DEVELOPMENT OF COASTAL WAVE ESTIMATION MODEL FOR TOYAMA BAY USING ENSEMBLE LEARNING NEURAL NETWORK
-
- MASUDA Kazuki
- 東洋建設株式会社 総合技術研究所鳴尾研究所
-
- NINOMIYA Junichi
- 金沢大学 理工研究域地球社会基盤学系
-
- SAITO Takehisa
- 金沢大学 理工研究域地球社会基盤学系
Bibliographic Information
- Other Title
-
- アンサンブル学習ニューラルネットワークを用いた富山湾沿岸波浪推算モデルの開発
- アンサンブル ガクシュウ ニューラルネットワーク オ モチイタ トヤマワン エンガン ハロウ スイサン モデル ノ カイハツ
Search this article
Abstract
<p> The risk of Yorimawari-wave disasters will increase due to the strengthening of extreme events caused by climate change. Statistical evaluation of low-frequency high waves is important and requires long-term wave data. However, it is difficult to reproduce high waves on the coast of Toyama Bay by the numerical wave model. The authors have developed the neural network (NN) model to estimate wave height at the NOWPHAS observation point of Toyama Bay, but there was a problem in its applicability related to the characteristics of low pressure. In this research, we construct a versatile NN model by incorporating multi-time learning of teacher data and stacking ensemble learning. Then, new NN model estimates the wave height and period for the period without observation data and the statistical evaluation of the low-frequency waves are performed. As a result, a general-purpose and highly accurate NN was constructed, and the statistical evaluation of low-frequency waves at NOWPHAS Toyama point revealed that the observation period was insufficient.</p>
Journal
-
- Journal of Japan Society of Civil Engineers, Ser. B2 (Coastal Engineering)
-
Journal of Japan Society of Civil Engineers, Ser. B2 (Coastal Engineering) 76 (2), I_235-I_240, 2020
Japan Society of Civil Engineers
