Boiling point elevation estimation of electrolyte solutions by means of two methods: thermodynamic model and artificial neural network
Volume Title: 1
Chemical Engineering Dept., Faculty of Engineering, University of Guilan, Rasht, Iran
Boiling point elevation estimations of KBr, NaCl, NaOH, LiCl, KI, NaBr, Sr(NO3)2, KCl, RbCl, BaCl2 electrolyte solutions have been studied in this paper by two thermodynamic model and artificial neural network methods. In thermodynamic model, the activity of water is required and the semi-empirical Pitzer equation has been selected. Four solutions, KBr, NaCl, NaOH, LiCl, have indicated lesser deviations and errors from experimental BPEs. Since their experimental molalities to measure BPEs are in the range of Pitzer equation molalities, the AAD% and MSE are low. Another method is about application of artificial neural network. Since the experimental BPEs of KBr and NaCl in different boiling points of pure water and various molalities are published in another literature, by trusting to the thermodynamic model, BPEs of NaOH and LiCl solutions have been computed in different pure water boiling points (60, 70, 80, 90, 100) and different molalities in the range of Pitzer equation. The back-propagation artificial neural network has been trained by input ( molalities, Boiling points of pure water, molecular weights) and output (BPEs) information. Levenberg-Marquardt algorithm has been selected and epochs and goal criteria are respectively equal to 104 and 10-7 with three different structures. Each structure has different neurons in hidden layer (15, 18 and 20). Some tested data have been examined. A good agreement has been seen between tested data and expected ones since ranges of deviations and errors are respectively between 1.48% and 3.06% and about 10-5 in each three structure.