Abstract: Based on the BP neural network, the 4×24×12×2 four-layer structure automobile magnesium alloy wheel hubs low-pressure casting optimization model was established, the input layer was the casting mold preheating temperature, pouring temperature, pouring speed and filling pressure, and the output layer was the grain size and yield strength. The low-pressure casting molding process of automobile magnesium alloy wheel hubs was optimized through the BP neural network optimization model. The results show that the grain size of the optimized automobile magnesium alloy wheel hub castings is reduced from 146.8 μm to 123.2 μm, the grain size reduction rate is 16.0%, the yield strength is increased from 169.7 MPa to 185.6 MPa, and the yield strength improvement rate is 9.4%. The optimal process parameters for low-pressure casting of automobile magnesium alloy wheel hubs are casting mold preheating temperature of 380℃, pouring temperature of 690℃, pouring speed of 0.50 m/s and filling pressure of 6.5 kPa.