Estimation of vapour liquid equilibria for the system carbon dioxide–difluoromethane using artificial neural networks Swati Mohanty * Regional Research Laboratory (C.S.I.R.), Bhubaneswar 751013, India Received 11 March 2004; received in revised form 11 May 2005; accepted 25 May 2005 Available online 19 August 2005 Abstract In this paper, an alternate tool, i.e. the artificial neural network technique has been applied for estimation of vapour liquid equilibria (VLE) for the binary system, carbon dioxide–difluoromethane, which is an attractive alternative to chlorofluorocarbons and hydrochlorofluorocarbons, normally used as refrigerants. The model can satisfactorily estimate the vapour liquid equilibrium pressure and mole fraction carbon dioxide in vapour phase in the temperature range 222.04–343.23 K and in the pressure range 0.105–7.46 MPa. The average absolute error for the system in the estimation of vapour phase mole fraction is 0.0086 and 0.056 MPa for the pressure. q 2005 Elsevier Ltd and IIR. All rights reserved. Keywords: Research; Equilibrium; Liquid-vapour; Binary mixture; Carbon dioxide; Modelling; Neuronal network Evaluation des e ´quilibres vapeur-liquid d’un syste `me au dioxyde de carbone-difluorome ´thane a ` l’aide de re ´seaux neuronaux artificiels Mots cle ´s : Recherche ; E ´ quilibre ; Liquide-vapeur ; Me ´lange binaire ; Dioxyde de carbone ; Mode ´lisation ; Re ´seau neuronal 1. Introduction Vapour–liquid equilibria (VLE) play a vital role in designing and modelling of process equipments. Very often equations of state (EoS) are used for estimating the VLE. Although EoS are derived based on strong physical principles, there is still certain amount of empiricism involved in terms of several adjustable parameters that are required in mixing rules. Using EoS for estimating the VLE is tedious and requires an iterative method that may sometimes pose problem for real time control of an operating plant. In such cases other faster alternative methods would be more attractive. The development of numerical tools, such as artificial neural network (ANN), has paved the way for alternative methods to estimate the VLE [1–5]. It has attracted considerable interest because of its ability to capture with relative ease the non-linear relationship between the independent and dependent variables. In addition, once the network has been trained, calculation of VLE can be done in a single step. From its initiation in the early forties till today there are hundreds of ANN architecture developed, however, there are a few which are more popular and find wide applications. Details have been dealt with elsewhere [6,7]. Several authors have reported application of ANN for estimation of thermodyn- amic properties such as estimation of viscosity, density, International Journal of Refrigeration 29 (2006) 243–249 www.elsevier.com/locate/ijrefrig 0140-7007/$35.00 q 2005 Elsevier Ltd and IIR. All rights reserved. doi:10.1016/j.ijrefrig.2005.05.007 * Tel.: C91 674 2581635x535; fax: C91 674 2581637. E-mail address: swati_mohanty@yahoo.com.