Local mesh refinement within a multi-block structured-grid scheme for general flows

Abstract A multi-level flow-adaptive mesh-refinement strategy has been formulated and implemented within a structured-grid, multiblock scheme for 2D laminar, transitional and turbulent recirculating flow. Although the locally refined irregular grid-portions are organised as block-structured patches within coarse-grid blocks, a special data structure allows high storage economy to be achieved. Issues of particular interest in the study include the performance of different flow-adaptation sensors based on solution properties and error, and the response of the solution accuracy to inter-grid interpolation practices based on bi-linear, bi-quadratic and compact interpolation. The effectiveness of the local refinement strategy, especially in terms of CPU and memory resources, is examined by reference to four test cases, ranging from an analytically prescribed scalar field over a square domain to a transitional flow over a highly-loaded compressor blade. The study demonstrates that CPU and memory savings with a three-level refinement for the compressor-blade case are of order 70 and 90%, respectively, relative to the requirements of globally fine grids returning similar accuracy levels.