20.12.2013 Views

The Finite Element Method for the Analysis of Non-Linear and ...

The Finite Element Method for the Analysis of Non-Linear and ...

The Finite Element Method for the Analysis of Non-Linear and ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

<strong>The</strong> <strong>Finite</strong> <strong>Element</strong> <strong>Method</strong> <strong>for</strong> <strong>the</strong> <strong>Analysis</strong> <strong>of</strong><br />

<strong>Non</strong>-<strong>Linear</strong> <strong>and</strong> Dynamic Systems<br />

Pr<strong>of</strong>. Dr. Eleni Chatzi<br />

Lecture 9 - 27 November, 2012<br />

Institute <strong>of</strong> Structural Engineering <strong>Method</strong> <strong>of</strong> <strong>Finite</strong> <strong>Element</strong>s II 1


State Space Equation Formulation<br />

2d<strong>of</strong> Mass Spring System<br />

k 1<br />

c 1<br />

m 1<br />

( t 1 )<br />

x 1 ( t )<br />

F k<br />

2<br />

m<br />

2<br />

c 1<br />

F ( t 2 )<br />

x 2 ( t )<br />

FBD<br />

k1x1<br />

cx<br />

1 1<br />

m 1<br />

( t 1 )<br />

k<br />

( x2 − x1)<br />

k2( x2 − x1)<br />

2<br />

m<br />

2<br />

F<br />

c2( x2 − x1)<br />

c2( x2 − x1)<br />

(Lumped Mass System)<br />

F ( t 2 )<br />

We introduce <strong>the</strong> augmented state vector:<br />

x = [ x 1 x 2 ẋ 1 ẋ 2<br />

] T (controllable <strong>for</strong>m equivalent). <strong>The</strong>n,<br />

ẋ =<br />

⎡<br />

⎢<br />

⎣<br />

⎤<br />

0 0 1 0<br />

0 0 0 1<br />

[<br />

−m −1 k ] [ −m −1 c ] ⎥<br />

⎦<br />

⎡ ⎤ ⎡<br />

x 1<br />

⎢ x 2<br />

⎥<br />

⎣ ẋ 1<br />

⎦ + ⎢<br />

⎣<br />

ẋ 2<br />

⎤<br />

0 0<br />

0 0<br />

[<br />

m<br />

−1 ] ⎥<br />

⎦<br />

[<br />

F1<br />

F 2<br />

]<br />

Institute <strong>of</strong> Structural Engineering <strong>Method</strong> <strong>of</strong> <strong>Finite</strong> <strong>Element</strong>s II 2


State Space Equation Formulation<br />

2d<strong>of</strong> Mass Spring System<br />

where u = [ F 1<br />

F 2<br />

] T<br />

⇒ ẋ = Ax + Bu<br />

Assume you would like to monitor <strong>the</strong> displacement x 1 , x 2 . <strong>The</strong>n<br />

<strong>the</strong> “observation vector” is:<br />

y =<br />

⎡<br />

⎢<br />

⎣<br />

1 0 0 0<br />

0 1 0 0<br />

0 0 0 0<br />

0 0 0 0<br />

⎤ ⎡<br />

⎥ ⎢<br />

⎦ ⎣<br />

⎤<br />

x 1<br />

x 2<br />

x 3<br />

x 4<br />

⎥<br />

⎦ + O 4×2 u(t)<br />

⇒ y = Cx + Du<br />

Institute <strong>of</strong> Structural Engineering <strong>Method</strong> <strong>of</strong> <strong>Finite</strong> <strong>Element</strong>s II 3


