Process variables in mixture experimental design applied to wood plastic composites

Abstract

[EN] The inclusion of process variables in mixture experimental design is crucial for optimizing final products with precision. Unlike standard response surface designs, which are limited by the requirement that proportions must sum to 100%, mixture-process experiments enable a thorough evaluation of how operational factors, such as particle size and mixing time, interact with mixture components. This approach enhances the understanding of how processing conditions affect product properties and leads to more accurate predictive models, thereby improving production consistency and reliability. Regression analysis reveals that interactions between PET and both particle size and mixing time significantly impact the response variable. The model demonstrates strong predictive accuracy, with R-squared and adjusted R-squared values of 92% and 86%, respectively, and a low root mean square error (S) of 0.2818. The PRESS value of 3.38 confirms the model s ability to accurately predict new data. The absence of high multicollinearity, as indicated by variance inflation factor (VIF) values below 5, further supports the model's stability and interpretability. Contour plots illustrate the effect of varying mixture proportions on the response, such as tensile strength, showing a positive impact of both particle size and mixing time. The highest tensile strength is achieved at maximum levels of these variables, indicating a synergistic effect. Response variable optimization identifies the optimal mixture composition as 90% PET, 10% wood, and no coupling agent. To maximize tensile strength, the largest particle size and longest mixing time should be used, though extrapolating beyond the studied parameters should be done with caution.