20.04.2014 Views

design space pruning heuristics and global optimization method for ...

design space pruning heuristics and global optimization method for ...

design space pruning heuristics and global optimization method for ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

aBiB <<br />

25%, <strong>and</strong> 15% <strong>for</strong> Leg 1, Leg 2, <strong>and</strong> Leg 3, respectively. The number of asteroid<br />

sequences is reduced from 3072 to 131, <strong>and</strong> only 19 feasible sequences were eliminated.<br />

th<br />

The best sequence eliminated has a final mass of 653 kg, which ranks 37P<br />

P. There<strong>for</strong>e,<br />

none of the top solutions are eliminated <strong>and</strong> the <strong>design</strong> <strong>space</strong> is reduced by a factor of 23.<br />

Table 4: Pruning <strong>method</strong>ology applied to sample problem.<br />

Pruning<br />

Metric<br />

Trajectory Leg % Sequences<br />

Eliminated<br />

# Sequences<br />

Eliminated<br />

aBi+1B All N/A 2560<br />

θBwedgeB Leg 1 30% 128<br />

θBwedgeB Leg 2 25% 96<br />

θBwedgeB Leg 3 15% 42<br />

∆VBoptB Leg 1 30% 41<br />

∆VBoptB Leg 2 25% 46<br />

∆VBoptB Leg 3 15% 28<br />

2.5.2 Evaluation of Global Optimization Methods on the Sample Problem<br />

The <strong>global</strong> <strong>optimization</strong> <strong>method</strong>s under consideration are all tested on the full<br />

sample problem (instead of the pruned <strong>design</strong> <strong>space</strong>), in order to evaluate them on the<br />

largest possible <strong>design</strong> <strong>space</strong>. For the genetic algorithm, two approaches are considered,<br />

as was presented in Figure 11 <strong>and</strong> Figure 12: (1) a single-level approach, where all <strong>global</strong><br />

<strong>design</strong> variables are solved <strong>for</strong> simultaneously, <strong>and</strong> (2) a two-level approach, where an<br />

outer loop optimizer solves <strong>for</strong> the asteroid sequence <strong>and</strong> an inner loop optimizer solves<br />

<strong>for</strong> the time variables. The branch-<strong>and</strong>-bound is used solely in the two-level<br />

configuration, with the branch-<strong>and</strong>-bound <strong>method</strong> solving <strong>for</strong> the asteroid sequence <strong>and</strong> a<br />

genetic algorithm solving <strong>for</strong> the time variables.<br />

The basic genetic algorithm is applied to the sample problem in a variety of ways.<br />

In each case, the following settings must be chosen:<br />

72

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

Saved successfully!

Ooh no, something went wrong!