kang-min choi, graduate student, kaist, korea young-jong moon , graduate student, kaist, korea

27
Kang-Min Choi, Kang-Min Choi, Graduate Student, KAIST, Korea Young-Jong Moon Young-Jong Moon, Graduate Student, KAIST, Korea Ji-Eun Jang Ji-Eun Jang, Graduate Student, KAIST, Korea Woon-Hak Kim, Professor, Hankyong National University, Korea In-Won Lee In-Won Lee, Professor, KAIST, Korea Algebraic Method for Eigenpair Derivatives of Damped System with Repeated Eigenvalues

Upload: matthew-snow

Post on 03-Jan-2016

30 views

Category:

Documents


1 download

DESCRIPTION

Algebraic Method for Eigenpair Derivatives of Damped System with Repeated Eigenvalues. Kang-Min Choi, Graduate Student, KAIST, Korea Young-Jong Moon , Graduate Student, KAIST, Korea Ji-Eun Jang , Graduate Student, KAIST, Korea - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Kang-Min Choi,  Graduate Student,  KAIST, Korea Young-Jong Moon ,  Graduate Student,  KAIST, Korea

Kang-Min Choi, Kang-Min Choi, Graduate Student, KAIST, Korea

Young-Jong MoonYoung-Jong Moon, Graduate Student, KAIST, Korea

Ji-Eun JangJi-Eun Jang, Graduate Student, KAIST, Korea

Woon-Hak Kim, Professor, Hankyong National University, Korea

In-Won LeeIn-Won Lee, Professor, KAIST, Korea

Algebraic Method for Eigenpair Derivatives of Damped System

with Repeated Eigenvalues

Page 2: Kang-Min Choi,  Graduate Student,  KAIST, Korea Young-Jong Moon ,  Graduate Student,  KAIST, Korea

Structural Dynamics & Vibration Control Lab., KAIST, KoreaStructural Dynamics & Vibration Control Lab., KAIST, Korea 22

OUTLINE

INTRODUCTION

PREVIOUS METHODS

PROPOSED METHOD

NUMERICAL EXAMPLE

CONCLUSIONS

Page 3: Kang-Min Choi,  Graduate Student,  KAIST, Korea Young-Jong Moon ,  Graduate Student,  KAIST, Korea

Structural Dynamics & Vibration Control Lab., KAIST, KoreaStructural Dynamics & Vibration Control Lab., KAIST, Korea 33

INTRODUCTION

determination of the sensitivity of dynamic response optimization of natural frequencies and mode shapes optimization of structures subject to natural frequencies

Applications of sensitivity analysis are

Typical structures have many repeated or nearly equaleigenvalues, due to structural symmetry.

The second- and higher order derivatives of eigenpairsare important to predict the eigenpairs, which relieson the matrix Taylor series expansion.

Page 4: Kang-Min Choi,  Graduate Student,  KAIST, Korea Young-Jong Moon ,  Graduate Student,  KAIST, Korea

Structural Dynamics & Vibration Control Lab., KAIST, KoreaStructural Dynamics & Vibration Control Lab., KAIST, Korea 44

)( 2 0KCM

Problem Definition Problem Definition

(1)

K

C

M

Eigenvalue problem of damped system

shape) (moder eigenvecto :

frequency) (natural eigenvalue :

definite semi-positive matrix, Stiffness :

matrix Damping :

definite positive matrix, Mass :

Page 5: Kang-Min Choi,  Graduate Student,  KAIST, Korea Young-Jong Moon ,  Graduate Student,  KAIST, Korea

Structural Dynamics & Vibration Control Lab., KAIST, KoreaStructural Dynamics & Vibration Control Lab., KAIST, Korea 55

,

,,, K ,C ,M

K C, M,

,, ,

Given:

Find:

* represents the derivative of with respect design parameter α (length, area, moment of inertia, etc.)

