Active Control of Combustion Instabilities

1
Active Control of Combustion Instabilities

• Aimee Morgans
• RAEng/EPSRC Research Fellow
• Cambridge University

2
Overview

• Personal research background
• Background to Active Control of Combustion
  Instabilities
• Research
• - Model based control (experimental)
• - Model-based control for annular combustors
• - Adaptive control
• 4) Conclusions and Future work

3
Overview

• Personal research background
• Background to Active Control of Combustion
  Instabilities
• Research
• - Model based control (experimental)
• - Model-based control for annular combustors
• - Adaptive control
• 4) Conclusions and Future work

4
Personal background

• BA/MEng at Cambridge 1996-2000
• Engineering (specialised in Fluid Mechanics)
• MEng project on Active Flow Control
• PhD at Cambridge 2000-2003
• Aeroacoustics (Transonic Helicopter Noise,
  supervised by Ann Dowling, spent 1 term at
  Caltech)
• RAEng/EPSRC Fellowship 2004-present
• Cambridge, Active Control of Combustion
  Instabilities
• Currently supervising 3 PhD students

Aeroacoustics
Control of Combustion Instabilities
Active Flow Control
5
Overview

• Personal research background
• Background to Active Control of Combustion
  Instabilities
• Research
• - Model based control (experimental)
• - Model-based control for annular combustors
• - Adaptive control
• 4) Conclusions and Future work

6
Combustion Instabilities
unsteady heat release generates acoustic waves

• Caused by a coupling between unsteady heat
  release and acoustic waves

these reflect from the combustor boundaries
and further perturb the heat release rate
7
Combustion Instabilities
unsteady heat release generates acoustic waves

• Caused by a coupling between unsteady heat
  release and acoustic waves

these reflect from the combustor boundaries
and further perturb the heat release rate
8
Context

• Observed by Lord Rayleigh (over 100 years ago)
  who suggested instability occurs when heat
  release in phase with pressure.
• Current interest driven by susceptibility of low
  NOx gas turbines (lean premixed combustion).

Damaged gas turbine transition piece
9
Some insight
Rate of change of Rayleigh source losses
across total energy boundaries
10
More insight
11
Controlling Combustion Instabilities

• Passive control e.g. Helmholtz resonators,
  geometry changes.
• Active control input to system based on sensor
  measurement.

Active control first demonstrated in 80s,
loudspeaker actuation, pressure or heat release
sensing. Has been demonstrated at full-scale
using pressure sensing, fuel actuation. Control
approaches still tend to be trial-and-error.

combustion system (unstable)
actuator
sensor
-
controller
12
Aims of Fellowship

combustion system (unstable)
actuator
sensor
-
controller

• Develop more systematic approaches to active
  control
• - Robust model-based control
• - based on specific models
• - robust to small changes in operating
  conditions
• - Adaptive control (collaborating with Adaptive
  Control Group, MIT)
• - applicable to wide class of combustion
  systems
• - automatically tracks changes in plant
  operating conditions

13
Overview

• Personal research background
• Background to Active Control of Combustion
  Instabilities
• Research
• - Model based control (experimental)
• - Model-based control for annular combustors
• - Adaptive control
• 4) Conclusions and Future work

14
The combustion rig
Experimental work. Pressure sensing, fuel flow
rate actuation
cooling air
combustor (quartz tube)
fuel in
fuel valve
ethylene cylinder
15
The unstable system
Instability occurs at different equivalence
ratios depending on fuel back pressure, acoustic
damping etc. Same modes always present.
16
Controller design

• SISO control Nyquist techniques
• 1-pair of open loop unstable poles ? 2
  encirclements of -1 point needed on Nyquist
  diagram

17
Implementing control

• Control across a range of operating conditions
• Stabilising even when operated from within limit
  cycle

18
Overview

• Personal research background
• Background to Active Control of Combustion
  Instabilities
• Research
• - Model based control (experimental)
• - Model-based control for annular combustors
• - Adaptive control
• 4) Conclusions and Future work

19
Model-based control for annular combustors

• Many gas turbines have annular combustors.
• Single-sensor single-actuator approach no longer
  sufficient

