DESIGN METHOD FOR LAYERED BED ADSORBER FOR SEPARATION OF MOLECULAR SIEVE AND ACTIVATED CARBON. A Thesis

Transcription

1 DESIGN METHOD FOR LAYERED BED ADSORBER FOR SEPARATION OF CO 2 AND N 2 FROM NATURAL GAS USING ZEOLITE13X, CARBON MOLECULAR SIEVE AND ACTIVATED CARBON A Thesis Submitted to the Faculty of Graduate Studies and Research In Partial Fulfillment of the Requirements for the Degree of Master of Applied Science in Process Systems Engineering University of Regina By Mohammad Rokanuzzaman Regina, Saskatchewan February, 2015 Copyright 2015: Mohammad Rokanuzzaman

2 UNIVERSITY OF REGINA FACULTY OF GRADUATE STUDIES AND RESEARCH SUPERVISORY AND EXAMINING COMMITTEE Mohammad Rokanuzzaman, candidate for the degree of Master of Applied Science in Process Systems Engineering, has presented a thesis titled, Design Method for Layered Bed Adsorber for Separation of CO 2 and N 2 from Natural Gas Using ZEOLITE13X, Carbon Molecular Sieve and Activated Carbon, in an oral examination held on December 16, The following committee members have found the thesis acceptable in form and content, and that the candidate demonstrated satisfactory knowledge of the subject material. External Examiner: Supervisor: Committee Member: Committee Member: Dr. Daoyong Yang, Petroleum Systems Engineering Dr. Amornvadee Veawab, Process Systems Engineering Dr. Stephanie Young, Process Systems Engineering Dr. Adisorn Aroonwilas, Industrial Systems Engineering Chair of Defense: Dr. Doug Durst, Faculty of Social Work

3 ABSTRACT Natural gas (NG) is a low-carbon fossil fuel that carries impurities such as carbon dioxide (CO 2 ) and nitrogen (N 2 ). These two impurities reduce the heating value of NG. Also, CO 2 causes corrosion in the pipeline and N 2 produces nitrogen oxide (NO x ) when combusted. These facts have forced NG transmission and distribution companies to limit the concentrations (mole percent) of CO 2 ( 3%) and N 2 ( 4%). Consequently, selective separation of CO 2 and N 2 from NG has gained considerable importance. There are many technologies that are in use for separation of these two constituents. Most of them are suitable for single component separation: either CO 2 or N 2. In the context of multicomponent separation common in industries, adsorption is an emerging technology that offers low-cost and energy-efficient separation for small- to medium-sized industries. The technology lacks commercial availability due to its dependency of design methodologies on experimentation, simulation or both. This work focuses on easy-to-use design methodology for the design of a double bed adsorber. This easy-to-use methodology is tailored for separation of CO 2 and N 2 from NG using zeolite13x and a carbon molecular sieve (CMS3K) or activated carbon (ACB). These adsorbents are commercially available, and they offer easy and energy efficient regeneration for repeated uses. The product will meet the specified concentration limit for NG transmission and distribution systems. To achieve this goal, two layered bed adsorbers, zeolite13x-cms3k and zeolite13x-acb, were simulated using Aspen Adsorption. Simulation requires a trustworthy mathematical framework i.e. model. Therefore, a model was developed in i

4 Aspen adsorption by selecting relevant equations and submodels. Inputs for the model were collected from literature, calculated using various equations, and obtained by fitting experimental data. A numerical solution method was specified and, finally, the model was validated against experimental measurements. A parametric study was performed for a wide range of operating conditions. Data generated through parametric study were correlated. The correlations, the first of this kind, can be used to predict required amounts of adsorbents for 100% CO 2 separation and 50 to 90% N 2 separation. Finally, a procedure was outlined to transform the amount of adsorbent into the physical dimensions of an adsorber. ii

5 ACKNOWLEDGEMENTS I would like to express my sincere gratitude to my supervisor, Dr. Amornvadee Veawab, for giving me the opportunity to carry out this interesting research under her enthusiastic supervision. Her enormous financial and technical support, valuable guidance, and encouragement were a great source of inspiration and the driving force throughout the entire course of this research. I would also like to express my gratitude to Dr. Adisorn Aroonwilas for his great support, guidance, and encouragement. I would like to thank the Natural Sciences and Engineering Research Council of Canada (NSERC), SaskEnergy, and the Faculty of Graduate Studies and Research (FGSR) for their financial support. I would also like to thank the Faculty of Engineering and Applied Science at the University of Regina for their help and support. Finally, I am sincerely thankful and grateful to my parents, family, and friends for their unconditional love, prayers, and support to fulfill my dreams. iii

6 Table of contents Abstract Acknowledgement Table of contents List of Tables List of Figures Nomenclature i iii iv vii ix xii 1. Introduction Natural gas Industrial separation processes for removal of N 2 and CO 2 from natural gas Adsorption process and adsorbents for CO 2 and N 2 removal Modeling and simulation Research motivation, objectives and scope of work 9 2. Literature review Scope of review Adsorption fundamentals Adsorbents Zeolite13X Carbon adsorbents Adsorption modeling Multicomponent separation Numerical Solution of partial differential equations Modeling and simulations of a gas adsorber 25 iv

7 3.1 Adsorption modeling Model equations Solution of model equations Calculation procedure Model validation Nitrogen separation using activated carbon Methane separation from hydrogen using Zeolite 5A Carbon dioxide separation using Zeolite 13X Results and Discussion Description of simulated gas adsorption systems Simulation results for zeolite13x Parametric study Correlation to determine amount of zeolite13x Simulation results for zeolite13x-cms3k system Parametric Study Correlations based on simulated results Simulation results for zeolite13x-acb system Parametric study Correlations based on simulated results Determination of column dimensions using correlations Conclusions and recommendation for future work Conclusions Recommendation for future work 94 v