State Space Equation Formulation<br />

Using <strong>the</strong> state space representation we have converted a 2nd order ODE<br />

into an equivalent 1st order ODE system. We can now use any 1st order<br />

ODE integration method to convert <strong>the</strong> continuous system into a<br />

discrete one <strong>and</strong> obtain an approximate solution:<br />

1st order ODE Integration <strong>Method</strong>s<br />

Assume dy<br />

dt = f(t, y(t)), y(t 0) = 0<br />

Forward Euler <strong>Method</strong><br />

y n+1 = y n + hf(t n , y n )<br />

where h is <strong>the</strong> integration time step. This explicit expression is<br />

obtained from <strong>the</strong> truncated Taylor Expansion <strong>of</strong> y(t n + h)<br />

Backward Euler <strong>Method</strong><br />

y n+1 = y n + hf(t n+1, y n+1 )<br />

This implicit expression (since y n+1 is on <strong>the</strong> right h<strong>and</strong> side) is<br />

obtained from <strong>the</strong> truncated Taylor Expansion <strong>of</strong> y(t n+1 − h)<br />

Institute <strong>of</strong> Structural Engineering <strong>Method</strong> <strong>of</strong> <strong>Finite</strong> <strong>Element</strong>s II 4


State Space Equation Formulation<br />

2nd Order Runge Kutta (RK2)<br />

k 1 = hf(t n , y n ), k 2 = hf(t n + 1 2 h, y n + 1 2 k 1)<br />

y n+1 = y n + k 2 + O(h 3 )<br />

4th Order Runge Kutta (RK4)<br />

k 1 = hf(t n , y n ), k 2 = hf(t n + 1 2 h, y n + 1 2 k 1)<br />

k 3 = hf(t n + 1 2 h, y n + 1 2 k 2), k 4 = hf(t n + h, y n + k 3 )<br />

y n+1 = y n + 1 6 k 1 + 1 3 k 2 + 1 3 k 3 + 1 6 k 4 + O(h 5 )<br />

Institute <strong>of</strong> Structural Engineering <strong>Method</strong> <strong>of</strong> <strong>Finite</strong> <strong>Element</strong>s II 5


Mode Superposition <strong>Method</strong><br />

Direct Integration <strong>Method</strong>s<br />

Assuming diagonal mass matrix <strong>and</strong> no damping, implicit methods<br />

require on average 2nm k calculations per time step<br />

where n is <strong>the</strong> <strong>the</strong> dimension <strong>of</strong> <strong>the</strong> stiffness matrix <strong>and</strong> m k <strong>the</strong><br />

half-b<strong>and</strong>width<br />

<strong>The</strong> central difference method (explicit scheme) requires significantly<br />

fewer operations per time step however has o<strong>the</strong>r drawbacks (requires<br />

T < T cr ⇒ more time steps!)<br />

<strong>The</strong> total number <strong>of</strong> operations in a Direct Integration <strong>Method</strong> are<br />

directly proportional to <strong>the</strong> number <strong>of</strong> load steps ⇒ only effective <strong>for</strong><br />

short durations (small number <strong>of</strong> time steps)<br />

Mode Superposition <strong>Method</strong>s<br />

<strong>The</strong>se methods work by trans<strong>for</strong>ming <strong>the</strong> equilibrium equations into a<br />

<strong>for</strong>m in which step by step solution is less costly<br />

In particular we aim at a reduction <strong>of</strong> m k which would proportionally<br />

decrease <strong>the</strong> step by step solution<br />

Institute <strong>of</strong> Structural Engineering <strong>Method</strong> <strong>of</strong> <strong>Finite</strong> <strong>Element</strong>s II 6


Mode Superposition <strong>Method</strong><br />

Equilibrium Equations Trans<strong>for</strong>mation - Change <strong>of</strong> Basis<br />

<strong>The</strong> following trans<strong>for</strong>mation is used on <strong>the</strong> n finite element nodal<br />

point displacements U:<br />

U(t) = PX(t)<br />

where P is an n × n square matrix <strong>and</strong> X(t) is <strong>the</strong> “generalized<br />

displacement” vector (time dependent)<br />

<strong>The</strong> dynamic equation <strong>of</strong> motion <strong>the</strong>n can be written as:<br />

