Numeric Model For Saturated Curve Caculation Of Carbon Dioxide

The thesis widely investigates the research approachs of the saturated curve calculation of carbon dioxide. Focusing on the disadvantages of conventional approaches, methods based on numeric simulaton and ANN (artifial neural network) is propsed.A comprehensive CO2 saturated curve database is built based on the collected experimental data. Reliability of collected data is reviewed; data is combined based on the reliabilty. Saturation border curve is fitted using polynomial interpolation methods of Lagerange, Hermite, segmental cubic spline. Lagrange interpolation is the simplest, but has oscillation at higher order to guarantee the smoothness. Hermite and spline methods don't meet that difficulty. Continuous 1st and 2nd order derivatives provide better smoothness. Cubic spline may reach over-smooth, which degrades the flatness of the entire data set.ANN based fitting is proposed to improve the flexibility of saturated curve calculation. It makes possible to calculate more than one thermodynamic parameter at same time with selected single basic state paratemters, which is more helpful in the 3-dimensional saturation curve fitting in the space of p-v-T. Principles of saturated curve calculation software design are proposed based on query computation cost, model expansion, model transplant, model storage space, and computation precision.