8 References 95 Appendix A: Adjustments of transport parameters 106 vi

9 List of Tables Table 1.1 Table 1.2 Table 1.3 Quantity of air pollutants produced from fossil fuel combustion in lbs/billion Btu (U.S. Energy Information Administration (EIA), 1999) Composition of natural gas observed in different reservoirs as mole percentage (Kidnay and Parish, 2006) Specification of pipeline natural gas (modified from Kidnay and Parish, 2006) Table 3.1 Model equations 26 Table 3.2 Model input (N 2 -ACB system) 40 Table 3.3 Model inputs (CH 4 -H 2 -zeolite5a system) 43 Table 3.4 Model inputs (CH 4 -CO 2 -N 2 -zeolite13x system) 46 Table 3.5 Isotherm parameters (CH 4 -CO 2 -N 2 -zeolite13x system) 47 Table 4.1 Physical properties of zeolite13x and properties of column (Cavenati et al., 2006) 53 Table 4.2 Parameters used in simulation for zeolite13x 54 Table 4.3 Table 4.4 Parameters of correlation for determination of amounts of zeolite13x Physical properties of double bed adsorber (zeolite13x-cms3k) (Cavenati et al., 2006) Table 4.5 Parameters used in simulation of zeolite13x-cms3k system 66 Table 4.6 Parameters for correlations vii

10 Table 4.7 Physical properties of double bed adsorber (zeolite13x-acb) (Cavenati et al., 2006 and Shen et al., 2010) 78 Table 4.8 Parameters used in simulation of zeolite13x-acb system 79 Table 4.9 Parameters for correlation viii

11 List of Figures Figure 3.1 Calculation procedure 37 Figure 3.2 Figure 3.3 Figure 3.4 Breakthrough concentration profiles of N 2 in pitch-based AC beads (0.5% N 2 in helium at 303K and 1 bar) under isothermal conditions Breakthrough concentration profile of methane in Zeolite 5A (8.8% Methane in Hydrogen at 303K and 20.2 bar) under isothermal conditions Breakthrough concentration profiles of 70% CH 4, 20% CO 2 and 10% N 2 in Zeolite 13X at 300K and 2.5 bars Figure 3.5 Temperature profile at bed exit (70% CH 4, 20% CO 2 and 10% 49 N 2 in Zeolite 13X at 300K and 2.5 bars) Figure 4.1 Double bed adsorber 51 Figure 4.2 Required amount of zeolite13x for complete separation of CO 2 55 Figure 4.3 Figure 4.4 as a function of feed pressure Adsorption capacities and selectivity for CO 2 -N 2 -zeolite13x system Effect of concentration (%) of CO 2 on required amount of zeolite13x for 100% separation of CO 2 Figure 4.5 N 2 separation efficiency of zeolite13x 60 Figure 4.6 Comparison of simulated and predicted amounts of zeolite13x 63 Figure 4.7 Total amount (kg/mol of feed gas) of adsorbents for N 2 and CO 2 67 ix

12 Figure 4.8 separation from natural gas for zeolite13x-cms3k adsorber Effect of feed pressure on N 2 separation efficiency (%) for zeolite13x-cms3k system 68 Figure 4.9 Effect of feed gas pressure on total amount of adsorbents for Figure 4.10 Figure 4.11 Figure 4.12 Figure 4.13 Figure 4.14 Figure 4.15 Figure 4.16 to 90% N 2 separation efficiency for zeolite13x-cms3k system Effect of feed concentration on nitrogen separation efficiency at 2.5 bars for zeolite13x-cms3k system Effect of N2 separation efficiency on total amount of adsorbent for zeolite13x-cms3k system Comparison of simulated result with the results obtained from correlation 4.2 for feed composition of 75% CH 4, 15% N 2 and 10 % CO 2 for zeolite13x-cms3k system Total amount of adsorbents for N 2 separation at different feed pressures and compositions (zeolite13x-acb system) Effect of feed pressure on total amount of adsorbent for different N 2 separation efficiency (zeolite13x-acb system) Effect of concentration on total amount of adsorbents at different feed pressures for zeolite13x-acb system Effect of N 2 separation efficiency on total amount of adsorbent for feed pressures of (a) 2.5 bars and (b) 30 bars (zeolite13x- ACB system) Figure 4.17 Comparison of simulated and predicted (correlation 4.3) results 88 Figure 4.18 Calculation procedure for determination of column dimension 91 x

13 Figure A.1 Figure A.2 using correlations Breakthrough of CO 2 in zeolite13x for various mass transfer resistances Breakthrough of CO 2 in zeolite13x with modified macropore resistance Figure A.3 Effect of conductivity (gas and solid) on breakthrough dynamics 111 xi

14 Nomenclature C 0 Initial concentration (kmol/m 3 ) C i Molar concentration (kmol/m 3 ) C pa C ps C pw C vg D a D AB D b D k D m D p H amb H w K K H M Nu P Pe H Pr Specific heat capacity of adsorbed phase (MJ/kmol/K) Specific heat capacity of adsorbent (MJ/kmol/K) Specific heat capacity of column wall (MJ/kg/K) Specific gas phase heat capacity at constant volume (MJ/kmol/K) Axial dispersion coefficient (m 2 /s) Binary diffusivity (cm 2 /sec) Bed diameter (m) Knudsen diffusivity Molecular diffusivity Pore diffusivity (m 2 /s) Wall-ambient heat transfer coefficient (MW/m 2 /K) Gas-wall heat transfer coefficient (MW/m 2 /K) Dimensionless Henry s constant Henry s constant Molecular weight (kg/kmol) Nusselt number Feed pressure (bar) Peclet number for gas wall heat transfer Prandtl number xii

