Gas turbine combustor thermoacoustics

An introduction

Marco Zedda
Combustion Aerothermal Methods Tech Lead
Combustion and Turbines Subsystems
Civil Large Engines, Derby
Rolls-Royce plc
Problem statement

 Drive towards low emissions => lean burn combustion

 Lean AFRs => combustor is more likely to be prone to heat release fluctuations
and thermoacoustic problems
Types of combustion noise
 Different types of combustion noise:
• High Frequency Rumble (HFR): ~500Hz, mainly circumferential mode, can affect
integrity
• Low Frequency Rumble (LFR): ~200Hz, axial mode due to entropy waves (non-
uniformity of temperature field)
• Buzz: ~100Hz, axial mode affecting afterburners, can couple with LP shaft vibration
• Screech: ~2kHz, radial mode, affecting afterburners integrity through enhanced heat
transfer to liner
• Indirect noise: entropy waves hit choked nozzle introducing noise sources in turbines,
can be a problem because noise sources coming from other components are
reducing
 Combustion noise can be more than just annoying

Rumble Damaged RB211 DLE discharge nozzle

TA at work
 Approaches to TA instabilities
 Design it out:
• Identify TA-desirable design features – hard to do
• Carry forward only designs showing acceptable level of
pressure fluctuations – problem: TA instabilities may manifest
themselves only at relatively late stages of the design Fuel injector
 Damp it:
• Introduce dampers:
• frequency targeted dampers (e.g. Helmholtz resonators)
work well only on a narrow frequency band
• broadband dampers (e.g. viscous dampers) may not
reduce the pressure fluctuation enough
• Volume can be at a premium – solution is inherently more
Helmholtz resonator
suitable for industrial gas turbine combustors
• Cooling is a challenge: both types rely on introducing flow
fluctuations at the wall
 Control it out:
• Apply active control: impose frequency and amplitude
dependent fluctuations of the fuel – reliability and durability of
Control system
the actuators are issues
 TA modelling
 The problem has to be simplified to make it tractable
 Starting point: Navier-Stokes equations
 Assumptions:
• Thin annular combustors or long cannular combustors with no significant radial
fluctuations => modes are azimuthal, axial or a combination thereof
• Linear regime: looking for susceptibility of fluctuations to become unstable
• Viscous effects can be neglected
 Because of linearity:

 NS equations can be reduced to the inhomogeneous wave equation:

 Equations are usually resolved in the frequency domain

 Frequency is complex: imaginary part linked to the growth rate, real part is instability
frequency
Flame transfer function
 The inhomogeneous wave equation can be solved if a model of the heat release
fluctuations is available
 The Flame Transfer Function (FTF) is defined as the ratio of heat-release
perturbations to flow perturbations as a function of frequency:

 The flame transfer function can come from analytical models, experiments or CFD.
 A simple analytical model is

where t can be interpreted as the fuel convection time.

 If instead of a single time delay, a uniform distribution is assumed (t - Dt to t +

Dt):

 Such models can be fitted to the measured or simulated flame response at a range
of frequencies
Flame transfer function vs flame describing function

 The flame transfer function approach applies in the linear domain (i.e. stable vs unstable)
 As such, it does not allow calculating the amplitude of the fluctuations
 In order to capture non-linear effects, a flame describing function has to be available:

T  T( A,)
 Unstable modes will not grow indefinitely: the amplitude of fluctuations will stop when
offset by damping effects
 The resulting TA instability is called limit cycle
 In conventional rigs, measurements of TA instabilities (e.g. pressures, velocities, heat
release, etc) are usually associated to limit cycles, hence the flame transfer function
cannot be directly derived. The resulting amplitudes are referred to as saturation
amplitudes
 Using a FDF allows calculating the amplitude of oscillations
 However, to be of use in modelling, FDFs should be defined for a range of amplitudes
going from zero to the saturation value
How to model a combustor
 The system’s acoustics
is approximated by
using acoustic paths

 Large volumes have to

be modelled directly

 Acoustic boundary
conditions have to be
realistic (e.g. choked,
open ended, impedance
based, etc)
Typical results

