18th European Symposium on Computer Aided Process Engineering – ESCAPE 18 Bertrand Braunschweig and Xavier Joulia (Editors) © 2008 Elsevier B.V./Ltd. All rights reserved. Composite zeolite membranes characterization by using a transient state experimental technique and a parameter estimation procedure Lucile Courthial,a Arnaud Baudot,a Mélaz Tayakout-Fayolle,a Christian Jallutb a IFP-Lyon, BP 3, 69390 Vernaison, France Université de Lyon, Université Lyon 1, Laboratoire d’Automatique et de Génie des Procédés-UMR CNRS 5007,ESCPE, 43 Bd du 11 Novembre 1918, 69622 Villeurbanne, France b Abstract We describe a non-destructive transient state method for the determination of thermodynamic and transport properties of zeolite composite membranes. The properties under consideration are included as parameters of a linear dynamic model of the membrane. The parameters values are obtained thanks to an estimation procedure based on the comparison between simulated and measurements realized in the linear domain. A step-by-step experimental procedure is derived from a structural identifiability analysis of a simplified linear model of the membrane. A more detailed linear model is then used for the estimation procedure. Due to its transient nature, this technique allows the zeolite selective layer effective thickness determination as well as that of thermodynamic and transport properties of sorbing hydrocarbons. Some examples of results are given to illustrate the capabilities of the technique. Keywords: membrane, identifiability dynamic modelling, parameter estimation, structural 1. Description of the system A scheme of the composite membranes to be characterized is represented in the figure 1. These membranes are composed of a very thin and dense zeolite layer deposited on the inner side of a -alumina porous support. The length of this tubular support is 150 mm, its internal diameter 6 mm and its external diameter 10 mm. Enameled sections and connections tubes Macroporous support L2 0 Zeolite layer L4 LZ z1 L3 z2 z Figure 1: the composite membrane The support global porosity is 0,48 but it is made of three layers having respectively the three following mean pore diameters: 10 m, 20 m and 1700 m. The zeolite active layer is obtained by using an in situ crystallisation process (Chau et al., 2003) at the surface and within the inner layer of the support having a mean pore diameter of 10 m. As a consequence, the only way to characterize this active layer with a non-destructive 2 L. Courthial et al. macroscopic method is to perform experiments by using the membrane as it is represented in the figure 1. In this paper, we describe the way transient experiments similar to the transient version of the well-known Wicke-Kallenbach diffusion cell can be done according to the principle that is represented in figure 2 (Bakker et al., 1996; Sun et al., 1996; Tayakout-Fayolle et al., 1997; Courthial et al., 2006; Courthial, 2007). zeolite layer inner compartment outer compartment macroporous support Figure 2: principle of the transient experiments A fixed composition gaseous mixture flows through the outer and inner compartments of the membrane. The initial conditions are established when the thermodynamic equilibrium is reached. The inlet composition of the inner compartment feed gas is then suddenly modified and the time evolution of the two outlet compositions is measured. We use a concentration pulse injection and we verify that its intensity is sufficiently low so that experiments are performed in the linear domain of the system. The abovedescribed experiments can be performed with pure components or mixtures. Furthermore, one can vary the concentrations of the gases at the initial equilibrium state in order to study the evolution of the parameters with the zeolite active layer loading and composition according to the approach that is used in inverse chromatography (Tondeur et al., 1996; Jolimaitre et al., 2001). We use the capabilities of a gas phase chromatograph and an infrared analyzer to perform the experiments (see Courthial, 2007 for more details about the experimental set-up). 2. Structural identifiability study for pure component experiments The estimation procedure is based on a linear state model of the system. One has to check for the structural identifiability of its parameters. It has been previously shown that when linear or non linear state models are derived for chromatographic columns (Tayakout-Fayolle et al., 2000; Couenne et al., 2005), the solid phase composition should not be represented by the ordinary concentration q but by the concentration of a * gas phase C that would be at equilibrium with the solid phase at each point and at each time. A simple examination of a state model written according to the classical choice of state coordinate q cannot lead to the conclusion that it is overparametrized and to strategy. In order to check for the propose an adapted experimental estimation pertinence of the proposed change of state coordinate in the case of the composite membrane, a simplified version of the model is used in order to apply the transfer function technique (Walter and Pronzato, 1997). Since the result is based on the way the adsorbent composition in the zeolite is represented, we assume that the main conclusion of the structural identifiability study will be valid for the more realistic model that is used to perform the parameter estimation. 