˜MẌ(t) + ˜CẊ(t) + ˜KX(t) = ˜R(t)<br />

where ˜M = P T MP ˜C = P T CP ˜K = P T KP ˜R = P T R<br />

are <strong>the</strong> trans<strong>for</strong>med matrices corresponding to a smaller b<strong>and</strong>width<br />

than <strong>the</strong> original system.<br />

Institute <strong>of</strong> Structural Engineering <strong>Method</strong> <strong>of</strong> <strong>Finite</strong> <strong>Element</strong>s II 7


Mode Superposition <strong>Method</strong><br />

Choice <strong>of</strong> Trans<strong>for</strong>mation Matrix P<br />

P is established using <strong>the</strong> displacement solution <strong>for</strong> Undamped, Free<br />

vibrations:<br />

MÜ + KU = 0 (1)<br />

<strong>The</strong> solution <strong>of</strong> this 2nd Order ODE is <strong>of</strong> <strong>the</strong> <strong>for</strong>m:<br />

U = φsinω(t − t 0 )<br />

where φ is a vector <strong>of</strong> order n, t is <strong>the</strong> time variable, t 0 is a time<br />

constant corresponding to <strong>the</strong> phase <strong>and</strong> ω is <strong>the</strong> vibration frequency<br />

<strong>of</strong> vector φ, <strong>the</strong>n<br />

Ü = −ω 2 φsinω(t − t 0 )<br />

<strong>and</strong> (1) becomes:<br />

−ω 2 Mφ + Kφ = 0 (eigenproblem)<br />

Institute <strong>of</strong> Structural Engineering <strong>Method</strong> <strong>of</strong> <strong>Finite</strong> <strong>Element</strong>s II 8


Mode Superposition <strong>Method</strong><br />

Matrix Form<br />

Defining a matrix Φ whose columns are <strong>the</strong> eigenvectors φ i <strong>and</strong> a<br />

diagonal matrixΩ 2 which stores <strong>the</strong> eigenvalues ωi 2 on its diagonal,<br />

i.e:<br />

⎡<br />

⎤<br />

ω 1<br />

Φ = [ ]<br />

φ 1 , φ 2 , ... φ n ; Ω 2 = ⎢ ω 2 ⎥<br />

⎣ ... ⎦ ;<br />

ω n<br />

we can write <strong>the</strong> n solutions to <strong>the</strong> eigenproblem as:<br />

MΦΩ 2 = KΦ<br />

Institute <strong>of</strong> Structural Engineering <strong>Method</strong> <strong>of</strong> <strong>Finite</strong> <strong>Element</strong>s II 9


Properties <strong>of</strong> Eigenvectors<br />

Solution <strong>of</strong> <strong>the</strong> Eigenproblem<br />

−ω 2 Mφ + Kφ = 0<br />

(eigenproblem)<br />

<strong>The</strong> above homogeneous linear system <strong>of</strong> equations can only have a<br />

solution if <strong>the</strong> determinant is equal to 0:<br />

∥<br />

∥−ω 2 M + K ∥ ∥ = 0<br />

<strong>The</strong> solution <strong>of</strong> <strong>the</strong> above equation will yield:<br />

n eigenvalues ωi 2,<br />

i = 1, ..., n<br />

with 0 ≤ ω 1 ≤ ...ω n (<strong>the</strong> eigenfrequencies) <strong>and</strong><br />

<strong>The</strong> solution <strong>of</strong> ω 2 Mφ = Kφ (eigenproblem) will yield:<br />

n eigenvectors (or modal vectors) φ i<br />

Institute <strong>of</strong> Structural Engineering <strong>Method</strong> <strong>of</strong> <strong>Finite</strong> <strong>Element</strong>s II 10