Mode instability analysis

Visualisation of
azimuthal mode

Modeshape analysis
 How to obtain a FTF

 There is no “one size fit all” FTF

 FTFs are strongly dependent of geometry and operating conditions: thermoacoustics can
be very sensitive to even small changes in geometry/operating conditions (“cliff edge”
behaviour)
 Changes to the fuel injector are a classical cause for appearance of TA problems
 Analytical models can be useful to fit to experimental or numerical data, but are unlikely
to provide accurate guidance on the susceptibility to TA instabilities
 FTFs can be derived:
• Experimentally: by forcing a flame in the linear regime for a range of frequencies at
which the heat release and velocity fluctuations are measured
• Numerically: CFD can be used to simulate the flame forcing and derive the flame
response in a similar manner to experiments
How experiments can help
 Cold flow single sector: in-depth characterisation of steady/unsteady fuel injector

COST and REPRESENTATIVENESS

aerodynamics – no reaction, no acoustics
 Forced cold flow single sector: in-depth characterisation of unsteady aerodynamic
response to acoustic forcing – no reaction
 Single sector reactive (low to high P): characterisation of flame response to single
sector rig’s acoustics (possibly with advanced laser diagnostics for measuring
heat release and fuel placement) – limited representation of acoustics
 Single sector siren rig: characterisation of flame response to axial harmonic
forcing up to cruise pressure (possibly with advanced laser diagnostics) – siren rig
 Full annular: characterisation of flame response to actual combustor geometry up
to max pressure achievable by the rig for the combustor size – high Technology
Readiness Level (TRL) test required for verification of other combustion
parameters, but limited diagnostics can be applied. Need for restricted build

Siren rig layout

Combustor
Rotating
shaft
 Combustor Technology Development
TRL
Low order modelling & CFD
1/2

LP combustion, spray diagnostics, aerodynamics

and fuel stability.
3

2 sector sub-atmospheric & HP Single Sector

4

HP and sub-atmospheric annular & HP

multisector
5

E3E
Engine Demonstrators
6
ANTLE / POA EFE
 The challenge of siren rig testing
 Often TA problems surface at a late stage of the combustor verification process
 A low TRL filter is required
 Issues to be addressed for siren rig testing:
• Unsteady heat release difficult to measure for partially premixed and diffusion flames
• Response of flame to axial forcing may be unrepresentative of circumferential modes
• Pressure effects: engine MTO pressure are getting higher and higher
• Amplitude effects: siren has to be designed to be able to modulate the amplitude of the forcing to
remain in linear regime
• Contribution of potential coupling between entropy waves and HFR not accounted for before full
annular testing
• Even full annular testing may not be sufficiently representative of the forcing the combustor will
be subject to in an engine (e.g. forcing by the rotors)
• System effects: geometry upstream and downstream of the fuel injector may not be
representative of engine
• Rig effects: rig will have its own acoustic resonance, which may mask the injector response

Siren
position
 How CFD can help
 CFD can be used to derive the FTF
 Approach is based on forcing the CFD for a range of frequencies to calculate heat release and velocity
fluctuations

 Different numerical approaches are being pursued for derivation of FTF, all based on unsteady CFD:
• Turbulence modelling:
• Large Eddy Simulation (LES)
• Unsteady RANS (uRANS)
• Compressibility effects:
• Incompressible to assess “hydrodynamic” behaviour of flame
• Compressible to account for density changes with pressure
• Effect of fuel break up on time delay:
• Simplified modelling of prefilmer through Volume of Fluids (VoF)
• Estimate of time delay due to break up based on first principles
 The challenge of the numerical siren
1/2
 Issues:
• Impact of compressibility on flame response:
• Compressibility effects on density combine with combustion effects on density
• Flow inertia requires explicit modelling of fuel injector in reactive compressible
simulation
• Boundary conditions/extent of computational domain:
• Explicit modelling of fuel injector may be required to simulate impact of pressure
waves on wakes from swirlers
• Anechoic or impedance-based boundary conditions to be used for compressible
solution
• Time delay contribution of spray break up:
• For airblast injectors with long prefilmers, time to break up is comparable to
convective time from end of film to flame front
• Role of fuel passage dynamics:
• Instability can be affected by fuel passages not running full
• Fuel passages can respond to air pressure oscillations
• Flame response to different types of forcing (axial vs circumferential):
• HFR’s mode is azimuthal in annular combustors, whereas forcing is axial in
experimental and numerical sirens => flame response could be different
• Injector-to-injector interaction may be important
The challenge of the numerical siren
2/2
• Contribution of entropy waves:
• At the moment, 1D modelling assumes either perfect or no mixing of entropy non-
uniformities in acoustic calculations. Some reasonable level of mixing would be
required
• Turbulence modelling:
• Injector flows are better predicted by LES, however turn around time can be very
long => uRANS may be a more pragmatic solution, but predicted spectra will be
much less rich of peaks
• Multi-modal behaviours:
• injectors can show propensity to relax on different flow configurations depending
on operating conditions and geometry changes
• Acoustic modelling approach:
- FTF is a fundamentally 1D quantity linking unsteady heat release rate to unsteady
velocity, whilst acoustics is modelled in 2D (radial component is neglected in thin
annular combustors). However, flame response may be 2 - 3D
• Forcing:
- Broadband is most appropriate, but time consuming
- Harmonic can be misleading for wide range of frequencies
- Impulse perturbs the system but potentially in a non representative way
• Rumble and structures:
- Clear definition of acceptable levels of rumble
Deriving LFR TFs from forced CFD 18

