基于CFD和RBF神经网络的潜艇水动力预报方法研究(英文)

Prediction of Submarine Hydrodynamics using CFD-based Calculations and RBF Neural Network

为实现CFD技术在潜艇操纵性优化设计中的应用,文章结合粘性求解器和RBF神经网络预报了潜艇的水动力。通过引入首部和尾部肥瘦指数,确定了潜艇主艇体线型表达的五参数模型。采用均匀试验设计方法,给出了30条潜艇模型的五参数表达。针对每个模型,分别计算了9个漂角下的纵向力、横向力和摇首力矩,得到共计270组数据。为提高计算效率和精度,利用ANSYS ICEM CFD脚本文件和ANSYS FLUENT journal函数实现了从模型建立、网格划分到数值模拟的自动化操作。在多漂角计算过程中,采用"漂角扫掠"方法加快收敛速度。利用上述计算结果训练RBF神经网络,得到了潜艇水动力预报的神经网络模型。以SUBOFF为例,采用该网络预报了其水动力,并与文中数值方法计算结果、试验结果和文献值进行对比,符合较好,说明该方法可应用于工程实践。

To explore the usage of CFD techniques into the optimization design process of submarine maneuverability, CFD-based calculations and RBF neural network were combined to predict the sub-marine hydrodynamics. The fullness of the nose and stern index was introduced to the geometric de-scription of submarine axisymmetric hull, thus creating a five-parameter model for the hull geometry expression. A series of 30 similar hull bodies was adopted by the uniform design approach. For each of the models, 9 different drift angle cases were calculated, and 270 groups of data were achieved con-sisting of the longitudinal force, the lateral force and the yaw moment. To improve the efficiency and accuracy of the computation, automatic mesh and computation using the ANSYS ICEM CFD scripts and ANSYS FLUENT journal functions were used, as well as the drift sweep procedure. A RBF neu-ral network was adopted and trained by the computation results to predict the hydrodynamics of oth-er submarines. For the SUBOFF case, the hydrodynamics were predicted and compared with the CFD-based calculation results, the experimental results and literature values. The results agreed well with each other. It indicates that the method used in this paper is suitable for the practical application in engineering and has a better accuracy and higher efficiency.