一种基于迁移学习的双代理辅助船型优化方法

doi:10.19678/j.issn.1000-3428.0069366

摘要/Abstract

摘要：

在昂贵目标函数的代理辅助优化中，获取足量样本在流体仿真中会相当耗时。为了减少所需样本量并提高问题的求解性能，提出了一种基于迁移学习双代理辅助的船型优化算法(TLDSAO)。首先，在代理构建阶段采用了迁移学习来辅助建模，利用迁移源域知识到目标域来减少对船型样本数量的需求。其次，针对样本数据构建了粗代理模型和细代理模型，用以进行双代理的辅助优化，该过程通过引入外部公共池来进行双代理间种群信息的交换，以提高优化算法的搜索性能。最后，将TLDSAO算法应用在KCS的船型优化实验上，实验结果表明在相同样本量下，迁移学习的加入显著提高了代理模型的准确性，在达到相同精度的情况下相比未加入迁移学习时能减少约1/2的样本量。通过TLDSAO算法最终使得KCS的总阻力系数降低了10.85%，预测误差为2.93%，相较于同条件下的3种对比方法其优化结果进一步减少了0.61%、6.11%和1.56%，并且对比方法在增加40个样本后的效果才接近未增加样本的TLDSAO算法。因此，即使在更少样本量下，TLDSAO算法也能获得更优的解和更低的预测误差。

关键词: 优化算法, 迁移学习, 代理辅助优化, 船型设计, 计算流体力学

Abstract:

In the surrogate-assisted optimization of expensive objective functions, obtaining enough samples is time-consuming in fluid simulations. This paper proposes a Transfer Learning based Dual-Surrogate-Assisted hull form Optimization (TLDSAO) algorithm to reduce the required number of samples and improve the solving performance of the problem. Firstly, transfer learning is used to assist modeling during the surrogate construction phase, utilizing transfer of source domain knowledge to the target domain to reduce the demand for hull form samples. Secondly, the coarse surrogate and fine surrogate models are constructed for the sample data to perform the dual-surrogate-assisted optimization. For this purpose, an external public pool is introduced to exchange population information between the two surrogates to improve the search performance of the optimization algorithm. Finally, the proposed method is applied to a hull form optimization experiment of KCS. Experimental results show that under the same sample size, the addition of transfer learning significantly improves the accuracy of the surrogate model and can reduce the sample size by approximately half compared to the algorithm without transfer learning with the same accuracy. The TLDSAO algorithm reduces the total drag coefficient of KCS by 10.85%, with a prediction error of 2.93%. Compared with three comparison methods under the same conditions, the optimization results are further reduced by 0.61%, 6.11%, and 1.56%, respectively. Comparison methods achieve a performance comparable to that of the TLDSAO algorithm only when the sample size is increased by 40. Despite fewer samples, the TLDSAO algorithm achieves better solutions and lower prediction errors.

Key words: optimization algorithm, transfer learning, surrogate-assisted optimization, hull form design, computational fluid dynamics

安畅, 毛力. 一种基于迁移学习的双代理辅助船型优化方法[J]. 计算机工程, 2025, 51(9): 317-327.

AN Chang, MAO Li. A Transfer Learning Based Dual-Surrogate-Assisted Hull Form Optimization Method[J]. Computer Engineering, 2025, 51(9): 317-327.

https://www.ecice06.com/CN/Y2025/V51/I9/317

图/表 21

图1 源域与目标域特征空间对齐

Fig.1 Aligning feature spaces between the source domain and the target domain

图2 构建代理模型流程

Fig.2 The process of constructing the surrogate model

图3 TLDSAO算法工作流程图

Fig.3 Flowchart of TLDSAO algorithm

图4 KCS船体形状图

Fig.4 KCS hull form

图5 KCS船体FFD方法变形的控制点示意图

Fig.5 Schematic of control points for deformation using the FFD method on the KCS hull

