Civilian Aircraft Based On Neural Network Methods Advanced Aerodynamic Wing Design

As the rapid development of computer technology and database technology, civil aircraft aerodynamic design faces many new directions. The whole design process demands more efficient, intelligent and highly credit. This thesis brings forwards a civil aircraft wing design process based on artificial neural network technology and database technology. An expert database is built by neural network method, which can automatically give the reference geometry according to the design target along with the optimize direction by the certainty factor inference method. During the optimization circle, another neural network is used as the assistant for CFD calculation, so that we can inspect a large number of designed geometry in a short time, which greatly increases the design efficiency.The main research issues and achievement in this thesis are as follows:(1) Build up an expert database according to different geometry and aerodynamic characteristics in airfoils and wings. A SOM neural network is used to classify all the aerodynamic geometry and select most suitable ones as the design reference to aim the design target.(2) Define the optimize direction by certainty factor inference method. Relevance is built between the geometry and aerodynamic characteristics which clearly indicates the optimize direction.(3) Based on the reference geometry and optimize direction mentioned above, the network gives a large number of optimized geometry. A BP network is built up as the assistant for CFD calculation to inspect all the designed geometry in a limited time.In this thesis, we build up a database including 117 supercritical airfoils and 150 supercritical wings. The neural network finally gives the design results which can generally meet the design target. Result shows the neural network method can achieve highly credit aerodynamic geometry as well as high efficiency.