Optimal Design of Plasticizing Screw Using Artificial Intelligent Approach

Abstract

This study integrated plasticizing screw analysis software with neural-network in the design of a screw for injection molding application. The qualities of the plasticizing screw selected in this study are output rate, melt temperature variation at the end of metering zone, the specific mechanical energy (SME), and the melting distance. The Taguchi orthogonal array is implemented to carry out the experiment and to obtain the test data for training the neural network. The Back-propagation Neural Network (BPNN) was then used for screw quality predictor, and optimal design was solved with Genetic Algorithm (GA). The optimal screw design for a diameter of 25 mm screw for molding PC resin in this study is 5.37D in solid conveying zone, 9D in compression zone, metering zone depth of 2.44 mm, and flight width of 3 mm. The performance of this screw with the preset processing condition can have the temperature difference at the end of metering (ΔT) of 5.67°C, the output rate Q of 20.12 kg/h, the SME of 520.80 (kJ/kg), and the plastics completely melted at 17.39D.