<p>A three-dimensional computational fluid dynamics (CFD) model is developed to simulate the turbulent flow, heat transfer, and solidification behavior of 5A06 aluminum alloy during the direct chill (DC) casting process. Four criteria—Niyama, KC, Lee, and Suri—are incorporated into the model to predict microporosity, and the grain size distribution is predicted based on the uniform nucleation termination principle. The morphology of the porosity is characterized using X-ray computed tomography (X-CT), and the simulation results are well compared with experimental data. Results indicate that high cooling rate significantly refines the grain structure and reduces the size of micropores, while simultaneously increasing microporosity, and promoting more spherical micropores. The largest grain was observed near the narrow face at approximately 3/8 of the thickness, rather than the center of the ingot, which is caused by a local thermal stagnation zone determined by the geometry of the shunt pocket. Among the criteria, the Niyama criterion exhibited superior stability and predictive accuracy under complex thermal-fluid conditions for Al–Mg alloys in DC casting. Increasing the casting speed deepens the melt pool and enhances convection, resulting in grain coarsening and local porosity. Enlarging the ingot size intensifies central heat accumulation, lowers the cooling rate, and promotes the formation of coarse grains and central porosity. In contrast, the influence of superheat is mainly confined to the sidewall region and has minimal effect on the central structure.</p>