无人机涡喷发动机的自适应神经网络控制研究

该文介绍了一种新型的神经网络自适应控制方式，它由基于多层BP网络的近似PID构成的神经网络控制器和基于带自反馈的Elman神经网络构成的模型辨识器共同组成。Elman网络是一种新型的动态递归神经网络，具有很好的逼近能力和性能；改进的自反馈网络具有更大的灵活性。为加快收敛速度，文中采用了共轭梯度算法，选择共轭方向作为最小化方向；在用于无人机涡喷发动机的不同状态的控制中被证实是非常有效的，具有鲁棒性好、响应速度快、稳态误差小等优点。

Research on Self-adaptive Neural Network Control of Pilotless Turbojet Engine

This thesis introduces a new self-adaptive controlling method of the neural network, which is made up by the neural network controller based on similar PID of the multiple BP network and the model identifier of the self-adaptive neural network. Elman network is a new dynamic recursive neural network, and it has good approaching capability and function; and the ameliorated self-adaptive network is of better flexibility. In order to accelerate the speed of convergence, the conjugation grads arithmetic is adopted with choosing the conjugation direction as the minimized direction. The method is proved effective in the control of the Pilotless Aircraft Turbojet Engine and has such merits as good robustness , sensitive response , and minimal stable error.