15 Q R c Re S Sc Sh T 0 T amb T g T s T w V W T Z a p d p h f k k f k g q 0 q Amount of adsorbents (g or kg per mol of feed gas) Crystal radius (m) Reynolds Number Separation factor (mol/mol) Schmidt Number Sherwood Number Wall temperature (K) Ambient temperature (K) Gas phase temperature (K) Solid phase temperature (K) Wall temperature (K) Atomic diffusion volume Width of column (m), Height of adsorbent bed (m) Specific particle surface per unit volume bed (m 2 (Particle area)/m 3 (Bed)) particle diameter (m) Gas-solid heat transfer coefficient (MW/m 2 /K) Effective mass transfer coefficient Film mass transfer coefficient (m/s), Conductivity of gas phase Initial loading (kmol/kg) Loading (kmol/kg) q* Instantaneous equilibrium concentration (kmol/kg) xiii

16 r mac r p t w y i ε b ε i ν g Macropore radius Particle radius (m) Time (second) Solid phase concentration (kmol/kg) Mole fraction of component i Bed voidage Interparticle voidage Gas velocity (m/s) μ Dynamic viscosity (N s/m 2 ) η Separation efficiency (%) μ mix ψ Viscosity of gas mixture Shape factor ρ g Gas phase molar density (kmol/m 3 ) ρ mix Density of gas mixture ρ s Adsorbent bulk density (kg/m 3 ) ρ w Wall density (kg/m 3 ) φ ij τ ΔH i Binary viscosity of gas mixture Tortuosity Heat of adsorption (MJ/kmol). xiv

17 1 Introduction 1.1 Natural gas More than 80% of energy comes from carbon-based fossil fuel (coal, oil, and natural gas), and such contribution is expected to remain the same until 2030 (World Energy Council, 2010). Combustion of these fossil fuels produces carbon dioxide (CO 2 ), which has already raised concern regarding global warming and climate change. Some other pollutants that are also associated with fossil fuel combustion are carbon monoxide (CO), nitrogen oxides (NO x ), sulfur dioxide (SO 2 ), particulate matters (PMs), formaldehyde (CH 2 O), and mercury (Hg) (Table 1.1). These pollutants pose adverse impacts on human health and the environment. Among the fossil fuels, natural gas is considered to be the cleanest fossil fuel as it produces fewer quantities of CO 2, NO x, SO 2, PMs, and Hg than coal and oil (EIA, 1999). Natural gas is a complex mixture of hydrocarbons (such as methane, ethane, propane, butane, and heavier hydrocarbons) and nonhydrocarbons (such as N 2, CO 2, and hydrogen sulfide (H 2 S)). It may be present as free gas (bubbles) or dissolved in either crude oil or brine under reservoir conditions in hundreds of different components with various concentrations. Even two wells in the same reservoir may yield different natural gas compositions (Younger, 2004). For example, as shown in Table 1.2, the concentrations of CH 4 vary from to 96.91%, while the concentrations of N 2 and CO 2 vary from 0.68 to 26.1% and 0.82 to 42.66%, respectively. Some reservoirs other than those shown in Table 1.2 may have extreme contents of CO 2 (92%), N 2 (86%), and H 2 S (88%) (Hobson and Tiratso, 1985). 1

18 Table 1.1: Quantity of air pollutants produced from fossil fuel combustion in lbs/billion Btu (U.S. Energy Information Administration (EIA), 1999) Pollutant Natural Gas Oil Coal Carbon dioxide Carbon monoxide Nitrogen oxide Sulfur dioxide Particulates Formaldehyde Mercury Table 1.2: Composition of natural gas observed in different reservoirs as mole percentage (Kidnay and Parish, 2006) Component Canada (Alberta) Western Colorado Southwest Kansas Bach Ho Vietnam Miskar Tunisia Cliffside Texas Rio Arriba New Mexico Methane Ethane Propane Butane Pentane and heavier Helium Nitrogen Carbon dioxide Hydrogen sulfide

19 Natural gas is typically processed to meet pipeline specifications (Table 1.3), which are intended to deliver the natural gas with high heating value to the end users and also to protect pipeline from corrosion and plugging. For example, to prevent corrosion, the concentrations of CO 2, H 2 S, and mercaptans or total sulfur are limited to less than 3% (mole), 6 7 mg/m 3, and mg/m 3, respectively, while to prevent liquid dropout, the concentrations of butane and heavier hydrocarbons are limited to less than 2.0% (mole) and less than 0.5%, respectively. This study focuses on separation of CO 2 and N 2 from natural gas for the purpose of compliance to the pipeline gas specification. These two gases are considered to be the contaminants of natural gas since they have no heating value and occupy transport volume. The CO 2 corrodes pipelines in the presence of water and the N 2 produces NO x when natural gas is combusted. 1.2 Industrial separation processes for removal of N 2 and CO 2 from natural gas Four gas separation techniques, namely, cryogenic distillation, membrane separation, absorption, and adsorption are in practice for natural gas purification. Of these, the cryogenic and absorption processes are economically viable at high gas throughputs (> 15 MMscfd), while the membrane and adsorption processes are viable at a gas throughputs of MMscfd and 2-15 MMscfd, respectively (Kidnay and Parish, 2006). The absorption process is widely used for CO 2 capture while the cryogenic process is established for N 2 removal from natural gas. Neither of these two processes is suitable for removal of both CO 2 and N 2 simultaneously. The cryogenic process for N 2 separation requires extensive pretreatments that eliminate CO 2 from feed gas. 3

