Neural Network Control Of Aeroengine

Aero-engine is a complex mechanical device with the characteristics of time-varying, nonlinear and many control variables and so on. So it is difficult to establish the mathematical model of the aero-engine. The control demand cannot be able to meet simply by the traditional control method. At present, intelligent control algorithm is one of the hot topics in the field of controlling aero-engine. The traditional PID control algorithm is introduced in the paper, and the basic principles and several parameter tuning methods of PID control algorithm are also presented. Combining with the small deviation state space model of the aero-engine at a steady operating point, cut-and-trial method can be used to set PID parameters, a PID steady controller is designed based on the control objective. The artificial neural network is one of the most representative algorithms in the field of intelligent control. The basic structures and work mechanism of neural network are introduced in the paper. In order to introduce the control method of neural network, several neural network control models are presented. In addition, the application of neural network in aero-engine is also introduced in the paper. At last, a reinforcement learning NN-based controller for aero-engine is proposed in the paper. The presented controller design has two interrelated entities: an action NN and a critic NN. The action NN is designed to produce a signal for aero-engine and the critic NN approximates certain strategic utility function which evaluates the performance of the action network. The simulation results show that the presented controller has excellent adaptability to the external environment, good anti-interference ability, and also has excellent robustness to the parameter perturbations.