An experimental and theoretical study is performed for bulk separation of gas mixture by pressure swing adsorption, a process widely used commercially for gas purification. By cycling the pressure of a bed of activated carbon between 3 and up to 500 psig (0.021 to 3.445 MPa), at the ambient temperature, a 50/50 H2/CH4 mixture is separated into two products with well over 90% purity and recovery at high throughputs. All process characteristics can be predicted by a pore‐diffusion model. A fundamental understanding of the function of each step in the cyclic process is given.

No abstract

A theoretical and experimental study is performed for the bulk separation of a ternary mixture by pressure swing adsorption. Three concentrated products can be obtained by cycling the pressure in the adsorber. Three models are formulated for the cyclic process: equilibrium, Knudsen diffusion, and Knudsen plus surface diffusion. The latter model provides the best results when compared with the experimental data, due to the important contribution of surface flux to the total flux in the sorbent pores. SCOPEAn increasing number of commercial applications are being discovered for pressure swing adsorption (PSA) because of its low energy requirements and costs. The applications are, however, somewhat limited to gas purification processes, with the only bulk separation being for 0, or N, from air. Multicomponent bulk separation by PSA using one sorbent and a single PSA unit, despite its promising commercial value, has not appeared in the literature. In this study, an H,/CH,/ CO, mixture (one-third each by volume) is separated by PSA using activated carbon and involving five basic cyclic steps: I repressurization, I 1 adsorption, 111 cocurrent depressurization, IV countercurrent blowdown, and V purge. Following the order of increasing adsorption strength, H, is produced in step II and in the early stage of step 111. The later cut in step 111 yields CH,, and steps IV and V produce CO, . A basic understanding of the cyclic process is obtained by modeling. Three models are formulated and compared with experimental results: equilibrium model, Knudsen diffusion model, and Knudsen plus surface diffusion model. CONCLUSIONS AND SIGNIFICANCEBulk separation of the ternary mixture H,/CH,/CO, is accomplished by PSA using a single sorbent. Highpurity products of H, and CH, are obtained, whereas the product purity for CO, only reaches 60% due to the low selectivity between CO, I CH, on activated carbon. Of the three models for the cyclic process, the Knudsen plus surface diffusion model provides the best results when compared to the experimental data. Due to the high surface coverage, surface diffusion significantly contributes to the total flux in the sorbent pores, generally over 50% in the PSA process. Accounting for the strong dependence of surface diffusivity on surface coverage, the effects of purge/feed ratio, pressure ratio, feed rate, and end pressure of depressurization on the separation are predictable by the model. A basic understanding of the bed dynamics also is presented.

The Dubinin-Astakhov equation, which is based on the micropore volume filling theory, is extended to mixed-gas adsorption by using the concept of maximum available pore volume. The proposed model requires only single-gas data for predicting mixed-gas adsorption; no additional equation, such as the Lewis relationship, is needed. The proposed model is not iterative, and the calculation involved is extremely simple. The model has been tested against experimental data for 16 binary mixture systems involving a wide variety of adsorbents and a wide range of pressures, with satisfactory results. It also compares favorably or equally satisfactorily with the ideal adsorbed solution theory, the Grant-Manes model, and an extension of the DA equation proposed by Bering et al. However, all three of these methods require iterative (numerical) computation.

The published theoretical models for pressure swing adsorption (PSA) are of either the equilibrium type, i.e., instantaneous equilibrium is assumed between the gas and adsorbed phases, or the diffusion type considering only a monodisperse pore structure (Yang and Doong, 1985; Doong and Yang, 1987). There is reason for doubt that either type of model is applicable to adsorption processes using zeolite sorbent, which has a bidisperse pore structure. Commercial zeolite sorbents contain crystals of the size 1-9 microns that are pelletized with a binder. Sorption is entirely within the crystals, which contain micropores, whereas the binder contains macropores with a negligible sorption capacity. This paper presents a general PSA model for zeolite sorbents. Both micropore and macropore diffusion are considered. The mathematical complexity of the pore diffusion equations for the two types of pores is reduced by assuming parabolic concentration profiles in both crystals and pellets. Thus the two partial differential equations are converted into ordinary differential equations containing only time derivatives, and the burden of integration along the radial distance is completely eliminated. The model is general enough to be applied to bulk, multicomponent separations using any PSA cycle. The specific separation discussed in this work is the bulk separation of a hydrogenmethane mixture using 5A zeolite. The boundary conditions in the model depend on the PSA cycle. A wide variety of PSA cycles has been commercialized (Yang, 1987). Under consideration here is the most widely used five-step cycle in which each adsorber undergoes the following:I. Repressurization with the light product 11. High-pressure feed 111. Cocurrent depressurization IV. Countercurrent blowdown V. Low-pressure purge. The model is formulated for an n-component mixture. TheIdeal gas behavior 0 Negligible pressure drop across the bed assumptions made in the model are:Thermal equilibrium between the gas flow and the solid sorNo radial temperature and concentration gradients in bed Spherical pellets and crystals The dimensionless mass balance equations for the gas flow in the bed are, respectively, for the individual component and for the mixture (Doong, 1986): If molecular diffusion is assumed to be dominant in the binder phase of zeolite, the mass balance equations for the macropores will be similar to those for the monodisperse pore structure in Yang and Doong (1985). The resulting dimensionless equations are: ( 1 Ob) Since the crystals account for all of the adsorbed amount, the P b 910 6 I 621 --p c * &---first overbar on Q denotes the volume-averaged amount adsorbed over the crystal, and the second overbar denotes that averaged over the pellet. The mole fraction on the surface of the pellet Yip is the same as that in the bulk flow, V,, since the film resistance is negligible. is required:The local adsorption rate, s, through which the bulk flowThe heat capacity of the wall is not negligible in the experimental unit. Therefore, an energy balance equation...