Hull form optimization based on multi-fidelity deep neural network

Table 1. Parameter setting of multi-fidelity deep neural network (MFDNN)
View table
View in Article
Table 1. Parameter setting of multi-fidelity deep neural network (MFDNN)
输入层神经元数量宽度深度输出层神经元数量
$ N{N_{\rm{L}}} $ 1 32 5 1
$ N{N_{{\rm{H}}1}} $ 2 32 4 1
$ N{N_{{\rm{H}}2}} $ 2 32 4 1

Table 2. Main particulars of DTMB 5415
View table
View in Article
Table 2. Main particulars of DTMB 5415
参数数值
垂线间长$ L_{{\mathrm{pp}}}/\mathrm{m} $ 5.72
最大船宽B/m 0.77
吃水T/m 0.25
型深D/m 0.77
排水体积$ \nabla /{{{\mathrm{m}}^3}} $ 0.55
湿表面积$ S /{{{\mathrm{m}}^2}} $ 4.86

Table 3. Comparison between potential flow calculation result and experimental result
View table
View in Article
Table 3. Comparison between potential flow calculation result and experimental result
$Fr$ ${C_{\rm{w}}}$ ${C_{\rm{t}}}$ ${C_{\text{t-EFD}}}$ 计算值和试验值总阻力系数误差/%
0.28 1.20×10⁻³ 4.16×10⁻³ 4.23×10⁻³ 1.65

Table 4. Comparison between viscous flow calculation result and experimental result
View table
View in Article
Table 4. Comparison between viscous flow calculation result and experimental result
$Fr$ ${C_{\rm{w}}}$ ${C_{\rm{t}}}$ 计算值和试验值总阻力系数误差/%
0.28 4.21×10⁻³ 4.23×10⁻³ 0.47

Table 5. The range of values for the design variables

View table

View in Article

Table 5. The range of values for the design variables

方法	设计变量	取值范围	备注
平移法	${\alpha _{1{\mathrm{f}}}}$	[−0.016, 0.016]	前半体变化幅值
平移法	${\alpha _{1{\mathrm{a}}}}$	[−0.016, 0.016]	后半体变化幅值
FFD	${{\boldsymbol{x}}_1}$	[−0.01, 0.01]	沿船长方向移动
	${{\boldsymbol{y}}_1}$	[−0.006, 0.006]	沿船宽方向移动
	${{\textit{z}}_1}$	[−0.012, 0.012]	沿吃水方向移动

Table 6. Error analysis of Kriging model

View table

View in Article

Table 6. Error analysis of Kriging model

	零次确定性多项式	一次确定性多项式	二次确定性多项式	误差指标
高斯相关函数	0.233 2	0.218 8	0.167 1	AAE
	0.980 2	0.592 8	0.403 3	MAE
	0.099 4	0.070 4	0.037 8	MSE
样条相关函数	0.259 1	0.192 6	0.162 9	AAE
	0.743 8	0.524 6	0.399 8	MAE
	0.087 8	0.053 8	0.036 4	MSE

Table 7. Error analysis of MFDNN model
View table
View in Article
Table 7. Error analysis of MFDNN model
样本点数目
10 15 20 25 30 35 40 45 50
MSE 0.009 2 0.006 7 0.011 5 0.013 0 0.015 7 0.019 6 0.021 0 0.023 2 0.018 7

Table 8. Parameter settings for genetic algorithm
View table
View in Article
Table 8. Parameter settings for genetic algorithm
单目标遗传算法参数数值
种群数量 200
迭代次数 500
变异率 0.2
交叉率 0.8

Table 9. Comparison of values taken for design variables with optimization results

View table

View in Article

Table 9. Comparison of values taken for design variables with optimization results

	设计变量取值					总阻力${R_{\rm{t}}}$
	${\alpha _{1{\mathrm{f}}}}$	${\alpha _{1{\mathrm{a}}}}$	${{\boldsymbol{x}}_1}$	${{\boldsymbol{y}}_1}$	${{\textit{z}}_1}$	预报结果/N	CFD/N	预报误差/%	优化百分比/%
Initial	0	0	0	0	0	−	44.66	−	−
Opt-K	0.016	0.016	−0.01	−0.006	0.012	41.31	42.16	2.02	5.59
Opt-MF	0.016	0.016	−0.01	0.006	0.012	40.57	41.65	2.59	6.73

Tools

Get Citation

Copy Citation Text

Yabo WEI, Yangjun WANG, Decheng WAN. Hull form optimization based on multi-fidelity deep neural network[J]. Chinese Journal of Ship Research, 2024, 19(6): 74

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: Theory and Method of Intelligent Design for Ship and Ocean Engineering

Received: Jul. 12, 2024

Accepted: --

Published Online: Mar. 14, 2025

The Author Email:

DOI:10.19693/j.issn.1673-3185.04062

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology

Table 1. Parameter setting of multi-fidelity deep neural network (MFDNN)

Table 1. Parameter setting of multi-fidelity deep neural network (MFDNN)

Table 2. Main particulars of DTMB 5415

Table 2. Main particulars of DTMB 5415

Table 3. Comparison between potential flow calculation result and experimental result

Table 3. Comparison between potential flow calculation result and experimental result

Table 4. Comparison between viscous flow calculation result and experimental result

Table 4. Comparison between viscous flow calculation result and experimental result

Table 5. The range of values for the design variables

Table 5. The range of values for the design variables

Table 6. Error analysis of Kriging model

Table 6. Error analysis of Kriging model

Table 7. Error analysis of MFDNN model

Table 7. Error analysis of MFDNN model

Table 8. Parameter settings for genetic algorithm

Table 8. Parameter settings for genetic algorithm

Table 9. Comparison of values taken for design variables with optimization results

Table 9. Comparison of values taken for design variables with optimization results

微信扫一扫：分享