Investigation Of Overset Unstructured Grids Algorithm

The unsteady flow field past multiple bodies in relative motion is a common-faced engineering problem, in which there is a strong aerodynamic interaction among bodies. The aerodynamic interaction which is, in nature, always unsteady and nonlinear can make a serious impact on aerodynamic characteristics of flying vehicle and sometimes can threaten the safety of flight. Recently, with the rapid development of numerical simulation methods and computer technologies, computational fluid dynamics (CFD) method has been used to handle these problems. Overset unstructured grids method which was introduced by Nakahashi has the advantages of unstructured grid method and Chimera grid method and is very suitable to deal with unsteady flow fields past complex geometries and/or multiple bodies in relatively moving. However, the overset unstructured grids method is relatively unexplored and it has many problems such as automatic inter-grid definition, efficient algorithm for donor searching, interpolation between subgrids, general type element overlapping grids, dynamic overset grids assembling, parallel computation algorithm, etc. In this thesis, a new efficient and reliable dynamic overset unstructured grids method is proposed and a distributed memory parallel unsteady solver based on the method is developed to predict unsteady flow fields with multiple bodies in relative motion.First of all, the numerical methods which are used in this thesis are discussed. The unsteady compressible Euler/Navier-Stokes equations are solved using the finite volume method in spatial discretization and the explicit multistage dual time-stepping scheme in temporal discretization. The low speed preconditioning method is applied to inner iteration of the dual-time stepping to accelerate the solution of the governing equations. In the simulation of turbulence flows, the Spalart-Allmaras one-equation turbulence model is implemented to evaluate the turbulent viscosity and the wall distance of grid nodes is evaluated by an efficient advancing front method.Secondly, the inter-grid boundary definition method of overset unstructured grids is investigated. Nakahashi's method is extended to any type element and any number of overlapping grids. To improve the computational efficiency, a neighbour-to-neighbour searching algorithm coupled with advancing front method and ADT algorithm is used to search donor cells for all grid nodes in other component grids and as a consequence, the time used in inter-grid boundary definition is reduced. A general interpolation method which can deal with any element type is presented to realize information transfer between the overlapping grids. Many test cases were considered to validate the developed overset unstructured grids method and the numerical results indicate that the method is efficient and robust.Then, a new dynamic overset unstructured grids algorithm is proposed and by coupling the six degree of freedom motion equations, it can be used to handle the unsteady flow problems with multiple bodies in relatively moving. To evaluate the capability of the dynamic overset unstructured grids algorithm for these problems, it was applied to a number of numerical test cases (3D store separation, Caradonna rotor in hover and flight forward, and Robin helicopter with a four-blade rotor). The results show that the dynamic overset unstructured grids method developed in the thesis is not only efficient and reliable but also automatic.At last, a distributed memory parallel computation strategy for unsteady Euler/Navier-Stokes solver based on dynamic overset unstructured grids method is proposed and a parallel software was developed to predict unsteady flow fields with moving geometries. The parallel computation is based on the dynamic domain decomposition method which is performed by using METIS system at each physical time step. Load balance is achieved by respectively partitioning each component grid according to the number of CPUs and weighting the active nodes and inactive nodes. Communication between processors is implemented by calling MPI standard library. Numerical test results on the developed distributed memory parallel strategies indicate that a good parallel performance is achieved.