Effects of Input Parameter Range on the Accuracy of Artificial Neural Network Prediction for the Injection Molding Process

Abstract

Artificial neural network (ANN) is a representative technique for identifying relationships that contain complex nonlinearities. However, few studies have analyzed the ANN's ability to represent nonlinear or linear relationships between input and output parameters in injection molding. The melt temperature, mold temperature, injection speed, packing pressure, packing time, and cooling time were chosen as input parameters, and the mass, diameter, and height of the injection molded product as output parameters to construct an ANN model and its prediction performance was compared with those of linear regression and second-order polynomial regression. Following the preliminary experiment results, the learning data sets were divided into two groups, i.e., one showed linear relation between the mass of the final product and the range of packing time (linear relation group), and the other showed clear nonlinear relation (nonlinear relation group). The predicted results of ANN were relatively better than those of linear regression and second-order polynomial for both linear and nonlinear relation groups in our specific data sets of the present study.