11.12.2012 Views

Chapter 5 Robust Performance Tailoring with Tuning - SSL - MIT

Chapter 5 Robust Performance Tailoring with Tuning - SSL - MIT

Chapter 5 Robust Performance Tailoring with Tuning - SSL - MIT

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.

the eigenvalue equation, the derivatives of the frequencies are obtained by differenti-<br />

ating Equation 2.5:<br />

∂ωj<br />

∂x<br />

1<br />

= φ<br />

2ωj<br />

T j<br />

�<br />

−ω 2 ∂M<br />

j<br />

∂x<br />

�<br />

∂K<br />

+ φj<br />

∂x<br />

(2.21)<br />

Obtaining the derivatives of the eigenvectors, or mode shapes, is slightly more in-<br />

volved. Using Nelson’s method [98], it is assumed that the j th eigenvector derivative<br />

is written as a linear combination of the j th eigenvector and a linearly independent<br />

vector, ψj:<br />

∂φj<br />

∂x = ψj + ajφj<br />

The scalar aj is obtained through derivation of the mass normalization equation:<br />

aj = −φ T j<br />

�<br />

Mψj + 1<br />

2<br />

∂M<br />

∂x φj<br />

�<br />

The vector ψj is found by solving the following matrix equation:<br />

where Kj and bj are defined as:<br />

bj = Kjψj<br />

(2.22)<br />

(2.23)<br />

(2.24)<br />

Kj = −ω 2 j M + K (2.25)<br />

bj = ∂ω2 j<br />

∂x Mφj<br />

�<br />

− −ω 2 �<br />

∂M ∂K<br />

j + φj<br />

(2.26)<br />

∂x ∂x<br />

The matrix Kj is singular <strong>with</strong> rank of n−1 if the eigenvalues are distinct. This issue is<br />

addressed by arbitrarily removing one element from bj as well as the corresponding row<br />

and columns from Kj. The corresponding element in ψj is set to zero. Equations 2.21<br />

and 2.22 provide expressions for the derivatives of the modal quantities <strong>with</strong> respect<br />

to the design variables as a function of the mass and stiffness matrix derivatives, ∂M<br />

∂x<br />

and ∂K.<br />

These final quantities will differ depending on how the design variables enter<br />

∂x<br />

the stiffness and mass matrices. These gradients are derived in the following section<br />

for the specific design variables chosen for the SCI model PT optimization.<br />

50

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

Saved successfully!

Ooh no, something went wrong!