Interaction of Turbulence, Chemistry, and Radiation in Strained Nonpremixed Flames

1
Interaction of Turbulence, Chemistry, and
Radiation in Strained Nonpremixed Flames

• Chun Sang Yoo, Hong G. Im
• Department of Mechanical Engineering
• University of Michigan
• Yi Wang, Arnaud Trouvé
• Department of Fire Protection Engineering
• University of Maryland
• Sponsored by the DOE SciDAC Program
• http//purl.org/net/tstc

2
Outline of Presentation

• Introduction
• Role of DNS in Combustion Science (a brief
  version)
• Overview Terascale High-Fidelity Simulations of
  Turbulent Combustion with Detailed Chemistry
  (TSTC)
• Research Highlights (work led by U. Michigan)
• Computational Improved Navier-Stokes
  Characteristic Boundary Conditions (NSCBC)
• Science Counterflow Diffusion Flames with Soot
  and Radiation Models
• Ongoing/Future Work
• More TSTC Research Highlights Poster
  Session
• WED21 Trouvé and Wang (Maryland)
• WED22 Rutland and Wang (Wisconsin)

3
DNS A Computational Microscope

• A diagnostic tool to study the fundamental
  physics of turbulent reacting flows
• Full access to temporally/spatially resolved
  information.
• Allows identification of key paths for relevant
  phenomena, such as turbulence-chemistry
  interaction
• A benchmark tool to develop and validate physical
  submodels used in macro-scale simulations of
  engineering-level systems (LES with
  embedded DNS)

Formation of edge flames in a turbulent
counterflow
Physical Models
Engineering-level CFD Codes
DNS
A KIVA-3V engine simulation
4
Terascale High-Fidelity Simulations of Turbulent
Combustion with Detailed Chemistry (TSTC)
http//purl.org/net/tstc
Software architecture
Numerical algorithms
SciDAC CFRFS
. S3D0 F90 MPP 3D . S3D1 GrACE-based . S3D2
CCA-compliant
. IMEX ARK . IBM . AMR
Physical models
SciDAC CCA
. Thermal radiation . Soot formation . Spray
dynamics
MPP S3D
SciDAC CMCS SDM
Hong G. Im, University of Michigan Arnaud Trouvé,
University of Maryland Chris Rutland, University
of Wisconsin Jackie Chen, Sandia National Labs
Post-processors In-situ visualization Feature
tracking
5
S3D MPP DNS Code

• S3D code characteristics
• Compressible reacting Navier-Stokes, total
  energy, species equations
• Fortran 90, MPI domain decomposition
• Highly scalable and portable on all modern
  architectures
• Numerical algorithms
• 8th order non-dissipative spatial finite
  difference, 10th order dealiasing filter
• 4th order explicit RK integrator with error
  monitoring
• Additive 4th order RK integrator for stiff
  chemistry
• Improved boundary conditions to allow transverse
  velocity, flame passage through boundary, or
  solid walls

• Physical models
• Lewis number, mixture averaged, or
  multi-component transport
• Detailed gas-phase chemical kinetics
• (Chemkin-compatible)
• All thermodynamic properties are functions of T,
  p, and Yi
• Radiative heat transfer (discrete ordinate /
  discrete transfer method)
• Soot formation
• Lagrangian spray model
• Recent Contributions from the SciDAC TSTC
  Project

6
Characteristic Boundary Conditions

• A pre-requisite issue for high-quality
  turbulent combustion DNS
• Historical Development
• General nonreflecting outflow boundary conditions
• (Engquist and Majda 1977, Hedstrom 1979)
• Pressure damping for Navier-Stokes equations
• (Rudy Strikwerda 1980, 1981)
• Inviscid characteristic theory for Euler
  equations (Thompson 1987,1990)
• Navier-Stokes characteristic boundary conditions
  (NSCBC) - Viscous conditions (Poinsot Lele
  1992)
• Multi-component reacting flows (Baum et al. 1994)

• Applications to turbulent and reacting flows have
  revealed problems of spurious pressure waves,
  numerical instabilities.
• Reaction source terms (Sutherland Kennedy 2003)

7
Characteristic Waves

• Li characteristic wave with ?i
• (wave velocities, ?1 (u?c), ?2?3?4u, ?5
  (uc))

Computational domain
inflow
outflow
flow
8
Locally One-Dimensional Inviscid (LODI) Relations

• Neglecing transverse convection, viscous, and
  reactive terms
• The incoming Lis can be determined at both
  inflow and outflow boundaries using LODI
  relations
• Hard inflow boundary conditions yield large
  spurious wave reflections nonreflecting
  conditions are needed

• Inflow boundary

• Outflow boundary

9
Generalized NSCBC for Transverse, Viscous,
Reacting Flows

• LODI relations are no longer valid transverse,
  viscous, reaction terms must be considered in Lis

Outflow boundary conditions (at x lx)
Conventional LODI
Improved BC
Spatial
Temporal
Low-Ma asymptotic expansion yields
10
Test 1 Vortex-Convection

• Incompressible inviscid vortex
• Conditions
• Three different boundary conditions
• BC1 conventional LODI with
• BC2 keep all the transverse terms (a 0.0)
• BC3 improved BC with pressure and transverse
  damping (a M 0.05)

11
Vorticity and Pressure
??
LODI
BC2 (a 0.0)
Improved BC (a 0.05)
P
12
Velocities
u
LODI
BC2 (a 0.0)
Improved BC (a 0.05)
v
13
Temporal Pressure Variation

• Examine how the solution approaches the steady
  state
• The L2-norm

Temporal variations of the L2-norms of pressure
difference
14
Test 2 Ignition H2-O2 Mixture

