APPLICATION OF CFD MODELING IN COAL UPGRADING PROCESSES FOR HIGHER ENERGY EFFICIENCY AND LOWER ENVIRONMENTAL IMPACTS S.P. Kaldis, G. Skodras, G. Pantoleontos and G.P. Sakellaropoulos Chemical Process Engineering Research Institute and Department of Chemical Engineering, Aristotle University of Thessaloniki, P.O. Box 1520, 54006, Thessaloniki, Greece D. Sofialides and O. Faltsi Simtec Ltd., 2, Paleon Patron Germanou, 54622, Thessaloniki, Greece Abstract Three computational fluid dynamics (CFD) mathematical models are examined in order to explore complex geometries as well as complex phsical and chemical phenomena. In all three cases it was desirable to include User-Defined Functions in FLUENT6.1 standard code. An extensive number of numerical simulations with different geometries, initial and operating conditions were performed in order to validate the proposed models and determine the sensitivity of various operating parameters. It was shown that the proposed models are appropriate for the description of flow behavior in membrane modules, electrostatic precipitators and rotary kilns. Keywords CFD modeling, Membranes, Electrostatic precipitators, Thermal Desorption Technology Introduction Advanced CFD technology lets you handle complex geometries, while capturing the complex physics involved in laminar and turbulent flows. A CFD mathematical model implemented in FLUENT6.1 reliably simulates problems arising in chemical processes, offering an easy-to-use interface with powerful graphics and optimization procedures. In the present paper, CFD simulation analysis is described in three case studies in coal upgrading processes for higher energy efficiency with lower environmental impact. The first one refers to gas separation membrane module design for more efficient CO2 removal technology. The second, is the analysis of operation of electrostatic precipitators for particles removal from the coal combustion flue gases. Finally CFD analysis is used for the optimization of a typical coal combustor for higher efficiency. Research Objectives CO2 separation in hollow fiber membrane modules The use of hollow fiber polymer membrane modules for gas separations has expanded in recent years, to many industrial applications. Analogous research effort has been given for the modeling of these membrane modules, for the prediction of optimum operating conditions and the investigation of various design options [Kaldis et al., 1997]. However, little attention has been given in the fluid phenomena inside and outside of the hollow fibers, although it is clear, especially in high flux industrial membranes, that they influence strongly their separation behavior. Mass transfer limitations in the shell region, and pressure drop inside the hollow fiber [Lim et al., 2000], are two of the various phenomena that are ought to the fluid flow development. A CFD code for the operation of an asymmetric hollow fiber membrane module is constructed and its results are directly compared, under the same operating conditions, with the one-dimensional model. With this procedure, the direct investigation of the fluid phenomena on membrane operation will be accomplished. The one-dimensional code is based on first order differential equations, which describe the permeate and the residue concentrations and the pressure profiles, inside the hollow fiber and on the shell side, along length of the module [Kaldis et al., 1997]. In the CFD code, appropriate user-defined subroutines were incorporated for the complete description of the process in two axial dimensions. In such a way, the flow field calculations are performed together with the simulation of permeation phenomena. The user-defined subroutines constructed for this specific process describe: (i) the mass transfer across the membrane due to permeation, (ii) the HagenPoiseuille description of flow inside the fibers and (iii) the Navier-Stokes equations which describe the flow outside the fibers inside the shell side of the module. The pressure drop in the shell side was calculated with the pore media model of the basic CFD code. The separation of a binary CO2/N2 mixture is examined in a counter current hollow fiber membrane module with 4500 fibers and 1.4 m length. The mixture composition is CO2 21.7 wt% and N2 78.3% w/w, simulating thus a typical coal combustion effluent stream. A parametric study is conducted in terms of two process parameters: a) the feed pressure (50, 30, 10 and 5 bar pressure) and b) the feed flow rate (nF,o and 1/4nF,o, where 51 2004, Workshop of CPERI nF,o=36 mol s−1). All the other flow parameters and mixture characteristics are kept constant throughout the study. In Table 1 the permeate purity is presented as it was obtained with the use of both models for all the examined cases. Both, the 1D and CFD models demonstrate the same qualitative behavior. They both exhibit a local maximum in the range PF=10÷25 bar and indicate that for low pressures the effect of feed flowrate becomes negligible. The 2D model always predicts higher values for purity, than the 1D one. This difference is very exaggerated for small and moderate feed pressures (5, 10 bar) and becomes negligible for higher values (30, 50 bar). This trend is explained by a corresponding deviation between the two models in terms of the driving force. As the feed pressure drops (and consequently the density of the mixture), the mean inlet velocity is increased in order to maintain the feed molar flow rate constant. At the shell side the local velocities at the membrane inlet area exhibit a strong acceleration region near the side wall. Therefore, by