2.1. A simplified linear model of the system In this model, we have neglected the axial dispersion in the gas flows as well as the zeolite layer curvature. Furthermore, the macro-porous support is not taken into Composite zeolite membranes characterization by using a transient state experimental technique and a parameter estimation procedure consideration and we take into account only the part of the membrane situated between z1 and z2 (see figure 1). The equilibrium condition between the zeolite layer and the gas phase is represented by the Henry law qr,t A C r,t with C r,t the adsorbate concentration of a gas phase that would be at equilibrium with the zeolite. Once the r Ri z z1 dimensionless coordinates are defined – ie – and where Lz is the Lz ez zeolite active layer length, ez its thickness and R i its internal radius, the dimensionless adsorbate balance equations for the inner and outer gas phases as well as for the zeolite are as follows: * C i 1 t L i C o 1 L o t C i z C o z C 0, t * a i Bz Boundary conditions : C i 0,t C i0 t C o 0,t 0 a o k o C o C 1, t * * (1) Boundary conditions : C t * 2 * 1 C z 2 C * 0,t C i ,t * C k 1,t o C o C * 1, t Bz C i and C o are the inner and outer compartments gas phase concentrations, Bz (2) A Dz ez included in the boundary condition of the zeolite layer inner is a mass transfer resistance 2 surface. z ez is the diffusion time constant. a i and a o are the specific surface areas Dz of the active layer and k o a mass transfer coefficient between the outer gas flow and the active layer surface. As far as the inner surface of the active layer is concerned, it is assumed to be at equilibrium with the gas phase (see figure 1). function analysis 2.2. Transfer The derivation of the transfer function is very long and we only give the result for the transfer function between the inlet concentration of the inner compartment and the outlet concentration of the outer compartment (see figure 2) as an example (see the appendix): L Cˆ o s G s 0 Cˆ i s z Cˆ o 1,s Ds r s r s e e ˆ C i 0,s r1 s r2 s 1 2 (3) These expressions show that the same macroparameters including the initial microparameters (Walter and Pronzato, 1197) appear in the state model and in the corresponding transfer functions. As a consequence, the state model is correctly 3 4 L. Courthial et al. 2 parametrized. As far as the active layer is concerned, only z ez Dz and Bz A Dz ez are structurally identifiable. 2.3. A step-by-step experimental procedure ez and the diffusion layer thickness In order to get the Henry constant A , the zeolite coefficient Dz , we propose to first use a tracer that is not adsorbed in the zeolite. In this case, A 1 and the layer effective thickness can be determined as well as the tracer diffusion coefficient in the zeolite. The latter can be used to check for the results. Once z can be obtained for a given adsorbate from the estimations of ez is known, A and D 2 z ez Dz and Bz A Dz ez . of results for a pure adsorbate 3. Examples We present here some results concerning membranes that has been obtained at IFP from anaqueous solution containing the following species: SiO2/tetrapropyle ammonium hydroxyde/H2O (Chau et al., 2003). 3.1. The linear model of the system for the parameter estimation We use a more realistic model than the one used for the identifiablity study for the membrane as it is represented in the figure 1. The main features of this model are the followings: the diffusion within the macroporous support is taken into account; the curvature of the crystal layer is taken into account; axial dispersion is included in the gas flows model; the tubes connected to the system are included in the model. They are represented by CSTR’s or plug flow- axial dispersion models. The parameters of the flow models have been estimated from experiments where the membrane was replaced by a stainless steel tube. We don’t present this part of the study here (see Courthial, 2007). The volume of the connection tubes is about 9 % of the total volume of the system and we have checked for the consistency of the estimated axial dispersion coefficients with respect to literature. 3.2. Estimation of the active layer effective thickness by using hydrogen as tracer One can see in the figure 3a an example of time domain fitting that is obtained in this case. In the figure 3b is shown the stability of the effective thickness estimation according to different operating conditions (gas flow rates and temperature). From the results represented in the figure 3b, the effective thickness of the active layer of the membrane under consideration is ez 26 4 m . This result is consistent with previous knowledge about this membrane. 