Calculation and prediction of binary mixture flash point using correlative and predictive local composition models

Abstract Flash point (FP) is a key property and play an important role in hazard classification, safe handling, transportation, and it is used to characterize the fire hazards of liquids. The prediction of this important property of mixtures, leads to design process to avoid the occurrence of fire or explosion. Activity coefficient models are vital to impose the non-ideality of the mixture and accurately predict the flash point of the system. In the present work, correlative local composition activity coefficient models such as Wislon, NRTL, UNIQUAC, and predictive models such as UNIFAC, UNIFAC-LBY, UNIFAC-DMD, and NIST modified UNIFAC (UNIFAC-NIST) together with the general flash point model of Liaw et al. were used to predict the flash point of 15 binary mixtures. The deviations on the flash point using the Ideal, Wilson, NRTL, UNIQUAC, UNIFAC, UNIFAC-LBY, UNIFAC-DMD, and UNIFAC-NIST, are 1.68, 1.72, 1.73, 1.80, 2.04, 2.11, 2.36, and 4.02 K, respectively. Finally, one can conclude that the results of predictive model are superior respect to the correlative models. UNIFAC-LBY model gives the minimum deviation amongst the predictive models, even rather than UNIFAC-NIST model, which in it development, the quality of experimental data critically evaluated. So, UNIFAC-LBY model successfully can be used to predict the unreported FP data, and those materials that are toxic, explosive, and radioactive. Finally, in the absence of phase equilibria data, the local composition parameters regressed through the flash point data.