Local thermal warpage deformation of polypropylene injection molded flat part and neural network prediction model

Abstract

Warpage deformation is a typical phenomenon for polymer injection-molded parts, mainly caused by unbalanced cooling, and it is inevitable. Complex process parameters usually lead to uncontrollable thermal behavior of the polymer materials during injection molding and significant experimental errors. This work presents an experimental mold with a flat mold cavity and nine local heating sections to determine the exact effect of temperature difference on the thermal deformation of injection molded parts. Through local heating at different positions, different warpage deformation was caused. Experimental results demonstrated the relationship between the local temperature and the local thermal warpage. The predicted results of local temperature distribution by numerical simulation presented a strong negative correlation with the experimental results (R2 = 67%); however, the warpage prediction results by numerical simulation were moderate (R2 = 35%). Machine learning with neural networks was further conducted based on the experimental results. When more data was given with a suitable neural network structure, the model prediction accuracy of warpage could be up to 97%, while for the extrapolation test, the prediction accuracy could also be up to 89%. This local thermal heating technique and neural network modeling method can be applied in further theoretical investigation of warpage of injection molded parts and support the development of new models with high accuracy in predicting warpage deformation.