Proposing an Anisotropic Adaptive Refinement-Based Grid Optimization Method for Solving Geometric Non-Conservation Issues in Cartesian Adaptive Mesh Generation Traditional methods rely on localized refinement to capture object surface layer information, but face challenges such as inconsistent refinement densities across complex models, extensive adjustment testing, and inefficiency in large-scale computations due to excessive computational loads. The proposed method leverages the geometric attributes of surface triangles within grid cells (e.g., flux area and centroid of cut surfaces) to dynamically adjust cell properties through iterative refinement with stepwise reduction, effectively avoiding degenerate cells caused by abrupt adjustments.