Abstract
The objective of the present study is to model the separation of nitrogen from air by PSA using carbon molecular sieves. These processes rely on the different rates at which oxygen and nitrogen gain access to the adsorption sites within the carbon, and not solely on the adsorption equilibria between the gaseous and adsorbed phases. The model of diffusion controlled pressure swing adsorption in this study consists of a set of inhomogeneous, non-linear, first order, partial differential equations (PDEs). These PDEs are derived from convective mass balance equations, mass transfer equations and the equilibrium isotherm equations. The Method of Characteristics is used to reduce the derived set of hyperbolic PDEs to a set of ordinary differential equations (ODEs) in an exact manner. The adsorption rate and equilibrium isotherm parameters for oxygen and nitrogen are estimated from data published by Hassan, Ruthven and Raghavan (1986). This study also compare their published theoretical and experimental results against the DCPSA model's result. Such comparisons, however are of limited value because the exact conditions and the manner in which the experiment is conducted are not known. The original idea of performing implicit integration along the characteristics has made the DCPSA model more robust. This means it is more stable and less computer processing time is required when high but realistic values of the overall mass transfer coefficients are used. The DCPSA model is able to predict the qualitative behaviour of the simple, product release, purge, backfill and bed pressure equalisation cycles. A series of sensitivity analysis of the process parameters on a simple purge cycle are carried out. The model predicts an extraordinary 'reverse flow phenomenon' during the product release step when low feed rate is introduced at the feed end. This is because when the feed rate is low, the rate of adsorption at the product end is much greater than the gas feed velocity, and additional gas is required to be drawn in from the product end. No published models have reported this phenomenon. Finally, it is also the intention of the author to develop the DCPSA model such that the model can also be used in future to simulate any other kinetically based, binary adsorbable gas separating process.