The present investigation focused on developing an improved process development methodology which involves a neural network based intelligent control unit. The objective was to utilize current material processing models in an inverse manner and in real-time manufacturing environments. This method eliminates the trial and error procedure that is typical with current process development methods by utilizing the material processing models in an inverse manner. Furthermore, the intelligent control scheme is capable of acting very fast and thus can be used in real-time manufacturing environments. This capability enables the methodology to make up for the inaccuracies in the process model and/or part to part variations that inevitably exist on the factory floor.; The ability of this neural network based control method to achieve such an intelligent control strategy for manufacturing processes was demonstrated on injection molding and resin transfer molding (RTM) processes. The training and test data for the neural network was obtained from process models rather than actual molding processes. The focus of these applications was to control the progression of flow along with some predetermined desired flow progression schemes by controlling the inlet flow rates at multiple inlet gates of a mold cavity. A simulation scheme was developed that involved back-and-forth usage of a molding process model and the neural network based control method in such a way to mimic actual mold filling experiments. Through such simulations, the neural network based control strategy was shown to successfully steer the flow front along the desired flow front progression path. The inlet flow rate profiles obtained from such simulations were utilized during mold filling experiments and the numerical results were shown to agree very well with the experimental results.
展开▼