,)( )(

Objective of this study: Objective of this study:

,, ,

Page 6: Kang-Min Choi,  Graduate Student,  KAIST, Korea Young-Jong Moon ,  Graduate Student,  KAIST, Korea

Structural Dynamics & Vibration Control Lab., KAIST, KoreaStructural Dynamics & Vibration Control Lab., KAIST, Korea 66

PREVIOUS STUDIES

Damped system with distinct eigenvalues K. M. Choi, H. K. Jo, J. H. Lee and I. W. Lee, “Sensitivity

Analysis of Non-conservative Eigensystems,” Journal of Sound and Vibration, 2001. (Accepted)

jjj

jj

j

j

jjjj

jjjj

)CM2(5.0

)KCM(

M)CM2(

)CM2(KCM

,,

,,,

,

,2

T

T

TT

- The coefficient matrix is symmetric and non-singular.

- Eigenpair derivatives are obtained simultaneously.

- The algorithm is simple and guarantees stability.

Page 7: Kang-Min Choi,  Graduate Student,  KAIST, Korea Young-Jong Moon ,  Graduate Student,  KAIST, Korea

Structural Dynamics & Vibration Control Lab., KAIST, KoreaStructural Dynamics & Vibration Control Lab., KAIST, Korea 77

Undamped system with repeated eigenvalues R. L. Dailey, “Eigenvector Derivatives with Repeated

Eigenvalues,” AIAA Journal, Vol. 27, pp.486-491, 1989.

- Introduction of Adjacent eigenvector

- Calculation derivatives of eigenvectors by the sum of homogenous solutions and particular solutions using Nelson’s algorithm

- Complicated algorithm and high time consumption

Page 8: Kang-Min Choi,  Graduate Student,  KAIST, Korea Young-Jong Moon ,  Graduate Student,  KAIST, Korea

Structural Dynamics & Vibration Control Lab., KAIST, KoreaStructural Dynamics & Vibration Control Lab., KAIST, Korea 88

Second order derivatives of Undamped systemwith distinct eigenvalues

M. I. Friswell, “Calculation of Second and Higher Eigenvector Derivatives”, Journal of Guidance, Control and Dynamics, Vol. 18, pp.919-921, 1995.

j,jjjj,jjj

jjjjjj

)MMK()MMK(

)MMMK(

,,,,,,

,,,,,,,

TT

T

(3)

jjjj cv,

where j,jj,jjjj ,,,T )MK()MK(M5.0c

(4)

- Second eigenvector derivatives extended by Nelson’s algorithm

Page 9: Kang-Min Choi,  Graduate Student,  KAIST, Korea Young-Jong Moon ,  Graduate Student,  KAIST, Korea

Structural Dynamics & Vibration Control Lab., KAIST, KoreaStructural Dynamics & Vibration Control Lab., KAIST, Korea 99

PROPOSED METHOD

First-order eigenpair derivatives of damped systemwith repeated eigenvalues

Second-order eigenpair derivatives of damped systemwith repeated eigenvalues

Numerical stability of the proposed method

Page 10: Kang-Min Choi,  Graduate Student,  KAIST, Korea Young-Jong Moon ,  Graduate Student,  KAIST, Korea

Structural Dynamics & Vibration Control Lab., KAIST, KoreaStructural Dynamics & Vibration Control Lab., KAIST, Korea 1010

First-order eigenpair derivatives of damped systemwith repeated eigenvalues

Basic Equations

• Eigenvalue problem

(5)

(6)

0KCM 2 mmmmm

mmmm I)CM2(T

• Orthonormalization condition

mm

m

seigenvalue repeated ofy mutiplicit ofnumber the:

rseigenvecto ofmatrix the

( seigenvalue ofmatrix the

:

)I: mmm

Page 11: Kang-Min Choi,  Graduate Student,  KAIST, Korea Young-Jong Moon ,  Graduate Student,  KAIST, Korea

Structural Dynamics & Vibration Control Lab., KAIST, KoreaStructural Dynamics & Vibration Control Lab., KAIST, Korea 1111

TX mm

mITTT

(7)

where T is an orthogonal transformation matrixand its order m

(8)

Adjacent eigenvectors

Page 12: Kang-Min Choi,  Graduate Student,  KAIST, Korea Young-Jong Moon ,  Graduate Student,  KAIST, Korea

Structural Dynamics & Vibration Control Lab., KAIST, KoreaStructural Dynamics & Vibration Control Lab., KAIST, Korea 1212

• Rearranging eq.(5) and eq.(6) using adjacent eigenvectors

0KXXCXM 2 mmmmm

mmmm IX)CM2(XT

(9)

(10)

