X型航空发动机神经网络控制研究

【作者】 刘延峰；

【导师】 郭迎清；

【作者基本信息】 西北工业大学 ， 控制理论与控制工程， 2003， 硕士

【摘要】 对于发动机这种工作机理复杂，特性随工作状态和外界条件变化很大的控制对象，常规的反馈控制理论或者最优控制理论很难或无法保证系统性能最优。本文研究了新型神经网络的结构和学习算法，并在此基础上，研究了将神经网络用于航空发动机单回路和双回路稳态控制问题，并初步研究了航空发动机单回路过渡态控制中神经网络的适用性。在一些典型的工作状态进行了仿真。为对比起见，还就经典PID控制器的设计和参数整定给出了完整的过程。 通过仿真可以看出，利用神经网络进行在线学习、在线控制，它可以在发动机对象参数不断变化的情况下，对单回路稳态控制可以达到控制要求，对双回路稳态控制，可以实现系统的解耦控制，使系统变量之间的耦合基本消除。对单回路加速过程控制，设计出的控制器具有良好的跟踪性能。而且该网络结构简单规范，便于实现，对于工作范围宽广的航空发动机控制系统，仿真效果令人满意。更多还原

【Abstract】 Aero-engine was a controlled object with such features as complicated working mechanism , characteristics varied greatly with working state and environment condition. For such controlled object , classical feedback control theory and optimal control theory can’t guarantee optimal performance. In this paper, a new type neural network construction and algorithm was studied . Based on this, the neural network was applied as controller to the single loop and the double loops of aero engine in steady state .By the way, elementary study was applied for the aero engine transient state control . For contrast, the full process for the classical PID controller design and the parameter determination was given.Based on the simulation result , a conclusion was drawn: the neural network can satisfy the control requirement for the single loop steady state under the parameters varying . Realized the decouple and control requirement for the double loops steady state, the coupling between the parameters was decoupled. For the single loop transient state control , the controller possess perfect tracking performance. Moreover, the network construction was simple and normative, easy to be realized. The simulation result was satisfied for the aero-engine control system with broad working range.更多还原