Computational Transport Phenomena Laboratory
S.L. Yılmaz Department of Mechanical
Engineering and Material Science
Ph.D. ProposalDecember 5th 2007
APPLICATION OF RANS/PDF & LES/FDF METHODS TO PREDICTION OF PREMIXED TURBULENT FLAMES
Outline•Objectives summary•PDF/FDF methodologies•Flame configurations•Preliminary Results•Scalable parallelization•Tasks summary
Concluding remarks [Drozda 2005]• Sandia Flame DSandia Flame D and and Sydney/Sandia bluff-body stabilized flamesSydney/Sandia bluff-body stabilized flames are are
simulated via the LES/SFMDFsimulated via the LES/SFMDF• Transport equation for the FDF is solved via the Transport equation for the FDF is solved via the hybrid Eulerian-hybrid Eulerian-
Lagrangian methodLagrangian method. . • MKEVMKEV and and SmagorinskySmagorinsky models are considered for the SGS stresses and models are considered for the SGS stresses and
fluxes.fluxes.• Flamelet chemistry modelFlamelet chemistry model relates the thermo-chemical variables to the relates the thermo-chemical variables to the
mixture fraction.mixture fraction.
Future/Ongoing workFuture/Ongoing work• Prediction of non-equilibrium flames via scalar FMDF Prediction of non-equilibrium flames via scalar FMDF
and ISAT.and ISAT.• Prediction of turbulent flames via joint velocity-scalar Prediction of turbulent flames via joint velocity-scalar
FDF and FMDF.FDF and FMDF.• Optimization of the solver to reduce computational
requirements.
Objectives• Extend the boundaries of two novel
methodologies for prediction of turbulent flames,
• Predict a bluff-body burner with a RAS/PDF methodology,
• Predict a pilot-stabilized burner with a LES/FDF methodology,
• Employ non-equilibrium kinetics,• Develop scalable LES/FDF
computational code.
DNS, LES, and RAS
Large Eddy Simulation
(LES)
Reynolds Average
Simulation (RAS)
Direct Direct Numerical Numerical Simulation
(DNS)(DNS)
Starting Equations
Averaging versus Filtering• Statistical description: Turbulent fields are
Random Signals in time and space. • RANS: Define one-point, one-time joint-PDF
• LES: Define spatial low-pass filter
Transported PDF Methods• Solve the one-point one-time Joint PDF
directly• Closed quantities
• Unclosed quantities
PDF Transport Equation• From full-set of NS equations:
• Solves for
LHS closed, RHS needs model
RANS Connection• Ensemble-mean equations derived from JPDF
transport equation:
• Define Favre averages:
Filtered Density Function (FDF)• Formal definition:
• Satisfying the properties of PDF
Transport of FDF• Identically to PDF transport:
Scalar FDF• Transport equation of the marginal FDF of
scalars
Lagrangian PDF Models• Closure via Stochastic Differential Equations • e.g. The Velocity-Scalar-Frequency SDEs
Modeled PDF Transport Equation• Corresponding Fokker-Planck equation
Lagrangian FDF Model• SDEs for closure
• Fokker-Planck equation
Simulation Procedure• Monte-Carlo simulation of the SDEs• Finite Difference solution of Eulerian Filtered
or Averaged equations
Bluff Body Lean Premixed Flame• Pan (1991);
Velocity measurements
• Nandula (2007); Temperature, species concentrations measurements
• Re = 43,400• Φ = 0.59
Simulation Details• Axisymmetric, adiabatic walls, uniform inlet.
Comparison of Mean Velocities
Mean and RMS Velocities
Temperature and Major Species
Temperature and Minor Species
Premixed Bunsen Burner• Y.C. Chen 1996. Velocity, temperature,
species measurements.• Three Flames of Re = 24K, 40K and 52K• Φ = 1.0
Cold Flow Computations• Using only FD solver with MKEV turbulence
model
Parallelization
Monte Carlo solver consumes 95% of total time.
Load Balancing Problem
Region of High Activity
Region of Low Activity
Stiff Term
Load Balancing Problem
Uniform Partitions
Balanced Partitions! (with METIS)
3D Arbitrary Partitions
Task Summary• Modify RAS/PDF computational code for
Bluff Body configuration• Perform RAS/PDF simulation of the Bluff
Body• Implement detailed kinetics into LES/FDF
solver• Implement scalable FDF computational code• Perform LES/FDF simulation of the Bunsen
Burner
Also thanks to, • CTPL members!• Cornell Combustion• NETL• PSC
THANK YOU.