Title :
Aerodynamic parameter fitting based on BP neural network and hybrid optimization algorithm
Author :
Chao Tao ; Dong Chen ; Wang Songyan ; Yang Ming
Author_Institution :
Control & Simulation Centre, Harbin Inst. of Technol., Harbin, China
Abstract :
A new aerodynamic parameter fitting approach is proposed to avoid online aerodynamic parameter interpolation for advanced flight vehicle trajectory generation, guidance and control. Due to its ability to fit any nonlinear function and simple structure, BP neural network was chosen as the tool to fit the aerodynamic parameters which are the function of Mach number, angle of attack and other variables. A weight value learning method based on hybrid genetic algorithm and support vector machines optimization algorithm is presented in order to overcome the shortcoming of reaching local minimal values of the BP neural network. Simulation results show that aerodynamic parameter fitting time is less than aerodynamic parameter interpolation and the proposed approach is a way to save computation time during trajectory design, guidance and control, and numeric simulation process.
Keywords :
aerodynamics; aerospace control; backpropagation; genetic algorithms; neural nets; support vector machines; trajectory control; BP neural network; Mach number; aerodynamic parameter fitting approach; aerodynamic parameter interpolation; angle-of-attack; backpropagation; flight vehicle trajectory control; flight vehicle trajectory generation; flight vehicle trajectory guidance; hybrid genetic algorithm; hybrid optimization algorithm; numerical simulation; support vector machines optimization algorithm; weight value learning method; Aerodynamics; Fitting; Genetic algorithms; Neural networks; Optimization; Support vector machines; Training; Aerodynamic Parameter; Genetic Algorithm; Neural Network; Support Vector Machines;
Conference_Titel :
Intelligent Control and Automation (WCICA), 2014 11th World Congress on
DOI :
10.1109/WCICA.2014.7053555
