Abstract: The inner surface quality of the space elbow in aero-engine pipeline systems directly affects medium conveying performance. Magnetic particle grinding technology is used for precision machining. To address the challenge of the unobservable grinding process, a simulation system including contact mechanics, wear mechanism and magnetic field force model is established based on discrete element method. The influence of yoke speed on grinding force and material removal is analyzed. Simulation results replace the traditional empirical method to determine the yoke speed range. Subsequently, the PSO-ELM intelligent prediction model was constructed, which was trained by L₁₆(4⁴) orthogonal test data, with process parameters as input and surface roughness as output. The prediction performance and reliability of the model are further verified. Results show that the optimized combination of process parameters effectively reduces the roughness of the inner surface of the elbow, which a deviation of only 1.48% from the predicted value, and completely eliminates material defects, providing a reliable digital solution for the precision machining of aero-engine elbow inner surface.