Pages 672-675

When the melted temperature or mold temperature of plastic injection molding process was altered, the mass of injected plastic part would be caused to rise and fall, adequate adjustment of molding process variables would be required, thus, the very accurate predictions of molded mass are indispensable. The paper put forward a approach that the fluctuation of injected plastic part mass can be predicted using artificial neural networks quite accurately when key molding process variables are changed, which mainly includes two sections: the experiment model to obtain the weights trained by the back-propagation network (BPN) which apply to real-time control model; the process control model based on the weights and molding process variables to realize the adaptive molding process. The experiment modeling must satisfy the practical request and incarnate the dynamic character when it establishes the relationship between process variables and the mass of molded part. The BPN can be trained by the site data of the modeling, and after it has been trained successfully, it can be used to control the fluctuation of molded mass caused by the change of molding process variable by means of rule-based reasoning.