24.07.2014 Views

COMPIT 2010 in Gubbio - TUHH

COMPIT 2010 in Gubbio - TUHH

COMPIT 2010 in Gubbio - TUHH

SHOW MORE
SHOW LESS

Create successful ePaper yourself

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

The value of impact acceleration and R T of the basis hull at sea state 3 is 2.11g and 13545.90 N<br />

respectively, while all optimized designs have impact acceleration 1.5g and R T of 21830.67 N,<br />

21829.52 N and 21833.40 N for NSGA-II, EASDS and IDEA, respectively. The optimized hull forms<br />

found by the optimizers have similar characteristics where larger values of L, B and T result <strong>in</strong> a larger<br />

displacement <strong>in</strong> order to satisfy the vertical impact acceleration constra<strong>in</strong>t. The details on the<br />

dimensions are highlighted <strong>in</strong> Tables V to VII.<br />

The results for the optimization without the impact acceleration constra<strong>in</strong>t are shown <strong>in</strong> Fig. 6. Total<br />

resistance values of the best run of each algorithm (EASDS, IDEA and NSGA-II) are plotted aga<strong>in</strong>st<br />

function evaluations for typical sea states of 1, 2 and 3. In all sea-states, EASDS was able to<br />

converge faster than IDEA and NSGA-II. One can observe by compar<strong>in</strong>g Fig. 5(b) and Fig. 6(d) that<br />

an <strong>in</strong>crease of R T as high as 61% is necessary to satisfy the impact acceleration constra<strong>in</strong>t, as<br />

compared to a reduction by 2.97% when the impact acceleration constra<strong>in</strong>t is ignored.<br />

Shown <strong>in</strong> Fig. 7 is the progress for the median designs obta<strong>in</strong>ed us<strong>in</strong>g NSGA-II, EASDS and IDEA<br />

for sea-state 1. Given the approximately same number of function evaluations, IDEA converges faster<br />

than the other two algorithms while EASDS converges better than NSGA-II. A more comprehensive<br />

comparison between algorithms can be observed <strong>in</strong> Tables III and IV where the best values of median<br />

designs at all sea-states are depicted <strong>in</strong> bold type. EASDS consistently performs better than NSGA-II<br />

and IDEA <strong>in</strong> solv<strong>in</strong>g the m<strong>in</strong>imization problem with the impact acceleration constra<strong>in</strong>t, while IDEA<br />

outperforms the other two algorithms for the problem without the impact acceleration constra<strong>in</strong>t.<br />

1.38 x 104 Function Evaluations<br />

1.36<br />

1.34<br />

EASDS<br />

IDEA<br />

NSGA-II<br />

1.32<br />

Total Resistance<br />

1.3<br />

1.28<br />

1.26<br />

1.24<br />

1.22<br />

1.2<br />

1.18<br />

0 100 200 300 400 500 600<br />

Fig 7: Median design obta<strong>in</strong>ed us<strong>in</strong>g EASDS, IDEA and NSGA-II<br />

Table III: Comparison between R T (N) of median designs with I a constra<strong>in</strong>t<br />

Sea-state 1 (0.4m) Sea-state 2 (0.6m) Sea-state 2 (0.8m) Sea-state 3 (1.1m)<br />

NSGA-II 12060.96 12460.96 13858.59 21844.52<br />

EASDS 11955.90 12415.75 13314.61 21842.02<br />

IDEA 11933.19 12549.24 13709.51 21871.29<br />

Table IV: Comparison between R T (N) of median designs without I a constra<strong>in</strong>t<br />

Sea-state 1 (0.4m) Sea-state 2 (0.6m) Sea-state 2 (0.8m) Sea-state 3 (1.1m)<br />

NSGA-II 12065.10 12460.96 12968.60 13506.20<br />

EASDS 11999.19 12415.75 12916.91 13468.44<br />

IDEA 11962.04 12445.15 12880.76 13315.37<br />

14

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

Saved successfully!

Ooh no, something went wrong!