Application of Reinforcement Learning to Dyeing Processes for Residual Dye Reduction

Abstract

Sustainability has become a prominent theme in the manufacturing industry, with an emphasis on optimal process configurations that enable environmentally friendly and economically viable operations. Particularly, the textile dyeing and finishing industry has garnered special attention due to its substantial water consumption and consequential wastewater generation. Moreover, dye residues in textile wastewater contain a multitude of chemical substances, posing a serious threat to environmental pollution. Therefore, there is a pressing need for effective decision-making tools to reduce dye residues. In this study, we introduce a reinforcement learning-based model to predict waste discharge in the textile dyeing and finishing industry and recommend dyeing process variables to minimize such waste. Leveraging manufacturing data collected from real production facilities, we constructed a Gradient Boosting model for waste prediction and developed a Q-learning-based process variables recommendation model for dye residue reduction. The recommendation model demonstrated high predictive performance with an R-value of 0.96, and through process configuration recommendations, achieved an average reduction of 66.58% in dye residue. These results have been validated through the collection of on-site information and experiments. This study proposes an innovative approach to effectively predict and reduce residual dyes generated in the dyeing and processing industry. However, a limitation of the developed dyeing process recommendation model is that it was tested on only two out of 124 formulations, making it challenging to generalize the model's performance. More extensive training data is necessary. These facts suggest that, if addressed in future research, improvements can overcome practical constraints and contribute to enhancing the prospects for future decision-making. It is anticipated that such advancements will strengthen the sustainability of the dyeing and processing industry, fostering environmentally friendly operations and contributing to a sustainable future.