图6 TLDSAO算法迭代收敛图

Fig.6 Convergence plot of the TLDSAO algorithm iterations

图7 优化前后船体前视图对比

Fig.7 Comparison of the front view of the hull before and after optimization

图8 优化前后船体局部形状对比

Fig.8 Comparison diagram of local hull shape before and after optimization

图9 优化前后船体型线对比

Fig.9 Comparison of the hull lines before and after optimization

图10 粗代理样本子集上各方法的MSE对比

Fig.10 Comparison of MSE on coarse surrogate sample subsets for each method

图11 细代理样本子集上各方法的MSE对比

Fig.11 Comparison of MSE on fine surrogate sample subsets for each method

图12 粗代理样本子集上模型MSE对比

Fig.12 Comparison of model MSE on coarse surrogate sample subsets

图13 粗代理样本子集上模型盒图对比

Fig.13 Box plot comparison of model performance on coarse surrogate sample subsets

图14 细代理样本子集上模型MSE对比

Fig.14 Comparison of model MSE on fine surrogate sample subsets

图15 细代理样本子集上模型盒图对比

Fig.15 Box plot comparison of model performance on fine surrogate sample subsets

参考文献 37

1	ZHANG B J , ZHANG S L . Research on ship design and optimization based on simulation-based design (SBD) technique. Berlin: Springer, 2019.
2	MIAO A , ZHAO M , WAN D . CFD-based multi-objective optimisation of S60 catamaran considering demihull shape and separation. Applied Ocean Research, 2020, 97, 102071.
3	GARG V V , STOGNER R H . Hierarchical Latin hypercube sampling. Journal of the American Statistical Association, 2017, 112 (518): 673- 682.
4	AN G , XIANG G , XIANG X , et al. Resistance prediction and optimization of Tri-SWACH using hybrid surrogate model with particular consideration of outrigger layout. Ocean Engineering, 2023, 285, 115239.
5	SIMPSON T W , MAUERY T M , KORTE J J , et al. Kriging models for global approximation in simulation-based multidisciplinary design optimization. AIAA Journal, 2001, 39 (12): 2233- 2241.
6	KUDELA J, MATOUSEK R. Combining Lipschitz and RBF surrogate models for high-dimensional computationally expensive problems[EB/OL]. [2024-01-08]. https://arxiv.org/abs/2204.14236.
7	BUHMANU M D . Radial basis functions. Acta Numerica, 2009, 9, 1- 38.
8	YAO C J , HAN D W , Tinkle C , et al. Data-driven evolutionary optimization: an overview and case studies. IEEE Transactions on Evolutionary Computation, 2018, 23, 442- 458.
9	赵昳. 代理模型辅助的高维多目标进化优化方法研究[D]. 太原: 太原科技大学, 2024.
	ZHAO Y. Surrogate assisted high-dimensional multi-objective evolutionary optimization methods[D]. Taiyuan: Taiyuan University of Science and Technology, 2024. (in Chinese)
10	于成龙, 付国霞, 孙超利, 等. 全局与局部模型交替辅助的差分进化算法. 计算机工程, 2022, 48 (3): 115- 123. doi: 10.19678/j.issn.1000-3428.0060693
	YU C L , FU G X , SUN C L , et al. Differential evolution algorithm alternately assisted by global and local models. Computer Engineering, 2022, 48 (3): 115- 123. doi: 10.19678/j.issn.1000-3428.0060693
11	MCKAY M D , BECKMAN R J , CONOVER W J . A comparison of three methods for selecting values of input variables in the analysis of output from a computer code. Technometrics, 1979, 21 (2): 239- 245.
12	SEDERBERG T W, PARRY S R. Free-form deformation of solid geometric models[C]//Proceedings of the 13th Annual Conference on Computer Graphics and Interactive Techniques. New York, USA: ACM Press, 1986: 151-160.
13	MIAO A , WAN D . Hull form optimization based on an NM plus CFD integrated method for KCS. International Journal of Computational Methods, 2020, 10, 17.
14	DIEZ M , CAMPANA E F , STERN F . Design-space dimensionality reduction in shape optimization by Karhunen-Loeve expansion. Computer Methods in Applied Mechanics and Engineering, 2015, 283, 1525- 1544.
15	LI R , XU P , PENG Y , et al. Multi-objective optimization of a high-speed train head based on the FFD method. Journal of Wind Engineering and Industrial Aerodynamics, 2016, 152, 41- 49.
16	ZHOU H C , JIAO D Q , WANG Q Y , et al. Aerodynamic shape optimization of passenger car fender based on the FFD method. Proceedings of the Institution of Mechanical Engineers, Part D: Journal of Automobile Engineering, 2025, 239 (2/3): 722- 735.
17	WEISS K , KHOSHGOFTAAR T M , WANG D . A survey of transfer learning. Journal of Big Data, 2016, 3, 1- 40.
