(320k) Experimental Validation of the Hybrid Membrane/Psa Principle for Gas Separation

Abstract. An extensive effort has been done on developing theories and industrial practices of gaseous separations by PSA and membrane permeation (Yang, 1897; Ruthven, 1994; Karger, 1992; Drioli, 2001; Baker, 2002; Koros, 2003, 2004). Membrane separations are generally unfavorable when a high-purity product is required and are usually considered to be more suitable for bulk separation. Often, membranes provide a moderately pure product at low cost that may be inexpensively upgraded by a subsequent process (Pan, 1998). This fact has motivated active research on the integration of membranes with other separation processes. On the other hand, gas adsorption processes are well-established separations in the chemical and petrochemical industries. Since the pioneering works of Skarstrom (1958) and Guerin de Montgareuil and Domine (1957, 1964) on PSA process, many schemes have been developed and commercialized in order to increase energy efficiency, improve product purity, and enhance operation flexibility (Tondeur, 1985). Although there are several published works on hybrid membrane/PSA systems (Baker, 1998, 2000, 2001; Sirkar, 1992), truly synergistically concepts have only been proposed by Pan et al. (1998), Esteves and Mota (2002, 2006) and Esteves (2005). An important conclusion drawn from these works is that membrane permeation can be an affective aid in the pressurization and high-pressure adsorption steps of a typical PSA process. The results also indicate the feasibility of incorporating membrane permeation into the blowdown step of the PSA cycle, so that the pressure difference available from the PSA can be used for operating the membrane. Therefore, complete theoretical understanding and commercial integrated applications for gas separations has yet a long field of research to be developed.

This paper describes the experimental validation of novel integrated schemes combining membrane permeation and PSA that were developed for gas separation. By taking advantage of the separation ability of the membrane and the intrinsically dynamic periodic operation of the PSA, the hybrid schemes enhance separation performance when compared to the two stand-alone units. Instead of constant-composition regular feed, the PSA is fed with a mixture progressively enriched in the more adsorbed component during the pressurization (PR) and high-pressure adsorption (HPA) steps of a cycle. This results in sharper concentration fronts. The hybrid schemes were successfully applied to the separation of H2/CH4, CO2/CH4 and CO2/N2 mixtures, on activated carbon. The performance was reported in terms of productivity, product recovery and purity at cyclic steady-state. The effect of different operating parameters on process performance was considered. The numerical simulations are complemented with experimental work on a composite membrane and a lab-scale PSA unit (Esteves, 2005).

The operating pressure of the PSA unit is used as the driving force for permeation, which assists the PR and HPA steps of the PSA process. To fully explore the synergy between both units for a wide range of separation scenarios, two cases are considered (Fig. 1): ? CASE A in which the more permeable component is the least adsorbed, as for example H2/CH4 separation; ? CASE B in which the more permeable component is the more adsorbed; examples of applicability are CO2/N2 and CH4/CO2 separations. Instead of a regular binary feed mixture of A (least adsorbed) and B (more strongly adsorbed), the PSA is fed with a mixture which is progressively enriched in the more adsorbed component during the pressurization and high pressure adsorption steps of a PSA cycle.

Experimental Work. Adsorption equilibrium measurements of N2, CO2 and CH4 were performed on a coal-based, high activity, extruded carbon. A gravimetric method was used with real-time acquisition of temperature, pressure and weight of the adsorbent sample. The Sips isotherm extended to multi-component adsorption is adopted to predict the experimental equilibrium data. Experimental permeation runs were also performed for those gases, in a composite PES/PI (polyethersulfone/polyimide) membrane. Single-component permeances and selectivities were measured and found to be in agreement with those reported by other authors.