increasing the mean feed velocity, the velocity difference in the radial direction is enhanced, causing a corresponding increase in the radial variation of the mixture flow rate. Consequently, as the feed pressure is reduced, the absorption rate at the inlet region is augmented for the 2D model, while this effect cannot be captured by the 1D one, resulting in the amplification of the difference between the two models. Table 1: Permeate Purity for all Cases PF (bar) 50 30 10 5 50 30 10 5 52 nF (mol/s) 36 (nF,o) 9 (1/4nF,o) Purity (% w/w) 1D model CFD 83.05 85.32 85.34 88.58 82.14 89.03 68.98 82.73 60.46 71.27 72.16 78.10 83.83 67.74 81.40 Figure 2. Two-dimensional profiles of CO2 permeate concentration inside the fiber. Feed pressure: 10 bar, Feed flowrate : 9 mol/s... The effect of the driving force on the membrane separation performance is more clear when the concentration profiles inside the membrane are examined with CFD analysis. As shown in Fig. 1 at the fiber inlet region, substantial radial gradients of CO2 concentration are observed a behavior which can not also be predicted by the 1D model Electrostatic precipitators for gas cleaning An electrostatic precipitator (ESP) is a device to separate fine particles from a flue gas by charging the particles and driving them toward the collecting plate using electrostatic forces. ESPs have been commercialized in modern pulverized-coal fired power stations and the cement industry since the beginning of this century. Industrial ESP has very complex interaction mechanisms between the electric field, the fluid flow, and the particulate flow [Choi and Fletcher, 1998]. The numerical simulation of ESP is challenging especially when dust particles are heavily loaded in the gas stream. Electrostatic body forces can produce a secondary gas flow, well known as `electric wind' or `corona wind' in an ESP. Charged dust particles migrate to the collecting plate due to Coulomb forces, but are also under the influence of momentum interaction with the gas flow in terms of aerodynamic drag. The strong coupling of the governing equations describing the motion of ions, gas and particles, including the effects of particle space charge, and a novel description of the particle charging process are employed to predict accurate particle motion in a representative industrial precipitator. The particulate two-phase flow is described basically in two ways, the Lagrangian and the Eulerian method. The Langrangian approach treats the fluid phase as a continuum and calculates the trajectory of a discrete single particle from the balance forces acting on the particle. The Eulerian approach treats the particulate phase as a continuum, as it also does for the gas phase. The conservation equations of mass and momentum are solved for both phases. Fly ash particles are accelerated by the electrostatic force and the aerodynamic drag, while ions generated by the electric breakdown adhere to the particles suspended in the gas stream and charge them. The ion charge density and the strength of the electric field are determined by Poisson equation and current continuity equation. The differential equations are solved simultaneously, in order to derive the solution of the gas, particulate and electric field inside an ESP. The solution procedure in the standard FLUENT6.1 CFD solver is iterative, i.e. the solution is progressively approached Advanced Software Tools iteration by iteration. The standard version of the code includes gas and particulate flow equations but the electric field equations were programmed through the UDF (User Defined Functions) capability. The operating gas was ambient air (density, ρ=1.225 [kg/m3], viscosity, µ=1.7894×10–5 [kg/m/s]), flowing with uniform velocity, Uinlet=1.0 [m/s]. The ash particles were also flowing with equal velocity, Up,inlet=1.0 [m/s] and had a uniform diameter, dp=10 [µm]. The density of the particles, ρp, was equal to 1550 [kg/m3] and their mass flow rate was equal to 0.001225 [kg/s], while air mass flow rate was equal to 0.1225 [kg/s], i.e. the Particle Mass Loading, PML, was equal to 1%. The particles were injected uniformly from all 32 computational cells of the inlet face, and each injection was represented by 10 stochastic streams to account for the turbulent effects (hence the total number of injections was 320). The ion charge density, ρion, was taken constant and equal to 0.00003 [C/m3], while the electric potential of the wires was φ=70 [kV]. All the design parameters were measured i.e. overall mass collection efficiency, percentage of the collected particles at the three ESP zones, contour plots of electric potential and current, particle density and concentration, over the whole ESP domain and the particle tracks. Indicative results are shown in Figs. 3 and 4, which describe the contours of electric potential and particle tracks inside an ESP. The efficiency is measured by performing, after the solution is well–converged, 10 successive particle tracking samples, i.e. 320×10=3200 injections and counting how many of them are collected by the upper plate electrode, how many escape from the exit and how many are incomplete, i.e. are trapped in local flow recirculations or other ambiguities. This is done because of the stochastic/statistical nature of the particle tracking, so the differences between samples are smoothed out and a representative average is obtained. The efficiency of this case was calculated η=97.22 [%], i.e. 3111 injections were collected and the rest 89 escaped. Regarding the spatial distribution of the collected particles, 70.32 [%] were