Sustainable Dyeing Process Modeling for Recycled PET/PCT Microfibers via Gaussian Process Regression

Area of Science:

• Textile Chemistry
• Materials Science
• Sustainable Manufacturing

Background:

• The textile industry faces pressure to adopt sustainable and resource-efficient practices.
• Minimizing experimental iterations in dyeing processes is crucial for eco-friendly manufacturing.
• Developing predictive models for dyeing outcomes on novel fiber blends is essential.

Purpose of the Study:

• To develop a Gaussian process regression (GPR) framework for predicting colorimetric outcomes in dyeing.
• To apply the GPR model to ecofriendly fiber blends of recycled polyethylene terephthalate (rPET) and polycyclohexylene dimethylene terephthalate (PCT).
• To assess the model's accuracy and uncertainty quantification capabilities in a low-data scenario.

Main Methods:

• Gaussian Process Regression (GPR) was employed as the predictive modeling technique.
• The model was trained using a limited dataset of 52 experimental points.
• Input variables included dyeing temperature, time, and dye concentration.
• Output predictions focused on CIELAB color coordinates (L*, a*, b*) and the Kubelka-Munk (K/S) value.

Main Results:

• The GPR model demonstrated high prediction accuracy for colorimetric properties.
• Coefficients of determination (R²) reached 0.96 for L*, 0.96 for a*, 0.73 for b*, and 0.95 for K/S.
• The GPR framework provided uncertainty quantification via posterior predictive distributions, including 95% confidence intervals.

Conclusions:

• The GPR-based framework effectively predicts dyeing outcomes for rPET/PCT fiber blends with high accuracy.
• The model's ability to quantify uncertainty is valuable for decision-making in dyeing process design and quality control.
• This data-efficient approach significantly reduces experimental burden, supporting sustainable and resource-efficient textile coloration.