基于神经网络的绕组电流闭环控制方法

在解决闭环消磁绕组电流优化计算问题时,会面临将外部磁场推算误差带入电流反演计算或完备的基函数难以确定等问题。为了降低这些因素对舰船最终补偿效果的影响,从智能优化的角度出发,在讨论散布常数对模型预测误差的影响后,确定了适宜的散布常数,建立了内部磁场与补偿电流之间的径向基函数神经网络预报模型。该方法通过样本对网络进行训练,无须推算内外磁场,就能直接得到使绕组磁场与目标磁场拟合误差最小的补偿电流向量。对比其他数值建模方法,其换算精度有所提高,且选择不同的同维向量作为基函数对补偿结果影响较小。船模实验验证了该方法的有效性。

Optimal control method of degaussing currents based on radial basis function neural network

As the errors from off-board magnetic field evaluation and difficulties in determining basis functions tend to affect the result of calculating the degaussing currents,an intelligent control method was introduced.After discussing the influence from spread coefficient,a radial basis function(RBF) neural network model was established for predicting optimal currents from onboard measurements directly.The magnetic field produced by degaussing coils is very similar to ship′s object field.The method can avoid many problems from the numerical model.Its high accuracy and effectiveness were verified by mockup experiments.