Applying artificial neural network for binary vapor liquid equilibrium prediction

Most solvents used in the semiconductor industry are toxic and costly. Thus, the component of these solvents should be recovered for re-use in these processes by distillation methods. Vapor liquid equilibrium (VLE) data are basic information for design and operation of distillation columns. VLE data can be estimated using several thermodynamic models (Wilson and Tan-Wilson) based on calculation activity coefficients. For ideal systems, thermodynamics can be applied easily. However, it is difficult to apply for non-ideal systems, especially for azeotropic systems. In this work, artificial neural networks (ANNs), were applied to predict and estimate VLE data for binary systems without and with salts. The databases were collected from some authors (Tan et al., 1988; Iliuta et al., 1996; Pham, 2005; Munoz et al., 2005). The results obtained from ANNs prediction were compared with published and theoretical results. The predicted data showed good agreement with data