APPLICATION OF NEURAL NETWORKS IN PREDICTING TEMPERATURE DISTRIBUTION IN MOLD CAVITIES

Abstract

This study investigates the application of artifcial neural networks (ANN) in predicting temperature distribution in injection mold cavities during the pre-heating stage. A database of 25 experimental samples was established using a Taguchi L25 orthogonal design, considering four input factors: heating water temperature, water flow rate, cooling channel thickness, and blocking plate thickness. The ANN model was developed in MATLAB with a multilayer perceptron architecture, trained on pre-processed experimental data. The results demonstrate that the ANN accurately predicts mold temperature evolution, with prediction errors as low as 2.06% and not exceeding 4.75% across test cases. The fndings confrm that ANN-based models can capture nonlinear heat transfer behavior in mold heating and provide a reliable tool for process optimization, reducing experimental effort and improving mold design effciency.