基于神经网络代理模型和遗传算法的适伴流最佳环量对转桨设计方法  

Design method for wake-adapted optimal contra-rotating propellers via neural network surrogate model and genetic algorithm

作　　者：薛颖[1] 黄永生 杨晨俊[1] XUE Ying;HUANG Yong−sheng;YANG Chen−jun(State Key Laboratory of Ocean Engineering,Marine Numerical Experiment Centre,Shanghai Jiao Tong University,Shanghai 200240,China;School of Naval Architecture and Maritime,Zhejiang Ocean University,Zhoushan 316022,China)

机构地区：[1]上海交通大学海洋工程国家重点实验室,船海工程数值试验中心,上海200240 [2]浙江海洋大学船舶与海运学院,浙江舟山316022

出　　处：《船舶力学》2025年第4期517-527,共11页Journal of Ship Mechanics

基　　金：中央高校基本科研业务费专项资金资助(AF0100142/005)。

摘　　要：基于升力面理论涡格法提出了一种适伴流最佳环量对转桨的设计方法。该方法首先建立了对转桨径向环量分布与推力、扭矩之间非线性关系的神经网络代理模型,其数据样本由涡格法计算得到;然后采用遗传算法,在给定总推力和扭矩平衡的约束条件下,以总效率为目标对前、后桨的径向环量分布进行优化;最后根据优化得到的最佳环量分布及指定的弦向负荷分布形式设计前、后桨的螺距分布及拱弧面。以高速水下航行体的对转桨为例进行研究,并用非定常RANS方法进行了自航模拟,验证结果表明,设计桨的自航点转速与原型桨基本相同,总效率和扭矩平衡度都有所提高。A design method for wake-adapted contra-rotating propellers(CRPs)with optimal circulation distribution was presented based on the vortex lattice model in lifting surface theory.The implicit relations of propeller thrust and torque with the radial circulation distributions of CRPs were modelled via neural network,where the input data for training of the neural network were yielded from an in-house vortex-lattice code.Subject to the requirements for total thrust and torque-balance,a genetic algorithm was employed to optimize the radial circulation distributions of the forward and aft propellers to maximize the total efficiency.Taking the optimal radial circulation distributions and a prescribed chordwise distribution of circulation as the objective,the camber surface geometries and pitch distributions of the forward and aft propellers were designed.Numerical example was presented with the CRPs for a high-speed underwater vehicle,and the design results were then validated in self-propulsion simulation by solving the unsteady RANS equations.While the efficiency and torque-balance of the designed CRPs are slightly improved against the prototype,the minimum pressure values on blade surfaces are significantly increased,which is favorable for retarding the inception of cavitation.

关 键 词：对转桨 适伴流 最佳环量分布 神经网络 遗传算法 涡格法 

分 类 号：U661.313[交通运输工程—船舶及航道工程]