Statistical modeling of tensile properties of talc-filled polypropylene based on multivariate regression and neural network analyses

In this paper, tensile properties of homo polypropylene (PP) with respect to talc filler content were predicted using regression model and neural network model. Talc content, tensile speed, Differential Scanning Calorimeter (DSC), Gel Permeation Chromatography (GPC) and rheometer data were used as modeling input factors. 2 different multiple regression models and 1 neural network model were established and the models were compared quantitatively by average error rate (AER). The results showed high reliability for all models but neural network models were determined as the most meaningful model.