The primary objective of this thesis is to develop a framework for efficiently simulating unsteady turbulent flows involving multiple moving bodies. To accomplish this, the research first presents the development of exact and efficient wall distance-b...
The primary objective of this thesis is to develop a framework for efficiently simulating unsteady turbulent flows involving multiple moving bodies. To accomplish this, the research first presents the development of exact and efficient wall distance-based algorithms. These include an algorithm for computing the wall distance and a pre-classification overset grid assembly (OGA) algorithm that utilizes the wall distance. Both algorithms are formulated based on the same underlying methodology. The proposed approach attains high computational efficiency while satisfying the essential requirement of exactness, fully accounting for the geometry of the discretized surface.
A conservative relationship between the exact wall distance and the approximate distance computed from a set of reference points representing the discretized surface is established through a rigorous examination of the underlying geometric relations. This relationship is directly incorporated into the search-based wall distance calculation process, resulting in substantial improvements in algorithmic efficiency. It is also applied to the wall distance pre-classification OGA, in conjunction with a conservative reformulation of the pre-classification methodology. In addition, an intuitive load-balancing strategy is proposed to enhance parallel scalability in a mesh-partitioning-based distributed-memory environment.
To validate the performace of the developed algorithms, various numerical tests are conducted. For the wall distance calculation algorithm, the performance demonstrates a speed-up of three orders of magnitude compared to exhaustive search and one to two orders of magnitude compared to existing search-based approaches. In the case of the wall distance-based pre-classification OGA, the method consistently outperforms conventional OGA and shows robustness even in the presence of high proximity objects. The resulting efficiency demonstrates the suitability of the proposed algorithms for large-scale problems—such as unsteady turbulent flows involving multiple moving bodies—which require wall distance calculation and overset grid assembly at every time step.
Secondly, this research presents the development of an in-house flow solver, the Density-based flow solver with Optimized GEometric algorithms (DOGE3D), which integrates the wall distance algorithms and enables efficient multi-body unsteady flow simulations. DOGE3D is a three-dimensional unstructured flow solver based on the cell-centered finite volume method and is designed to solve the compressible Navier-Stokes equations for complex aerospace applications. The supported physics models include inviscid, laminar, and turbulent flow analyses. In addition, the solver incorporates a two-temperature non-equilibrium chemistry model for hypersonic flow simulations.
The implemented numerical capabilities include distributed-memory parallelization based on mesh partitioning, OGA, adaptive mesh refinement, and CFD-coupled six-degree-of-freedom analysis. To validate the solver, a series of numerical tests are performed, demonstrating that it accurately reproduces reference solutions in all cases. In addition, the solver efficiency is demonstrated by confirming that wall distance computation and overset profiling can be performed on the order of seconds for multi-body unsteady simulations involving meshes with tens of millions of cells.