摘要
以清漂无人船质量最轻为设计目标,将船舶设计规范要求的板厚尺寸及最大等效应力、剪应力作为约束条件,通过灵敏度分析筛选确定设计变量。将设计变量进行正交试验,得到神经网络训练和测试样本。构建双隐含层的BP神经网络模型替代耗时的有限元模型,确定应力与设计变量之间的关系,并与不同隐含层的神经网络进行比较,得出双隐含层BP神经网络模型性能更好的结论,从而用其对清漂无人船进行轻量化优化。优化后的船体质量降低10.3%。优化后的船体经有限元分析,满足船舶设计规范要求,说明双隐含层BP神经网络模型在船舶结构设计优化上具有可行性。
The design objective is to achieve the lightest weight of the Garbage Cleaning Unmanned Ship,and the design variables are determined through sensitivity analysis by taking the plate thickness size,maximum equivalent stress and shear stress as constraint conditions required by ship design specifications.The design variables are orthogonal tested to obtain neural network training and test samples.Double hidden layer of BP neural network model was constructed to replace the time-consuming finite element model,determine the relationship between the stress and the design variables.Compared with different hidden layer of neural network,the conclusion that double hidden layer BP neural network model performance is better is drawn.With its Garbage Cleaning Unmanned Ship for lightweight optimization,the optimized hull quality decreases by 10.3%.The results show that the BP neural network model with double hidden layer is feasible in ship structural design optimization.
作者
张楚鹏
陈铭
张道德
ZHANG Chupeng;CHEN Ming;ZHANG Daode(School of Mechanical Engineering,Hubei Univ.of Tech.,Wuhan 430068,China)
出处
《湖北工业大学学报》
2020年第2期10-14,共5页
Journal of Hubei University of Technology
基金
国家科技部重点研发计划(2016YFC0401702)。
关键词
清漂无人船
灵敏度分析
轻量化
双隐含层BP神经网络
Garbage Cleaning Unmanned Ship
sensitivity analysis
lightweight
double hidden layer BP neural network
