A Neural Network Adaptive Controller Considering Expert System for Aero-Engine

摘要

Many linear model based control methods for engine are difficult to ensure desired control performance in many cases due to the large unmodeled dynamics. To deal with the unmodeled dynamics of the aero-engine, this paper presents a model reference adaptive control approach that consider model uncertainty of the system. In this method, neural network is introduced to online estimate the unmodeled dynamics in the system matrix and control gain matrix due to its excellent nonlinear approximation property to improve the performance of the controller. By linearizing the nonlinear model of the aero-engine in the neighbourhood of the stable equilibrium points, a nominal linear model can be achieved. For a given desired trajectories of the controlled variables, a set of ideal control trajectories can be calculated using the obtained nominal linear model of the engine. Then together with the estimation of the unmodeled dynamics, error dynamics of the controlled variables can be built. Consequently, using feedback linearization technique, an adaptive control law is derived to ensure the asymptotic convergence of the closed loop system. For limit protection purpose, several parameter tuning strategies are designed to ensure the stall margin. The simulation results indicate the effectiveness of the proposed control approach.