Analysis of intelligent models in prediction nitrous oxide (N2O) solubility in ionic liquids (ILs)

Abstract

Accurate prediction of nitrous oxide (N2O) solubility in ionic liquids (IL) gives an in-depth insight into the effectiveness of these fluids as an absorbent for the removal and control of this deleterious gaseous agent in the atmosphere. The precise estimation of this key parameter is necessary for the design of a prospective IL-based separation processes at large scales. To this end, three intelligence methods including artificial neural network (ANN), support vector machine (SVM) and least square support vector machine (LSSVM) have been applied to forecast the N2O solubility in 25 ILs. The shuffled complex evolution (SCE) was employed to acquire the optimal magnitudes of hyper parameters (σ2 and γ) which are embedded parts of SVM and LSSVM models, and the trial and error was employed to obtain the optimal numbers of neutron and layers for ANN intelligent model. Gathering and using 627 solubility data, the comparison between the capability of applied intelligent models in giving the solubility has also been made between the aforementioned intelligent models so that the most effective predictive one can be presented for the use in chemical engineering software packages in conjunction with existing PVT thermodynamic models for the proper design of absorption process of this greenhouse gas (N2O) in ILs. The findings are indicative of a good agreement between the estimation from intelligent models and the experimental data. Comparison between these three investigated models reveals that the performance of SVM in prediction of solubility is somewhat better than other intelligent models (i.e., ANN and LSSVM) in prognosticating the N2O solubility in those 25 studied ILs as the coefficient determination (R2) and root mean squared error (RMSE) are respectively 0.9970 and 0.0104 for test sets of data. This is likely due to the existence of structural risk minimization principle of SVM which is embodied in SVM algorithm and effectively minimizes upper bound of the generalization error, rather than minimizing the training error.