The Mixture Critical Nature Of The Mathematical Model And Calculation

The study on the critical properties of mixtures has the significance of theory and practicability. The usual methods for calculation and prediction of the critical properties of mixtures are complicated or empirical, so these methods don't satisfy the demands of engineering calculation very well. In order to overcome the disadvantages of classical methods, in this paper, three improved calculation and prediction models of critical properties of mixtures are prompted.First, the repulsive force item of the original PR equation of state is modified with Carnahan-Starling hard sphere equation, and the modified CS-PR equation of state is introduced. The modified CS-PR equation combined with the rigorous critical state criterion enunciated by Gibbs is applied to calculate the critical properties of binary mixtures, and the same is done with PR equation as well. The calculation results indicate CS-PR equation can calculate the critical properties of such systems better than the original PR equation, so CS-PR equation is a more desirable modification of PR equation in the field of critical properties calculation.For the sake of more calculation accuracy and less complexity, the improved BP network with heuristic learning rules, BFGS algorithm and Levenberg-Marquardt algorithm individually is applied to predict the critical properties of binary mixtures. A group of critical properties of certain mixture as train examples are used to train the BP network, then the trained network is applied to predict the critical properties of other component points of this mixture. The calculation and prediction results of six types of binary mixtures according to van Konynenberg-Scott phase diagram classification indicate this method is more convenient, versatile and has higher calculation accuracy than other classical methods.Two methods above are only adapt to the mixtures whose several groups of critical properties are known. In order to release from the limitation, the prediction model of critical properties based on the parameters charactering the moleculars interaction which determines the thermodynamic properties of the mixture is proposed. The critical property of certain binary mixture is looked as a function of critical properties, mole fraction, molecular weight, acentric factor and susceptibility of each pure component of the mixture, and BP network is applied to build the corresponding relation of the input parameters and the critical properties. The BP network is trained with a group of input parameters and expected output as train examples, and the critical properties of other mixtures are predicted with the trained BP network, and more desirable prediction results are gained.