Research On The Aerodynamic Shape Optimization Of Vehicle Based On Surrogate Model And MOEA/D

Aerod ynamic shape significantly influe nces a vehicle’s general performance, and aerod ynamic op timization de sign(ASO) make s contributions to improve the level of vehicle integrated design. Conside ring the conflict between optimization objectives, expe nsive computation of ob jective functions and complex constraints, multiobjective opt imization algor ithm ba sed o n surrogate mode l and its app lications on ASO of vehicle are researched, and some useful conclusions are ob tained.To describe the configuration of vehicles exactly and conveniently, configuration parameter method and C lass function/Shape function Transformation(CST) based section-outline method are proposed. The former is used for representing simple three dimensional axis symmetric configurations and the latter is for three dimensional symmetric ones. These two parameterization methods have definite geometric meaning, less design variables, and powerful ability for shape description, and are the impor tant foundation of ASO. By summarizing and analyzing the turbulence model, boundary conditions and computational grid, the computational fluid dynamics(CFD) model used for ana lysis of flow field is built, and its accuracy is validated by solving the flow field around a hypersonic experimental vehicle.After introd ucing the basic theory of Kriging model, the influence of number and distribution of samples and pa rameters of app roximate mode l on mode l accuracy is investigated taking one dimensional test function and aerodynamic force of two dimensional airfoil as examples. As a result, the general rules of constructing surrogate mod el are ob tained. The basic concept and algorithm frame of multiobjective evolutionary algorithms based on decomposition(MO EA/D) are discussed, and the ability of handling nonlinear constrains of MO EA/D is promoted by introd ucing the penalty function method. The ASO research of airfoil is implemented using MO EA/D and nondo mina ted sorting genetic algorithm II(NSGA-II) respectively for compa rison of performance of the two opt imization algorithms. According to the result, MO EA/D has the same accuracy as NSGA-II while it is more efficiency and the distribution of Pareto sets is more uniform.In the optimization algorithm based on surrogate model,the so-called infill criterion is very pivotal to improve the accuracy of Kriging model,reduce the times of computation of obj ective functions and search for the op timal results.Three infill criterions are proposed in this thesis.(1)The hybrid infill criterion which is based on第页three single point methods including expected improvement(EI)is proposed for taking the advantages of parallel computation.(2)As to the collision be tween the predicted value and its standard error,a multi-po int infill criterion based on Pareto front of multiobjective optimization is proposed.(3)To overcome the contradiction between efficiency and precision in multiobjective opt imization,the two-step strategy is proposed,in which the first step is to add points at the place with high uncertainty of the Kriging model to improve the global accuracy and the second step is to add point on the coarse optimization Pareto front based on MO EA/D to improve the accuracy around opt imal solutions.The three criterions are tested by test function and ASO of lifting body,and the results reveal that,compared to EI method,the for mer two multi-point infill method s can significantly reduce iteration times,while the total times of computation of ob jective functions are like ly to increase slight ly.The two-step method can avoid to find the local opt imal solutions and distribute the limited computational resources on improving the accuracy of opt imal result.For validation of the proposed expensive optimization algorithm based on Kriging mod el and MOEA/D, it is app lied to the ASO research of typical bod y of revolut ion and hypersonic glide lifting body.As for the ASO research of body of revolut ion, first the geometric discrimina nt on stability of body of revolution under the condition of high-angle-of-attack is deduced and validated. Then the influences of five typical shroud cur ves of launch vehicle on drag, stability and volume characteristic under different conditions of flow are investigated. According to the result, configuration of shroud curve makes great influence on vehicle performance and some of them are conflict each other, so the multiobjective optimization designed should be necessary. And then, the single and multi objective op timization of launch vehicle with bicone shroud curve and capsule is implemented. The result reveals that the opt imization solution is credible and the availability of expensive op timization algor ithm based o n Kriging mode l is va lidated.As for the ASO research of hypersonic glide lifting body, first orthogonal test analys is method and correlation analysis method are adopted to analyze the sensitivity of design variables, and the influence of shape parameters on performance of lifting body is obtained. Kriging model, multi-point infill methods and two-step strategy are adopted to improve model accuracy. Since both global and local search are taken into account, the optimization process is accelerated and the accuracy around Pareto sets is improved. NSGA-II and MO EA/D are adopted to implement multiobjective optimization of two lifting bodies. The performance of the two optimization methods is compared by the ASO research of lifting bod y with ten parameters. The properties of Pareto sets and flow field of typical op timization solution are thoroughly analyzed and the influence of shape parameters on lifting body performance is explained from the view of flow mechanics. The research shows that the expe ns ive op timization method based on Kriging model and MOEA/D has fine efficiency and reliability, and the design result can make contribution to vehicle integrated design.Aiming to the prob lem arising from ASO research of aircraft, opt imization design methods including configuration parameter method, Kriging surrogate model and corresponding infill criterion, MOEA/D and the method handling nonlinear constraints are researched. These methods are also applied to ASO research of typical body of revolution and hypersonic glide lifting body. The research work in this thesis can greatly advance the development of expensive optimization algorithm based on surrogate model and enhance application of MOEA/D on ASO research.