20 Table 1.3: Specification of pipeline natural gas (modified from Kidnay and Parish, 2006) Components Quantity (mole % or as mentioned) Methane 75.0% (minimum) Ethane 10.0% (maximum) Propane 5.0% (maximum) Butane 2.0% (maximum) Pentanes and heavier 0.5% (maximum) Nitrogen 4.0% (maximum) Carbon dioxide 3.0% (maximum) Hydrogen sulfide 6 to 7 mg/m 3 Total sulfur mg/m 3 Water vapor mg/m 3 4

21 Additionally, these two processes incur high operating costs compared to membrane or adsorption processes (Robertson, 2007). Conventional membrane (cellulose acetate/polysulfone) separation technologies use kinetic diameters of molecules as the separation criterion. The kinetic diameters of CH 4, CO 2, and N 2 are 3.8Å, 3.3Å, and 3.6Å, respectively (Do, 1998). These diameters are too close to offer a favorable selectivity for membrane. Silicone membrane separation uses equilibrium affinity as the separation criterion. In CH 4 -N 2 separation using this membrane, CH 4 comes out at low pressure end and, hence, leads to additional recompression costs. Another critical element of the process is the pretreatment of the feed gas since particulates block the membrane openings and liquids cause swelling, resulting in decreased performance and even physical damage. Membranes can be highly efficient mass-separating mediums, especially when the species that are to pass through the membrane are present in a large concentration (Choi et al., 2009). Adsorptive separation is a process where certain fluid particles are bonded to the surface of an adsorbent by physical/chemical bonding. It is based on three distinct mechanisms: steric (dimension: pore and molecule size), equilibrium (accommodation ability), and kinetic (diffusion rate) mechanisms (Do, 1998). The first step in separation is adsorption during which species are preferentially picked up from the feed by adsorbent/adsorbents (porous solid), and the second is regeneration or desorption during which the species are removed from the adsorbent. There are two types of adsorption processes: physical adsorption and chemical adsorption. Of them, the physical-adsorption process is an energy efficient and low cost technology (Siriwardane et al., 2001). 5

22 1.3 Adsorption process and adsorbents for CO 2 and N 2 removal The separation efficiency of the adsorption process depends on the quality of the adsorbent, a porous solid. Ideally, an adsorbent should have large adsorption capacity, fast adsorption and desorption kinetics, infinite regenerability, and a wide yet tunable range of operating conditions (Choi et al., 2009). However, in practice, it is rare to find such an ideal adsorbent. Another adsorption behavior to consider is the competitive adsorptions, known as selectivity, of components (CO 2 /N 2 /CH 4 ) of a gas mixture (natural gas, considered to be a mixture of CH 4, CO 2, and N 2 for simplicity). Thus, optimizing the trade-off between beneficial and non-beneficial features is the key in process design and operation. The adsorbents that have been used for CO 2 separation are zeolites (crystalline aluminosilicates), activated carbons, calcium oxides, hydrotalcites, and supported amines. A review of these materials can be found in Choi et al. (2009). The zeolite-based adsorbents were reported to yield relatively high adsorption capacities (Ding and Alpay, 2000). Harlick and Tezel (2004) carried out an experimental screening study of various synthetic zeolite adsorbents and reported that zeolite13x possesses a maximum CO 2 adsorption capacity of 4.5 mol/kg at 1 bar and 295K. Typically, the zeolites recover fresh adsorption capacity when regenerated, though little irreversible behavior was reported by Brandani and Ruthven (2004). Zeolite13X also provides high selectivity for CO 2 over CH 4 and N 2 (Cavenati et al., 2004). The adsorption of N 2 on several commercial adsorbents was studied by many researchers. Of these, carbon-based adsorbents, such as the carbon molecular sieve (CMS) and activated carbon bead (ACB), showed greater adsorption capacity (0.27 6

23 mol/kg at 303K and 100 kpa) (Shen et al., 2010). The notable difference between these two adsorbents is in pore distribution. ACB carries both micropores and transitional pores ranging from 10 to 500 Angstroms (Å), while CMS contains uniform pores of less than 10 Å (Do, 1998). The porous structures of ACB and CMS lead to equilibrium-based separation and kinetic-based separation, respectively. In equilibrium-based separation, the equilibrium selectivity of these carbonaceous adsorbents favors CH 4 over N 2, which eventually renders more adsorbed CH 4 at the surface of the adsorbents. In kinetic-based separation, the kinetic selectivity of N 2 over CH 4, offered by CMS, leads to less adsorption of CH 4 in adsorbents. An advantage of equilibrium-based separation is that it offers longer cycle time than its counterpart: kinetic separation. Both ACB and CMS have greater affinity for CO 2 than either CH 4 or N 2, which leads to the necessary removal of CO 2 from the natural gas to facilitate optimum nitrogen separation. In this study, zeolite13x was used to separate CO 2 from a ternary mixture of CH 4, CO 2, and N 2. ACB and CMS were used to separate N 2 from a binary mixture of CH 4 and N Modeling and simulation In adsorption separation systems, the process variables are strongly coupled, resulting in complex interrelationships. Therefore, the effect of any single variable on separation efficiency is simply unpredictable by simple reasoning or empiricism (Hassan et al., 1986). Furthermore, a change in adsorbent material adds additional complexity due to their unique adsorbate-adsorbent behaviour under the same operating conditions (Flores-Fernandez and Kenney, 1983). Thus, the design and optimization requires either 7

