Turbulent Combustion Modelling and Simulation Sustainable Combustion Laboratory
Turbulent Combustion Modelling and Simulation Sustainable Combustion Laboratory Studying Turbulent Combustion Physics with DNS, LES, RANS Numerical Combustion Team CFD Codes : Hea. RT (in house) for DNS/LES, ANSYS FLUENT for LES/RANS N. M. Arcidiacono, G. Calchetti, D. Cecere, A. Di Nardo, E. Giacomazzi, F. R. Picchia Hea. RT ‘s key features Turbulent Combustion Physics Scenario q Implementation q Fortran 95 with MPI parallelization. q Genetic algorithm for domain decomposition. Heat: Fourier, species enthalpy transport due to species diffusion; q Mass diffusion: differential diffusion according to Hirschfelder and Curtiss law; q Radiant transfer of energy: M 1 diffusive model from CTR [Ripoll and Pitsch, 2002]. q Fluid dynamics – Turbulence q q Numerics structured grids with possibility to use local Mesh Refinement (in phase of validation); q conservative, compressible, density based, staggered, (non-uniform) FD formulation [S. Nagarajan, S. K. Lele, J. H. Ferziger, Journal of Computational Physics, 191: 392 -419, 2003]; q 3 rd order Runge-Kutta (Shu-Osher) scheme in time; q 2 nd order centered spatial scheme; q 6 th order centered spatial scheme for convective terms (in progress); q 6 th and 10 th order compact spatial schemes; q 3 rd order upwind-biased AUSM spatial scheme for convective terms; q 5 th-3 rd order WENO spatial scheme for convective terms for supersonic flows (S-Hea. RT); q finite volume 2 nd order upwind spatial scheme for dispersed phases (Hea. RT-MPh); q explicit filtering of momentum variables (e. g. , 3 D Gaussian every 10000 time-steps); q selective artificial wiggles-damping for momentum, energy and species equations; q extended NSCBC technique at boundaries considering source terms effect; q synthetic turbulence generator at inlet boundaries [Klein M. , Sadiki A. , Janicka J. , Journal of Computational Physics, 186: 652 -665, 2003]. Chemical kinetics Acoustics Multi-phase flows Both LES and DNS require high spatial resolution, order of 10 -4 -10 -5 m at least, to capture large spatial gradients and small scales of turbulence. Besides, unsteady simulations require small time steps, ranging from 10 -6 s down to 10 -9 s depending on the integration scheme (implicit or explicit, mainly) and on the inclusion of acoustics. Hence, several millions of grid points and time steps are needed to solve a problem. These make the time to solution large and supercomputing absolutely necessary. The simulations reported here required nearly three months of computation each, even using supercomputing. q q q Alternative fuels q CCS Molecular Properties kinetic theory calculation and tabulation (200 -5000 K, T=100 K) of single species Cpi, i (20% saving in calculation time with respect to NASA polynomials); q Wilke’s law for mix; Mathur’s law for mix; Hirschfelder and Curtiss’ law for Di, mix with binary diffusion Di, j estimated by means of stored single species Sci or via kinetic theory; q supercritical transport properties and real gas equation. q q Turbulence and Combustion Models subgrid kinetic energy transport equation; q Smagorinsky model; q Fractal Model (modified) for both turbulence and combustion closures; q flamelets - progress variable - mixture fraction - flame surface density - pdf approaches; q Germano’s dynamic procedure to estimate models’ constants locally; q Eulerian Mesoscopic model for multi-phase flows. q q Complex Geometries Chemical Approach single species transport equation; q progress variable and its variance transport equations; q reading of chemical mechanisms also in CHEMKIN format. q (3 rd Immersed Boundary and Immersed Volume Methods order for the time being). IV is IB rearranged in finite volume formulation in the staggered compressible approach. Performance evolution of Hea. RT from CRESCO 2 to CRESCO 4 Importance of Combustion Dynamics EU Energy Road. Map 2050 q q Radiant transfer of energy Diffusive Transports Test Case Three slot premixed burners q Decarbonization Ø Clean and efficient § Stoichiometric CH 4/Air § Central Bunsen flame § Flat flames at side burners § 2 mm side walls separation q Computational domain § 10 x 7. 