School of Aerospace Engineering MITE School of Aerospace Engineering MITE Active and Passive Control of Compressor Flow Instabilities Active and Passive

School of Aerospace Engineering

MITE


MITE

OverviewOverview

Future Work

Research Team

Problem Statement

Objectives

List Of Accomplishments

Significant Findings


MITE


MITE

Research TeamResearch TeamPI’s: Dr. J.V.R. Prasad

Dr. Y. NeumeierPost Doctoral Fellows:

Dr. N. MarkopoulosDr. M. Lal (Took up a position

in ME School)Graduate Students:

Mr. A. Krichene, AE, Ph.D. studentDr. C. Rivera, AE (graduated)Dr. T-Y. Ziang, AE (graduated)Mr. R. Swaminathan, AE (graduated)Mr. S. Bae, AE (graduated)Mr. A. Meehan, ME (graduated)


MITE

Problem Statement Problem Statement

Rotating stall and surge limit the operation of modern day turbine engine compressors due to associated severe loss of performance, component failure, etc.

Current practice is to limit operation with roughly 20% stall margin and limitations on fuel flow authorityduring acceleration and decelerations, representingloss of opportunity

Active and/or passive control strategies can result in reduced stall margin that will correspond to reducedweight and fuel savings


MITE


MITE

ObjectivesObjectives

Improved understanding of compressor stall and surge phenomena through modeling, simulation and experimentation

Investigation of Passive and active control mechanisms for reducing compressor stall and surge

Development of hybrid control methods by combining control-theoretic and decision-theoretic techniques


MITE


MITE

List of AccomplishmentsList of Accomplishments

Using theoretical extensions to Moore-Greitzer model to include finite duct effects, analytically showed that the inlet shape affects the stall inception point in axial compressors. This finding has animportant bearing on the design of appropriate inlets for passive control of rotating stall. (Presented a paper at the 1999 JPC)

Further experimental evaluations of passive control schemes forsuppression of rotating stall. (Presented a paper at the 1999 IEEEConference on Control Applications)

Combined the backstepping control method from the literature with the adaptive neural net/fuzzy logic scheme for improving robustness of the controller and evaluated the scheme in simulations. (Presented papers at the 1999 JPC and 1999 AIAA GNC)


MITE


MITE

Implemented the observer scheme for on-line identification of stall precursor waves and experimentally evaluated a novel active control scheme based on stall precursors for active surge control in the centrifugal compressor experimental facility at Georgia Tech.

List of Current Year Accomplishments (Continued)List of Current Year Accomplishments (Continued)


MITE


MITE

List of AccomplishmentsList of Accomplishments


Experimental evaluations of passive control schemes forsuppression of rotating stall. (Presented a paper at the 1999 IEEEConference on Control Applications)



MITE


MITE

MODELING OF COMPRESSOR ROTATING STALL

AND SURGE - RECENT PROGRESS

by

N. Markopoulos

MODELING OF COMPRESSOR ROTATING STALL

AND SURGE - RECENT PROGRESS

by

N. Markopoulos


MITE


MITE

PREVIOUS WORK…

Complete stability analysis of the Moore-Greitzer model under stall amplitude feedback

REASONS FOR THE MODELING WORK…

Moore-Greitzer model highly approximate - does not predict correct r.s. frequency, does not include effects of finite compressor length

Moore has suggested in a patent that a separator would eliminate rotating stall – we showed experimentally that this is not true

To our knowledge, there is no model that takes into account at a fundamental level of compressibility effects is available in the open literature - Mach numbers between 0.4 – 0.6

No control oriented models available for centrifugal compressors

Bottom line: Develop a basic physical understanding of the phenomena that we are trying to control


MITE


MITE

Inlet

B

Plenum TOutlet

Schematic of a CompressorSchematic of a Compressor


MITE


MITE

Our model:

Moore-Greitzer model:

When Q = 0 the two models become qualitatively the same

Moore-Greitzer model is obtained as a limiting case from our model as the inlet and outlet duct lengths go to infinity

For our model stall inception occurs slightly before or beyond the peak – depending on the sign of Q, representing the effect of the inlet

AQLUdsinsinAUT

AQT R2

2

0

22

TLUdsinsinAUA

QQT R

22

0

22

mLR2R

LU;dsinsinAU

mLR2

1A R

2

0


MITE


MITE

COMPARISON WITH THE MOORE-GREITZER MODEL…COMPARISON WITH THE MOORE-GREITZER MODEL…

Stall inception point

M-G: Ours:

Stable operation

M-G: Ours:

Unstable operation

M-G: Ours:

