1. For the solution shown in Fig.9, 181 sets of subspace optimization were performed. Initialized from a fixed initial guess (cold-start), a call to the subspaces requires approximately 1-3 hours. At this rate, a cold-start solution requires approximately 1-3 weeks. Using warm-start subspaces (restarting from the previous domain-specific solution with knowledge of the prior optimum active set,Lagrange multipliers, and Hessian ofthe Lagrangian), the solution presented in Fig. 9 required approximately 4.5 days of computer time. Hence, warm-starting the subspaces provides a dramatic advantage. This level of efficiency gain is possible because the subspaces are tasked with solving a related sequence of subproblems.
2. 301- 400 -AMtude I'] l' I// ' / ^~-Transitionpropulsion ' ' / systemtosinglefuel mode
3. Distinctions among these four different objectives (gross weight, dry weight, development cost, and AV) are examined through application of the collaborative architecture. In each of these designs, the vehicle liftoff T/W is allowed to vary in the range 1.0-1.5. The minimum development cost solution, described in the previous section, is used to normalize the other optimal results.
4. Design characteristics ofthese four optimal concepts are listed in Table 3. As shown by the fourth column of this table, at the vehicle level, the minimum AV concept stands apart from the other 3 designs. This system is over 50% more expensive than the minimum development cost concept and more than 25%heavier than the minimumdry weight case (while yielding only a 4-6% improvement in AV). Normalized