No abstract

No abstract

A new procedure for the estimation of parameters in two-component vapor-liquid equilibrium is presented. It allows for measurement errors in all variables and gives better fits than most other procedures because of the use of two relations between the variables rather than one as is commonly used. An approximate and an exact solution are described and the importance of using all thermodynamic and statistical information is illustrated for the Wilson and UNIQUAC models. New alternatives for analysis of residuals are also discussed. SCOPEThe importance of phase equilibrium prediction in chemical process design has promoted the development of various semimechanistic models. An important aspect in the successful application of these models is the optimal estimation of their unknown parameters from limited experimental information. In this process of estimating parameters, it is important to use all the information that is available.One source of information is the thermodynamics of the system. This supplies the equations relating the several variables. In the past, not all equations have been used; this work illustrates the improvement that can be achieved by using simultaneously all the equations involved.Another problem with most previous attempts has been the use of estimation criteria which are basically arbitrary. Some of these assume some of the variables to be perfectly known. Our solution is a statistical one, based on the error-in-variables model (EVM), which gives optimal parameter estimation and allows for the presence of experimental error in all measured variables, giving general and flexible solutions, easy to apply.Once the parameter estimates have been found, it is important to have an idea of how much uncertainty is involved in them, and how well the model represents the system. We present