Research On Blade Optimization Design Of Analysis Code Using Artificial Neural Network And Genetic Algorithm

As one of the three "core" components of the large thrust jetengine,the axial compressor has a very important position in engine pre-research and model development.With the rapid development of the modern aviation industry,higher requirements are placed on the engine.The engine is designed to have a high thrust-to-weight ratio,low fuel consumption and a wide stable operating range.Correspondingly,the difficulty faced by high-performance compressor designs has increased dramatically.As far as the compressor aerodynamic design is concerned,the empirical dependence is still large in each design step.For the optimization,the influence factors and the degree of freedom of the optimization variables increase the complexity of the compressor optimization design.How to improve the accuracy requirements of the empirical design model and fully combine the high-precision design program with the optimization algorithm to achieve the pneumatic rapid optimization design of the compressor has become the bottleneck of the aerodynamic optimization design of the multi-stage axial compressor.For the high dependence of the design process of the multi-stage axial compressor on the expe-rience,this paper uses the artificial simulation results of the existing mature leaf type and the accu-mulated experimental data and empirical parameters formed by the induction of the leaf parameter information,using artificial neural network.Strong learning ability and generalization ability,estab-lished a neural network structure suitable for the prediction of backward angle and loss in the S2 flow surface flow program,and established the blade geometric parameters and aerodynamic parameters through the perfect training sample data under various working conditions.A mapping model with backward angle and total pressure loss coefficients.After replacing the model with the traditional empirical formula,comparing the full three-dimensional CFD numerical calculation and experimental results,the overall accuracy of the replaced S2 flow surface positive problem analysis program is greatly improved.In view of the problem that the computational resources occupied by numerical optimization calculation are too large and time-consuming,and it is difficult to apply to multi-stage axial compressor design,this paper combines the above S2 flow surface high-precision estimation program with genetic algorithm to carry out the compressor.Design parameter analysis optimization.Taking a three-stage axial fan as an example,the blade angle of each radial section of the axial fan and the stator blade is taken as the optimization parameter,and the flow field performance parameter is used as the optimization objective function value.Get the best design leaf shape you need.The three-dimensional CFD numerical calculation and analysis of the optimized three-stage axial flow fan is carried out.The calculated values of the design state working point are compared with the experimental values,and the effectiveness of the optimized design program is verified.Because the aerodynamic optimization design utilizes the neural network mapping model to quickly obtain the flow field parameter information,instead of the iterative calculation process of the flow field parameters in the traditional numerical optimization process,the optimization design efficiency is greatly improved,and the artificial experience dependence and blindness are reduced.And in the whole process,only the design parameters need to be initialized,which saves human resources and shortens the development cycle.It is of great significance for realizing the intelligent design of multi-stage axial compressor.