Reconstruction of Ship Propeller Wake Field Based on Physics-Informed Neural Networks

doi:10.16183/j.cnki.jsjtu.2023.101

Abstract

Abstract:

Physics-informed neural networks (PINN) are applied to the reconstruction of the ship propeller wake field. First, the principle and basic framework of PINN were introduced. Then, the Burgers equation was selected to verify the feasibility of PINN in solving partial differential equations. After that, the propeller of KVLCC2 in open water is simulated using computational fluid dynamics (CFD) software STAR CCM+, and the flow field information of the KVLCC2 propeller is obtained. Based on the simulated flow field information data, the training sample set was constructed to train PINN. The trained PINN was used to infer the approximate solution of the governing equation at any time and space. Finally, the velocity and pressure distribution obtained by PINN were compared with the velocity and pressure distribution simulated by STAR CCM+. The results validate the reliability of PINN in propeller wake field reconstruction, which can be concluded that PINN can be applied to the reconstruction of the ship propeller wake field.

Key words: physical-information neural network (PINN), partial differential equation, flow field information, flow field reconstruction, propeller

CLC Number:

U661

HOU Xianrui, ZHOU Xingyu, HUANG Xiaocheng. Reconstruction of Ship Propeller Wake Field Based on Physics-Informed Neural Networks[J]. Journal of Shanghai Jiao Tong University, 2024, 58(11): 1654-1664.

