【摘要】 用所获得的飞行数据训练前馈神经网络,可直接估计飞机的飞行参数。对于飞机的短周期运动方程,网络的输入变量是飞机的仰角、俯仰角速度和舵面偏转角,输出变量是气动力和力矩系数。训练过程中,时间点的输入由输入节点表示,相应时刻的输出在输出节点上获得。获得的值与对应的期望值对比,误差使用反传学习算法传回网络,用以更新连接权值。模拟飞行数据仿真证明,该方法有效并具有较好的鲁棒性。更多还原

【Abstract】 The feed-forward neural network was trained with obtained flight parameter, the value of parameters can be directly estimated. In motion equation of aircraft in a short period, input variable of network is elevation, speed of pitch angle and deflection angle of control piston for aircraft, and that output variable is aerodynamic force and moment coefficient. In training process, input point for every-time was expressed with input node, output of corresponding time was obtained from output node. Compared obtained value with corresponding expected value, error was sent to network with reverse learning algorithm, so that renovate connect weighting value. The flight data simulation shows that the method is validated and has better robustness.更多还原