Answers are grounded in this paper and cite their sources.
This study presents a comparative investigation of advanced modeling techniques for predicting the solubility behavior of triazine (C₃H₃N₃) reactive disperse dyes in supercritical CO₂ (ScCO₂). Accurate prediction of solubility behavior in such complex systems is of great importance for optimizing dyeing processes and advancing environmentally friendly applications of supercritical fluids. Two approaches were examined: Response Surface Methodology (RSM), a statistical tool widely applied in process optimization, and Multi-Layer Perceptron Artificial Neural Networks (MLP-ANN). The dataset used in this work was adapted from Huang et al. (2026), covering a wide operating range of temperatures (343.15–383.15 K) and pressures (11–23 MPa). The results demonstrate that while both RSM and MLP provide accurate predictions, the MLP model consistently outperforms RSM across all dyes investigated. RSM achieved strong correlations with R² values between 0.9773 and 0.9960 and MSE values in the range of 0.0385–0.0575. In contrast, the MLP model exhibited superior predictive power, yielding higher R² values (0.9910–0.9976) and significantly lower Mean Squared Error (MSE) values, with dye D showing the best performance. Moreover, a comparison with Huang’s cubic equation of state models (SRK and PR with vdW2 mixing rule) highlights the substantial accuracy gains achieved by the MLP approach, as the machine learning model reduced Average Absolute Relative Deviation (AARD%) from 6.0–8.2 (EOS) to as low as 0.9159.
No takes yet. Share an insight, caveat, or question.
Zohreh Khoshraftar (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: