Abrasive Flow Machining (AFM) process can not only promote product’s inner surface quality but also economize the labor cost. However, the issue of inner surface quality includes not only the surface roughness but also the geometric variability before/after the AFM process. Based upon the experimental data of the effects of AFM process parameters, e.g., length of stroke, extrusion pressure, number of cycles, percentage of abrasive concentration, and abrasive grain size, Taguchi experimental design concept, L18 (61×34) mixed orthogonal array, is used to determine the S/N ratio and optimize the AFM process parameters. The statistical model could predict about 91.39%, and 96.4% for surface roughness (Ra), and material removal respectively. Length of stroke was more effect on surface roughness while percentage of abrasive concentration was more effect on material removal.