Properties <strong>of</strong> Eigenvectors<br />

M-Orthonormality<br />

From Matrix Properties we know that (AB) T = B T A T ,<br />

Thus, <strong>for</strong> two eigenvectors φ n , φ r we obtain:<br />

(φ T nKφ r ) T = φ T rK T φ n<br />

Since <strong>the</strong> Stiffness matrix is symmetric, K = K T , hence<br />

(φ T nKφ r ) T = φ T rKφ n<br />

<strong>The</strong> same applies <strong>for</strong> <strong>the</strong> Mass matrix, M = M T , yielding:<br />

(φ T nMφ r ) T = φ T rMφ n<br />

Institute <strong>of</strong> Structural Engineering <strong>Method</strong> <strong>of</strong> <strong>Finite</strong> <strong>Element</strong>s II 11


Properties <strong>of</strong> Eigenvectors<br />

M-Orthonormality<br />

<strong>The</strong> eigenproblem <strong>for</strong> vector n <strong>the</strong>n can be written as:<br />

ωrφ 2 T nMφ r = φ T transpose<br />

nKφ r −→<br />

ω 2 rφ T rMφ n = φ T rKφ n<br />

<strong>and</strong> <strong>the</strong> eigenproblem <strong>for</strong> vector n is: <strong>The</strong> eigenproblem <strong>for</strong> vector n<br />

<strong>the</strong>n can be written as:<br />

ω 2 nφ T rMφ n = φ T rKφ n<br />

By subtracting <strong>the</strong> two previous <strong>for</strong>mulas we obtain:<br />

(ω 2 r − ω 2 n)φ T rMφ n = 0<br />

We conclude that <strong>for</strong> n ≠ r ⇒ φ T rMφ n = 0, φ T rMφ n = 0,<br />

Institute <strong>of</strong> Structural Engineering <strong>Method</strong> <strong>of</strong> <strong>Finite</strong> <strong>Element</strong>s II 12


Properties <strong>of</strong> Eigenvectors<br />

M-Orthonormality<br />

We choose φ n , such that:<br />

Φ T MΦ = I<br />

<strong>the</strong>re<strong>for</strong>e from MΦΩ 2 = KΦ we have that:<br />

Φ T KΦ = Ω 2<br />

<strong>The</strong> principle <strong>of</strong> M-orthonormality can <strong>the</strong>n be written as:<br />

{ 1, n = r<br />

φ T rMφ n =<br />

0, n ≠ r<br />

{ ω<br />

2<br />

φ T rKφ n = n , n = r<br />

0, n ≠ r<br />

Institute <strong>of</strong> Structural Engineering <strong>Method</strong> <strong>of</strong> <strong>Finite</strong> <strong>Element</strong>s II 13


Mode Superposition <strong>Method</strong><br />

It is now obvious that we can use Φ as <strong>the</strong> trans<strong>for</strong>mation matrix P.<br />

Using:<br />

U(t) = ΦX(t)<br />

we obtain <strong>the</strong> trans<strong>for</strong>med equilibrium equation:<br />

Ẍ(t) + Φ T CΦẊ(t) + Ω 2 X(t) = Φ T R(t)<br />

where using <strong>the</strong> property <strong>of</strong> M-orthonormality, <strong>the</strong> initial conditions<br />

will be:<br />

0 X = Φ T M 0 U;<br />

0 Ẋ = Φ T M 0 ˙U<br />

Institute <strong>of</strong> Structural Engineering <strong>Method</strong> <strong>of</strong> <strong>Finite</strong> <strong>Element</strong>s II 14


Dynamic Response with Damping Neglected<br />

If we neglect <strong>the</strong> velocity dependent damping effects <strong>the</strong> equilibrium<br />

equation reduces to:<br />

Ẍ(t) + Ω 2 X(t) = Φ T R(t)<br />

i.e n individual equations <strong>of</strong> <strong>the</strong> <strong>for</strong>m (since Ω 2 is diagonal):<br />