5 x 5 cm 3 (Z x Y x X) q BIG case § 534 x 432 x 207 47752416 nodes q Aims § Single zone performance analysis. § Validation of a new SGS turbulent combustion model. power generation q Security of energy supply Ø Safe operation Ø Availability and reliability q Renewables Lack of a gas quality harmonization code Electricity grid fluctuations q Power 2 Gas q H 2 -blends Fuel-flexibility Load-flexibility Shaheen (Blue Gene/P) 222 TFlops ENHANCED COMBUSTION DYNAMICS Development team: 16384 Single-Proc 4 cores 32 -bit Power. PC 450 850 MHz 4 GB/node 64 TB 3 D “torus” E. Giacomazzi, F. R. Picchia, D. Cecere, F. Donato, N. M. Arcidiacono Hea. RT’s LES APPLICATIONS Here, some examples of Hea. RT code simulations are reported. Topics cover both theoretical and applied aspects of turbulent combustion. On theoretical side, the research group is interested in analysing and modeling turbulence / combustion interaction (e. g. , VOLVO Flig. Motor), and hence in understanding the role and dynamics of turbulent structures in a reactive flow and the effects of chemical reactions on vortices. On the application side, interest is focused on premixed combustion of natural gas and air (e. g. , DG 15 -CON) and on combustion of hydrogen blends (in particular, syngas and hydrogen enriched natural gas) at low (e. g. , SANDIA Flame A) and high (e. g. , PSI) pressure, in premixed and non-premixed conditions. Some studies aims also at identifying the dynamic behaviour of new combustor concepts (e. g. , TVC). Besides, some activities are devoted to the general development of the code, i. e. , to the implementation of numerical integration schemes and numerical techniques aiming at enhancing its accuracy, efficiency (e. g. , Mesh Refinement), its capability of modeling complex geometries (e. g. , IVM and PRECCINSTA) and of simulating supersonic flows (Hy. Shot II). Combustion Dynamics in VOLVO Flig. Motor C 3 H 8/Air Premixed Combustor 65536 Immersed Volume Method for Complex Geometry Treatment Using Structured Cartesian Meshes and a Staggered Approach Hea. RT’s DNS APPLICATION Mesh Refinement in LES Compressible Solvers [E. Giacomazzi et al. , Comb. and Flame, 2004] CH 4/Air Premixed Comb. in DG 15 -CON [ENEA] [D. Cecere et al. , Flow Turbul. and Comb. , 2011] [D. Cecere et al. , submitted to Computer Methods in Applied Mechanics and Engineering, 2013] Acoustic Analysis in a TVC [D. Cecere et al. , in progress] [G. Rossi et al. , in progress] Thermo-Acoustic Instabilities in the PRECCINSTA Combustor PSI Pressurized Syngas/Air Premixed Combustor SANDIA Syngas Jet Flame “A” H 2 Supersonic Combustion in Hy. Shot II SCRAMJET [E. Giacomazzi et al. , Comb. Theory & Modelling, 2007 Comb. Theory & Modelling, 2008] [D. Cecere et al. , Int. J. of Hydrogen Energy, 2011 Shock Waves, 2012] [D. Cecere et al. , in progress] ANSYS-FLUENT’s LES/RANS APPLICATIONS Thermo-acustic instabilities in a labscale burner Lean premixed combustion in gas turbines (GT) is widely used in order to meet stringent low NOx emissions demands. If this technique allows the achievement of a quite homogeneous temperature distribution, thermo-acoustic instabilities are a common problem in gas turbine combustors operating in lean premixed mode. Pulsations, caused by resonant feedback mechanism coupling pressure and heat release, can lead to strong perturbations in the gas turbine. Equivalence ratio fluctuations is one of the major cause of flame instability. In this study the experimental campaign conducted at the German Aerospace Center (DLR) was chosen as test case. The simulations were conducted using the commercial CFD code ANSYSFLUENT. The computational grid consists of about 4. 000 of computational cells [E. Giacomazzi et al. , in progress] Φ = 0. 7 (25 k. W) Instantaneous (left) and mean (right) temperature (a) and OH mass fraction (b). + 10 o 15 < 40 > 60 m m m m m Temperature (top) and O 2 mole fraction (bottom) radial profiles Pressure signal in the plenum and in the chamber Axial velocity profiles Trapped-vortex approach for syngas combustion in gas turbines The trapped vortex technology offers several advantages as gas turbines burner and the systems experimented so far have limited mainly this technology at the pilot part of the whole burner. Aim of the work was to design a combustion chamber completely based on that principle, investigating the possibility to establish a MILD combustion regime, in case of syngas as fuel. The simulations, performed with the ANSYS-FLUENT code, were carried out according to a steady RANS approach. The models adopted for chemical reactions and radiation are the EDC , in conjunction with a reduced mechanism and the P 1, respectively. NOx were calculated in post-processing. In order to save computational resources, the simulations were conducted only on one sector of the whole prototype, imposing a periodicity condition on side walls. A structured hexahedral grid, with a total number of about 2 million cells, was generated. Reynolds 35000 -swirl number 0. 6 EXP * 6 Direct Numerical Simulation (DNS) of a turbulent premixed slot CH 4/H 2 -Air flame at Re=2586 and equivalence ratio of 0. 7. Isosurfaces of x-velocity at -/+ 5 m/s and temperature snapshot. The DNS is performed for studying the effects of H 2 on methane flames and obtaining an ENEA turbulent flame database for Large Eddy Simulation model validation. Temperature field(K)
- Slides: 1