SPEECH COMBUSTION Helsinki FINAL - prace

monkeyresultMécanique

22 févr. 2014 (il y a 3 années et 3 mois)

65 vue(s)

What we will learn on turbulent combustion
from simulations on the Exascale level
Thierry POINSOT
Université de Toulouse, CNRS
and CERFACS
2
COMBUSTION OVERVIEW
Just remember two equations:
3
ENERGY ON EARTH TODAY =
COMBUSTION
ENERGY ON EARTH = COMBUSTION
COMBUSTION IS PRODUCING MORE THAN 90 PERCENT OF THE
ENERGY TODAY. THIS WILL DECREASE... BUT NOT TOMORROW
4
COMBUSTION OVERVIEW
Just remember two equations:
5
ENERGY ON EARTH TODAY =
COMBUSTION
ENERGY ON EARTH TOMORROW =
COMBUSTION
CLIMATE CHANGE AND
ENERGY MARKET: 2010/2030

TO CONTROL CLIMATE CHANGE,
RENEWABLE ENERGIES MUST INCREASE
FASTER THAN ALL OTHER SOURCES

BUT THE GLOBAL DEMAND FOR ENERGY
ALSO GROWS (typically 2.6%) !

THE ENERGY PRODUCTION BY
COMBUSTION MUST ALSO INCREASE

COMBUSTION SCIENCE MUST
ALLOW THIS WITHOUT
INCREASING EMISSIONS, WASTING
FOSSIL FUELS OR MAKING
CLIMATE CHANGE WORSE (!...)
6
COMBUSTION IS ALSO DANGEROUS
7
In this global energy strategy,
gas turbines
have a special role:
8
The gas turbine market grows. The regulations
become tougher => Optimization is mandatory !
1/ No other way to propel aircraft
2/ Highly efficient (60%), cheap,
flexible (easy to start/stop)
systems to produce electricity
(Dec 2010 in France: no sun and
no wind -> for each windmill
plant, you need a gas turbine)
OPTIMIZING SOMETHING WHICH HAS
BEEN HERE FOR 70 YEARS IS DIFFICULT

Compromises (efficiency,
pollution, noise, stability, cost)
are difficult to find and
experimental costs too large to
test all possible designs
9
Sir F. Whittle
W2 engine

Simulation has become essential: gas
turbine companies now perform
advanced CFD, including Large Eddy
Simulation.

Since these simulations must optimize
complete real systems, multiphysics
must be integrated (see ASCI CITS at
Stanford)
Copyright Schlüter et al - Stanford CTR
!
WHAT IS A GAS TURBINE ?
10
Stanford Asci CITS project
NEED TO WORK ON THE
COMBUSTION CHAMBER ITSELF
11
AIR FROM
COMPRESSOR
TO TURBINE
TURBINE AXIS
SOME TERMINOLOGY:
12
Full combustion chamber
(10 to 24 ‘sectors’ or
‘burners’)
One sector :
FOR EACH SECTOR:
A SWIRLER AND A HPS
13
LPP injector
!
A swirler (one per
sector)
A HPS (High Pressure Stator)
TWO THINGS A «GOOD» COMBUSTION
CHAMBER SHOULD DO:

1 - NOT BURN THE TURBINE
BLADES OR THE HIGH
PRESSURE STATOR (HPS)
--> NOT SO EASY TO DO OR TO
PREDICT (requires
COMBUSTION + HEAT
TRANSFER + RADIATION)
14
High Pressure stator
2300 K
1700 K
The HPS can not survive
without cooling
A difficult choice: how warm
can the HPS function ?
15
From combustion
chamber
When the HPS temperature goes up:

the engine efficiency goes up (can increase
pressures and temperatures in the combustor)

lifetime of HPS goes down: increasing HPS
temperature by 25 K means 50 percent smaller
lifetime

Need to be precise ! (25 K ?)
TWO THINGS A «GOOD» COMBUSTION
CHAMBER SHOULD DO:

1 - NOT BURN THE TURBINE BLADES
--> NOT SO EASY (COMBUSTION +
HEAT TRANSFER)

2 - BE «STABLE»
--> DIFFICULT (COMBUSTION + HEAT
TRANSFER + ACOUSTICS)
16
WHAT IS A COMBUSTION INSTABILITY ?
17
STABLE
UNSTABLE
AIR
AIR
CH4
CH4
Isocontour of
methane
Experimentally:
18
STABLE
UNSTABLE
Berkeley backward
facing step
2000 im/s
WHEN A BURNER BECOMES UNSTABLE:
19
We want to predict this before it happens...
Today we cant -> this is a major danger for most
companies building gas turbines worldwide
COMBUSTION: MULTISCALE -
MULTIPHYSICS
20
Contrails
Engines
!
Heat transfer, radiation, flow, turbulence,
chemistry, fatigue, vibration, acoustics
COMBUSTION: MULTISCALE -
MULTIPHYSICS
21
Within the combustion chamber:
nanoseconds and nanometers
1 cm
1 cm
Fields of density in a H2-O2 engine
COMBUSTION: MULTISCALE -
MULTIPHYSICS
22
Outside the engine: miles and days !
2 miles
AND AN ADDITIONAL DIMENSION:

