Abstract
There are high demands for reliable hydrogen-alcohol phase equilibria in separation and conversion-related industrial processes. Since experimental measurements cannot be directly included in the computer-aided handling of these processes, this study utilizes various computational techniques for estimating hydrogen solubility in seven alcoholic solvents (methanol, ethanol, 1-propanol, 2-propanol, allyl alcohol, 1-butanol, and furfuryl alcohol). Ranking analysis shows that the adaptive-neuro fuzzy inference system having genfis2 (ANFIS2) is the best choice for this purpose. The model predictions are in excellent agreement with the 194 laboratory measurements (RAD = 3.32%, MSE = 6.9 × 10−4, and R2 = 0.998896). Statistical uncertainty analysis confirms that the ANFIS2 model is superior to the previously proposed equations of state and empirical correlations in the literature. Simulation results confirm that 1-butanol and furfuryl alcohol has the highest and lowest hydrogen absorption tendency, respectively. Furthermore, the ANFIS2 justifies that the solubility of hydrogen in all alcohols obeys Henry's law and decreases by decreasing temperature and pressure.
•Hydrogen/alcohol phase equilibria accurately simulated by machine learning methods.•Hybrid neuro-fuzzy is determined as the most precise model for this objective.•The model outperforms empirical correlations and PR, SRK, PC-SAFT equation of state.•The designed model approved that hydrogen-alcohol phase equilibria obey Henry's law.•1-butanol and furfuryl alcohol show the highest and lowest H2 absorption capacity.