Support Vector Regression Prediction of Dye Saturation in Recycled PET/PCT Microfibers for Deep Black Shade

Abstract

The textile dyeing industry faces increasing pressure to reduce chemical usage and improve resource efficiency. This study applies Support Vector Regression (SVR) to predict the dye saturation region of recycled PET/PCT fabrics dyed with a single disperse black dye. The objective is to identify the optimal dye concentration range using a machine learning-based approach. K/S values obtained from fabrics dyed at various concentrations were used to train and evaluate SVR models. Model performance was assessed using standard regression metrics such as coefficient of determination (R2), mean absolute error (MAE), and mean squared error (MSE). The SVR model demonstrated high predictive accuracy and effectively captured gradient transitions near the saturation region. From the predicted K/S curves, the point at which further dye addition yields no significant increase in color strength was quantitatively determined. Despite the limited number of training samples, the model successfully learned the nonlinear characteristics of the dyeing response and showed greater robustness to outliers than conventional regression methods. These findings support the potential of SVR as a reliable tool for predicting dye saturation points and optimizing dye use in deep black shade development, contributing to reduced experimental workload and improved resource efficiency.