Neural-Based Manufacturing Process Optimization of Short Glass Fiber Reinforced Polycarbonate Composites in Injection Molding by Genetic Algorithm and Extension Set Analysis

Abstract

This paper presents a fast and effective methodology for optimization injection molding process parameters of short glass fiber reinforced polycarbonate composites. Various injection molding parameters, such as filling time, melt temperature, mold temperature and ram speed were considered. The methodology combines the use of genetic algorithm and Extension Set method processes modeled by multi-layer neural networks and a CAE flow simulation software which was used to simulate the injection molding process and to predict the fiber orientation. This method can replace the traditional ”change-one-parameter-at-a-time” approach which is very inefficient, costly, time consuming and almost impracticable to yield an optimum solution. In the mean while, the fiber orientation were examined by CAE simulation to forecast shear layer thickness, simultaneously to check the accuracy of Extension Set. The results indicated that three distinct layers (frozen layer, shear layer and core layer) are observed from surface to core at various injection molding conditions. The fiber orientation is perpendicular to the melt flow direction in frozen layer and core layer, but it has opposite direction in shear layer. From the CAE analysis, we have got the optimum process parameters to obtain the thickest shear layer that is our target.