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
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
metric to the cost function. Note, however, that the uncertainty parameters do not<br />
appear in the constraint equations. The RPT optimizations considered in this thesis<br />
do not include constraints that are a function of the uncertainty parameters. For<br />
example, the uncertainty parameters chosen for the sample problem affect only the<br />
stiffness matrix and not the constraint on total system mass. In the most general<br />
form of Equation 3.3, uncertainty does affect the constraints and can be incorporated<br />
by requiring that the tailoring parameters satisfy the constraint equations for the<br />
uncertainty values that are worst-case in relation to the constraints. Examples of<br />
such problems are found in [42] and [103].<br />
There are many possible formulations for JRP T since there are numerous ways to<br />
account for uncertainty and penalize design sensitivity. In the following sections three<br />
known techniques are presented and discussed in detail.<br />
3.2.1 Anti-optimization<br />
Optimization <strong>with</strong> anti-optimization (AO) is a design approach for structural opti-<br />
mization <strong>with</strong> bounded uncertainty formulated by Elishakoff, Haftka and Fang in [42].<br />
The authors discuss both of the formulations described below and demonstrate the<br />
design method on a standard ten-bar truss optimization problem using sequential lin-<br />
ear programming. The truss is subjected to uncertain loads, and the cross-sectional<br />
areas of the truss members are optimized for minimum mass subject to stress and<br />
minimum gage constraints.<br />
Anti-optimization can be defined as a min-max optimization problem that includes<br />
the process of identifying the critical uncertainty parameter values.<br />
min<br />
�x<br />
JAO<br />
� �� �<br />
max<br />
�p∈P<br />
s.t �g(�x) ≤ 0<br />
f (�x, �p) (3.4)<br />
At each iteration of tailoring parameters, �x, an uncertainty analysis is run to de-<br />
termine the values of uncertainty parameters, �p, that result in the worst-case per-<br />
83