A lab-PSA unit was built and experiments were performed in order to validate the modeling work. The real-time acquisition and control of the unit, developed in LabView, allowed the execution of breakthrough and dynamic pressure swing cyclic runs. Several breakthrough experiments followed by complete desorption were done in order to analyze the unit and to study its kinetics. The adsorbate intraparticle diffusion coefficients were obtained by fitting an LDF model to the breakthrough and blowdown runs at different temperatures. The mixtures studied are CO2/He, CO2/N2, N2/He and CH4/He at several flow rates ranging from 0.84 to 5.05 slpm (STP: 0ºC, 1 atm), pressures of 5 and 10 bar, and temperatures of 26, 36 and 46ºC (±1ºC). Seven feed compositions were tested: 0.7, 6, 14, 30, 50, 70 and 80% (v/v). Gas composition was monitored on-line using mass spectrometry. Reproducibility was assessed by repeating some of the experimental runs. A typical PSA cycle of ~12 min. was developed for separation of CO2/N2 mixtures. The periodic cycle involves five steps: 1) PR up to pressure PH (6 bar), 2) HPA at PH, 3) Co-current blowdown (HBD) up to an intermediate pressure PM (3.2 bar), 4) Countercurrent blowdown (LBD) up to the minimum operating pressure PL (0.3 bar), 5) and Purge (LPG) at PL. During HPA and HBD, N2 product is obtained, while during LBD and LPG, CO2 is produced. The single-PSA model validation with the experimental dynamic runs allowed the process performance study, i.e., the product purity and recovery analysis. The temperature, component gas- and adsorbed-phases concentrations, and velocity profiles along the adsorbent bed were studied, as well as the histories of those variables during the PSA cycle and until the periodic steady-state was attained.

In order to confirm the enhanced performance of the hybrid process, this system was modelled both as a conventional PSA process and as an equivalent hybrid unit at exactly the same experimental operating conditions. Except for the integrated boundary conditions imposed to perform the membrane/PSA linkage, all the operating conditions are maintained on both systems. Both models reached the CSS at the 11th cycle. Therefore, the membrane addition did not slow down the convergence to the CSS. The results obtained confirm the integrity of the hybrid concept and the improvement of the separation performance when compared with the stand-alone units. In general, for the stand-alone process and depending on the feed available, the CO2 purities and recoveries are in the range 54-92% and 83-99%, respectively. For N2, the purities and recoveries are 51-99% and 66-69%, respectively. Depending on the membrane area available and the total amount fed to the system, the equivalent hybrid scheme (CASE B) enhances the CO2 purity and recovery up to 16% and 50%, respectively. For the N2, the respective values are 14% and 5%. Obviously, these enhancements have to be balanced with the need to compress the permeate in this case, before feed this stream to the PSA unit.

Finally, in order to validate the global principle developed, the integrated process was experimentally reproduced at cyclic steady-state with a single PSA column. This was the ending piece of the work needed to finalize the research developed and presented over the last years. The on-line gas composition profile at column inlet was imposed by a total mass flow rate profile that satisfies the outlet streams of the supposed coupled membrane module. In addition to on-line monitoring of the outlet gas composition, the experimental setup requires two mass flow controllers, which feed the column at variable mass flow rate with the two gases. This flow imposition simulates the outlet membrane streams: the permeate and the retentate, obtained from the cyclic steady-state operation of the MEMPSA process, and that are simultaneously fed to the coupled column at different steps of the cycle. The two mass flow rates are continuously manipulated so that the composition and mass flow rate of the combined inlet stream are the same as those for the hybrid membrane/PSA model.

Conclusions. The overall experimental and computational work done allows to conclude that, the inclusion of a membrane module into a periodic PSA process, by a truly synergistically scheme, is indeed an innovative method to improve separation performance of several gas systems when compared with the more common stand-alone units. A pre-established PSA process, already in operation, can be eventually coupled with membrane technology, through the alternative systems described, giving rise to enhanced product purities and recoveries. Whether these enhancements are large enough to payoff the energy and investment requirements for a particular separation, must be assessed in conjunction with process development. The proof of concept was successfully demonstrated. The final experimental results obtained validate the integrity of the hybrid concept and the improvement of the separation performance when compared with the conventional stand-alone units.