基于插值型径向基神经网络的弹丸落点预报方法

A Projectile Impact-point Prediction Method Based on Interpolation Radial Basis Function Neural Network

针对数值积分法预报弹丸落点存在的解算时间较长、迭代过程中易产生大的累积误差等问题,提出了基于插值型径向基(RBF)神经网络的弹丸落点预报方法。该方法首先建立了落点预报数学模型,然后采用严格插值的方法离线训练RBF神经网络,分别逼近射程和横向偏差(横偏)非线性方程,最后利用预报模型进行落点预报仿真测试。仿真结果表明,该方法预报弹丸射程和横偏的平均误差分别为0.060 2m和0.001m,其预报落点的时间在40ms以内,并与数值积分法相比,预报落点的时间少1 268.22ms。因此,提出的插值型RBF算法预报弹丸落点是有效可行的,可为实际工程应用提供参考。

Aiming at problem of numerical integral algorithm for impact-point prediction （IPP）, a method based on inter- polation radial basis function （RBF） neural network was presente＆ Firstly, the new mathermatical motel is built for IPP. Then this model based on exact interpolation RBF neural network was trained, and the motel was used for predicting im- pact point. The results show mean range deviation is 0. 0602m and lateral deviation is 0. 001m So the precision for IPP can reach the requirement of error precision. And prediction time of RBF algorithm is under 40ms At last, compared with prediction time of numerical integral algorithm, prediction time of RBF algorithm is much shorter. 511 in all, this predic- tion method based RBF neural network is effective and feasible.