Abstract: The phenomenon of flow separation has a critical impact on the flow field structure near the surface of aircraft and their overall performance. Based on two typical separated flows—hump flow and transonic bump flow, this study evaluates and analyzes the predictive performance of the shear-stress transport (SST) turbulence model and its various modifications, which were widely used in aerospace engineering, using the CFL3D solver. By comparing both mean flow fields and turbulence fields with high-accuracy experimental data and large eddy simulation results, it was found that the SST model tends to underestimate turbulence production in the shear layer of the separation zone, leading to insufficient turbulent mixing and frequently overpredicting the extent of two-dimensional separation regions. Modifications incorporating rotation/streamline curvature effects to the SST model showed no significant improvement. While modifications accounting for turbulence anisotropy improved the prediction accuracy of Reynolds stresses, they failed to effectively enhance mean flow field predictions. However, the introduction of a separation flow correction increased turbulence production in the shear layer of the separation zone, thereby enhancing turbulent mixing and significantly reducing the separation zone size. Furthermore, combining the separation flow correction with the turbulence anisotropy modification notably improved the prediction accuracy for both mean flow and turbulence fields. This study provides guidance for the selection and further refinement of turbulence models in engineering simulations.