An abductive neural network approach to the design of runner dimensions for the minimization of wrappage in injection mouldings

In this paper, the finite element and abductive network methods were used to obtain a runner-system design to reach minimum warppage on an injection moulded parts with free-form geometry. This study uses the diameter and length of the runner-system as the major control parameters. In the model creation stage, FEM simulation data were used to derivate an accurate abductive network model for predicting wrappage in injection moulded parts corresponding to different control parameters. In the optimization stage, the simulated annealing (SA) method was used. This abductive methodology allows user to efficiently discover an optimal set of parameters without involving complex iteration between the optimization process and FEM simulation.