Predicting Dyeing Properties and Light Fastness Rating of Recycled PET by Artificial Neural Network

Abstract

Global car manufacturers have been actively exploring strategies to incorporate eco-friendly fiber materials into automotive interiors, with a particular focus on replacing traditional PET (polyethylene terephthalate) fabrics with recycled PET. Light fastness stands out as one of the most crucial features for fabrics utilized in automobile interiors, especially when exposed to sunlight. This is because the interior temperature of a vehicle subjected to sunlight can escalate rapidly, accelerating dye fading due to degradation caused by UV (ultraviolet) radiation. In this study, we undertook the dyeing of recycled PET by adjusting the concentration of anthraquinone-based dispersion dye, renowned for its exceptional light fastness, alongside varying the concentration of a light fastness enhancer. Subsequently, leveraging the dyeing data obtained from the fabric, we developed an Artificial Neural Network (ANN), a machine learning model. This model facilitates the computation of dyeing properties both before and after light fastness tests, enabling the prediction of color differences and light fastness ratings. The generated model underwent training and optimization, achieving R2 values exceeding 0.98 for all dependent variables. To verify the model's accuracy, computations were conducted on data not included in the dataset. The outcomes indicated that the model could predict dyeing qualities with an average absolute error of approximately 6% compared to the actual values.