Abstract
The objective of this paper is to investigate the optimization schemes for intelligent process control based on neural networks in injection molding. To achieve the goal of intelligent process control, performance indexes, formulating by multi-losses functions, are adaptively optimized for reverse deducing the process control parameters from the quality factors of parts. In addition, the requirements on quality factors such as dimensions, shrinkage, and warpage are predicted by making use of the Computer-Aided Engineering software, namely C-MOLD, with the process window pre-screened by the Design of Experiments procedure. Hereby, a model consisting of Radial Basis Functions Networks (RBFN) is employed for representing the causal factors between the process control parameters and the quality factors. And, the RBFN model is then trained for optimizing the given performance indexes with an adaptive optimization scheme. Finally, two example cases based on numerical simulations on process control are demonstrated for verifications. It is observed that the proposed intelligent process control in injection molding could automatically achieve stable and nearly optimal process conditions within a short period of time for the given quality requirements. Therefore, the intelligent expert controller could be applied for practical uses on the shop floor in the future.