ẍ i (t) + ωi 2x }<br />

i(t) = r i (t)<br />

r i (t) = φ T i R(t) i = 1, 2, ..., n<br />

with<br />

0 x i = φ T i M 0 U;<br />

0ẋ i = φ T i M 0 ˙U<br />

Institute <strong>of</strong> Structural Engineering <strong>Method</strong> <strong>of</strong> <strong>Finite</strong> <strong>Element</strong>s II 15


Dynamic Response with Damping Neglected<br />

SDOF Response<br />

Each equation <strong>of</strong> <strong>the</strong> previous system describes a single degree <strong>of</strong> freedom<br />

system with unit mass <strong>and</strong> stiffness ωi 2 . <strong>The</strong> solution to this equation <strong>for</strong> a<br />

r<strong>and</strong>om input excitation can ei<strong>the</strong>r be obtained by using <strong>the</strong> direct<br />

integration methods or by using <strong>the</strong> Duhamel Integral:<br />

x i (t) = 1 ∫ t<br />

r i (τ)sinω i (t − τ) dτ + α i sinω i t + β i cosω i t<br />

ω i<br />

0<br />

where α i , β i are determined from <strong>the</strong> initial conditions. <strong>The</strong>re<strong>for</strong>e, <strong>the</strong><br />

SDOF response id owed to two contributions<br />

A dynamic (steady - state) response obtained by multiplying <strong>the</strong> static<br />

response by a dynamic load factor (this is <strong>the</strong> particular solution <strong>of</strong><br />

<strong>the</strong> governing differential equation), <strong>and</strong><br />

An additional dynamic response called <strong>the</strong> transient response<br />

Institute <strong>of</strong> Structural Engineering <strong>Method</strong> <strong>of</strong> <strong>Finite</strong> <strong>Element</strong>s II 16


Dynamic Response with Damping Neglected<br />

Complete Response<br />

<strong>The</strong> solution <strong>of</strong> all n SDOF equations are calculated <strong>and</strong> <strong>the</strong> finite<br />

element nodal point displacements are obtained by superposition <strong>of</strong><br />

<strong>the</strong> response in each mode:<br />

U(t) = ΦX(t) ⇒ U(t) =<br />

n∑<br />

φ i x i (t)<br />

i=1<br />

<strong>The</strong>re<strong>for</strong>e <strong>the</strong> solution scheme is:<br />

Solve <strong>for</strong> <strong>the</strong> eigenvalues <strong>and</strong> eigenvectors <strong>of</strong> problem<br />

Solve <strong>for</strong> <strong>the</strong> response <strong>of</strong> <strong>the</strong> decoupled SDOF equations<br />

Use superposition to find <strong>the</strong> total response.<br />

Institute <strong>of</strong> Structural Engineering <strong>Method</strong> <strong>of</strong> <strong>Finite</strong> <strong>Element</strong>s II 17


Superposition Principle<br />

An alternative View<br />

Superposition<br />

A dynamic load can be designed as<br />

Fourier series <strong>of</strong> harmonic sine <strong>and</strong> cosine<br />

contributions<br />

<strong>The</strong> total solution <strong>of</strong> such a problem is<br />

equal to <strong>the</strong> superposition <strong>of</strong> solution <strong>of</strong><br />

<strong>the</strong> Fourier terms.<br />

A dynamic load can be designed as Fourier series<br />

<strong>The</strong> total solution <strong>of</strong> such a problem is equal to<br />

<strong>the</strong> superposition <strong>of</strong> solution <strong>of</strong> <strong>the</strong> Fourier terms.<br />

U(t) = ΦX(t) ⇒ U(t) =<br />

n∑<br />

φ i x i (t)<br />

i=1<br />

13.11.2009 Mode Superposition 11<br />

Institute <strong>of</strong> Structural Engineering <strong>Method</strong> <strong>of</strong> <strong>Finite</strong> <strong>Element</strong>s II 18


