Data-Driven Identification of Saturation Regimes for Low-Reflectance Design in Recycled PET-Based Microfibers

Abstract

This study proposes a data-driven framework for searching dyeing recipes to achieve a deep-black shade, using mean reflectance as the target metric. Dye concentrations were used as input variables, and the mean reflectance calculated from the reflectance spectrum was defined as the output variable to build a Gaussian process regression model. A two-stage grid search identified candidate recipes with low-reflectance performance, and experimental validation confirmed that the selected recipe produced both low reflectance and low lightness. In addition, dyeing tests conducted before and after the best recipe showed that mean reflectance, lightness, color space distribution, and color difference from ideal black converged within similar ranges, suggesting that the derived recipe is near the center of the saturation region. These results demonstrate that the proposed data-driven approach can be effectively used for low-reflectance recipe design and saturation–region exploration in deep-black dyeing.