24 extensive experimentation or the guidance of a predictive model (Farooq and Ruthven, 1991). An adsorption model requires in-depth mechanistic knowledge of the kinetics and equilibria of adsorption process and their impact on the dynamic response of an adsorption column (Ruthven, 2000). The pioneer studies of kinetics and equilibria include, but are not limited to, the works of Habgood (1958), Barrer et al. (1963), and Mayers and Prausnitz (1965). These studies reveal that the adsorption separation efficiency is controlled by either equilibrium or kinetics. The simplest adsorption model uses the equilibrium theory. Thomas (1944) can be credited to be the pioneer of the use of equilibrium theory in an ion exchange column. His work was later shaped by Glueckauf (1955) and Rosen (1952) in a general form for application in gas adsorption processes. Such an equilibrium model accommodates the analytical solution of the governing material balance equations and provides useful behavioral insights. However, the theory does not consider real situations such as partial equilibrium and dispersive flows observed in an industrial setup. This model also ignores the mass transfer resistances (Hassan et al., 1986). Failure of incorporating such important process characteristics resulted in outsized deviations in the adsorption of CO 2 on silica gel (Mitchell and Shendahnan, 1972) and that of ethylene on zeolite 4A/5A (Hassan et al., 1985). The limited success of equilibrium models necessitates consideration of kinetic models as well. A kinetic model requires adequate representation of mass transfer kinetics. Mitchell and Shendahnan (1973) adopted this approach with one mass transfer resistance and constant velocity. They laid the foundation of a dynamic model that, later 8

25 on, comprehensibly described the mass transfer kinetics as well as resistances (Hassan et al., 1986). This dynamic model accounts for realistic scenarios, such as axial mixing and mass transfer resistances, which are always likely to be present in the practical systems. The model is, therefore, more realistic and sufficiently general to be applied for detailed optimization studies of both systems. The dynamic model equations then become very complicated and, hence, were solved numerically. Various numerical solution procedures were facilitated by the use of computers in the early 1980s. This trend gradually paved the way for the development of process simulators (Ruthven, 2000). The research performed by Liapis and Crosser (1982), one of the earliest examples, served as the foundation of commercial simulators such as Aspen Adsorption (Nilchan and Pantelides, 1998). The availability of such simulators made it possible to simulate the adsorption process with more rigorous mathematical models by greatly reducing the burden of the manual handlings of complex equations and their numerical solutions. The use of such simulators is not limited to merely solving some equations but rather has expanded into the design and optimization of commercial processes. 1.5 Research motivation, objectives and scope of work The separation of CO 2 and N 2 from NG is a two-step separation process as two adsorbents, CO 2 selective and N 2 selective, are required. This can be done in a single column using layers of different adsorbents (Chlendi et al., 1995; Chlendi and Tondeur, 1995; Malek and Farooq, 1998; Yang and Lee, 1998; Lee et al., 1999; Jee et al., 2001; Takamura et al., 2001; Cavenati et al., 2006; Rebeiro et al., 2008) or columns in series 9

26 carrying different adsorbents (Sircar, 1979; Kumar, 1990). Of these two types of adsorber combination, the layered bed adsorber offers compact design and operation flexibility. Layered bed adsorption systems were studied by many researchers, as mentioned above, for the separation of various components, such as CO 2, N 2, CH 4, CO, on various adsorbents. A discussion on the layered bed adsorption system is included in Chapter 2. Of the studies mentioned, the most relevant study for the separation CO 2 and N 2 from NG was published by Cavenati et al. (2006). They studied a layered bed adsorber containing zeolite13x and CMS3K in terms of product purity and separation efficiency and the effects of the ratio of bed height on separation efficiency. It was concluded that the bed ratio has an insignificant effect on product purity and separation efficiency. For their study, they used a single feed pressure (2.5 bars) and two compositions (mole %) of feed gas (70% CH 4 /20% CO 2 /10% N 2 and 60% CH 4 /20% CO 2 /20% N 2 ). No conclusions were drawn for other concentrations or other feed pressures. This study is good for conveying an understanding of the adsorption behavior of CO 2 and N 2 in a mixture of CH 4 -CO 2 -N 2. No methodology for the design of a double bed adsorber was outlined. To this end, this study aims to develop an easy-to-use design methodology for a layered bed adsorber using commercial adsorbents. The method shall cover a wide range of operating conditions consistent with CO 2 and N 2 concentrations observed in various reservoirs and typical feed pressure of NG distribution pipelines. Also, the product shall meet the concentration limit of CO 2 and N 2 in an NG distribution network to maintain pipeline integrity and product purity. The method shall significantly reduce the need for extensive experiment and simulation. 10

27 To achieve such an objective, a standalone simulation study was performed on a two-layered bed adsorption system (zeolite13x-cms3k and zeolite13x-acb) in Aspen Adsorption, a commercial simulator (discussed in Chapter 2). Since such study requires a trustworthy mathematical framework or model, a model was developed using the resources of the simulator. Required inputs were gathered from the literature, calculated using various equations/correlations, and obtained through fitting the experimental data. To justify the reliability of the model, it was validated against several adsorption processes that covered various operating conditions on different adsorbents. Parametric studies were performed for a wide range of operating conditions that covered concentrations observed in various NG reservoirs and feed pressures of typical NG distribution networks. The data generated through parametric study were correlated. The correlation, the first of this kind, predicts the amount of adsorbents for 100% CO 2 separation and 50-90% N 2 separation. A step-by-step procedure was outlined to transform the amount of adsorbent into physical dimensions of the adsorption column. 11

28 2 Literature Review 2.1 Scope of review Charles W. Skarstorm was awarded the first patent on a commercial adsorptive separation process for air fractionation in Since then, the technology has gained a phenomenal growth in commercial applications and process concepts (Sircar, 2006). This chapter addresses process fundamentals and essential components of process design that has led the technology to the present state. 2.2 Adsorption fundamentals Adsorption is a surface phenomenon that refers to enrichment (or rise in density) of material at the vicinity of fluid-solid interfaces through physical or chemical bonding (Rouquerol et al., 1999). The fluid is referred as an adsorbate and the solid is referred to (porous and permeable) as adsorbents. An adsorbent selectively adsorbs a component or components from a mixed feed. Such selective adsorption may depend on the difference in adsorption at equilibrium or on a difference in adsorption rates. There are three distinct mechanisms through which adsorption separation takes place (Yang, 2003). They are (i) steric or molecular sieving effects, (ii) kinetic or diffusional effects, and (iii) equilibrium or selective adsorption effects. The steric effect allows small and properly shaped molecules to diffuse into adsorbent where the molecules are consequently adsorbed while other molecules are barred from entering the pores. The success depends on the pore diameter of adsorbents and the kinetic diameter and shape of fluid particles. Examples of steric separation include gas drying with 3A 12

