases, the error becomes smaller - the higher modes have larger errors
12
Finite Element Programs General Purpose Programs ADINA: Prof. K-J Bathe, MIT SAP- 4, 6, 8, 2000: Prof. E. L. Wilson, UC. Berkeley NASTRAN: NASA ANSYS, ABAQUS, COSMOS Special Purpose Programs STRUDL: Structural Analysis NEABS: Nonlnear Earthquake Analysis Bridge System SHAKE: Prof. Schnable, EERC, U.C., Berkeley TAB/SAP86, ALGOR, GIFTS, IMAGES-3D, MSC/PAL, RM *MIDAS: Korea
13
Comments derived from the equation of motion under free load, but those for other elements are assumed.
(1) The displacement functions for truss and beam elements are
(1) The static analysis of a structure consisted of truss or beam elements by the FEM gives the exact solution. The dynamic analysis does not. (2) Numbering nodes (3) Storage methods Controls all sub-programs - half-bandwidth method: Building - skyline method: Bridge(2-Bay) (4) Structural analysis programs compute det[ K aa] the stiffness matrix after the boundary conditions are imposed, to check the input data.
L:
Methods
Finite Element Method (FEM): Physical approximation in domain Boundary Element Method (BEM): Physical approximation on boundary Efficient for problems in infinite domain Finite Difference Method (FDM): Mathematical approximation in domain
14
dv dx
=
x = xi
1 ( v i + 1 - v i -1 ) 2h
h=
L n
d 2v dx 2
=
x = xi
1 (v i -1 - 2v i + v i +1 ) h2
d v dx 3 d 4v dx 4
=
x = xi
1 ( - v i - 2 + 2v i -1 - 2v i +1 + v i + 2 ) 2h 3 1 (v i - 2 - 4v i -1 + 6v i - 4v i +1 + v i + 2 ) h4
=
x = xi
15
16
17
18
19