Predicting the tensile strength of materials applying machine learning for polymeric manufacturing

Abstract

This study compares and evaluates three Machine learning Algorithm in the field of polymeric film manufacturing. Compression moulding and extrusion-blow moulding were considered under the film manufacturing techniques. In this article, we are investigating how Principal Component Analysis (PCA) may be used to minimize the complexity of predictive and classification models such as k-nearest neighbors, artificial neural networks, and random forests. Using PCA, the provided dataset with 8 input parameters for Experiment 1 and 6 input parameters for Experiment 2 would be reduced and applied Machine learning algorithms for predicting the tensile strength of the material. The k-nearest neighbours (kNN) algorithm showed higher predictive abilities, with a coefficient of determination and Root mean square error of 96 percent and 1.09, respectively, for extrusion-blow moulded films. For compression molding, the ANN achieved 91 percent coefficient of determination and 1.11 as root mean square error. These algorithms can predict or categorize quality indices of commodities generated in multi-parameter manufacturing processes, such as the polymer manufacturing processes under study.