The NVH problem of electric vehicles requires a revolution in gear precision machining technology. Abrasive flow machining is one of the effective methods for polishing complex surfaces such as gears, and fixture design is a crucial part of the abrasive flow machining process. In the process of optimizing the design of abrasive flow fixtures, the selection of physical models contradicts the accuracy of simulation results and the computational expense of simulation. Different viscosity media were selected, and simulation experiments were conducted using different viscosity models and flow models. The fluid pressure distribution, velocity vector, wall shear, and streamline distribution cloud maps reflecting machining uniformity were analyzed to explore the steady-state simulation results of abrasive flow in the slit model. It was found that the distribution trend of simulation results from different physical models has strong similarity, which can achieve consistency in the processing area streamline in the simulation results, proving the feasibility of replacing complex physical models with simple physical models for fixture optimization simulation. By applying simulation results and using the simplest Newtonian fluid - water as the medium, and utilizing the streamline distribution in the machining area, the design optimization of the gear shaft abrasive flow fixture was carried out, achieving the removal of gear “ghost frequencies” after abrasive flow machining.