Abstract: Aero-engine is a kind of multi-variable plant，which is difficult to control with the possible interaction between input variable and output variables.This paper introduces a multi-variable decoupling control with single neuron，adopting the modified Hebb learning algorithm to fast convergence.The turbojet engine is further simulated on-the-ground and in-the-air，which proves several advantages of this control system such as complete decoupling，fast response，small static error，and easy algorithm.And the author concludes that using two single neurons as Dual-Variable controller is effective to reduce the number of controller all over the flight envelope.

Key words: aircraft turbojet engine, single neuron, dual-variable decoupling control, modified Hebb algorithm

摘要： 针对航空发动机这样的多变量控制对象，要解决的突出问题是输入变量对输出变量的交叉影响，介绍了单神经元进行多变量系统解耦控制的基本方法，采用改进的Hebb学习算法以加速收敛。对某涡喷发动机的数学模型进行了双变量单神经元PID控制仿真研究，结果表明：采用此算法构成的神经网络PID控制对地面模型和高空模型都具有完全解耦、响应速度快、稳态误差小、算法简单的优点；用两个神经元作为双变量控制器，可以使整个飞行包线内的控制器数目明显减少。

关键词: 涡喷发动机, 单神经元, 双变量解耦控制, 改进的Hebb学习算法