利用改进Elman神经网络预测舰船感应磁场

Application of improved Elman neural network in predicting ships′induced magnetic field

在舰船感应磁场的所有分量中,垂向感应磁场垂直分量很难依靠传统方法进行测量,测量精度也容易受到地磁模拟场均匀度的影响。为避免这一缺陷,提出了粒子群优化Elman神经网络的感应磁场预测方法。在硬件设施良好的磁场监测站测量不同地磁环境下的舰船感应磁场垂直分量、地磁场信号等信息,并建立数据库,利用粒子群优化Elman神经网络学习舰船垂向感应磁场垂直分量与这些磁场信号之间的非线性关系,进而对舰船在未知区域的垂向感应磁场垂直分量进行预测。仿真分析和实验室物理模型实验均验证了该方法的有效性。

Of all the components of induced magnetic field,ship′s vertical component of vertical induced field is always hard to be measured by traditional geomagnetic simulation method,which is li-mited to the low measurement accuracy.To avoid this deficiency,a hybrid predicting algorithm of particle swarm optimization and Elman neural network was proposed.By collecting ships′induced magnetic signatures in magnetic measuring stations with high hardware facilities,the improved Elman neural network can learn the non-linear relationship between vertical component of vertical induced field and other magnetic signatures,and then it can predict ships′induced magnetic signature in unknown areas.Both the simulation analysis and experimental results verify the effectiveness and accuracy of the improved Elman neural network.