摘要
基于有限元法的船体舱段整体结构优化优于分块优化,但由于其实现存在很多困难,目前国内研究较少。提出一种综合有限元法、均匀设计、神经网络等方法和理论的优化方案,对大型油船立体舱段结构优化进行了研究。建立了舱段结构强度评估的参数化有限元平台;构建了用于试验的均匀设计表,设计了171次试验,将之代入有限元平台进行计算,得到了用于神经网络训练的样本对,从而建立了设计变量和目标函数及应力响应的神经网络响应面。按照一般的优化流程,采用离散粒子群算法,对某30万t超大型油船舱段整体结构进行了优化设计。优化结果表明该方案是可行的,合理的,可以大大缩短直接计算消耗的机时。
Structure optimization for whole tank based on FEM is better than one for the separated. However, now there are not many researches on it because of much difficulty. A new approach which includes FEM, uniform design and neural network, researching on the optimization of large oil cargo tanks is presented. A parameter FEM platform for evaluation on stresses of cargo tanks is established. Then, 171 experiments are designed using uniform design table on FEM platform. Samples through experiments are put into neural network which built a NN respond surface. At last, based on the usual optimization method, discrete particle swarm algorithm is used for a 300 000 DWT oil tank. The result shows that the whole tank optimization is feasible and reliable, which can short much time expended by FEM.
出处
《中国造船》
EI
CSCD
北大核心
2008年第2期41-51,共11页
Shipbuilding of China