A control parameter partition mapping method based on aerodynamic parameter linearity

Abstract

A control parameter partition mapping method based on aerodynamic parameter linearity includes: generating a feature point set covering a nonlinear full airspace envelope according to the flight speed and airspace of an aircraft; establishing a state space equation according to the feature points, constructing a quadratic cost function, and solving the state feedback control parameters when the performance index is optimal; calculating the aerodynamic parameter linearity of each group of feature points, and dividing the nonlinear full airspace envelope into multiple linear sub-regions according to the linearity; in each linear sub-region, a double-implicit-layer BP neural network is constructed, the feature points in the linear sub-region and the corresponding state feedback control parameters are taken as the input and output, a comprehensive loss function including a regional fitting error term and a boundary smoothing constraint term is constructed, and the corresponding double-implicit-layer BP neural network is trained; the trained neural network is placed in the control system to form a control parameter calculation model; the real-time acquired flight Mach number, dynamic pressure and synthetic angle of attack are input into the neural network, and the optimal control parameters are output.