Airfoil Optimization Design Method Based On Neural Networks

During the process of aerodynamic optimization design, aerodynamic analysis usually requires a large number of evaluations for high-fidelity. In this thesis, an aerodynamic analysis model based on neural networks is built for aerodynamic shape optimization design. In addition, a new optimization design method is proposed here, which combines genetic algorithm with neural networks model. By using neural networks model instead of N-S equations solver for aerodynamic analysis, this method can achieve high-fidelity design results as well as reduce the expensive computational cost to improve optimization efficiency.The main research issues and achievements in this thesis are as follows:1 From neural networks theory, the effect of BP neural networks factors on nonlinear mapping performances is studied.2 Aiming at characteristics of aerodynamic optimization design, an aerodynamic analysis model based on neural networks is built for aerodynamic shape optimization design. During the process of aerodynamic optimization design, instead of N-S equations solver, this model can meet the demand of high-fidelity and efficiency. Combined with genetic algorithm, an aerodynamic optimization design method based on neural networks is formed here.3 This optimization design method has been applied to practical airfoils in this thesis. Results show that this method can achieve high-fidelity design results effectively as well as reduce the expensive computational cost to improve optimization efficiency.