THE COMPUTATION OF EXISTING
COMBUSTION SYSTEMS IS NOT OUR
ONLY OBJECTIVE: WE MUST COMPUTE
FUTURE DESIGNS AND OPTIMIZE THEM IN
TERMS OF IMPACT ON OUR SOCIETY

IT IS NOT ABOUT COMPUTING ONE
COMBUSTOR, IT IS ABOUT EXPLORING A
MULTIDIMENSIONAL SPACE OF ALL
POSSIBLE COMBUSTOR DESIGNS
23
WHICH EQUATIONS ?

The Navier Stokes equations (5 + N
unknowns: density, velocities and energy, N
species). Partial differential equations ->
non local, intense communication required

Kinetics : N = 10 to 300 species reacting
through 3000 reactions (everything local)

Heat transfer through the walls, radiation,
noise, soot

All these flows are turbulent
24
WHAT DO WE COMPUTE ?

FINITE VOLUME CODES USING DOMAIN
DECOMPOSITION AND MPI.

SOLVE THE UNSTEADY NAVIER-STOKES
EQUATIONS IN A MIXTURE OF GASES
WITH REALISTIC CHEMISTRY

Typically 100 Mcells with 100 variables at
each cell (3 velocities, density, energy + 5
to 90 species) over 1000000 time steps. ->
10^16 unknowns
25
WHAT ABOUT PARALLELISM ?

THE NAVIER STOKES EQUATIONS ARE
‘EMBARRASINGLY
DIFFICULT
’ TO
PARALLELIZE -> THIS PROBLEM HAS
BEEN IDENTIFIED AND IS REMODELING
OUR COMMUNITY IN LARGER
COLLABORATIVE TEAMS.

THIS IS NOT A ‘ONE-PROFESSOR ONE-
CODE’ SHOW ANY MORE BECAUSE THE
CODES ARE USED FOR INDUSTRY
APPLICATIONS ON A DAILY BASIS
26
27
Example  of  organiza0on:  the  AVBP  ‘club’
CERFACS
Institut Français du Pétrole
Laboratories:
IMFT  (Toulouse)
CORIA  (Rouen)
EM2C  (Centrale  Paris)
IRPHE  (Marseille)
TUM  (Munich)
TU  Eindhoven
TWENTE
MADRID
ONERA
Univ.  Pau
Models
Code
Users:
SNECMA  (Villaroche)
TURBOMECA  (Bordes)
SNECMA  (Vernon)
SPS  (Bordeaux)
RENAULT
PSA
ALSTOM
ANSALDO
SIEMENS
FERRARI
AIR  LIQUIDE
GAZ  DE  FRANCE

Code
Needs
Computer  companies  +
Compu0ng  centers:  CEA,  Barcelona,  GENCI
INCITE
(Argonne)
STANFORD
16000
14000
12000
10000
8000
6000
4000
2000
0
Equivalent Performance
16000
14000
12000
10000
8000
6000
4000
2000
0
Cores

Ideal

Thomas Watson, Bluegene /L
(1)

ARNL , Sicortex
(1)

CRAY, CRAY XT5
(2)
CERFACS, BlueGene /L
(3)


CINES, GENCI, SGI Altix ICE
(4)

ARNL, Bluegene P, INCITE
(5)

(1) 40M cells case - 1 step chemistry

(2) 18M cells case - 1 step chemistry

(3) 75M cells case - 1 step chemistry

(4) 29M cells case - 7 step chemistry

(5) 93M cells case - 1 step chemistry


AVBP Strong scaling examples
28
V. Moureau, P. Domingo, L. Vervisch, D. Veynante
DNS analysis of a Re  40,000 swirl burner
020484096614481921024012288Number of cores002048204840964096614461448192819210240102401228812288Scale-uplinearYALES2YALES2 scale-up on Babel @ IDRIS (Blue Gene/P)Up to 12288 cores and 2.6 billion tetrahedrons2.6B tets878M tets329M tets41M tets14M tets
2008
V. Moureau
CORIA-CNRS
2009-2010
h=100 µm
YALES2 solver
(CORIA Rouen)
Juelich Workshop 2010
WHAT DO WE NEED TO DO ?
30
1/ Be more PRECISE: get SMALLER !
Most combustion phenomena take
place at very small scales. Our
present capacities do not allow us to
resolve these scales:
1 cm
WHAT DOES SOCIETY NEED FROM US ?
31
2/ Be more GLOBAL: get LARGER
Need to compute the full engine and
sometimes even more
1 meter
10 m
WHAT DOES SOCIETY NEED FROM US ?
32
3/ OPTIMIZE: propose better designs
1 meter
How many holes? where ? shape of the chamber to:

minimize pollutant and instabilities

minimize weight and consumption

homogeneize temperature field in outlet plane
OUTLINE:

GET SMALLER AND MORE PRECISE:
single codes with very large meshes and
CPU time

GET GLOBAL: multiple codes coupled
on the same machine

OPTIMIZE: repeat simulations and
couple them with optimization methods
33
OUTLINE:

GET SMALLER AND MORE PRECISE:
single codes with very large meshes and
CPU time: The PRECCINSTA (European
FP6 program) test case
-> Two ‘typical’ combustion codes:

AVBP (CERFACS/Institut Français du
pétrole Energies Nouvelles)

YALES2 (CORIA Rouen)
34
35

THE OLD DAYS (2005): THE FIRST LES OF SWIRLED BURNERS
3 millions cells
Roux, Lartigue, Poinsot, Meier and Bérat
Comb and Flame 141, 2005
36
FULLY COMPUTED: NO QUESTIONABLE
BOUNDARY !
37
COLD FLOW - PRECCINSTA
38
All LES are compared with measurements DLR (LDA):

Velocity profiles are compared at five stations along the burner.

Comparison for axial, tangential and radial velocities (mean and RMS)
X= 5 mm
X= 1,5 mm
X= 25 mm
X= 15 mm
X= 35 mm
FROM
SWIRLER
TO OUTLET
U
W
Comparison of mean velocity fields
39
-40
-20
0
20
40
Distance from axis [mm]
15
10
5
0
-40
-20
0
20
40
15
10
5
0
-40
-20
0
20
40
15
10
5
0
-40
-20
0
20
40
15
10
5
0
-40
-20
0
20
40
15
10
5
0
x=1.5 mm
x=5 mm
x=15 mm
x=25 mm
x=35 mm
Uxp
Profiles: Red solid: CDP - Black solid: TTGC - Black dotted: TTGC_SSS - Circles: Exp.
RMS velocity profiles: Stanford code (red), CERFACS code (black)
and DLR experiments (symbols)
PRECCINSTA:

First computed with LES in 2004: 1
Million cells (AVBP)

Repeated in 2007 with 10 Mcells (AVBP)

In 2009/2010, repeated with 100 Mcells,
500 Mcells, 2 billion and 21 billion cells
(YALES)
40
41
DLR PRECCINSTA BURNER, Experiments: W. Meier et al., Combust. Flame, 150(1/2):2–26, 2007
Visualisa'on  of  vor'ces  using  the  Q  criterion  in  
PRECCINSTA  (Dr  Moureau,  CORIA)
IS THIS ENOUGH ?:

At 20 billion cells, we are almost reaching
what we need in terms of resolution for the
large flow structures but near walls and
within the flame front, this is not enough

This is a single burner at atmospheric
pressure with gaseous fuel (no liquid
phase). Real combustors have 16 to 24
burners, working at 20 to 100 bars, with
liquid fuels. We still miss a 10^6 factor in
power
42
OUTLINE:

GET GLOBAL: multiple codes coupled
on the same machine. AVBP + other
solvers in a real helicopter engine
43
CONFIGURATION: Helicopter engine
45
Air
Kerosene
High Pressure Stator
NEED MULTIPHYSICS TOOLS:
46
Combustion LES
solver
AVBP
Conduction solver
AVTP
Radiation solver
PRISSMA
O-PALM
(open source coupling tool)
But  parallel  makes  it  difficult  when  it  
comes  to  coupling
47
+  Heat  transfer
PROCESSORS
Slow mechanisms are
computed fast and vice
versa… Unfortunate !
The synchronisation problem
48
Characteristic
time
Restitution
time
Heat conduction
Radiation
Combustion
Diffusion (energy, mass)
Phase change
Convection (natural, forced)
Chemistry
Acoustics
Temperature field
Hot spots produced by the chamber
impinging on the High Pressure Stator
11
TWO COUPLED SIMULATIONS
51
Combustion
Conduction
Radiation
Combustion
Conduction
A/ ‘TWO CODES’
SIMULATION:
B/’THREE CODES’
SIMULATION:
Temperature field on the HPS
14
Hot (stagnation point)
Hot (trailing edge, not cooled enough)
TEMPERATURES WITHIN BLADE:
53
200 K
Without radiation With radiation
At some places, the blade loses heat
through radiation towards the cooler walls
CONCLUSIONS

Combustion science needs more simulation power

But the power we need is almost there to perform one single
simulation.

Problems:

we need to perform thousands of simulations for optimization of
combustor shapes, regimes and fuels

we do not need to run only
one
code on a parallel machine but
multiple
simultaneous codes -> new coupling software needed
and developed as open source codes (O’PALM)
54