3.3. Properties of the membranes with respect to pure n-butane In the figure 4a is shown an example of time domain fitting that is obtained for the nbutane at 200 °C. One can see in the figure 4b the evolution of the n-butane diffusion coefficient through the zeolite layer as a function of the gas phase concentration of n- Composite zeolite membranes characterization by using a transient state experimental technique and a parameter estimation procedure 0.6 4.5E-5 4.0E-5 0.5 3.5E-5 Estimation: 0.4 0.3 D z = 6.12x10 -9 m 2 .s -1 3.0E-5 e z = 2.55x10 -5 m 2.s -1 2.5E-5 k k 0.2 s o -1 = 0.963 m.s = 2.70x10 ez (m) Concentration (mol.m- 3 )ZZZZ butane at the initial equilibrium state. The mean confidence interval that is obtained for these estimations is 10 %. -2 m.s -1 2.0E-5 1.5E-5 1.0E-5 0.1 5.0E-6 0.0E+0 0.0 0 2 4 6 8 0 10 Time (s) 1 2 Volumetric flow rate (cm (a) 3 4 3 .s-1) (b) Figure 3: estimation procedure for the zeolite layer thickness 1.0E-8 Estimation: 0.25 0.20 D z = 3.69x10 -10 e z = 2.58x10 -5 m k s = 0.700 m.s -1 k o = 7.09x10 -3 m 2.s -1 m.s -1 0.15 0.10 9.0E-9 Dz (m 2.s - 1 )zzzz Concentration (mol.m-3)zzzzz 0.30 8.0E-9 7.0E-9 6.0E-9 5.0E-9 4.0E-9 3.0E-9 0.05 2.0E-9 0.00 0.0E+0 1.0E-9 0 5 10 15 Volumetric flow rate (s) (a) 20 0 5 10 15 20 CI = C E (mol.m -3) (b) Figure 4: properties of the membrane with respect to n-butane at 200 °C The method turns to be sensitive since we find that the diffusion coefficient of the nbutane increases with the zeolite loading as it is well known in this domain (Jolimaitre et al., 2001). 4. Conclusion In this paper, we have shown how to derive a transient state technique for the characterization of zeolite composite membranes from a structural identifiability study. Since the three parameters of interest are included into two macroparameters that are structurally identifiable, we have designed a step-by-step procedure. The first step is the determination of the zeolite active layer by using an inert tracer. Then, the equilibrium and the diffusion coefficients can be obtained by performing transient experiments with sorbing hydrocarbons. We have shown some results that have been obtained for a pure adsorbate to illustrate the capabilities of the technique. 5 6 L. Courthial et al. Appendix The quantities present in the transfer function Eq. 3 are as follows: lo Bz ko As z s , r1 s iL z As C s s 2 s 2ai Bz 1 lo e Bs 2 iL ai Bz 1 lo e s z 2 and zs z z 1 lo e s z e s e s L Cs o s ao ko1 1 lo e s 1 lo e lo . Ds 2 oL ao ko s 1 lo e 1 lo e s z A s C s s s 1 lo e s zs z , r2 s z z z s z z z z References W. J. W. Bakker, F. Kapteijn, J. Poppe, J. A. Moulijn, 1996, Permeation characteristics of a metal-supported silicalite-1 zeolite membrane, Journal of Membrane Science, 117, 57-78 F. Couenne, C. Jallut, M. Tayakout-Fayolle, 2005, On minimal representation of heterogeneous mass transfer for simulation and parameter estimation: application to breakthrough curves exploitation, Computers and Chemical Engineering, 30, 42-53 L. Courthial, 2007, Caractérisation des propriétés physico-chimiques et morphologiques des membranes zéolithes par mesure de perméation en régime transitoire, PhD Thesis, Lyon I University L. Courthial, A. Baudot, E. Jolimaitre, M. Tayakout-Fayolle, C. Jallut, 2006, Moments method applied to the in-situ characterisation of normal butane mass transfer in MFI zeolite membranes, Desalination, 193, 215-223 C. Chau, I. Prevost, J. A. Dalmon, S. Miachon, 2003, Process for preparing supported zeolitic membranes by temperature-controlled crystallisation. US Patent, 6582495 B2 E. Jolimaitre, M. Tayakout-Fayolle, C. Jallut, K. Ragil, 2001, Determination of mass transfer and thermodynamic properties of branched paraffins in silicalite by inverse chromatography technique, IEC Res., 40, 914-926 M. S. Sun, O. Talu, D. B. Shah, 1996, Diffusion measurements through embedded zeolite cristal. AIChE Journal, 42, 3001-3007 M. Tayakout-Fayolle, C. Jallut, F. Lefèvre, J. A. Dalmon, 1997, Application of transient methods to measurements of mass transfer parameters in zeolitic membranes, ECCE1, First European Congress on Chemical Engineering, Florence, Italy, May 4-7-1997, 2, 1209-1212 M. Tayakout-Fayolle, E. Jolimaitre, C. Jallut, 2000, Consequence of structural identifiability properties on state-model formulation for linear inverse chromatography, Chemical Engineering Science, 55, 2945-2956 D. Tondeur, H. Kabir, L. A. Luo, J. Granger, 1996, Multicomponent adsorption equilibria from impulse response chromatography, Chemical Engineering Science, 51, 3781-3799 E. Walter, L. Pronzato, 1997, Identification of parametric models from experimental data. Springer