20
Model-based control for annular combustors

• Computational model (LOTAN - Simon Stow) used to
  investigate control strategies ? linear
  acoustics, non-linear flame model
• Two approaches
• 1) Control circumferential modes
• (Axi-symmetry decoupled modes.
• To resolve up to n 2, at least 5
• sensors/actuators needed)
• 2) Control matrix of actuator-to-sensor
• transfer functions
• (Always MIMO control, H8 loop shaping used,
• need 2 sensors, at least two actuators)

21
Model-based control for annular combustors

• Modal control results, axisymmetric combustor
  with instability in modes n 1), 3 sensors and
  3 actuators used

Control on at t 0.8s, off at 1.5s
Control of more complex systems too e.g multiple
instability modes, non-axisymmetry due to burner
variations
22
Overview

• Personal research background
• Background to Active Control of Combustion
  Instabilities
• Research
• - Model based control (experimental)
• - Model-based control for annular combustors
• - Adaptive control
• 4) Conclusions and Future work

23
Adaptive control - background

• Adaptive control controller parameters vary in
  time - able to track large changes in operating
  condition
• E.g. k, a, b vary in time
• Self-tuning regulator (STR) by Cambridge MIT
  one algorithm for a whole class of (longitudinal)
  combustors
• Algorithm is derived using Lyapunov function
  only possible if the open loop system is minimum
  phase (no RHP zeros)

24
Adaptive control RHP zeros?
Back to the longitudinal combustor
Zeros of F(s) are important. When is F(s) likely
to have RHP zeros?
25
F(s)

• F(s)pref(s)/Q(s) modelled
• using 1-Dwave approach
• When entropy wave neglected,
• get (1Rde-a2s) factor in F(s)
• RHP zeros occur if Rd gt1
• Open end theoretical
• Rd -(1M2)/(1-M2).
• E.g. Buzz rig reflection
• coefficient measurements

26
Another mechanism for RHP zeros in F(s)

• Entropy waves means hot-spots from unsteady
  combustion
• Entropy waves generate upstream travelling
  acoustic waves when they are accelerated

27
RHP zero thoughts

• RHP zeros may occur when
• - Rd gt 1, even in absence of entropy waves
• - Rd lt 1 if entropy waves persist and are
  accelerated downstream
• Systems with RHP zeros are not guaranteed to be
  stabilised by our self-tuning regulator ? try and
  remove them?
• Zeros are artefact of sensor location and RHP
  zeros occur when magnitude of upstream travelling
  acoustic wave is large
• ? Try 2-microphone type technique to reduce
  relative magnitude of upstream travelling
  acoustic wave?

28
New sensor signal strategy

• Use 2 pressure measurements
• at different locations, p1 and p2
• 2-mic technique provides difference equations for
  h and j
• Form a new sensor signal, pref (t)
  h(x1,t)aj(x1,t) with a lt 1
• Then pref(s)/Vc(s) has no RHP zeros and is same
  relative degree as p1(s)/Vc(s)

29
Demonstration of new sensor strategy

• Computational model, open downstream end, entropy
  waves neglected.
• 3 different geometries with 3 different flames
  responses
• Large downstream reflection coeffs used, Rd
  -1.5

30
Results for new sensor strategy 1

• Case 1 L 2.85m, xf 0.5m, H(s)
  -2.1x106/(s2500)2

31
Results for new sensor strategy 2

• Case 2 L 2.7m, xf 0.5m, H(s)
  -2.2x106/(s1900)2

32
Results for new sensor strategy 3

• Case 3 L 2.55m, xf 0.5m, H(s)
  -3.3x106/(s2500)2

33
Overview

• Personal research background
• Background to Active Control of Combustion
  Instabilities
• Research
• - Model based control (experimental)
• - Model-based control for annular combustors
• - Adaptive control
• 4) Conclusions and Future work

34
Conclusions

• Combustion instabilities are
• - caused by coupling between acoustics and
  unsteady heat release
• - important to eliminate to achieve emissions
  reductions from gas turbines
• Active control is a feasible way of eliminating
  them
• Systematic control approaches, like robust
  model-based control and adaptive control, ensure
  good performance across operating conditions.

35
Future work

• Adaptive control for annular combustors
• Adaptive control when the flame transfer function
  contains RHP zeros
• Tuned passive control vary geometry of
  Helmholtz resonator in response to measured
  frequency
• Interesting to apply closed-loop control to other
  fluid/acoustic instabilities e.g. cavity
  resonances, gap tones?

36
Thank you