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.
Table 4.1: <strong>Tuning</strong> parameters for SCI development model.<br />
Name Description y0<br />
m1 -x tuning mass 0 kg<br />
m2 +x tuning mass 0 kg<br />
m1 00 11 000 111<br />
00 11 d1 d2 000 111 d2 00 11<br />
000 111<br />
00 11<br />
000 111<br />
00 11<br />
000 111<br />
m 2<br />
E 1 E 1 E 2 E 2<br />
Y<br />
d1<br />
000 111<br />
000 111<br />
000 111<br />
000 111<br />
000 111<br />
To demonstrate the technique the worst-case uncertainty realizations (∆ = 10%)<br />
of the PT and RPT AO designs are tuned using the SA and SQP algorithms. The<br />
nominal, worst-case and tuned performances are listed in Table 4.2. The worst case<br />
PT performance is significantly higher than nominal, and tuning results in a large<br />
improvement. <strong>Tuning</strong> also improves the RPT worst-case designs, but the results are<br />
not as dramatic.<br />
Table 4.2: <strong>Tuning</strong> performance summary for PT and RPT designs.<br />
Algorithm <strong>Performance</strong><br />
<strong>Performance</strong> [µm]<br />
Design nominal worst-case tuned<br />
PT 100.53 1355.5 303.17<br />
RPT AO 263.87 306.86 273.32<br />
In order to evaluate the performance of the different optimization algorithms, tuning<br />
is performed <strong>with</strong> SA-SQP, MC-SQP and an exhaustive search (ES) of the solution<br />
space. The results for the PT and RPT designs are given in Figures 4-1 and 4-2,<br />
respectively. Each figure includes a table of the algorithm performance data as well<br />
as a two-dimensional surface plot of the tuning solution space. In SA-SQP, a feasible<br />
initial guess is chosen randomly for SA, and the resulting design is used as the initial<br />
guess for the SQP algorithm. In MC-SQP ten feasible initial guesses are randomly<br />
111<br />
X