摘要
为了使非高精度的近似模型在船型优化中发挥最大效用,引入不确定性优化方法,采用区间数描述反向传播(BP)神经网络构建的兴波阻力系数近似模型的不确定性,基于主尺度与船型系数联合组成的设计空间构建了最小总阻力的优化模型.采用双层嵌套优化体系,外层采用加入学习因子改进策略的粒子群算法对引入罚函数的总阻力目标函数进行优化,内层采用改进的快速模拟退火算法针对近似模型的不确定域求解目标函数区间.算例表明了不确定性优化体系的优越性与两种优化算法的适用性.
In the process of ship bull optimal design with SBD (simulation based design) technology, approximate model is necessary for high precision CFD (computational fluid dynamics) solver simula- tion. ()n one hand, establishment of very high precision approximation model costs computation time excessively, on the other hand, low precision leads to the uncertainty cognition of the objective function. In order to resolve this contradiction, uncertainty optimization method was introduced. Uncertainty of approximate model built by neural network was depicted by interval number. Then the mini- mum resistance optimization model with principal dimensions and form coefficients as design variables was established. Double nested optimization structure was used. The outer layer used IPSO (im- proved particle swarm optimization) algorithm to optimize the objective with penalty function, and the inner layer used MVFSA (modified very fast simulated annealing) to solve the objective function interval of the uncertainty region of wave resistance coefficient approximate model. Cases calculation shows the superiority of uncertainty optimization method and applicability of the two optimization al- gorithm.
出处
《华中科技大学学报(自然科学版)》
EI
CAS
CSCD
北大核心
2016年第6期72-77,共6页
Journal of Huazhong University of Science and Technology(Natural Science Edition)
基金
国家自然科学基金资助项目(51579022)
中央高校基本科研业务费专项资金资助项目(3132014318
3132016066
3132016215)
关键词
船型优化
近似模型
不确定性
区间规划
神经网络
ship hull optimization
approximation model
uncertainty
interval programming
neuralnetwork