Page 13: Kang-Min Choi,  Graduate Student,  KAIST, Korea Young-Jong Moon ,  Graduate Student,  KAIST, Korea

Structural Dynamics & Vibration Control Lab., KAIST, KoreaStructural Dynamics & Vibration Control Lab., KAIST, Korea 1313

• Differentiating eq.(9) w.r.t. design parameter α

mmm

mmmmmm

X)KCM(

X)CM2(X)KCM(

,,,2

,,2

(11)

• Pre-multiplying at each side of eq.(11) by and

substituting

Tm

TX mm

,TDT m

mmmm )KCM(D ,,,2T

(12)

where

Page 14: Kang-Min Choi,  Graduate Student,  KAIST, Korea Young-Jong Moon ,  Graduate Student,  KAIST, Korea

Structural Dynamics & Vibration Control Lab., KAIST, KoreaStructural Dynamics & Vibration Control Lab., KAIST, Korea 1414

• Differentiating eq.(9) w.r.t. design parameter α

(13)

(14)

mmm

mmmm

X)KCM(

X)CM2()KCM(

,,,2

2

,Xm ,m

• Differentiating eq.(10) w.r.t. design parameter α

mmm

mmmm

X)CM2(X5.0

XMX )CM2(X

,,T

TT

,Xm ,m

Page 15: Kang-Min Choi,  Graduate Student,  KAIST, Korea Young-Jong Moon ,  Graduate Student,  KAIST, Korea

Structural Dynamics & Vibration Control Lab., KAIST, KoreaStructural Dynamics & Vibration Control Lab., KAIST, Korea 1515

mmm

mmm

m

m

mmmm

mmmm

X)CM2(X5.0

X)KCM(

X

XMX)CM2(X

X)CM2(KCM

,,T

,,,2

,

,

TT

2

(15)

- It maintains N-space without use of state space equation.

- Eigenpair derivatives are obtained simultaneously.

- It requires only corresponding eigenpair information.

- Numerical stability is guaranteed.

- It maintains N-space without use of state space equation.

- Eigenpair derivatives are obtained simultaneously.

- It requires only corresponding eigenpair information.

- Numerical stability is guaranteed.

• Combining eq.(13) and eq.(14) into a single equation

Page 16: Kang-Min Choi,  Graduate Student,  KAIST, Korea Young-Jong Moon ,  Graduate Student,  KAIST, Korea

Structural Dynamics & Vibration Control Lab., KAIST, KoreaStructural Dynamics & Vibration Control Lab., KAIST, Korea 1616

Second-order eigenpair derivatives of damped systemwith repeated eigenvalues

• Differentiating eq.(13) w.r.t. another design parameter β

(16)

(17)

mmmmmmmmm

mmmmmmmm

mmmm

MX2X)G~

G~

F~~

(

X)GF~

(X)GF~

(

X)CM2()KCM(

,,,,,,,

,,,,,,

2

,Xm ,m

• Differentiating eq.(14) w.r.t. another design parameter β

mmmmm

mmmm

mmmmmmm

mmmm

X)M2M2G~

(X5.0

X)M2G~

(X

X)M2G~

(XXGX

MXX)CM2(X

,,,,,T

,,,T

,,,T

,T

,

TT

,Xm ,m

Page 17: Kang-Min Choi,  Graduate Student,  KAIST, Korea Young-Jong Moon ,  Graduate Student,  KAIST, Korea

Structural Dynamics & Vibration Control Lab., KAIST, KoreaStructural Dynamics & Vibration Control Lab., KAIST, Korea 1717

• Combining eq.(16) and eq.(17) into a single equation

mmmmm

mmmmmmmmmmm

mmmmmm,mm,m

mmmmmmmm

m

m

mmmm

mmmm

X)M2M2G~

(X5.0

X)M2G~

(XX)M2G~

(XXGX

MX2X)G~

G~

F~~

(

X)GF~

(X)GF~

(

X

MXXCM2(X

X)CM2(KCM

,,,,,T

,,,T

,,,T

,T

,

,,,,,

,,,,,,

,

,

TT

2

(18)

where

,,,2

,

,,,2

,

2

KCMF~~

KCMF~

KCMF

mmm

mmm

mmm

,,,

,,,

CM2G~~

CM2G~

CM2G

mm

mm

mm

mmm

mmm

m

m

mmmm

mmmm

X)CM2(X5.0

X)KCM(

X

XMX)CM2(X

X)CM2(KCM

,,T

,,,2

,

,

TT

2

Page 18: Kang-Min Choi,  Graduate Student,  KAIST, Korea Young-Jong Moon ,  Graduate Student,  KAIST, Korea