18	DAI W, QIANG Y, XUE G, et al. Boosting for transfer learning[C]//Proceedings of the Twenty-Fourth International Conference on Machine Learning (ICML 2007). Corvallis, Oregon, USA: Oregon State University, 2007: 193-200.
19	DRUCKER H. Improving regressors using boosting techniques[C]//Proceedings of the Fourteenth International Conference on Machine Learning. San Francisco, USA: Morgan Kaufmann Publishers Inc., 1997: 107-115.
20	HUANG S , WANG R , YU Y , et al. Construction and application of effluent quality prediction model with insufficient data based on transfer learning algorithm in wastewater treatment plants. Biochemical Engineering Journal, 2023, 191, 108807.
21	ZHAO C W , JUN H L , KE Y R , et al. Estimation of wheat kernel moisture content based on hyperspectral reflectance and satellite multispectral imagery. International Journal of Applied Earth Observation and Geoinformation, 2024, 126, 103597.
22	HE C L , ZHANG Y , GONG D W , et al. A review of surrogate-assisted evolutionary algorithms for expensive optimization problems. Expert Systems with Applications, 2023, 217, 119495.
23	AZZOUZ N, BECHIKH S, SAID L B. Steady state IBEA assisted by MLP neural networks for expensive multi-objective optimization problems[C]//Proceedings of 2014 Genetic and Evolutionary Computation Conference (GECCO 2014). New York, USA: ACM Press, 2014: 581-588.
24	IKEDA Y , KUBO O , KOBAYASHI Y . Forecast of business performance using an agent-based model and its application to a decision tree Monte Carlo business valuation. Physica A: Statistical Mechanics and Its Applications, 2004, 344 (1/2): 87- 94.
25	LIN Y , YANG Q , GUAN G . Automatic design optimization of SWATH applying CFD and RSM model. Ocean Engineering, 2019, 172, 146- 154.
26	刘子怡, 王宇嘉, 孙福禄, 等. 结合特征扰动与分配策略的集成辅助多目标优化算法. 计算机工程, 2022, 48 (6): 115- 123. doi: 10.19678/j.issn.1000-3428.0061885
	LIU Z Y , WANG Y J , SUN F L , et al. Ensemble-assisted multi-objective optimization algorithm combining feature perturbation and allocation strategy. Computer Engineering, 2022, 48 (6): 115- 123. doi: 10.19678/j.issn.1000-3428.0061885
27	MAO L X , DAN L , DONG W G , et al. Calibration of an agricultural-hydrological model (RZWQM2) using surrogate global optimization. Journal of Hydrology, 2017, 544, 456- 466.
28	SUN J , FENG B , XU W , et al. Quantum-behaved particle swarm optimization: analysis of individual particle behavior and parameter selection. Evolutionary Computation, 2012, 20 (3): 349- 393.
29	HAI J K , YAO W Z , HEOW P L . Dynamic optimization based on quantum computation-a comprehensive review. Computers & Structures, 2024, 292, 107255.
30	KHAN S , KAKLIS P . From regional sensitivity to intra-sensitivity for parametric analysis of free-form shapes: application to ship design. Advanced Engineering Informatics, 2021, 49 (4): 101314.
31	DAUMÉ H, KUMAR A, SAHA A. Co-regularization based semi-supervised domain adaptation[C]//Proceedings of Neural Information Processing Systems 2010. New York, USA: Curran Associates Inc., 2010: 478-486.
32	LI F , SHEN W M , CAI X W , et al. A fast surrogate-assisted particle swarm optimization algorithm for computationally expensive problems. Applied Soft Computing, 2020, 29, 106303.
33	KRISO. KRISO Container Ship (KCS)[EB/OL]. [2024-01-06]. https://t2015.nmri.go.jp/kcs_gc.html.
34	盛振邦, 刘应中. 船舶原理[D]. 上海: 上海交通大学出版社, 2003: 5-6.
	SHENG Z B, LIU Y Z. Principles of ship[D]. Shanghai: Shanghai Jiao Tong University Press, 2003: 5-6. (in Chinese)
35	LIU X , ZHAO W , WAN D . Multi-fidelity co-Kriging surrogate model for ship hull form optimization. Ocean Engineering, 2022, 243, 110239.
36	张乔宇. 代理模型在船舶阻力预报及优化中的应用[D]. 北京: 中国舰船研究院, 2021.
	ZHANG Q Y. Application of surrogate model in the prediction and optimization of ship resistance[D]. Beijing: China Ship Research and Development Academy, 2021. (in Chinese)
37	PERI D , ROSSETTI M , CAMPANA E F . Design optimization of ship hulls via CVD techniques. Journal of Ship Research, 2001, 45 (2): 140- 149.

