Abstract Artificial neural networks were applied to the design of rubble mound breakwater. Five neural networks with different network structures were trained with the same training data. Then they were compared with conventional empirical model and one another. It was found that the neural network technique gives more accurate results than conventional empirical model and the extent of accuracy can be affected by the structure of neural network. After that, how to integrate the trained neural network into reliability analysis technique is proposed. Since the neural network technique shows better performance than empirical model based approach in breakwater design, it is expected that the neural network integrated reliability analysis gives more improved results for probability of failure than it is done with empirical model.
[1]
V. Meer,et al.
Deterministic and Probabilistic Design of Breakwater Armor Layers
,
1988
.
[2]
Minoru Hanzawa,et al.
NEW STABILITY FORMULA FOR WAVE-DISSIPATING CONCRETE BLOCKS COVERING HORIZONTALLY COMPOSITE BREAKWATERS
,
1997
.
[3]
C. Willmott.
ON THE VALIDATION OF MODELS
,
1981
.
[4]
Hajime Mase,et al.
NEURAL NETWORK FOR STABILITY ANALYSIS OF RUBBLE-MOUND BREAKWATERS
,
1995
.
[5]
J. V. D. Meer,et al.
Rock slopes and gravel beaches under wave attack
,
1988
.
[6]
Nobuhisa Kobayashi,et al.
PROFILE CHANGES OF ROCK SLOPES BY IRREGULAR WAVES
,
1993
.