Research On Manage Of Aerodynamic Data Based On Neural Network And Software Design

The establishment of aerodynamic model is the key in the simulation of aircraft system and an important link in the process of aerodynamic data processing.Whether the aerodynamic model is correct determines the confidence of the entire aircraft simulation system,and the accuracy of aerodynamic parameters affects the stability of the aircraft and other controls.Therefore,it is very important to establish a high-precision aerodynamic model.Neural network has the characteristic of approaching arbitrary nonlinear function.In order to adapt to more and more complex aircraft,the research on the method of using neural network to establish aerodynamic model becomes more and more active.In this thersis,the establishment of high-precision aerodynamic model using neural network is taken as the research objective.First of all,by comparing the characteristics of the basic neural network and combining the nonlinearity of the aerodynamic parameters under unsteady conditions and the characteristics of the parameters changing with time,this thersis chooses Elman neural network to establish the aerodynamic model.However,Elman neural network still has some disadvantages,such as easy to fall into the local optimal solution and slow convergence rate.In this thersis,the Elman neural network is improved from three aspects: convergence speed,dynamic information processing ability and global search ability.After the modeling method is determined,it is necessary to establish a model with aerodynamic data to verify the effectiveness of the method.In this thersis,NACA 2410 subsonic airfoil and NLR 7301 transonic airfoil were taken as test examples to establish their CFD computational model.By using the CFD simulation function in Fluent software,the aerodynamic data of airfoil under steady and unsteady conditions were obtained,and a preliminary aerodynamic analysis was performed on the aerodynamic data obtained by CFD calculation.Then,this thersis introduces in detail the improved method of Elman neural network,how to determine the structure of neural network model combined with practical engineering,and how to use genetic algorithm to optimize the initial value of neural network to build aerodynamic model.After the establishment of the model,the prediction effect of the model is verified in this thersis.After verifying the reliability ofthe model,the aerodynamic model established by NFM-MLP and FSRBF neural network is compared and analyzed in this thersis.Finally,Anaconda and PyQt are used to develop a pneumatic data processing software.Aerodynamic data processing software is composed of outliers processing module,low-pass digital filtering module,data integration module and gas-dynamic identification module.The outliers processing module mainly completes the unreasonable data in the measurement process,removes them and fills in the reasonable value.The low-pass digital filter module mainly filters out the unreasonable high frequency components.The data integration module is mainly composed of data combination,output of the combined data file and graphical display of data.The aerodynamic identification module extracts the parameters of the neural network model completed in this thersis and loads them into the software,so that the software can be used for on-line aerodynamic parameter identification of aircraft.