Properties of fiber-reinforced composite are dominated by the microstructure of the fabricated part rather than the properties of constituent materials. As the microstructure of composite is related to the flow-processing route of the fiber-reinforced suspensions and the geometry of the mold, the microstructure of composite can be tailored in order to achieving high-performance composites by exercising control over the flow processing. Thus, numerical methods are used to model the resin flow, the fiber orientation, and mold design, and they become important challenges during molding process. In our previous research, a theoretical fiber-fiber interaction model with a global fiber interaction coefficient was developed. In this study, the three-dimensional (3D) fiber orientation distribution is predicted by combining our developed fiber interaction model and improved Anisotropic Rotary Diffusion - Retarding Principle Rate (iARD-RPR) model. The fiber orientation calculation started from the gate, and from a random orientation to alignment orientation is investigated at that location as the inlet boundary condition. Moreover, simplified deformation behavior of fountain flow was employed for fiber orientation in the flow front region. In experimental, short-glass fiber-reinforced polyamide 6 specimens produced using plate-shaped cavity having three different thicknesses ranging from 2 mm to 4 mm and with the fiber contents ranging from 30 wt% to 65 wt% are carried out using injection molding. The 3D fiber orientation observation and measuring are exactly and automatically examined by Micro-computed tomography system. Comparisons with experimental results, it is showed that this advanced prediction methodology can be effectively used for the 3D fiber orientation prediction in complex parts.