摘要
针对线性化法预报弹丸落点存在侧向速度、角速度计算复杂和适用范围小的问题,提出了基于神经网络补偿的线性弹道落点预报方法。该方法在线性假设下,对刚体六自由度弹道进行线性化处理,得到线性弹道模型;将弹丸的圆周运动方程组视为线性定常系统,利用系统的解得到圆周运动的解析式,并利用梯形近似法处理其他参数的导数,得到基于线性弹道的落点预报解析式;然后利用神经网络理论设计了补偿项,不仅解决了线性化法适用范围小的问题,还提高了线性弹道预报落点精度。数值仿真测试结果表明,该方法预报弹丸射程和横偏的最大误差分别约为4m和7m,预报落点时间约0.024ms,比解算6D弹道的时间少了1.451s。因此,该方法可为快速精确预报弹丸落点提供理论参考。
Aiming at problems of the linear trajectory impact point prediction,e.g.computation of lateral velocities and angular velocities is complex and scope of application is small,an impact point prediction method of linear trajectory based on neural network compensation was proposed.First,with the assumption of linearity,the rigid body six degree of freedom trajectory equation was made to be linear trajectory model;secondly,the circular motion of the projectile was regarded as linear constant system,and the formula of the circular motion was obtained.At the same time,the derivatives of the other ballistic parameters are handled by the trapezoidal approximation method;thirdly,neural network was designed to compensate the linear trajectory impact point prediction accuracy.The numerical simulation results showed that the maximum error of the range and lateral deviation prediction method were 4mand 7mrespectively.The impact point prediction time was about 0.024 ms.
出处
《探测与控制学报》
CSCD
北大核心
2017年第4期96-102,107,共8页
Journal of Detection & Control
关键词
神经网络
线性弹道
落点预报
改进型梯度下降法
neural network
linear trajectory
impact point prediction
improved gradient descent method