[1]	马涛, 佘世刚. 灰狼粒子群混合算法在群控电梯中的应用[J]. 计算机工程, 2025, 51(9): 373-378.
[2]	逯暄, 景路琪, 彭甫镕. 基于增量学习的结直肠息肉分割方法[J]. 计算机工程, 2025, 51(7): 284-293.
[3]	武小丰, 袁培燕. 一种改进蛇优化算法的边缘服务器动态放置策略[J]. 计算机工程, 2025, 51(6): 255-265.
[4]	张绍坤, 沈加明, 胡燕海, 傅挺, 王舟挺. 基于GAPSO优化的注塑机注射速度模糊PID控制器[J]. 计算机工程, 2025, 51(5): 239-248.
[5]	刘坤, 张鹏程, 金惠颖, 吉顺慧. 云网融合环境下组合服务的动态重构[J]. 计算机工程, 2025, 51(5): 206-218.
[6]	贾硕, 林士飏, 杨苗会, 孙滕. 改进鲸鱼优化GRU的窄路短时车流量预测[J]. 计算机工程, 2025, 51(2): 111-125.
[7]	张欢, 王晨, 单景东, 仇润鹤. 基于领域自适应与注意力机制的电梯安全风险预测[J]. 计算机工程, 2025, 51(2): 86-93.
[8]	姬晨晨, 陈永青, 韩孟之. 基于国产加速器的三维卷积前向算子优化[J]. 计算机工程, 2025, 51(2): 250-258.
[9]	易鹏, 杨晔, 严仕嘉. 基于MPCNN模型的sEMG快速迁移学习的手势识别应用研究[J]. 计算机工程, 2025, 51(1): 304-311.
[10]	火久元, 王虹阳, 巨涛, 胡军. 多场景下基于AHP-EWM的人体健康状态评估模型研究[J]. 计算机工程, 2024, 50(7): 372-380.
[11]	周春雷, 宋继勐, 沈子奇, 余晗, 雷杰, 林兵. 数联网标识解析系统中的标识数据布局策略[J]. 计算机工程, 2024, 50(6): 311-320.
[12]	秦灏, 戚志东, 于灵芝, 童新. 基于IECSDE算法的PEMFC改进分数阶子空间辨识模型[J]. 计算机工程, 2024, 50(6): 346-357.
[13]	李致金, 汤佳辉, 闫金凤. 基于边缘计算的轻量化识别方法[J]. 计算机工程, 2024, 50(6): 287-295.
[14]	高家豪, 胡创业, 丁男, 刘战东. 智能网联汽车中联合驾驶风格的交通流数据有效性分析[J]. 计算机工程, 2024, 50(6): 367-376.
[15]	葛非, 闵珊, 邱含, 代振阳, 杨智敏. 求解时间依赖型绿色车辆路径问题的算法研究[J]. 计算机工程, 2024, 50(4): 1-10.

选择文件类型/文献管理软件名称

选择包含的内容