Structural Dynamics & Vibration Control Lab., KAIST, KoreaStructural Dynamics & Vibration Control Lab., KAIST, Korea 1818

Numerical stability of the proposed method

)det()det()det()det( YAYYAY *T*T

Determinant property

x

]xxx[X]xxx

I0

0Y

A

m21m21mn2

m

r eigenvectoadjacent oft independen

be chosen to t vectorsindependenArbitrary :

when [

singular-non :

eq.(15) ofmatrix t coefficien The : where

i

1

*

(19)

Page 19: Kang-Min Choi,  Graduate Student,  KAIST, Korea Young-Jong Moon ,  Graduate Student,  KAIST, Korea

Structural Dynamics & Vibration Control Lab., KAIST, KoreaStructural Dynamics & Vibration Control Lab., KAIST, Korea 1919

00

0A~

)KCM( 2Tmm where

T

T

T

I

B~

)CM2(X

I

B~

X)CM2(

m

mm

m

mm

Then,

(20)

mmmm

mmmm

mmmmm

mmmm

m

XMX)CM2(X

X)CM2()KCM(

I

Ψ

XMX)CM2(X

X)CM2(KCM

IYAY

T

T2T

TT

2*T

T

(21)

Page 20: Kang-Min Choi,  Graduate Student,  KAIST, Korea Young-Jong Moon ,  Graduate Student,  KAIST, Korea

Structural Dynamics & Vibration Control Lab., KAIST, KoreaStructural Dynamics & Vibration Control Lab., KAIST, Korea 2020

• Arranging eq.(20)

mmm

m

MXXIB~

I00

B~

0A~

YAYTT

*T

0)A

~(det

B~

0A~

B~0

MXXI

I0det)A

~(det

Y)A(Ydet

1

TT

*T

mmm

m

(22)

• Using the determinant property of partitioned matrix

(23)

Page 21: Kang-Min Choi,  Graduate Student,  KAIST, Korea Young-Jong Moon ,  Graduate Student,  KAIST, Korea

Structural Dynamics & Vibration Control Lab., KAIST, KoreaStructural Dynamics & Vibration Control Lab., KAIST, Korea 2121

0)A(det *

Therefore

Numerical Stability is Guaranteed.Numerical Stability is Guaranteed.

(24)

Page 22: Kang-Min Choi,  Graduate Student,  KAIST, Korea Young-Jong Moon ,  Graduate Student,  KAIST, Korea

Structural Dynamics & Vibration Control Lab., KAIST, KoreaStructural Dynamics & Vibration Control Lab., KAIST, Korea 2222

80 : DOF ofNumber

20 : elements ofNumber

21 : nodes ofNumber

Data System

0.001 :dampingRayleigh

kg/m 10.887 :density Mass

N/m 102.10E : Modulus sYoung'33

211

Properties Material

)01.0w/ww ( : parameter Design*

z

y

z

mL 10 mw 5.0

mh 05.0x

x

z

y

1

2

3 4

Cantilever beam

NUMERICAL EXAMPLE

Page 23: Kang-Min Choi,  Graduate Student,  KAIST, Korea Young-Jong Moon ,  Graduate Student,  KAIST, Korea

Structural Dynamics & Vibration Control Lab., KAIST, KoreaStructural Dynamics & Vibration Control Lab., KAIST, Korea 2323

Results of analysis (eigenvalues)

