The 10th Asian Symposium on Computational Heat Transfer and Fluid Flow ( ASCHT2025)

An adaptive mesh refinement (AMR) method for the discrete unified gas-kinetic scheme (DUGKS) with the Bhatnagar-Gross-Krook (BGK) model is proposed in this paper. The method adopts a pointer-based refinement strategy and introduces a neighbor update operation to avoid the recursive tree traversal. Instead of traversing hierarchical data structures, each block stores direct pointers to its parent, children, and both orthogonal and diagonal neighbors. During the refinement process, neighbor connectivity is reassigned through explicit pointer operations, which significantly reduces the computational complexity compared to conventional tree-based AMR frameworks. Only leaf blocks are evolved during time marching and conservative flux reconstruction is applied at coarse-fine interfaces to ensure solution consistency and stability. To verify the performance and accuracy of the proposed method, two benchmark cases are studied: lid-driven cavity flow and flow around a square cylinder . In both cases, the adaptive mesh captures key flow features with high resolution while reducing the total number of cells, demonstrating the method's effectiveness in balancing computational cost and accuracy.
 

DUGKS,pointer-based refinement strategy,Adaptive mesh refinement,neighbor update operation