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.

attractive, and more conservative alternative, when there is insufficient data to build<br />

an accurate statistical model of parametric uncertainty. When these models are used<br />

in uncertainty propagation, bounds on the response are obtained in lieu of statistical<br />

output distributions.<br />

These uncertainty models and propagation techniques are ultimately combined<br />

<strong>with</strong> structural optimization to produce designs that can meet performance require-<br />

ments despite the model uncertainty. This practice is known as robust design and<br />

has been popular among control experts for some time [110, 10, 47, 48]. <strong>Robust</strong><br />

design optimization is a relatively new field in structural design, but has generated<br />

a lot of interest due to the rising complexity of systems [40]. Anderson considers<br />

the problem of robust actuator and damping placement for structural control using<br />

a model <strong>with</strong> known errors [5]. Park, Hwang and Lee use Taguchi methods to post-<br />

process gradient-based structural optimization to find discrete sizing variables [99].<br />

The Taguchi method [107, 102] is a quality-control technique originally developed for<br />

circuit design that has recently found application in structural optimization. In a<br />

later paper, Park et al. apply the Taguchi method to unconstrained structural op-<br />

timization to find robust optimal designs of three and ten-bar trusses subjected to<br />

applied loads [100]. The areas of the truss members are optimized <strong>with</strong> conventional<br />

methods first to minimize mass and then <strong>with</strong> the Taguchi method to minimize the<br />

sensitivity of the displacement of a given node to variations in the member areas.<br />

Constrained robust structural optimization problems are considered as well. Sand-<br />

gren and Cameron suggest a two-stage hybrid approach that combines the use of<br />

genetic algorithms for topology optimization <strong>with</strong> Monte Carlo uncertainty propa-<br />

gation to determine the statistics of the objective function and/or constraints [103].<br />

The method is computationally expensive, but is successfully applied to a ten-bar<br />

truss problem and a higher-fidelity automobile inner panel using probabilistic mod-<br />

els of uncertainty. Elishakoff, Haftka and Fang use convex uncertainty models to<br />

perform “anti-optimization” or min-max style robust design [42]. They demonstrate<br />

their technique on a ten-bar truss problem subjected to uncertain loading as well as<br />

stress and displacement constraints.<br />

34

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

Saved successfully!

Ooh no, something went wrong!