<strong>The</strong> effect <strong>of</strong> Damping<br />

Problems with neglected dampi<br />

<strong>The</strong> presence <strong>of</strong> damping reduces <strong>the</strong> dynamic load factor (which <strong>the</strong>n<br />

cannot be infinite) <strong>and</strong> damps out <strong>the</strong> transient response<br />

<strong>The</strong> response in <strong>the</strong><br />

modes with ˆω large is<br />

ω<br />

negligible<br />

For ˆω close to zero <strong>the</strong><br />

ω<br />

system follows <strong>the</strong> loads<br />

statically<br />

Effectively only <strong>the</strong> first p modes need to be used p ≤ n, in order to obtain<br />

a good approximate solution ⇒ Advantage over Direct Integration <strong>Method</strong>s<br />

13.11.2009 Mode Superposition<br />

Institute <strong>of</strong> Structural Engineering <strong>Method</strong> <strong>of</strong> <strong>Finite</strong> <strong>Element</strong>s II 19


Swiss Federal Institute <strong>of</strong> Technology<br />

Example - 2 DOF system<br />

Direct Integration ti <strong>Method</strong>s<br />

Calculate <strong>the</strong> displacement response <strong>of</strong> <strong>the</strong> system<br />

Pag<br />

k<br />

1<br />

= 4<br />

U , U , U<br />

1 1 1<br />

m = 2 R = 0<br />

1<br />

2<br />

k<br />

2<br />

= 2<br />

1<br />

0<br />

m<br />

2<br />

= 1 R<br />

2<br />

= 10<br />

⎡2 0⎤⎡U<br />

⎤ 6 2<br />

1 ⎡ − ⎤⎡U1<br />

⎤ ⎡0⎤<br />

⎢ ⎥⎢ ⎥+ ⎢ =<br />

⎥⎢ ⎥ ⎢ ⎥<br />

⎣0 1 ⎦<br />

<br />

⎣U<br />

2<br />

⎦<br />

⎣−2 4 ⎦ ⎣U<br />

2<br />

⎦<br />

⎣10<br />

⎦<br />

U , U , U<br />

2 2 2<br />

k<br />

3<br />

= 2<br />

<strong>Method</strong> <strong>of</strong> <strong>Finite</strong> <strong>Element</strong>s II<br />

Institute <strong>of</strong> Structural Engineering <strong>Method</strong> <strong>of</strong> <strong>Finite</strong> <strong>Element</strong>s II 20


Example - 2 DOF system<br />

Eigenproblem Setup<br />

ω 2 nMφ n = Kφ n ⇒ ω 2 n<br />

Eigenvalue Calculation<br />

[ 2, 0<br />

0 1<br />

]<br />

=<br />

[ 6, −2<br />

−2 4<br />

]<br />

φ n<br />

∥ K − ω<br />

2<br />

n M ∥ ∥ ∥∥∥ 6 − 2ω 2<br />

= 0 ⇒ n,<br />

−2<br />

−2<br />

4 − 1ωn<br />

2<br />

∥ = 2ω4 n − 14ωn 2 + 20 = 0 ⇒<br />

ω1 2 = 14 − √ 196 − 160<br />

4<br />

Eigenvector Calculation<br />

= 2, ω 2 2 = 14 + √ 196 − 160<br />

4<br />

= 5<br />

(<br />

K − ω<br />

2<br />

1 M ) φ 1 = 0 ⇒<br />

[ 2, −2<br />

−2 2<br />

] [ ]<br />

φ11<br />

= 0 ⇒<br />

φ 12<br />

[<br />

φ11<br />

φ 12<br />

]<br />

[<br />

1<br />

= λ ∗<br />

1<br />

]<br />

Institute <strong>of</strong> Structural Engineering <strong>Method</strong> <strong>of</strong> <strong>Finite</strong> <strong>Element</strong>s II 21


