PVT DATA-ANALYSIS USING NEURAL-NETWORK MODELS

PVT data analysis is performed for pure gases and vapors using neural network based models. The resulting equations of state (EOS) are expli cit forms of the compressibility factor as a function of the reduced t emperature and pressure, and the acentric factor. In order to represen t the whole domain of pressure and temperature, two EOS have been deve loped, one equation specifically covering the critical region. These t wo equations were obtained by fitting a large number of experimental d ata points (about 1000 and 1500, respectively) characterizing the beha vior of several pure components (5 and 8, respectively). The EOS have been applied successfully to various other components and were found t o give accurate predictions for a reduced pressure as high as 10. Deri vation of the fugacity coefficient is also presented.