Modelling an aircraft´s aerodynamics leads to a nonlinear system. The depiction of physical relations is a complex inaccuracy task, however the determination of aerodynamic parameters is even more complex. It takes several steps to obtain these parameters from wind tunnel tests and numeric methods. Remaining discrepancies can be reduced by flight test and parameter identification methods (PID). In this paper, this PID task is solved with a Modular Neural Network. Each aerodynamic parameter is represented by one module. Using so called input-weights allows the identification of nonlinear derivatives. A Neural Network offers some advantages towards comparable methods. For example, it is possible to integrate an adaptive model into a realtime flight simulator. The effects of optimized aerodynamic parameters on flight mechanics are assessed by objective comparison of flight test data and simulation results.
«Modelling an aircraft´s aerodynamics leads to a nonlinear system. The depiction of physical relations is a complex inaccuracy task, however the determination of aerodynamic parameters is even more complex. It takes several steps to obtain these parameters from wind tunnel tests and numeric methods. Remaining discrepancies can be reduced by flight test and parameter identification methods (PID). In this paper, this PID task is solved with a Modular Neural Network. Each aerodynamic parameter is re...
»