Modenumber

EigenvaluesFirst derivatives

of eigenvaluesSecond derivatives

of eigenvalues

1, 2-1.4279e-03

±j5.2496e+00-2.8057e-10

j3.5347e-104.3916e-09

±j1.0285e-08

3, 4-1.4279e-03

±j5.2496e+00-2.2756e-02

±j5.2494e+01-2.7553e-01

j6.1102e-02

5, 6-5.4154e-02

±j3.2895e+01-6.6265e-10±j2.3445e-10

1.0084e-08 j2.4918e-09

7, 8-5.4154e-02

±j3.2895e+01-1.0818e+00±j3.2886e+02

-1.0391e-08 j2.6913e+00

9, 10-4.2409e-01

±j9.2090e+016.9247e-10

j6.9600e-10-1.0391e-08±j1.1514e-08

11, 12-4.2409e-01

±j9.2090e+01-8.4753e+00±j9.2029e+02

-8.4535e+01 j1.8358e+01

±

±

±

±

±

±

Page 24: Kang-Min Choi,  Graduate Student,  KAIST, Korea Young-Jong Moon ,  Graduate Student,  KAIST, Korea

Structural Dynamics & Vibration Control Lab., KAIST, KoreaStructural Dynamics & Vibration Control Lab., KAIST, Korea 2424

Results of analysis (first eigenvectors)

DOFnumber

EigenvectorsFirst derivatives of eigenvectors

Second derivatives of eigenvectors

1 0 0 0

2 0 0 0

3-6.6892e+05-j6.6892e+05

3.3446e-04+j3.3446e-04

-5.0169e+03-j5.0169e+03

4-2.6442e+04-j2.6442e+04

1.3221e-03+j1.3221e-03

-1.9596e+02-j1.9596e+02

77 0 0 0

78 0 0 0

79-1.5577e+02-j1.5577e+02

7.7887e-02+j7.7887e-02

-1.1683e+00-j1.1683e+00

80-2.1442e+03-j2.1442e+03

1.0721e-02+j1.0721e-02

-1.6082e+01-j1.6082e+01

… … … …

Page 25: Kang-Min Choi,  Graduate Student,  KAIST, Korea Young-Jong Moon ,  Graduate Student,  KAIST, Korea

Structural Dynamics & Vibration Control Lab., KAIST, KoreaStructural Dynamics & Vibration Control Lab., KAIST, Korea 2525

Results of analysis (errors of approximations)

Mode number

ActualEigenvalues

Approximated eigenvalues

Variations of eigenpairs Errors of approximations

Eigenvalues Eigenvectors Eigenvalues Eigenvectors

1, 2-1.4279e-03

±j5.2496e+00-1.4279e-03

±j5.2496e+002.2281e-11 4.9628e-03 2.2283e-11 3.7376e-05

3, 4-1.4556e-03

±j5.3021e+00-1.4555e-03

±j5.3021e+001.0000e-02 9.9010e-03 2.6622e-08 1.0000e-04

5, 6-5.4154e-02

±j3.2895e+01-5.4154e-02

±j3.2895e+013.7084e-12 4.9628e-03 3.6899e-12 3.7376e-05

7, 8-5.5241e-02

±j3.3224e+01-5.5236e-02

±j3.3224e+019.9997e-04 9.9023e-03 1.6763e-07 1.0001e-04

9, 10-4.2409e-01

±j9.2090e+01-4.2409e-01

±j9.2090e+019.1400e-12 4.9628e-03 9.1432e-12 3.7376e-05

11, 12-4.3261e-01

±j9.3010e+01-4.3256e-01

±j9.3010e+019.9936e-03 9.9041e-03 4.6508e-07 1.0002e-04

Page 26: Kang-Min Choi,  Graduate Student,  KAIST, Korea Young-Jong Moon ,  Graduate Student,  KAIST, Korea

Structural Dynamics & Vibration Control Lab., KAIST, KoreaStructural Dynamics & Vibration Control Lab., KAIST, Korea 2626

CONCLUSIONS

Proposed Method Proposed Method

is an efficient eigensensitivity method for the damped system with repeated eigenvalues

guarantees numerical stability

gives exact solutions of eigenpair derivatives

can be extended to obtain second- and higher order derivatives of eigenpairs

Page 27: Kang-Min Choi,  Graduate Student,  KAIST, Korea Young-Jong Moon ,  Graduate Student,  KAIST, Korea

Structural Dynamics & Vibration Control Lab., KAIST, KoreaStructural Dynamics & Vibration Control Lab., KAIST, Korea 2727

An efficient eigensensitivity method for the An efficient eigensensitivity method for the

damped system with repeated eigenvaluesdamped system with repeated eigenvalues

Thank you for your attention!Thank you for your attention!