29 zeolite and the separation of normal paraffins from iso-paraffins and other hydrocarbons with 5A zeolite (Yang, 1987). Kinetic separation is achieved by virtue of differences in diffusion rates and, hence, the mechanism is also known as partial molecular sieving action. For effective separation, the pore size needed to be precisely controlled between the kinetic diameters of the molecules to be separated (Yang, 2003). Nitrogen-methane separation with 5A zeolite and nitrogen-oxygen separation with a carbon molecular sieve are examples of kinetic separations. An equilibrium effect uses the adsorbate-adsorbent interaction at the solid surface when all the components of a gas mixture are present. The strength and affinity of fluid particles determines selective adsorption of components. Separation of carbon dioxide and methane with zeolite is an example of equilibrium separation. In a practical process, any of the mechanisms or any combination of these mechanisms may play a significant role since all of the mechanisms depend on the geometry and topology of the adsorbent (Rigby et al., 2004). 2.3 Adsorbents The essential component of an adsorption separation process is the adsorbent. As for industrial applications, it is a structured solid with inter-connected voids that hold a certain fluid and, hence, separates a contaminant from the bulk of fluids. Characteristics of adsorbents have been described by various researchers. A summary can be found elsewhere (Rigby et al., 2004). The description includes the origin, size, structure, and inter-connectivity of pores. The portrayal of pores in terms of size, distribution, and interconnectivity has found widespread applications in industries. An adsorbent with interconnected pores of the same distribution is known as a homogeneous adsorbent 13

30 (silica gel, activated carbon, activated alumina, etc.). In contrast, heterogeneous (carbon molecular sieve, zeolite), also known as composite, adsorbent consists of pellets of microporous crystal that result in a bidispersity in pore networks. Several features illustrate the quality or usefulness of an adsorbent. In general, an ideal or hypothetical adsorbent should have large adsorption capacity, fast adsorption and desorption kinetics, infinite regenerability and stability, and a wide yet tunable range of operating conditions (Choi et al., 2009). In reality, no single ideal adsorbent is likely to exist and an effective separation process uses trade-offs of these features. Three adsorbents (zeolite13x, activated carbon, and a carbon molecular sieve) were considered in this study. Relevant discussion on these three adsorbents is included in next three subsections Zeolite13X Zeolites are tridimensional aluminosilicates: a periodic array of SiO 4 /AlO 4 composed of Si and Al tetrahedra linked through bridging oxygen atoms giving rise to a periodic distribution of pores and cavities of particular molecular dimensions. This microporous adsorbent represents a major breakthrough in the adsorption separation process due to their uniform pore structure (8 to 10 Å), wide topology, and high (thermal, hydrothermal, and chemical) stability. There are different criteria (pore aperture, shape of pores, dimensionality of channel, and channel connection) according to which the structure of zeolites can be classified. Zeolites can be found in nature or can be synthesized. In synthesized zeolites, the pore structures are controlled by replacing 14

31 negatively charged alumina with cations. Such replacement by sodium produces zeolite- 13X with a pore diameter of 8Å. Adsorption separation of CO 2 in Zeolite-13X has been studied by many researchers. These studies were focused on three important characteristics of the adsorption process: (i) nature of adsorption (Ward and Habgood, 1966), (ii) equilibrium adsorption capacity (Siriwardane et al., 2005; Cavenati et al., 2005), and (iii) breakthrough behavior (Cavenati et al., 2006). As per the study by Ward and Habgood (1966), the dominant adsorption process in zeolite13x is physisorption and, hence, it offers fresh adsorption capacity (when regenerated) and lower regeneration cost. As per Siriwardane et al. (2005), zeolite13x offers the highest equilibrium adsorption capacity among the adsorbents the tested. The separation of molecules by zeolites as adsorbents can take place because of a molecular sieve effect or selective adsorption. Though zeolites are known for molecular sieving actions, separation of CO 2 from a ternary gas mixture of CH 4 -CO 2 -N 2 occurs due to selective adsorption since kinetic diameter of CH 4 (3.8Å), CO 2 (3.3Å), and N 2 (3.6Å) are considerably less than the pore opening (8Å) of Zeolite-13X. All these three gases get adsorbed in zeolite13x showing the highest capacity and selectivity (CO 2 over CH 4 or CO 2 over N 2 ) for carbon dioxide (Cavenati et al., 2004) Carbon adsorbents Carbon adsorbents are employed to absorb non-polar or weakly polar organic molecules. They are roughly divided into four categories: (i) activated carbons (ACs), (ii) carbon molecular sieves (CMS), (iii) activated carbon fibers (ACFs), and (iv) carbon- 15

