基于神经网络的潜艇大机动水动力表达方式

摘要

In this paper, the BP neural network is used to build the submarine's hydrodynamics model at high angles of attack.The model is extremely nonlinear,and the net is trained based on few experiment data.The research shows that, the Bayesian regularization algorithm is reasonable to describe the system.The net is set to learn recurrently to improve itself.The submarine's hydrodynamics predicted by the BP network model shows good agreement with the experiment data.%基于BP神经网络技术对潜艇大攻角机动运动水动力的表达方式进行了研究.所建网络具有循环强化以及自适应设计最佳隐层单元数的特点.应用Bayesian正则化算法进行网络训练,结果表明,这种方法训练的网络具有较高的泛化能力和准确性,适合于表达样本数据较少且非线性强烈的潜艇大攻角机动运动水动力.