Jerzy Majewski

Investigation of blending-function-based overlapping-grid technique for compressible flows

The paper presents the development and investigation of a blending-function-based overlapping-grid technique for compressible Euler flows. In this technique, local solutions, obtained on overlapping subdomains, are combined into a global function using blending functions as weights. This global function is subsequently used to impose new boundary conditions on interfaces. The local solutions are obtained using the Riemann solver and an implicit-unfactored method, for spatial and temporal discretization, respectively. Several numerical experiments are carried out for flows around circular-arc and two-element airfoils, and demonstrate that the blending function approach improves the numerical convergence for grid configurations containing multiple overlaps.

Anisotropic solution-adaptive technique applied to simulations of steady and unsteady compressible flows

Grid adaptation is a very powerful tool for optimizing CFD calculations. Unfortunately the typical isotropic adaptation used for 3D flows may result in excessive number of elements. The reason for this is that during refinement of the grid around, e.g. the shock wave the grid cells becomes smaller uniformly in all directions. It means that after splitting N cells in the direction normal to the shock wave the number of the cells will be increased to NK3 (K describes how many new edges are created after splitting an old one - if an edge of a cell is split into two edges then K = 2). When the anisotropic adaptation is used the grid is refined only in the direction normal to the shock wave and the number of the cells would be increased to N K.

Anisotropic Mesh Adaptation in the Presence of Complex Boundaries

Grid adaptation is a very powerful tool for optimizing CFD calculations. However typical isotropic adaptation used for 3D flows still may result in excessive number of elements. This is especially the case when lower+dimensional features of the flow are dominant (boundary layers, shockwaves). The present paper investigates anisotropic adaptation for flows with complex boundaries. Of particular interest is the automatic adaptation for laminar / turbulent boundary layer for which case new error indicator is proposed.

Parallel performance of overlapping mesh technique for compressible flows

Abstract The paper presents parallelization of a blended-function-based overlapping mesh technique and its implementation in the case of inviscid compressible flows around complex geometries. The overlapping grid technique is based on the partition of the domain into overlapping subdomain. The solution in each subdomain is obtained by an implicit algorithm and a characteristics-based method, while the blended functions are implemented in order to allow the accurate interpolation of the boundary conditions in the case of multiple overlaps. The above are applied to flows around multi-element airfoil geometries and results from parallel computations on CRAY Server CS6400 and CRAY T3E computing platforms are presented. The load-balancing and parallel performance issues are investigated for several grid systems. A simple model has also been developed for describing the parallel efficiency of the simulations and it is shown that the model results agree well with the parallel experiments.

Anisotropic Adaptation for Simulation of High-Reynolds Number Flows Past Complex 3D Geometries

The paper presents the anisotropic adaptation algorithm applied to simulations of high-Reynolds turbulent compressible flows past complex 3D geometries. The adaptive algorithm relies on anisotropic grid-cell spacing definition provided by the error estimator in a form of a metric field. The error estimator is based on the Hessian of the solution with additional terms used to improve the grid spacing in the regions of high viscous shear forces. The adaptive algorithm is used for the two testcases the Onera M6 wing and the High Lift Prediction Workshop 1 trap wing.

Parallel Performance of Chimera Overlapping Mesh Technique

In the paper we present the Chimera overlapping mesh technique applied to the solution of compressible flow problems. This technique is used to facilitate the grid generation in complex geometries allowing at the same time for natural parallelisation of the problem. The presented algorithm is particularly suitable for cases with large number of meshes overlapping in almost arbitrary manner (including multiple overlaps). The parallel implementation is based on the PVM approach.

Dynamic load balancing for adaptive parallel flow problems

Large scale computing requires parallelization in order to arrive at solution at reasonable time. Today parallelization is a standard in fluid problems simulation. On the other hand adaptation is a technique that allows for dynamic modification of the mesh as the need for locally higher resolution arises. Adaptation used during parallel simulation leads to unbalanced numerical load. This in turn decreases the efficiency of parallelization. Dynamic load balancing strategies should be applied in order to ensure proper parallelization efficiency. The paper presents the potential benefits of applying the dynamic load balancing to adaptive flow problems simulated in parallel environments.

High-Order 3D Anisotropic Hybrid Mesh Generation for High-Reynolds Number Flows

This paper considers a problem of generation of high-order anisotropic hybrid grids to be used for simulation of high-Reynolds number compressible turbulent flows around 3D geometries. The algorithm relies on generating a curvilinear structural grid in the boundary layer region, separately from the usual low-order unstructured grid in the rest of the computational domain. A grid deformator based on an elastic analogy is used in order to curved unstructured elements. The whole process is driven by the global spacing described in a form of a metric field. The presented method is verified for the OneraM6 wing and the L1T2 high lift testcases.

Parallel Efficiency of an Adaptive, Dynamically Balanced Flow Solver

Computations in Fluid Dynamics require minimisation of time in which the result could be obtained. While parallel techniques allow for handling of large problems, it is the adaptivity that ensures that computational effort is focused on interesting regions in time and space. Parallel efficiency, in a domain decomposition based approach, strongly depends on partitioning quality. For adaptive simulation partitioning quality is lost due to the dynamic modification of the computational mesh. Maintaining high efficiency of parallelization requires rebalancing of the numerical load. This paper presents performance results of an adaptive and dynamically balanced in-house flow solver. The results indicate that the rebalancing technique might be used to remedy to the adverse effects of adaptivity on overall parallel performance.

Overlapping meshes technique for compressible flows—parallel implementation

In this paper, parallelization of the Chimera overlapping-mesh technique and its implementation in conjunction with an implicit Riemann solver is presented. The parallelization of the method is based on the PVM approach. Computations are performed for compressible flows over multi-element airfoils. Efficiency results are presented for fairly complex domains consisting of a large number of meshes overlapping each other in an almost arbitrary manner, including multiple overlaps. The parallel performance of the method is investigated on the Cray CS6400 and Cray T3E computing platforms.