From prediction to process control: a critical review of machine learning in reactive dyeing of cotton
Abstract
Reactive exhaust dyeing of cotton uses large amounts of water, salt, alkali and heat, partly because the recipe and programme are fixed when a batch starts. This review asks whether learned models, acting on the machine during a batch, have been shown to reduce that consumption. An OpenAlex search (2003 to September 2026, 862 records) and web searching identified 27 primary studies of learned, rule-based, model-based or search-based methods, or inline monitoring, in reactive dyeing of cotton. Each was coded by setting and by how far its output reached the process. Twenty-two stop at offline prediction, three only measure the bath, one advises an operator and one acts automatically; 24 are laboratory studies. No included study documents a learned model automatically changing an outcome-level decision (dye, salt or alkali quantity or timing, temperature programme, or rinse endpoint) during a batch. The one automatic controller, a fixed rule timing alkali injection at pilot scale, is not learned. A check on 135 studies of all fibres agreed. Reported fixation spans 40 to 70%, and no study has measured how much of the shortfall control could recover. Two priorities follow: measuring that share, and a matched-batch trial of learned, rule-based and operator control.
Related
- Bottom-Up Prediction of Amorphous Poly(1,3-dioxolane) using Ab Initio Reactive Machine-Learning Force Fields
- Conventional Ion-Exchange and Chelating Resins for Metal Recovery: A Critical Comparison of Selectivity, Dynamic Performance, Regeneration, and Process Integration
- Dynamic PET-Based Networks: A Critical Review of Reactive Strategies, Processing Windows, and Circular Applications
- Data-driven models for performance prediction in a microalgae-based wastewater treatment process
- Process effect prediction method based on visual generative models
BibTeX
@misc{mainuddin2026from,
title = {{From prediction to process control: a critical review of machine learning in reactive dyeing of cotton}},
author = {Sk. Mainuddin and Tinon Turja Majumder and Md Shajjad Khan Faisal},
howpublished = {ChemRxiv},
year = {2026},
month = sep,
doi = {10.26434/chemrxiv.15009673/v1},
url = {https://chemrxiv.org/doi/full/10.26434/chemrxiv.15009673/v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - From prediction to process control: a critical review of machine learning in reactive dyeing of cotton
AU - Sk. Mainuddin
AU - Tinon Turja Majumder
AU - Md Shajjad Khan Faisal
PY - 2026
DA - 2026/09/29
PB - ChemRxiv
DO - 10.26434/chemrxiv.15009673/v1
UR - https://chemrxiv.org/doi/full/10.26434/chemrxiv.15009673/v1
AB - Reactive exhaust dyeing of cotton uses large amounts of water, salt, alkali and heat, partly because the recipe and programme are fixed when a batch starts. This review asks whether learned models, acting on the machine during a batch, have been shown to reduce that consumption. An OpenAlex search (2003 to September 2026, 862 records) and web searching identified 27 primary studies of learned, rule-based, model-based or search-based methods, or inline monitoring, in reactive dyeing of cotton. Each was coded by setting and by how far its output reached the process. Twenty-two stop at offline prediction, three only measure the bath, one advises an operator and one acts automatically; 24 are laboratory studies. No included study documents a learned model automatically changing an outcome-level decision (dye, salt or alkali quantity or timing, temperature programme, or rinse endpoint) during a batch. The one automatic controller, a fixed rule timing alkali injection at pilot scale, is not learned. A check on 135 studies of all fibres agreed. Reported fixation spans 40 to 70%, and no study has measured how much of the shortfall control could recover. Two priorities follow: measuring that share, and a matched-batch trial of learned, rule-based and operator control.
ER -