• Stoichiometric H2-O2 mixture diluted with 50 N2
  by volume
• 2mm ? 2mm (200 ? 200 grid points)
• Initial temperature and pressure, 300K and 1atm
• Initial Gaussian temperature peak

• Three test cases
• Case A conventional LODI
• Case B include source terms in incoming Lis
  (Sutherland Kennedy 2003)
• Case C improved BC with a 0.125 (scaling
  analysis)

15
Temperature and HO2
T
Case A (LODI)
Case B (Sutherland Kennedy)
Case C (Improved BC)
YHO2
16
Test 3 Poiseuille Flow (Isothermal Wall)

• Viscous terms must be considered
• Test cases
• Case A conventional LODI B.C. with ?1,exact
• Case B including only pressure damping term (a
  0.0)
• Case C improved B.C. with a 0.1

• The pressure level of Case A is increased because
  ?1,exact does not cancel out all the viscous and
  heat flux effect
• The velocity at the outflow boundary in Case B is
  not accurate transverse damping term is needed

Temporal variation of pressure
17
Test 4 Turbulent Reacting Counterflow

• Transverse terms cannot be ignored
• a 0.01
• Use the steady laminar H2?air nonpremixed
  counterflow flame as the initial condition
• Turbulence inflow condition
• Velocity fluctuations are superimposed on the
  mean inlet velocities.
• Homogenous turbulence

(a) temperature
(b) vorticity
18
Strained Nonpremixed Flames with Soot and
Radiation

• Motivation
• Predictive tools for pollutant formation (soot,
  NOx)
• Thermal radiation plays an important role, but
  has not been incorporated in high-fidelity
  simulations
• Need better understanding of interaction between
  flow, chemistry, and heat transfer
• Objectives
• To develop high-fidelity DNS capabilities with
  advanced physical submodels for soot and
  radiation
• Validate and assess the impact of the advanced
  physical models in a canonical configuration
  (flame-vortex)
• Perform laboratory-scale simulations to answer
  science questions on turbulence-chemistry-radiatio
  n interaction (future work)

19
Radiation Models in S3D

• Based on gray gas assumption
• Radiative heat flux
• Optically thin model (OTM)
• Discrete ordinate method (DOM)
• RTE solved in n discrete directions (ordinates)
• Sn approx. ? number of equations n(n2)/2 (2-D)
• S2 4 eqs., and S4 12 eqs.
• Discrete transfer method (DTM)
• RTE solved for n rays (ray-tracing)

20
Performance of DOM/DTM

• MPI Scalability

• Total radiative power

• DOM is found to be overall superior for the
  desired accuracy.

Relative error
21
Soot Model (Two Equation Model)

• A semi-empirical two-equation model based on a
  flamelet approach (Young and Moss, 1995)
• Soot number density
• Soot volume fraction

• Parameters

22
Computational Configuration

• Calculation procedure
• Generate 1-D diffusion flame profile (Oppdif)
• Establish steady diffusion flame in counterflow
• Superimpose initial vortex pairs
• Velocity profile for a vortex

23
Parameters

• Three different vortex strength cases
• Weak vortex flame and soot are not extinguished
• Medium vortex extinguishes soot only
• Strong vortex extinguishes both flame and soot

24
Weak vs. Strong Vortex Cases
Temperature
Nsoot
fv
Vorticity
Case A
Case B
Case C
25
Integrated Nsoot and fv (Case B)
Volume-integrated Nsoot and flame volume
Volume-integrated fv in different temperature
regions

• Soot number density depends strongly on the
  high-temperature flame volume
• Soot volume fraction increases by surface growth
  at low temperature, fuel-rich regions

26
Effects of the Vortex Strength
Comparison of integrated fv for Cases A-C
Comparison of integrated Nsoot for Cases A-C

• As vortex strength increases, more soot particles
  are convected into fuel rich zone
• Case A fv is more directly affected by the soot
  nucleation.
• Case C fv does not change much even the the soot
  nucleation (Nsoot) is turned off.

27
Comparison of Radiation Models

• Radiative heat loss
• During transient period, OTM overpredicts the
  radiative heat loss by up to a factor of two
  compared to DOM
• Fidelity of radiation model is important in DNS

Total radiative heat loss with OTM and DOM for
Case B
28
Ongoing/Future Work

• Terascale Computing 3D Turbulent Nonpremixed
  Counterflow
• Flames with Radiation, Soot, and Water Spray
• Integration of all the developed physical
  submodels
• Test bench for numerical algorithms boundary
  conditions, acoustic speed reduction (ASR)
• Science issue partial/total extinction and
  pollutant formation due to water spray
  interaction
• Further To-Do List
• Computational Development
• Immersed boundary method
• Adaptive mesh refinement
• Chemistry reduction strategies
• Physical Models
• Detailed soot model
• Radiation model (spectral)
• Catalytic surface reaction

DOE INCITE Project 3D DNS of turbulent
nonpremixed jet flame, J. H. Chen et al. Sandia
National Labs

• Enabling Technologies
• Data-mining and visualization
• Object-oriented code architecture for efficient
  management

29
Acknowledgments

• SciDAC TSTC Program
• Hong G. Im (Michigan)
• Chunsang Yoo, Ramanan Sankaran (SNL)
• Christopher J. Rutland (Wisconsin)
• Yunliang Wang
• Arnaud Trouvé (Maryland)
• Yi Wang
• Jacqueline H. Chen (Sandia National Laboratories)
• Scott Mason, Chris Kennedy, James Sutherland,
  Evatt Hawkes
• Pittsburgh Supercomputing Center
• Ravishankar Subramanya, Raghurama Reddy
• DOE Computing Resources
• National Energy Research Scientific Computing
  Center
• Oak Ridge National Laboratory
• Pacific Northwest National Laboratory
• University of Oregon (the Tau Project)
• Sameer Shende, Allen Malony