Analysis and prediction of the mass of injection molded parts using factorial models and random forests

Authors:

Abstract
This work presents a combined approach using factorial design of experiments and machine learning models to analyze and predict the mass of parts manufactured by injection molding. A 2k factorial design was employed to evaluate the effect of four process factors—holding pressure, injection pressure, injection time, and injection speed—on the mass of molded parts. A factorial regression model was constructed and assessed using analysis of variance, showing a high coefficient of determination. Subsequently, a regression model based on random forests was trained, achieving remarkable predictive performance. Both approaches were compared using mean squared error and root mean squared error metrics. The results show that both models are highly accurate. This hybrid approach provides a solid foundation for optimizing the injection molding process in industrial applications.