This paper presents a methodology for obtaining an optimal parting line of an injection mould based on FEM analysis and managements. This research utilizes Taguchi's method to design a DATABASE for the gate position and warping of parts. Firstly, it sets up an injection model via a CAD system, and secondly, utilizes FEM to analyze the warp of the injection mould in different gate positions. Because of different gate positions, the parts experienced extreme differences in the amount of warping. This paper applies the neural network (abductive network) to build the relationship (database) between the gate position and amount of warp. Engineers can use this database without recourse to a CAD model and FEM solution to predict the amount of warping at different gate positions. This procedure saves time in building CAD models and on carrying out FEM analysis. The optimal parting line parameters can then be reached through a Simulated Annealing (SA) optimization algorithm. The methodology uses multi-objective function criteria, with a performance index to obtain perfect parts. This method of database management can offer various design fields of injection mould (e.g., design of gating and runner, and design of cooling system). It already achieves real optimization of design for injection moulding.