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
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
the truss material is a source of uncertainty and is allowed to vary asymmetrically in<br />
the two interferometer arms as in the development model. The resulting asymmetry<br />
in the two interferometer arms greatly affects the performance and models the effects<br />
of many uncertainty parameters while keeping the computational effort low. A third<br />
uncertainty parameter, the modal damping, is added to the parameter set. Damping<br />
is a mechanism that is not well-understood and most models of it are conservative<br />
approximations to the physical reality. Therefore it is a prime candidate for uncer-<br />
tainty. In this example, a singl modal damping ratio is applied globally to all modes.<br />
Table 6.12: TPF SCI model uncertainty parameters.<br />
p Description p0 Units ∆ [%]<br />
E1 Young’s Modulus of -X truss 111.7 GPa 25<br />
E2 Young’s Modulus of +X truss 111.7 GPa 25<br />
ξn modal damping ratio 0.001 none 40<br />
A bounded uncertainty model of the form in Equation 3.1 is used. The percent<br />
bounds on the uncertainty parameters are given by ∆ and are listed in Table 6.12.<br />
The truss Young’s Modulus is allowed to vary ±25% about its nominal value and the<br />
modal damping ranges in value ±40% about nominal. The range on the damping<br />
parameter is larger than that of the Young’s Moduli to capture the high uncertainty<br />
inherent in the modal damping model.<br />
6.3 Optimization Implementation<br />
It is shown in the previous chapters that a combination of SA and SQP finds the<br />
best design consistently and efficiently in the case of the development model. Unfor-<br />
tunately, running the SQP optimization on the TPF model is not straightforward.<br />
Recall from <strong>Chapter</strong> 2 that the SQP algorithm requires performance gradients, and<br />
that the calculation of the gradients requires the gradients of the eigenvalues and<br />
eigenvectors. The development model is relatively simple and as a result, these quan-<br />
tities can be derived directly. The TPF model, on the other hand, is much more<br />
200