Figures/Tables 18

Fig.1

Fig.2

Fig.3

Fig.4

Tab.1

Fig.5

Fig.6

Fig.7

Fig.8

Tab.2

Tab.3

Fig.9

Fig.10

Fig.11

Fig.12

Fig.13

Fig.14

Fig.15

References 19

[1]	郭海鹏, 邹早建, 李广年. 基于OpenFOAM的螺旋桨紧急倒车工况数值模拟[J]. 上海交通大学学报, 2023, 57(2): 168-176. doi: 10.16183/j.cnki.jsjtu.2021.305
	GUO Haipeng, ZOU Zaojian, LI Guangnian. Numerical simulation of crashback ccondition of a propeller based on OpenFOAM[J]. Journal of Shanghai Jiao Tong University, 2023, 57(2): 168-176.
[2]	ZHAO Y M, AKOLEKAR H D, WEATHERITT J, et al. RANS turbulence model development using CFD-driven machine learning[J]. Journal of Computational Physics, 2020, 411: 1-14.
[3]	崔永赫, 张文耀, 闫慧龙, 等. “硬”边界低阶导数型物理信息神经网络: 一种流动求解器[J]. 西安交通大学学报, 2022, 56(9): 123-133.
	CUI Yonghe, ZHANG Wenyao, YAN Huilong, et al. “Hard” boundary low-order derivative physics informed neural network: A fluid flow solver[J]. Journal of Xi’an Jiaotong University, 2022, 56(9): 123-133.
[4]	郑素佩, 靳放, 封建湖, 等. 双曲型方程激波捕捉的物理信息神经网络(PINN)算法[J]. 浙江大学学报(理学版), 2023, 50(1): 56-62.
	ZHENG Supei, JIN Fang, FENG Jianhu, et al. PINN-type algorithm for shock capturing of hyperbolic equations[J]. Journal of Zhejiang University (Science Edition), 2023, 50(1): 56-62.
[5]	JIN X W, CHENG P, CHEN W L, et al. Prediction model of velocity filed around circular cylinder over various Reynolds numbers by fusion convolutional neural networks based on pressure on the cylinder[J]. Physics of Fluids, 2018, 30(4): 1-16.
[6]	XU H, CHANG H B, ZHANG D X. DL-PDE: Deep-learning based data-driven discovery of partial differential equations from discrete and noisy data[J]. Communications in Computational Physics, 2021, 29(3): 698-728.
[7]	RAISSI M, PERDIKARIS P, KARNIADAKIS G E. Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations[J]. Journal of Computational Physics, 2019, 378: 686-707.
[8]	JAGTAP A D, KAWAGUCHI K, KARNIADAKIS G E. Adaptive activation functions accelerate convergence in deep and physics-informed neural networks[J]. Journal of Computational Physics, 2020, 404: 1-24.
[9]	CHEN Z, LIU Y, SUN H. Physics-informed learning of governing equations from scarce data[J]. Nature Communications, 2021, 12(1): 1-13.
[10]	JIN X W, CAI S Z, LI H, et al. NSFnets (Navier-Stokes flow nets): Physics-informed neural networks for the incompressible Navier-Stokes equations[J]. Journal of Computational Physics, 2021, 426: 1-26.
[11]	CHENG C, MENG H, LI Y Z, et al. Deep learning based on PINN for solving 2 DOF vortex induced vibration of cylinder[J]. Ocean Engineering, 2021, 240: 1-13.
[12]	SUN L N, GAO H, PAN S W, et al. Surrogate modeling for fluid flows based on physics-constrained deep learning without simulation data[J]. Computer methods in Applied Mechanics and Engineering, 2020, 361: 1-25.
[13]	WANG S, WANG H, PERDIKARIS P. On the eigenvector bias of Fourier feature networks: From regression to solving multi-scale PDEs with physics-informed neural networks[J]. Computer Methods in Applied Mechanics and Engineering, 2021, 384: 1-27.
[14]	ZHANG D, GUO L, KARNIADAKIS G E. Learning in modal space: Solving time-dependent stochastic PDEs using physics-informed neural networks[J]. SIAM Journal on Scientific Computing, 2020, 42(2): A639-A665.
[15]	WESSELS H, WEIβENFELS C, WRIGGERS P. The neural particle method—An updated Lagrangian physics informed neural network for computational fluid dynamics[J]. Computer Methods in Applied Mechanics and Engineering, 2020, 368: 1-15.
[16]	LU L, MENG X H, MAO Z P, et al. DeepXDE: A deep learning library for solving differential equations[J]. Siam Review, 2021, 63(1): 208-228.
[17]	曹艳华, 张姊同, 李楠. 时空多项式配点法求解三维Burgers方程[J]. 应用数学和力学, 2022, 43(9): 1045-1052.
	CAO Yanhua, ZHANG Zitong, LI Nan. A space-time polynomial collocation method for solving 3D Burgers equations[J]. Applied Mathematics and Mechanics, 2022, 43(9): 1045-1052.
[18]	任宏斌. 机器学习方法在第一性原理计算中的应用[D]. 北京: 中国科学院大学, 2021.
	REN Hongbin. Machine learning method in first principle calculation[D]. Beijing: University of Chinese Academy of Sciences, 2021.
[19]	谭康力. 带桨舵的船舶操纵水动力数值研究[D]. 武汉: 武汉理工大学, 2019.
	TAN Kangli. Numerical study of hydrodynamic forces on a hull with propeller and rudder in maneuvering motion[D]. Wuhan: Wuhan University of Technology, 2019.

参数	数值
直径/m	0.089 6
叶片数量	4
轮毂比	0.155
螺距比(0.7R)	0.721 2
盘面比	0.431
侧斜角/(°)	21.15
纵斜角/(°)	0

J	K_T		相对误差/%
J	NMRI^[19]	CFD	相对误差/%
0.2	0.234 7	0.218 2	7.03
0.3	0.199 4	0.185 7	6.87
0.4	0.161 2	0.159 6	0.99
0.5	0.117 3	0.111 4	5.02

J	10K_Q		相对误差/%
J	NMRI^[19]	CFD	相对误差/%
0.2	0.264 4	0.255 6	3.33
0.3	0.233 5	0.229 1	1.88
0.4	0.199 4	0.199 7	0.15
0.5	0.161 4	0.166 7	3.28