Applying Solution-Adaptive Mesh Refinement in Engine Simulations

A new solution-adaptive mesh refinement capability has been implemented in ANSYS Forte CFD. In this paper, we discuss its implementation and application to engine simulations. This feature is built upon the automatic meshing framework in Forte. The automatic and on-the-fly mesh generation capability in Forte generates a Cartesian cell mesh and the solver uses an immersed boundary method. The mesh generation employs an octree data structure to represent the computational cells inside the computational domain, and all computational cells are perfectly orthogonal. In many scenarios of engine application, high mesh resolution is required to resolve fine geometrical structures or sharp gradients in a physical model. It is impractical and often unnecessary to apply tiny cells everywhere inside the whole domain; thus adaptive mesh refinement and coarsening can be important for obtaining high local mesh resolution while keeping the computational cost low. There are three types of refinement controls implemented in Forte: 1) Fixed Mesh Refinement – Cells along certain boundary surfaces or in predefined volumes are refined to a userspecified level during specified time or crank-angle intervals. 2) Geometry-Adaptive Mesh Refinement – This control is mainly concerned with moving walls. For example, the cell size within valve gaps is dynamically controlled based on the valve lift profiles and user-specified minimum lift thresholds and minimum number of cells in the gap. 3) Solution-Adaptive Mesh Refinement – In this control, cells are adaptively refined or coarsened based on userspecified solution parameters (or their gradients) following certain refinement criteria. This control can help apply high resolution at locations where the high resolution is most needed. We refer to this control as SAM in this paper. In typical engine cases, the first two types of refinement are always used. In this work we focus on the implementation and application of solution-adaptive mesh refinement (SAM). Our first objective is to demonstrate the usage of SAM in different modeling components involved in engine simulations, including basic flow solution, spark ignition and flame propagation, and sprays. We explore different SAM control parameters, demonstrating that refinement is applied at expected locations, and then derive best practices for SAM application. A second objective is to take advantage of SAM’s flexibility to study the mesh sensitivity and mesh convergence of several key sub-models in Forte. A natural question about mesh refinement is: how much should the mesh be refined? The answer will certainly vary for different CFD implementations, mesh types, and sub-models. In this paper, we use SAM to explore the mesh sensitivity of the flame propagation model and spray model. From this we provide guidance for users to select cost effective refinement levels in their engine applications.