Prediction of CO Burnout using a CHEMKIN based Network Tool Engler-Bunte-Institut / Bereich Verbrennungstechnik Universität Karlsruhe
Contents Complex reaction schemes vs. complex transport models. Complex transport and simplified reaction: CFD. Simplified transport and complex reaction: Zonal models. Zonal modeling of complex combustion systems. ERNEST: a reactor network tool. Required network elements. Solution scheme. Calculation results using a complex network of ideal reactors. Demonstration of ERNEST. 2
Complex reaction schemes vs. complex transport models I. Complex transport models Turbulent transport is usually described within CFD using Reynolds-averaged Navier-Stokes (RANS) equations or even more costly methods (LES, DNS). The solution of RANS equations requires the detailed discretisation of the fluid domain with a computational mesh. (For realistic geometries 0 5-0 7 grid-points) CFD is computationally very expensive. Turbulent Mixture is still a matter of model development (velocity-species correlations) There are strong limitations concerning the description of chemical reaction for combustion models (number of reaction progress variables, species) Time-scales of intermediate formation and consumption are in many situations orders of magnitude smaller than the time-scale of turbulent mixture > resolution problem 3
Complex reaction schemes vs. complex transport models II 2. Complex reaction models and simple transport models The detailed chemical description of even the first alkene (methane) requires at least 20 to 00 species to be considered. Detailed reaction schemes comprise several hundred elementary reactions. The time-scales of these reactions span a range of several orders of magnitude. Anyhow, specialized solvers are able to solve the coupled ODE of many composition sets for simple, idealized reaction systems. Such systems are : - Homogeneous reactor system - Perfectly stirred reactor system - -dimensional laminar flame One approach to realistic combustion systems therefore is to compose a zonal network of idealized reactor elements that substitutes the turbulent mixture. 4
Reactor Network Model Substitute convection with network-element connections. Substitute turbulent diffusion with Internal mixing. Network element connections. Neglect details in flow patterns. Needs preceding knowledge of flow patterns. Network design criteria: turbulent mixing intensity. Strong turbulent Mixing Perfectly stirred reactor Low turbulent Mixing Plug flow reactor 5
Required Elements of a Reactor Network I Perfectly stirred reactor (PSR) 0-dimensional Perfect internal mixture Substitute for strong turbulent mixing intensity Concentration inside the reactor Y = k,reactor Y k,outlet Residence time Species conservation Energy conservation ρv τ = M& V = V& ( Y Y ) w& W V M& = = k k,inlet K τ K k k k ( hk hk,inlet ) + + = 0 k= k= k k h w& W ρ Q& ρv 6
Required Elements of a Reactor Network II Plug-flow reactor (PFR) -dimensional No internal mixture Substitute for low turbulent mixing intensity Time-dependent species concentration Time-dependent internal energy Position-dependent species concentration Position-dependent energy ρdy dt K k = w& k W k K dt dv c v,k + v e w& k kwk + p = 0 dt dt k= k= dy ρu dz k = w& k W K dt ρcp u + h w& k kwk = 0 dz k= k v = = ρ V M 7
Required Elements of a Reactor Network III Non-reactant mixer (MIX) 0-dimensional No reaction Required to merge flows T,M Y k, T 3 T 2,M 2 Y k,3 Y k,2 M 3 Species conservation Energy conservation M & & + K k= 3Yk,3 = MYk, M2Yk,2 & K K ( M & h ) = ( M& h ) + ( M& ) 3 k,3 k, 2 hk,2 k= k= 8
Required Elements of a Reactor Network IV Non reactant flow splitter (VZW) 0-dimensional No reaction Required to split flows T,M 2 T Y k,,m Y k, T,M 3 Y k, Mass conservation M & M & = α & 2 M ( α) & 3 = M 9
Additional Network Elements Inlet Defines mass flow rate, composition and thermo-dynamic state of inlet flows. Outlet Marks an element, that has no successor for network management. Equilibrium calculator Calculates the thermo-dynamical equilibrium state and ignores the mass flow rate. Under-relaxation Element required for stabilization of the solution procedure. 0
ERNEST A reactor network program Based on CHEMKIN II library for description of thermo-dynamic properties and chemical reaction systems. Incorporates re-engineered CHEMKIN II modules for idealized reactors PSR, Plug-flow. Post processing of results in various units of measure. Handles re-circulating flows without principle limitation in complexity. next foil. Console program with keyword based problem description. Tcl/Tk based GUI. more details: demonstration.
Examples for possible network structures Internal and external re-circulation: Overlapping re-circulations: Multi-staging re-circulations: 2
Stack based network management algorithm Take topmost network element from stack. If stack is empty: Stop Start Process the topmost network element. Lay all network inlet elements onto stack. Judge outlet parameter. Converged? Optimize stack. Lay the following network element onto stack. 3
Convergence control algorithm 4
Reactor network example for an aero-engine Reactor network and suppositional flow field, based on expert knowledge 5
NO X prediction of aero-engine network 65 60 55 Lower Limit Upper Limit First Reactor Network Newest Reactor Network 50 45 NO x [g/kg] 40 35 30 25 20 5 0 5 0 0 25 50 75 00 Thrust [%] 6
NO X prediction of aero-engine network II 7
CO prediction of aero-engine network Network is not sufficient for prediction of CO This result could be attributed to insufficient of the quench near the wall A more detailed description would require more knowledge of the mixing process near the wall 8
ERNEST - Demonstration 9
Summary CFD-Models are usually incapable of handling complex chemical reaction systems due to limitations in models and computational resources. A work-around can be given reducing the complexity of the considered convective/diffusive problem. This results in a zonal treatment of the problem. Within these zones an idealized reactor may be applied to get information about the chemical reaction output. ERNEST is capable of handling such a zonal treatment of realistic combustion systems. With this tool, general trends for emission data can be predicted. 20