中文摘要：

高超声速飞行器级间分离时，飞行速度约为6Ma，动压约为70kPa，前后体之间会有较强的气动干扰，造成飞行器出现姿态偏差。为了抑制这种气动干扰，提出了一种基于CMAC神经网络的预置舵偏设计方法。该方法利用CMAC神经网络的非线性映射作用，并对CMAC神经网络结构进行改进，不以网络输出量为网络自适应学习的输入，而是以分离后的攻角为网络学习的输入，计算不同的分离干扰所需的预置舵偏值。通过仿真验证，文中提出的预置舵偏设计能够有效抑制分离气动干扰对攻角和侧滑角的影响，能使角度偏差由4°减小到0．02°。

英文摘要：

Aim. The Introduction of the full paper reviews some papers in the open literature and then proposes the design method mentioned in the title, which is what we believe to be more effective than previously and which is explaimed in sections 1 through 3. We explain how to calculate the separation interference of hypersonic stage separation in section 1. A CMAC （cerebellar model arithmetic computer） based rudder deflection presetting method is proposed in sections 2 and 3. The rest of the core of sections 1 through 3 consists of： ＂Using the nonlinear mapping function of the CMAC neural network, whose structure is improved by us, the rudder presettings corresponding to different separation interferences are calculated. Instead of the output of the CMAC neural network, the angle of attack at the end of stage separation is the input of learning arithmetic. ＂ The simulation results, presented in Tables 1 and 2, show preliminarily that the rudder deflection presetting indeed effectively reduces the biases of the angle of attack and sideslip angle, which are hugely affected by separation interference, from 4 degrees to approximately 0. 02.