针对传统神经网络算法无法有效利用少量实际飞行数据的问题,提出基于置信度神经网络的机载武器攻击区拟合算法。为有效区分、利用不同置信度样本,算法通过计算样本误差容量区间并设置样本敏感系数,修改用于权值调整的网络预测误差,得到更准确的拟合网络。仿真结果表明,置信度神经网络算法相比传统神经网络算法,具有更高的估计精度,增加了拟合攻击区的成功发射概率。

For the problem that the traditional neural network algorithm cannot effectively use a small amount of actual flight data, a fitting algorithm of airborne weapon attack area based on confidence neural network is proposed. In order to effectively distinguish and use different confidence samples, a more accurate fitting network is obtained by calculating the sample error capacity interval and setting the sample sensitivity coefficient, modifying the network prediction error used for weight adjustment. The simulation results show that, compared with the traditional neural network algorithm, the confidence neural network algorithm has higher estimation accuracy and increases the success probability of fitting attack area.