本文提出一种基于神经网络和粒子群算法(Particle Swarm Optimization,PSO)的船舶板架动力学优化方法,用于板架布局的快速寻优。首先,分析船舶板架布局的特征参数,利用拉丁超立方采样和模态分析获得样本点的固有频率;然后,构建BP神经网络代理模型,用以反映板架特征参数和固有频率之间的非线性映射关系;最后,结合粒子群算法,以结构重量和一阶固有频率为目标,将代理模型应用于船舶板架结构的动力学优化,以确定较优的布局型式。结果表明,BP神经网络代理模型对板架固有频率的预测具有较高的精度,BP-PSO方法对不同尺寸和类型的板架均适用,具有广泛性、高效性、普适性的优势。因此,BP-PSO法能为板架优化设计提供较好的思路和方案。
In this paper, a dynamic optimization method of ship grillage based on neural network and particle swarm optimization (PSO) is proposed for rapid optimization of grillage layout. Firstly, the characteristic parameters of the ship grillage layout are analyzed, and the natural frequencies of the sample points are obtained by Latin hypercube sampling and modal analysis. Then, the BP neural network surrogate model is constructed to reflect the nonlinear mapping relationship between the characteristic parameters and the natural frequency of the plate frame. Finally, combined with the particle swarm optimization algorithm, the surrogate model is applied to the dynamic optimization of the ship grillage structure to determine the optimal layout type with the structural weight and the first-order natural frequency as the objectives. The results show that the BP neural network surrogate model has high accuracy in predicting the natural frequency of the plate frame. The BP-PSO method is applicable to different sizes and types of plates, and has the advantages of universality, high efficiency and universality. Therefore, the BP-PSO method can provide a better idea and scheme for the optimization design of the plate frame.
2025,47(15): 30-35 收稿日期:2024-10-23
DOI:10.3404/j.issn.1672-7649.2025.15.006
分类号:U663.5
作者简介:周俞(2000-),男,硕士研究生,研究方向为船体结构动力学分析
参考文献:
[1] 刘波, 林焰, 吕振望, 等. 基于量子行为遗传算法的船体局部结构优化设计[J]. 船舶力学, 2017, 21(4): 484-492.
LIU B, LIN Y, LV Z W, et al. Optimization design of local hull structure based on quantum-behaved genetic algorithm[J]. Ship Mechanics, 2017, 21(4): 484-492.
[2] 秦洪德, 石丽丽. PSO算法在油船双底结构优化设计中的应用研究[J]. 哈尔滨工程大学学报, 2010, 31(8): 1007-1011.
[3] 何小二, 王德禹, 夏利娟. 基于粒子群算法的多用途船结构优化[J]. 上海交通大学学报, 2013, 47(6): 928-931.
[4] 吕利叶, 鲁玉军, 王硕, 等. 代理模型技术及其应用: 现状与展望[J]. 机械工程学报, 2024, 60(3): 254-281.
[5] 陈松坤, 王德禹. 基于神经网络的蒙特卡罗可靠性分析方法[J]. 上海交通大学学报, 2018, 52: 687-692.
[6] 李永华, 陈鹏, 田宗睿, 陈志豪. 基于改进BP神经网络的结构可靠度计算方法[J]. 机械强度, 2021, 43(6): 1359-1365.
[7] 陈晨铭, 郭雪岩, 常林森. 基于BP神经网络代理模型的翼型优化及领域自适应研究[J]. 动力工程学报, 2022, 42(7): 657-663.
[8] 苏绍娟, 刘灿波, 张祥, 等. 基于RBF代理模型的NOBS油船货舱结构多目标优化设计[J]. 武汉理工大学学报, 2021, 43(11): 35-42.
[9] 郭海丁, 路志峰. 基于BP神经网络和遗传算法的结构优化设计[J]. 航空动力学报, 2003(2): 216-220.
[10] 张宇, 黄小平, 闫小顺. 基于神经网络和粒子群算法的环肋圆柱壳优化设计[J]. 舰船科学技术, 2016, 38(5): 5-9.
ZHANG Y, HUANG X P, YAN X S. Optimization design of ring-stiffened cylindrical shells based on neural network and particle swarm optimization[J]. Ship Science and Technology, 2016, 38(5): 5-9.
[11] 周松林, 茆美琴, 苏建徽. 基于主成分分析与人工神经网络的风电功率预测[J]. 电网技术, 2011, 35(9): 128-132.
[12] MAAS A L, HANNUN A Y, NG A Y. Rectifier nonlinearities improve neural network acoustic models[C]//Proc. icml. 2013, 30(1): 3.
[13] 张豪, 王贤琳. 自适应惯性权重优化的粒子群算法[J]. 智能计算机与应用, 2023, 13(9): 5-8.