基于深度学习的活塞式造波机传递函数获取
Acquisition of transfer function for piston-type wave generator based on deep learning
魏唯嘉 1赵西增 2谢玉林 1邓龙赐 1许诺1
作者信息
- 1. 浙江大学 海洋学院,浙江 舟山 316021
- 2. 浙江大学 海洋学院,浙江 舟山 316021;浙江大学 舟山海洋研究中心,浙江 舟山 316021
- 折叠
摘要
活塞式造波机广泛应用于波浪水槽中,但由于造波边界的复杂性,其造波传递函数的推导较为困难.基于深度学习技术,提出一种可适用于任意形状活塞式造波机的传递函数获取方法.通过收集造波过程中的流场数据训练神经网络,建立造波机运动速度与造波过程流场变量之间的映射关系,从而拟合出该造波机的传递函数,通过引入流体质量守恒约束项,提高神经网络的物理可解释性;并通过基于OpenFOAM的数值波浪水槽与自主搭建的物理波浪水槽,开展规则波、孤立波与畸形波的造波模拟,验证该传递函数的有效性.结果表明,深度学习能够降低活塞式造波机的传递函数推导难度,并使造波机能实现不同类型波浪的水槽造波.
Abstract
Piston-type wave generators are widely utilized in wave flumes,but deriving their wave generation transfer functions poses challenges due to the complexity of the wave-generation boundary.Leveraging deep learning technology,this study proposes a method for acquiring transfer functions applicable to piston-type wave generators of any shape.This method involves collecting flow field data during the wave-generation process to train a neural network,thereby establishing a mapping relationship between the wave generator's motion velocity and the flow field variables of the wave-generation process.Consequently,the transfer function of the wave generator is approximated.By incorporating a fluid mass conservation constraint,the physical interpretability of the neural network is enhanced.Additionally,wave generation simulations of regular waves,solitary waves,and freak waves are conducted using a numerical wave tank based on OpenFOAM and a physically built wave tank,validating the effectiveness of the derived transfer function.The findings indicate that deep learning can simplify the derivation of the transfer function for piston-type wave generators and enable wave generators to facilitate the wave generation of various wave types in tanks.
关键词
造波机/活塞式/深度学习/传递函数/物理约束Key words
wave generator/piston-type/deep learning/transfer function/physical constraints引用本文复制引用
出版年
2024
NETL
NSTL
万方数据