32 based nanomaterials such as single-walled carbon nanotubes (SWNTs) (Tagliabue et al., 2009). Among them, ACs and CMSs are the most employed material in industrial gas separations. Despite favorable properties, high costs of ACFs and SWNTs limit their uses to small units. ACs and CMSs were studied for nitrogen separation from NG by, for example, Shen et al. (2010) and Cavenati et al. (2006). They also included a comparison with other adsorbents with respect to N 2 rejection. Activated Carbon (AC): AC is a form of carbon processed to be riddle with small, low-volume pores that increase the surface area available for adsorption or chemical reactions. Their usefulness is undoubtedly derived from large pore volume as well as high surface area (Yang, 2003). These meso- or micro-porous carbonaceous materials offer advantages over other materials in terms of cost (Choi et al., 2009). Among practical adsorbents that are being used in industries, activated carbons are complex in terms of both pore structure and surface chemistry due to presence of slit-shaped micropores (3 to 10Å) and oxygen (Do, 1998). The adsorption properties of activated carbon depend on raw material and, also, on activation process (Choi et al., 2009) as well as adsorbateadsorbent interactions. AC performs adsorption separation by exploiting differences in equilibrium adsorption for the constituent of a gas mixture. Carbon Molecular Sieve (CMS): CMSs are nanoporous materials that separate adsorbing molecules on the basis of their size and shape. A noteworthy feature of the CMS materials is that they separate molecules on the basis of rates of adsorption (Foley, 1995). In terms of molecular sieving, CMSs are similar to zeolites with distinctive physical structures. For example, CMSs are amorphous solids that has no long-range 16

33 order while zeolites are crystalline ordered material. Another feature that sets CMSs apart from zeolites is its variable surface chemistry (acidic/basic/neutral/radical). 2.4 Adsorption modeling Mathematical exploration of adsorption processes traces back to work of Thomas (1944) who studied the mechanism of ion exchange with zeolite in an ion-exchange column. His analytical solution assumed a single solute solution. The work was then extended to binary and multicomponent systems by Glueckauf (1949) with an assumption of local equilibrium. With same assumption i.e. local equilibrium, Lapidus and Amundsen (1952) examined the effects of longitudinal diffusion and incorporated first order kinetics in their solution. In the same year (1952), Rosen published a study that included the exact same solution of an adsorption model that introduced rate of adsorption. He assumed that the rate of adsorption is linear. LDF Model: The linear rate of adsorption was then explored by Glueckauf (1955) in the form of a linear driving force (LDF) model. The LDF approximation founded the basis for the kinetic model. Glueckauf (1955) introduced a value of 15 for the LDF constant that provided satisfactory solutions for processes with large cycle times. For smaller cycle times, Nakao and Suzuki (1983) proposed a graphic correlation that provides the values of the LDF coefficient as a function of dimensionless time. Haynes and Sharma (1975) incorporated more realistic cases of mass transfer limitations such as film resistance, interparticle resistance, and intraparticle resistance in LDF approximation. Serbezov and Sotirchos (1996) formulated a general methodology for the 17

34 development of LDF approximations of different degrees of complexity for multicomponent mixtures. Particle-bed Model: The particle-bed models are the most complex models for adsorption-based separations as they combine equations for both the bed and the particle (Serbezov and Sotirchos, 1999). This coupled approach was first formulated by Yang and Doong in They assumed parabolic intraparticle concentration profiles in the solution scheme, which is equivalent to Glueckauf s LDF approximation (Serbezov and Sotirchos, 1999). The model equations were also solved by many others using different numerical approaches such as orthogonal collocation (Shin and Knaebel, 1987; Lu et al., 1992), finite element method (Kikkinides and Yang, 1993), finite difference method (Sun et al., 1996), and global collocation (Khrisnan, 1993). Mass Transport Model: The mass transport models used in the formulation of the equations for the adsorbent bed and the adsorbent particles are essential parts of the overall model. In general, there are four mechanisms of mass transport that have to be considered: bulk diffusion, Knudsen diffusion, Knudsen flow, and viscous flow. Serbezov and Sotirchos (1997a) showed that, in the adsorbent bed, the dominant mode of mass transport is typically viscous flow, which can be modeled by Darcy s law. In the adsorbent particles, however, depending on the operating conditions, all four mechanisms may be equally important (Serbezov and Sotirchos, 1997b). Therefore, the intraparticle mass transport model must accurately describe the multicomponent mass transport of the individual species over a wide range of conditions in order to be useful for simulations. The widely applied and accepted model for the intraparticle mass transport in the adsorption literature so far is the Fickian model, which provides a simple mathematical 18

35 expression for the molar fluxes of the species. However, the Fickian model does not account for intraparticle viscous transport and underestimates the Knudsen flow of each species caused by total pressure gradients. For mixtures of more than two components, the Fickian mass transport coefficient becomes an adjustable parameter that must be obtained by fitting experimental data, even for pore structures that can be represented as parallel pore bundles. The occurrence of viscous flow, Knudsen flow, Knudsen diffusion, and bulk diffusion during transport of gases in porous materials is accounted for in the dusty-gas model (Jackson, 1977; Mason and Malinauskas, 1983; Sotirchos, 1989). Pore Diffusion Model: The bulk separation of gas mixture was first addressed by Yang and Doong in 1985 with a pore diffusion model. They studied a 50/50 gas mixture of methane and hydrogen in activated carbon and, then, extended it for a ternary mixture of hydrogen, methane, and carbon dioxide (Doong and Yang, 1986). The new model, a pore diffusion model for mass transfer, was compared to the LDF model by Farooq and Ruthven (1990). They concluded that the pore diffusion model was complex and computationally cumbersome and was no better than the simple LDF model. Thermal Effects: Non-isothermal effects are intrinsic to every sorption process because of the heat associated with adsorption and desorption. When the process takes place in small-diameter beds with thick metallic walls, the heat is quickly transported to the walls where it is stored, and the operation is nearly isothermal despite the presence of heat effects. In large-diameter beds, such as the ones used commercially, the heat produced or consumed is not conducted fast to or from the walls, and the temperature fluctuation in the bed can sometimes be as high as 100 K (Yang, 1987). A comprehensive discussion and experimental evidence of heat effects in large adsorbent beds is provided 19

