基于深度神经网络的斜坡堤平均越浪量预测方法研究  

Study on prediction method of mean wave overtopping discharge of slopingbreakwater based on deep neural network

作　　者：解晶 张黎邦 刘晓[1] 刘勇[1] XIE Jing;ZHANG Li-bang;LIU Xiao;LIU Yong(Shandong Provincial Key Laboratory of Ocean Engineering,Ocean University of China,Qingdao 266100,China)

机构地区：[1]中国海洋大学山东省海洋工程重点实验室,青岛266100

出　　处：《水道港口》2023年第2期173-182,共10页Journal of Waterway and Harbor

基　　金：国家重点研发计划项目(2020YFE0201200);国家自然科学基金资助项目(52001294)。

摘　　要：合理评估结构越浪量对防波堤设计具有重要的指导意义。文章基于深度神经网络学习,对斜坡式防波堤的平均越浪量计算方法进行研究。基于欧洲CLASH项目(EurOtop ANN)模型数据库提取数据集,建立斜坡堤平均越浪量的深度神经网络计算模型,采用验证样本集和独立物理模型试验数据验证文章新方法的可靠性和适用性。分析结果表明:相较其他机器学习回归算法,文章建立的计算模型有效提高了斜坡堤越浪量的预测精度。研究结果可为实际工程设计提供重要参考。Reasonable evaluation of structural wave overtopping is of great guiding significance for the design of breakwater.The mean wave overtopping discharge calculation method of sloping breakwater based on deep neural network learning was studied.The dataset based on the European CLASH project database(EurOtop ANN)was extracted.Then,a deep neural network calculation model of the mean wave overtopping discharge of the sloping breakwater was established.Finally,the reliability and applicability of the new approach in this paper were verified based on validate sample sets and independent physical model test data.The analysis results show that the calculation model established in this paper effectively improves the prediction accuracy of wave overtopping on the sloping breakwater compared to other machine learning regression algorithms.This study provides an important reference for actual engineering design.

关 键 词：EurOtop 斜坡堤 平均越浪量 深度神经网络 

分 类 号：U656.2[交通运输工程—港口、海岸及近海工程] TV139.2[交通运输工程—船舶与海洋工程]