Design Of An Adaptive Inverse Fault-tolerant Controller Based On EMRAN For Hypersonic Vehicles

Hypersonic vehicles are generally powered by air-breathing or combine cycle engine and cruise at a speed above a mach number of 5 at the altitude from 20 km to 100 km. The dynamic character of hypersonic vehicles is highly nonlinear and coupled duo to the flight condition of hypersonic vehicles is of a large span and of high speed and altitude when operate, and it is hard to perform well in flight control using ordinary control methods. Towards the characters of hypersonic vehicles, and based on the concept of inverse control, a type of adaptive inverse controller whose filters are realized with improved Extended Minimum Resource Allocating Network(EMRAN) algorithm.EMRAN is an online training algorithm of Radial Basis Function(RBF) networks.EMRAN algorithm can be used to train RBF networks fast and online by adding and deleting hidden neurons dynamically and update the network parameters using Extended Kalman Filter(EKF) algorithm, and the network size can be minimum. But EMRAN algorithm doesn't make good use of training data, to overcome it, the EMRAN algorithm is improved that the amount of hidden neurons updated in every loop is determined by the compute ability of platform. Training result indicates that the improved algorithm can make the target RBF network convergence faster.Adaptive inverse control is a method that the controller is designed as same as the inverse system of the object. To overcome the disadvantage that the size of controller network is too large duo to the traditional adaptive inverse controller need several input and output of the past moments, the structure of adaptive inverse controller is improved that the controlled parameters are classified and processed distributed using main controller and compensators.The improved EMRAN algorithm is adopted to train the adaptive inverse controller, and used in a hypersonic vehicle 6-DOF nonlinear dynamic model which is set up according to wind tunnel aerodynamic derivatives. Simulation result indicates that under the flight condition of normal cruise and maneuvering, the adaptive inverse controller can guide the hypersonic vehicle trace the expected input rapidly and steadily. There is no notable decrease of effect when there is perturbation of 30% on aerodynamic derivatives. And when one wing has8% damage, the control effect is still steadily.