Genetic algorithm to estimate interaction parameters of multicomponent systems for liquid-liquid equilibria

Genetic algorithm (GA) has been utilized to estimate the binary interaction parameters from multicomponent liquid-liquid equilibria data; and its applicability for NRTL and UNIQUAC thermodynamic models has been demonstrated. These models being highly non-convex can have several local extrema. GA leads to globally optimum values; it does not require any initial guess but only the upper and lower bounds of the interaction parameters. It has also been shown to perform better than inside variance estimation method (IVEM) and the techniques used in ASPEN, DECHEMA. The objective function and the root-mean-square deviation for these four techniques have been compared for methanol-diphenylamine-cyclohexane system using UNIQUAC. The applicability of GA to six ternary, two quaternary and two quinary systems has also been undertaken. Invariably the rmsd values for GA are better than reported in the literature.