A Hybrid Adaptive Gridding Procedure for Recirculating Fluid Flow Problems

A hybrid adaptive gridding procedure combining the concepts of both local grid refinement and global grid moving has been developed for time-independent recirculating flow problems. The procedure starts with the global grid moving method which provides an initial adaptive solution. Base on this initial solution, large error regions are flagged and local refinement is then applied on the large error regions. The methods for error estimation and interface treatment are discussed. The procedure is assessed in a one-dimensional convection-diffusion equation, a driven polar cavity flow, and a laminar backward-facing step flow. Efficiencies of the various approaches are evaluated. Specifically, in the test problems, the hybrid adaptive grid solution requires less than one-tenth of the CPU time of that of the uniform fine grid solution to achieve the same accuracy, The procedure can be conveniently extended to three-dimensional, irregular geometry flow problems.