• Fuel is forced by step change

• Gain and phase of the FTF and TTF is computed and supplied to LOTAN

Outlet T Heat release

time
time
LOTAN predictions – growth rate and frequencies

 T30=450K AFR=62 point predicted to be unstable, all the

other points predicted as stable
 Reasonable (qualitative) tie up between predicted growth
rates and measured amplitudes
Conclusion on flame forcing

 TA tests and simulations are highly challenging

 Experimental and numerical approaches based on derivation of FTF have been

demonstrated to be successful for prediction of TA instabilities of some gaseous
flames

 The matter is very much a research topic

 Alternative numerical approaches
 Industrial combustors can be affected by
high frequency transverse modes to which
low order modelling is not applicable
 Alternative approaches have been
developed, e.g.:
• Full blown compressible simulations
with acoustic boundary conditions:
- Computationally expensive (large
grids, small timesteps)
- Limited to the time domain => no
information about the range of
modes present
• Linear Euler approaches: 3D
- Can be run in the frequency
domain Compressible LES of a full annular
combustor (courtesy of CERFACS)
- Flame model has to come from
forced CFD or analytical
approach
 Effect of spray on TA
 TA instabilities are linked to the interaction between pressure and heat release
fluctuations. Heat release is a function of the local air to fuel ratio.
 So, the response of the spray to pressure fluctuations can have a significant role by
increasing the time delay of the FTF, especially for airblast atomisers
 Different sensitivities to spray effects on TA depending on the fuel injector design

Side View

air flow
prefilmer length

air flow
Engine vs rig effects on TA

 Single sector test rig with forcing can be useful if:

• Damping effects of the wall are not dominating
• Injector P/D (pitch-to-diameter ratio) in full annular
configuration is high => no sector to sector interaction
Single sector rig
• Wavelength of TA instability is larger than injector pitch Single sector rig

 Full annular test rig is high TRL but:

• Inlet acoustics has to be set up to be representative of HPC
outlet
• Pressure levels reached may not be engine representative Full annular rig
• No effects due to rotor forcing

 Engine is the real thing, but:

• Too late to fix design problems
• Too expensive for design iterations

Engine
Effects of hydrodynamic instabilities on TA
 Incompressible aerodynamic phenomena can be a cause of TA instabilities:
 Vortex shedding from fuel injectors

 Flow separations

 These problems can be spotted by aerodynamic testing or LES

 Periodic phenomena can lock in with acoustics only if the frequencies are consistent
(i.e. the same or harmonics)
Effect of external aerodynamics on TA
 Cavities external to the flametube can have a resonant effect
 Unsteady aerodynamic phenomena can lock in with characteristic frequencies

Full system aerodynamics

Fuel flow effects on TA
 Fuel pumps can induce fuel flow fluctuations leading to TA instabilities
 Mechanical vibration can introduce forcing to the fuel delivery (fluid-structure
interaction)

Fluid-structure interaction
Afterburner TA instabilities

 Afterburners can suffer from buzz and screech

 Buzz: low frequency, axial model linked to propagation of entropy waves. It can be
modelled in 1D/2D
 Screech: high frequency transverse mode, requires introduction of dampers, can only
be modelled in 3D
 Afterburner TA instabilities can be a show stopper

EJ200 engine cutaway

Summary
 TA instabilities pose a serious risk to engine development
 Low order modelling can be used to understand problems, but FTF/FDF
is required
 Deriving FTF from experiments is possible, but challenging
 Deriving FTF from CFD requires very careful modelling
 Some success has been shown using CFD for gaseous flames, FTF for
liquid fuelled combustors is more difficult to derive
 There are many different possible causes for TA instabilities
 Rich and lean burn combustors can be affected, but lean ones are more
susceptible to TA instabilities
 Afterburners suffer from TA instabilities as well
 The topic is focus of intense research