Example - 2 DOF system<br />

Use <strong>of</strong> M - orthonormality<br />

φ T 1Mφ 1 = 1 ⇒ λ 2 (m 1 + m 2 ) = 1 ⇒ λ = 1 3<br />

⎡<br />

1<br />

⎤<br />

⎡ √ ⎤ 2<br />

√<br />

⎢<br />

Hence, φ 1 = ⎣ 3 ⎥<br />

1 ⎦ Similarly, φ 2 = ⎢ 2 √ 3<br />

√ ⎥<br />

⎣<br />

√ 2 ⎦<br />

−√ 3 3<br />

Trans<strong>for</strong>med Equilibrium Equation<br />

[ 2, 0<br />

Ẍ(t) +<br />

0 5<br />

Ẍ(t) + Ω 2 X(t) = Φ T R(t) ⇒<br />

⎡<br />

]<br />

X(t) = ⎢<br />

⎣<br />

⎤<br />

1 1<br />

√ √3 [<br />

√ 3 √ ⎥ 0<br />

2 2<br />

2 √ ⎦ 10<br />

−√ 3 3<br />

]<br />

Institute <strong>of</strong> Structural Engineering <strong>Method</strong> <strong>of</strong> <strong>Finite</strong> <strong>Element</strong>s II 22


Example - 2 DOF system<br />

Decoupled SDOF equations<br />

ẍ 1 + 2x 1 = 10 √<br />

3<br />

√<br />

2<br />

ẍ 2 + 5x 2 = −10<br />

3<br />

Initial Conditions<br />

U(0) = 0<br />

˙U(0) = 0<br />

<strong>and</strong><br />

x i (0) = φ T i MU(0)<br />

ẋ i (0) = φ T i M ˙U(0)<br />

⇒<br />

x 1 (0) = 0, ẋ 1 (0) = 0, x 2 (0) = 0, ẋ 2 (0) = 0<br />

<strong>The</strong>n <strong>the</strong> exact solutions to <strong>the</strong> ODEs are:<br />

x 1 = 5 √<br />

3<br />

(1 − cos √ 2t) x 2 = 2√<br />

2<br />

3 (−1 + cos√ 5t)<br />

And <strong>the</strong> total solution is:<br />

n∑<br />

U(t) = φ i x i (t)<br />

i=1<br />

Institute <strong>of</strong> Structural Engineering <strong>Method</strong> <strong>of</strong> <strong>Finite</strong> <strong>Element</strong>s II 23


Example - 2 DOF system<br />

Comparison <strong>of</strong> Direct Integration <strong>and</strong> Modal Superposition<br />

<strong>Method</strong>s<br />

Institute <strong>of</strong> Structural Engineering <strong>Method</strong> <strong>of</strong> <strong>Finite</strong> <strong>Element</strong>s II 24


Example - 2 DOF system<br />

Note<br />

In this case we dealt with a 2DOF example <strong>and</strong> used an analytical<br />

solution to solve <strong>the</strong> two decoupled SDOF system equations<br />

In practice <strong>for</strong> multi degree <strong>of</strong> freedom systems, higher modes are<br />

neglected <strong>and</strong> instead <strong>of</strong> an analytical solution we again use a Direct<br />

Integration scheme to solve <strong>for</strong> each SDOF ODE<br />

In essence <strong>the</strong> finally achieved accuracy might be <strong>the</strong> same between<br />

<strong>the</strong> two approaches, however teh mode superposition method usually<br />

requires <strong>for</strong> less computational cost<br />

Institute <strong>of</strong> Structural Engineering <strong>Method</strong> <strong>of</strong> <strong>Finite</strong> <strong>Element</strong>s II 25


