Study On High Fidelity Multidisciplinary Design Optimization Technique Of Vehicles

The application of high fidelity model becomes more widely available in modern aircraftdesign. Multidisciplinary Design Optimization (MDO) based on high fidelity model plays akey role in aircraft design and optimization because of its ability to improve productperformance and reduce development time. This article is based on the high fidelity aircraftdesign with characteristics of complex structure, numerous disciplines, high interdisciplinarycoupling and nonlinear performance. The MDO model, approximate technique, MDOframework, parametric modeling and disciplinary interface are studied for overcoming theshortage of traditional MDO modeling and solving with high fidelity model. On this basis theapplication pattern of high fidelity MDO in aerospace engineering is explored, and feasibleresearching ideas and solving schemes are proposed. The main contents and achievements ofthis dissertation are summarized as follows:(1) The MDO problem modeling method based on Disciplinary Relation Matrix (DRM)is presented, on this basis the MDO solving procedure is reconstructed and the MDOevaluation criteria is established. The coupling relationship recognition criterion is proposedbased on the DRM with respect to disciplinary interface of input and output variables. Thetraditional MDO solution methods are redefined and standardized, and then the MDOmodeling method and solving procedure with discipline model at the core are established fordesigners. The coupling degree and complexity degree of discipline and MDO system arestudied based on the disciplinary local impact and system impact in DRM, and finally theevaluation criteria of MDO system is established.(2) The anisotropic kernel parameter optimization method of RBF and Kriging surrogatemodels based on structure risk minimization principle is studied. The kernel parameteroptimization is introduced into kernel parameter designing of RBF and Kriging. According tothe objective function of smoothness as generalization abilities, the anisotropic kernelparameter optimization method is proposed. Several benchmark functions and NACA0012wing section optimization problem are used to verify its effectiveness, and the comparisonand analyses of these surrogate models are discussed based on sampling points with uniformdistribution and nonuniform distribution.(3) A MDO solving framework based on asymptotic sampling correction of disciplinarysurrogate model is presented. The disciplinary surrogate model is updated with dynamicadaptation based on the mode-pursuing sampling method. According to the MDO model with DRM, MDO solving procedure is decomposed into framework level, strategy level anddiscipline level for designing and managing independently, and a new interactive and parallelmethod named Disciplinary Model Pursuing Sampling-MDO Framework (DMPS-MDOF) isestablished. The above method is detail studied from dynamic acceleration, disciplinarycorrection, asymptotic convergence, parallel calculation, multi-methods selection andinteractive solving environment. For the lack of design points around the optimal discretedesign space region, the above method is improved by multivariate normal distributionsampling. Numerical results indicate these methods have the characteristics of parallelcalculation and rapid convergence, only less than half of the discipline evaluations are neededto get the proper results compared to traditional MDO solution method.(4) A novel aircraft geometry shape parameterization modeling method based on auniversal three-dimension Class/Shape Transformation (CST) technology is presented. Thethree-dimension CST geometry equation is established by expanding the originaltwo-dimension CST method using B-spline function and lengthways/lateral profile functionsas original CST. A full three-dimensional aircraft geometrical shape description method isproposed with components build-up technology based on aircraft characteristic componentlibrary. On this basis the multidisciplinary master model is established with three-dimensionCST geometry model. The integration modeling method and design process is also discussedon the aerodynamic, structure and aerothermodynamics discipline models.(5) The transformation method of field variables based on disciplinary interface withProper Orthogonal Decomposition (POD) is studied for reducing MDO problem dimension.Three states of MDO problem with coupling field variables are discussed and correspondingsolving models are proposed. This approach is verified with a classical static aeroelasticoptimization problem coupling with CFD and CSD models on M6wing, and the resultsdemonstrate this method can save lots of time for parallel calculation support.(6) Based on the aircraft system rapid design and optimization software (MCDesign)developed by national key laboratory of aerospace flight dynamics in northwesternpolytechnical university, the design process of MDO modeling and solving framework onhigh fidelity model aircraft is studied, and a MDO solving environment is developed forsupporting the abilities of MDO problem quick definition, automatic modeling and dynamicadapted solving. The X-37hypersonic aircraft optimization problem and relevant geometry,aerodynamic, structure, aerothermodynamics, trajectory, mass disciplines are modeled withhigh fidelity model, and the optimization problem is successful solved in the above MDOsolving environment.