36 by Leavitt (1962). The temperature variation during adsorption and desorption can dramatically change the performance of the adsorption-based process because the properties of most of the adsorbents exhibit a strong temperature dependence. Therefore, the mathematical models used for the design and optimization of adsorption-based processes must account for the temperature changes in the adsorbent bed. Meyer and Weber (1967) and Nagel et al. (1987) developed non-isothermal adsorption models in which both energy equations (for the adsorbent bed and for the adsorbent particle) were included. However, for adsorption systems with moderate heat effects and moderate temperature dependence of the adsorption isotherms, the temperature in the particle may be assumed to be uniform and the energy equation in the particle does not have to be included in the model (Serbezov and Sotirchos, 1998a). Such non-isothermal models were proposed by Chihara and Suzuki (1983), Yang and Doong (1985), and Farooq et al. (1988). 2.5 Multicomponent separation Separation of more than one impurity from a gas mixture requires more than one selective adsorbent and, hence, the bed models for multicomponent separation need to consider different adsorbents. These also create complexity for the inlet conditions for the second adsorbent bed. Such a model has been studied by several investigators. The arrangements of adsorbents led two types of adsorption systems: multiple adsorption column and layered bed adsorption column systems. Multiple Adsorption Columns: Sircar (1979) patented the first hydrogen purification system with impurities such as CO 2, CO, and CH 4. Two adsorption beds 20

37 were operated in series. A similar approach, a series of beds, was also patented by Kumar (1990) for the production of hydrogen from coke oven gas by using activated carbon and zeolite5a. Both of them kept options for independent operation of the beds, which made the process complicated in terms of operations. Layered Bed Adsorption Column: Chlendi and Tondeur (1995) were the first to study fixed-bed adsorption with two layers (activated carbon and molecular sieve 5A) of adsorbents in a single column for separation of carbon dioxide using an equilibrium model. Chlendi et al. (1995) extended the previous work for hydrogen purification from cracked natural gas. They neglected thermal effects for the system and studied the effects of some operating and design variables on the performance of PSA cycles. Yang and Lee (1998) studied adsorption dynamics of a layered bed adsorption system with activated carbon and zeolite5a for hydrogen recovery from coke oven gas. They used a single composition of the gas mixture and a simplified form of numerical simulation in their study. They found an intermediate breakthrough behavior. Later, Lee et al. (1999) extended the study and investigated the effect of the ratio of bed heights. They found the ratio to affect the separation purity for a given throughput. Malek and Farooq (1998) studied the removal of hydrocarbons from refinery fuel gas with a double layered (silica gel and activated carbon) adsorption system. They pointed out that the heat effect had a significant effect on the performance of PSA cycles. Methane, ethane, propane, and butane were considered to be major impurities. Jee at el. (2001) studied the effect of adsorption pressure, feed flow rate, and the ratio of adsorbents (activated carbon to zeolite 5A) for hydrogen PSA cycles with two adsorbents 21

38 (activated carbon and zeolite 5A). The notable difference with previous studies was the inclusion of energy balance. Takamura et al. (2001) studied a dual bed adsorption system for CO 2 separation from boiler exhaust gas. They also studied the effects of the adsorbent ratio and concluded that the ratio affects the separation efficiency of the process. Cavenati et al. (2006) studied separation of CO 2 and N 2 from a mixture of CH 4, CO 2, and N 2. They used a layered bed of zeolite13x and CMS 3K in their study. They investigated the breakthrough dynamics and temperature variation in the bed. They limited their study to two concentration combinations and a single feed pressure. They also used a fixed volumetric flow rate in their study. Rebeiro et al. (2008) studied five component separations in a dual bed of activated carbon and zeolite for hydrogen purification. They compared a reduced model based on controlling resistance with complete model and concluded that the effect of micropore resistance was not significant. Commercial Platforms: The next level of publications on adsorption modeling included all the features and criticality of adsorption processes together to produce a general platform with which any process can be explored. Notable examples of such an approach are Kumar et al. (1994) and Da Silva (1998). Based on these publications, commercial simulators, such as Aspen Adsorption and gprom, were built. They offer various modeling flexibility, which can be customized for a process of interest. Even procedures for numerical solutions can be chosen. 2.6 Numerical solution of partial differential equations 22

39 Several numerical methods that address the solution of partial differential equations (PDEs) with steep fronts and highly non-linear behaviour are available in literature. For example, Nilchan (1997) and Nilchan and Pantelides (1998) used finite difference and collocation methods in both time and space to discretize the PDEs, which were then solved using a non-linear solver in the gproms platform. The drawbacks of the technique include ineffective initialization of a large set of equations and a lack of guarantee of a real-valued solution (Biegler et al., 2005). Ko et al. (2003) also stated that complete discretization may lead to convergence failure due to the accumulation of errors, especially for complicated models. On the other hand, the method of lines (MOL) is a two-step technique that discretizes the space derivative first and then applies numerical integration to find approximate solution. The decoupling of the space and time variable can produce high-order accuracy (Biegler et al., 2005). Ko et al. (2003) found the technique to be easier and more reliable than complete discretization models. Discretization of Space Derivatives: Several discretization techniques (finite difference, finite element, and finite volume) have been applied with first order or higher order accuracy by different authors (discussed by Beigler et al., 2005) within the MOL framework. The problem is numerical smearing or oscillation, which were addressed by introducing a high resolution scheme (Finlayson, 1992), multiresolution scheme (Cruz et al., 2003) and flux corrected transport (Book, 1981). The flux corrected transport method has been modified in modern versions. For example, Van Leer used an anti-diffusion step to avoid excessive smearing. Hirsch (1988), Webley and He (2000), and Jiang et al. (2003) successfully used several flux limiter methods. The upwind finite differencing method uses an adaptive or solution-sensitive finite difference stencil to determine the 23