<strong>Analysis</strong> Including Damping<br />

Trans<strong>for</strong>med Equilibrium Equation<br />

Ẍ(t) + Φ T CΦẊ(t) + Ω 2 X(t) = Φ T R(t)<br />

Proportional Damping Assumption<br />

φ T i Cφ j = 2ω i ξ i δ ij (2)<br />

where ξ i is a modal damping parameter <strong>and</strong> δ ij is <strong>the</strong> Kronecker delta<br />

(δ ij = 1 <strong>for</strong> i = j, δ ij = 0 <strong>for</strong> i ≠ j)<br />

Institute <strong>of</strong> Structural Engineering <strong>Method</strong> <strong>of</strong> <strong>Finite</strong> <strong>Element</strong>s II 26


<strong>Analysis</strong> Including Damping<br />

<strong>The</strong>re<strong>for</strong>e, we still end up with Decoupled SDOF equations <strong>for</strong><br />

each x i :<br />

ẍ i (t) + 2ω i ξ i ẋ i (t) + ω 2 i x i (t) = r i (t)<br />

with <strong>the</strong> Duhamel Integral now being:<br />

x i (t) = 1¯ω ∫ t {<br />

r i (τ) e −ξ iω i (t−τ) sin¯ω i (t − τ) dτ<br />

i 0<br />

}<br />

+ e −ξ iω i t (α i sin¯ω i t + β i cos¯ω i t)<br />

where ¯ω i = ω i<br />

√1 − ξ 2 i<br />

Institute <strong>of</strong> Structural Engineering <strong>Method</strong> <strong>of</strong> <strong>Finite</strong> <strong>Element</strong>s II 27


<strong>Analysis</strong> Including Damping<br />

Rayleigh Damping<br />

If <strong>the</strong>re are only two different damping ratios ξ i ,<br />

Damping can be used:<br />

i = 1, 2 Rayleigh<br />

C = αM + βK (3)<br />

Eqns (2), (3) now yield:<br />

φ T i (αM + βK)φ i = 2ω i ξ i ⇒<br />

α + βω 2 i = 2ω i ξ i<br />

<strong>The</strong> 2 × 2 system can be solved to obtain α, β.<br />

In actual analysis it may well be that <strong>the</strong> damping ratios are known<br />

<strong>for</strong> many more than two frequencies. In that case two average values<br />

say ¯ξ 1 , ¯ξ 2 are used to evaluate α, β.<br />

Institute <strong>of</strong> Structural Engineering <strong>Method</strong> <strong>of</strong> <strong>Finite</strong> <strong>Element</strong>s II 28


Measuring Damping<br />

Logarithmic decrement method can be used to measure damping in time<br />

domain. In this method, <strong>the</strong> free vibration displacement amplitude history <strong>of</strong> a<br />

system to an impulse is recorded. Logarithmic decrement is <strong>the</strong> natural<br />

logarithmic value <strong>of</strong> <strong>the</strong> ratio <strong>of</strong> two adjacent peak values <strong>of</strong> displacement in free<br />

decay vibration.<br />

Institute <strong>of</strong> Structural Engineering <strong>Method</strong> <strong>of</strong> <strong>Finite</strong> <strong>Element</strong>s II 29


Example - Damping <strong>for</strong> an MDOF system<br />

Assume that <strong>the</strong> approximate damping to be specified <strong>for</strong> a multiple degree <strong>of</strong><br />

freedom system is as follows:<br />

Institute <strong>of</strong> Structural Engineering <strong>Method</strong> <strong>of</strong> <strong>Finite</strong> <strong>Element</strong>s II 30


Example - Damping <strong>for</strong> an MDOF system<br />

Damping as function <strong>of</strong> frequency<br />

A problem with Rayleigh<br />

damping is that higher<br />

modes are much more<br />

damped than lower modes.<br />

However, in general<br />

Rayleigh damping provides<br />

a good approximation<br />

Institute <strong>of</strong> Structural Engineering <strong>Method</strong> <strong>of</strong> <strong>Finite</strong> <strong>Element</strong>s II 31

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!