Machine-Learning-Based Prediction Of Tensile Strength In Injection-Molded Thermoplastics Using Multi-Parameter Process Data

This study analyses the influence of key injection-molding parameters on the tensile strength of molded thermoplastic parts using supervised machine-learning models. The dataset includes melt temperature, mold temperature, injection speed, injection pressure, holding pressure, cooling time, and other processing variables collected across 500 molding trials. Histogram and correlation analysis showed strong relationships between strength and pressure-temperature conditions, with holding pressure, melt temperature, mold temperature, and injection speed acting as the most influential factors. Four regression models were evaluated: Linear Regression, Decision Tree, Random Forest, and Gradient Boosting. Linear Regression achieved the highest accuracy with a test R² of 0.9507, RMSE of 3.82 MPa, and MAE of 3.14 MPa, indicating that tensile strength follows a predominantly linear pattern across the operating range. Gradient Boosting and Random Forest also delivered high accuracy with test R² values of 0.9348 and 0.9170, respectively. Residual analysis confirmed stable error behaviour with residuals centred around zero and following a near-normal distribution. The results show that controlled variations in melt temperature, mold temperature, injection speed, and holding pressure can predict tensile strength with high reliability, demonstrating the potential of machine-learning-based modelling for improving product quality in injection molding.