collected in the first zone (x=0÷0.15 [m]), 19.35 [%] in the second zone (x=0.15÷0.30 [m]) and 10.33 [%] in the third zone (x=0.30÷0.45 [m]). Thermal desorption treatment of raw coal materials The state of the art of the Thermal Desorption Technology (TDT) aims at removing hazardous air pollutants and greenhouse gases from raw coal materials, providing less corrosions in the combustion kiln walls and thus offering improved process efficiency. In this study kiln geometries in a rotating frame, multiphase flow behavior and the complex reactions taking place within the kiln were examined using a CFD mathematical model in order to check the technical characteristics of TDT. Figure 3. Spatial distribution of electric potential in an ESP. Figure 4. Tracks of particles in an ESP. Due to high Reynolds numbers within the kiln it was decided to use a simple, industry standard turbulence model, namely the “standard k-ε” one. This model belongs to the large family of “two-equation eddyviscosity models” and is included in the FLUENT6.1 commercial package. Eulerian approach was used instead of the Langrangian one, as more suitable in regions where the solid phase exhibits high volume fractions (dense bed). Multiphase flow of granular and gaseous phases was solved using the equations of conservation of mass and momentum, taking into account the angular velocity of the rotating frame. Equations of heat transfer between solid particles, gaseous phase and kiln walls were established, resulting in a highly non-linear set of partial differential equations in three spatial dimensions and time variable, which were defined and solved in FLUENT6.1 with the appropriate initial and boundary conditions. The conversion of raw coal involves two major steps (Williams et al., 2001): (i) Thermal decomposition (pyrolysis, devolatilization) during particle heating accompanied by physical and chemical changes and (ii) Combustion of the porous solid residue (char). Particle drying can be included in the overall thermal decomposition step. A kinetic model, namely the single reaction approach (Østberg et al., 1998), has been adopted for the devolatilization reaction showing an acceptably low computational cost for full-scale CFD calculations. Char combustion in the absence of 53 2004, Workshop of CPERI molecular oxygen in the kiln atmosphere has been assumed to proceed through reactions with carbon dioxide and water: (A) C(s) + CO2 → 2CO C(s) + H2O → CO + H2 (B) Eventually, four reaction were considered including the coal moisture evaporation. The kinetics of the above equations are defined by User-Defined Functions (UDF) written in C language and incorporated dynamically in FLUENT code. One of the main results of the CFD simulations is that the residence time is significantly increased when the angular velocity is not high, indicating that the rotation speed is the basic mechanism for the advancement of the fuel downstream. Steady-state condition (periodic) is reached after approximately 900 s (=15 min), as depicted in Fig. 5. Figure 5. Instantaneous coal mass flow rate for t=885960 s (negative values denote exit from the kiln). Figure 6. Reaction rates in [kgmol/m3s] (red denotes maximum and blue minimum values). A quite interesting aspect of coal conversion chemistry is the reaction rates of the four reactions considered. In particular, the rate of moisture evaporation is largely independent of kiln temperature and this is due to its very low (Ea=42 KJ/mol) activation energy. On the other hand, the devolatilization reaction is characterized 54 by very high activation energy (Ea=5470 KJ/mol) and thus, its rate is highly temperature dependent. Char combustion reactions (A and B) are also characterized by relatively low activation energies (Ea=138 KJ/mol), so one would also expect them to be temperature independent. This is clearly the case for the C(s) + H2O reaction, but not for the C(s) + CO2. This is attributed to the fact that water vapor levels do not substantially change since the moisture evaporation rate is largely temperature independent. Moreover, the increased devolatilization rate (Fig. 6) substantially increases CO2 levels, which in turn accelerate char combustion reaction A. Conclusions The CFD models performed in FLUENT6.1 were proven to be capable of describing complex chemical processes and handle different geometries such as rotating frames, electric fields and porous media. For all cases studied the governing equations were established and various initial and operating conditions were set in FLUENT6.1 commercial package. User-Defined Functions were also incorporated in the standard CFD code for best accuracy. The results showed that CFD models are powerful tools in the analysis of different cases in both simple and industrial configurations. References Choi B.S. and Fletcher C.A.J., (1998), Turbulent Particle Dispersion in an Electrostatic Precipitator, Appl. Math. Modelling 22, 1009. Kaldis, S. P., Kapantaidakis, G. C., Papadopoulos, T. I., and Sakelaropoulos, G. P., (1997), Simulation of binary gas separation in hollow fiber asymmetric membranes by orthogonal collocation, J. Memb. Sci. 142, 43. Lim, S. P., Tan, X, and Li, K. (2000) Gas/vapour separation using membranes: Effect of pressure drop in lumen of hollow fibres, Chem. Eng. Sci. 55, 2641. Østberg, M, Glarborg, P, Jensen, A., Johnsson, J. E., Pedersen, L. S., Dam-Johansen, K. (1998), A Model of the Coal Reburning Process, Proc. Combust. Inst. 27, 3027. Williams, A., Pourkashanian, M. and Jones, J. M. (2001). Combustion of pulverised coal and biomass, Progr. Energy Combustion Sc. 27, 587.