Parallel Unstructured Mesh Deformation Algorithms And Their Applications In CFD

Numerical simulations like the aerodynamic shape optimization and aeroelastic computation usually involve moving boundaries,and always carry out with transonic/supersonic compressible flows.Therefore,this thesis focuses on compressible flows with moving boundaries,and studies the core dynamic mesh algorithms and flow solvers based on unstructured grids and the finite volume method.A series of efficient parallel mesh deformation algorithms and parallel density-based implicit solvers are proposed,and are successfully applied to airfoil pitching and shape optimization problems.The main contributions of this thesis are as follows:1)This thesis aims at the widely used mesh deformation algorithm based on the radial basis function interpolation,and the greedy selection data reduction method is optimized and improved.An incremental approach that fully utilizes the solution of the previous selection step is proposed to speed up the coefficient linear system solving.While maintaining the robustness of the RBF algorithm and further improving its parallel computing efficiency,this thesis proposes two parallel strategies for pre-known moving boundaries and unpredictable moving boundaries.Then,an improved parallel RBF algorithm is implemented based on the OpenFOAM framework.The deformation ability and computational efficiency of the proposed algorithm are analyzed through typical cases,and the performance has doubled.2)By further loosening the constraints of the greedy RBF mesh deformation algorithm,a more efficient parallel mesh deformation algorithm based on the support vector machine is proposed.Then this thesis implements the novel mesh deformation algorithm in parallel based on the OpenFOAM framework.Moreover,the parameter setting strategy of this machine learning method is given while applying to the mesh deformation.After that,quantitative analysis of the performance of the SVM and RBF method in terms of deformability,control point selection and computational efficiency is investigated.The performance of SVM method increased by a factor of six.3)For the first time,an arbitrary Lagrangian-Eulerian method is introduced in the OpenFOAM framework to implement a density-based LU-SGS implicit parallel solver,which makes up for the application defects of the OpenFOAM software in compressible moving boundary problems.The proposed solvers can be easily coupled with different numerical methods,turbulence models and mesh deformation algorithms,and they can be applied to various flow problems such as steady and unsteady problems,laminar and turbulent flows,static and dynamic meshes.Furthermore,many cases are tested to verify the availability and parallel performance of the proposed solvers.4)This thesis first in-depth compares and analyzes the characteristics of the OpenFOAM and SU2 open source CFD software in terms of framework composition,algorithm type,data structure and application fields,and provides relevant guidance for research based on CFD frameworks.Based on SU2 software platform,this thesis combines CFD numerical simulation technology,geometry parameterization method,continuous adjoint method and the proposed SVM mesh deformation algorithm,and they are successfully applied to the aerodynamic shape optimization for airfoil design.