Loading...

Prediction of the selectivity coefficient of ionic liquids in liquid-liquid equilibrium systems using artificial neural network and excess Gibbs free energy models

Dehnavi, S. M ; Sharif University of Technology | 2010

1864 Viewed
  1. Type of Document: Article
  2. DOI: 10.1080/02726351.2010.496294
  3. Publisher: 2010
  4. Abstract:
  5. In this work, the selectivity coefficients of ionic liquids in liquid-liquid systems were correlated and predicted by the NRTL, UNIQUAC, and Wilson-NRF Gibbs free energy models and also by an artificial neural network system. The three thermodynamic models need six binary interaction parameters between solvent(1)-solvent(2), solvent(1)-ionic liquid, and solvent(2)-ionic liquid pairs in obtaining the selectivity of ionic liquid in liquid-liquid systems. Also, the selectivity coefficients of ionic liquids were modeled using an artificial neural network system. In the proposed neural network system, temperature, molecular weight of ionic liquid, molecular weight of solvents, and mole fractions of components (1) and (2) in the solvent-rich phase were considered as input data and the selectivity of ionic liquids in the liquid-liquid system was considered as output. The weights and biases were obtained using the quick propagation (QP) method. A total of 150 experimental data points were used for training and 30 experimental data for testing. The best network topology obtained was one neuron in a hidden layer with five nodes. The average absolute deviation percentages (AAD%) for the results obtained from the neural network system, the NRTL model, the UNIQUAC model, and the Wilson-NRF model are 0.0096, 0.0168, 0.0017, and 0.0178, respectively. Although the artificial neural network gave reasonably good results, the UNIQUAC model can more accurately predict the selectivity of ionic liquids in liquid-liquid equilibrium systems than the other models
  6. Keywords:
  7. Gibbs free energy model ; liquid-liquid system ; selectivity ; Average absolute deviation percentages ; Binary interaction parameter ; Excess Gibbs free energy ; Experimental data ; Hidden layers ; Input datas ; Liquid liquid equilibrium ; Liquid-liquid systems ; Mole fraction ; Network topology ; Neural network systems ; NRF model ; NRTL model ; Rich phase ; Selectivity coefficient ; Thermodynamic model ; UNIQUAC ; UNIQUAC models ; Data flow analysis ; Electric network topology ; Free energy ; Gibbs free energy ; Ionization of liquids ; Ions ; Molecular weight ; Neural networks ; Solvents ; Ionic liquids ; Ionic liquid ; Solvent ; Artificial neural network ; Data analysis ; Energy ; Equilibrium constant ; Liquid ; Priority journal ; Separation technique ; Temperature ; Thermodynamics
  8. Source: Particulate Science and Technology ; Volume 28, Issue 4 , 2010 , Pages 379-391 ; 02726351 (ISSN)
  9. URL: http://www.tandfonline.com/doi/abs/10.1080/02726351.2010.496294