研究了由一个中心刚体带有一对柔性梁的航天器的神经网络控制问题。使用广义卡尔曼滤波训练算法对辨识神经网络和神经控制器进行在线训练，使用间接模型参考自适应控制算法对受控对象进行控制，提出了一种当被控对象的某些状态变量无法量测时进行在线辨识、控制的解决办法。仿真结果表明，此控制方案可以实现刚体由静止到静止的姿态机动，并且对附件振动有很强的抑制作用。 The neural network control problem of a spacecraft with a pair of flexible beams with a rigid center is studied. The generalized Kalman filter training algorithm is used to train the neural network and the neural network controller. The indirect model reference adaptive control algorithm is used to control the controlled object. A new method is proposed when some state variables of the controlled object can not be measured When online identification, control solution. The simulation results show that this control scheme can make the rigid body move from static to static attitude, and it has a strong inhibitory effect on the attachment vibration.