Conclude: It is very desirable to have Q > 0 for delaying loss of stability

P q p n s g m

h

f d e k l

a c

b r

o U

0U

0U

0U

T

QLUU R

T

QLUU R

T

QLUU R

)lsinhklcoshk(

kkQ;0vkuk

i22

ui22

v

vuvu


MITE


MITE

Quantitative account of instability dynamics for axial compressors - extends well-known Moore-Greitzer model

Chief difference effect of finite inlet and outlet duct lengths

What happens at entrance to inlet slightly hastens or delays settling of instabilities before or beyond peak of compressor map

Predicted r. s. frequency higher than Moore-Greitzer and function of compressor inlet length

Needed: a more fundamental account of effect of inlet in terms of inlet design parameters - future work

Brings up practical questions for the design of inlets and control of instabilities - transition to industry

MAIN RESULTS ON MODELING SO FAR…MAIN RESULTS ON MODELING SO FAR…


MITE


MITE

Axial velocity in real compressors varies between 150 to 200 m/sec corresponding to Mach numbers between 0.4 and 0.6

Inclusion of compressibility effects into our model

Governing equation is Classical wave eq. rather than Laplace’s eq.

Implies two qualitatively different types of disturbances (bound and scattering)

CURRENT WORK – AXIAL COMPRESSORS…CURRENT WORK – AXIAL COMPRESSORS…

Disturbance analysis for purely radial flow

Disturbance theory and modeling for centrifugal compressors

CURRENT WORK – CENTRIFUGAL COMPRESSORS…CURRENT WORK – CENTRIFUGAL COMPRESSORS…


MITE


MITE

List AccomplishmentsList Accomplishments


Further experimental evaluations of passive control schemes forsuppression of rotating stall. (Presented a paper at the 1999 IEEEConference on Control Applications)



MITE

Schematic of Experimental Set-up(Flow Separators and Flow Recirculation)

Schematic of Experimental Set-up(Flow Separators and Flow Recirculation)

Controller

Pressure Measurements

Servomotor and bleed

Computer

Bleed/recirculation loop

Main Throttle


MITE

Flow Separators in the InletFlow Separators in the Inlet

•Moore predicted that one separator in the inlet should eliminate the rotating stall altogether (Patent No. 5,297,930 by Moore F, K. “Rotating Stall Suppression” )


MITE

Time [sec]

No separator One separator

Transducers' angular

locations

Pressure Oscillations with and without an Inlet Separator

Pressure Oscillations with and without an Inlet Separator

• The separator seems to have no apparent effect upon the traveling waves


MITE

0

0.002

0.004

0.006

0.008

0.01

0.012

0.014

0.016

0.018

0.02

35.0 40.0 45.0 50.0 55.0 60.0

Main Throttle Openning (%)

Rot

atin

g St

all A

mpl

itud

e (%

of

Pam

b) No Separator

8 Separators

8 Separators with active ambient bleed

No Separator with active ambient bleed

Effect of Eight Flow Separators on Rotating Stall Amplitude

Effect of Eight Flow Separators on Rotating Stall Amplitude


MITE

0

0.005

0.01

0.015

0.02

0.025

0.03

0 0.1 0.2 0.3 0.4 0.5

Normalized Mass Flow

Rot

atin

g St

all A

mpl

itud

e(%

of

Pam

b)

50% Bleed to ambient 50% Bleed to inlet

100% Bleed to ambient 100% Bleed to inlet

Effect of Flow Recirculation on Rotating StallEffect of Flow Recirculation on Rotating Stall


MITE

0

0.005

0.01

0.015

0.02

0.025

0.03

0 0.05 0.1 0.15 0.2 0.25 0.3


Rot

atin

g S

tall

Am

plitu

de(%

of

Pam

b)

100% Bleed to inlet 50% Bleed to inlet (Feedback) 50% Bleed to inlet

Effect of Flow Recirculation with Active Control on Rotating Stall

Effect of Flow Recirculation with Active Control on Rotating Stall


MITE

0.1

0.15

0.2

0.25

0.3

0.35

0.0 20.0 40.0 60.0 80.0 100.0

Main Throttle Opening (%)

Nor

mal

ized

Tot

al P

ress

ure

Bleed(0%) Bleed(50%) Bleed(100%)

Compressor Pressure Rise versus Main Throttle Opening for Different Ambient Bleed Openings

Compressor Pressure Rise versus Main Throttle Opening for Different Ambient Bleed Openings


MITE

0.1

0.15

0.2

0.25

0.3

0.35

0 0.1 0.2 0.3 0.4 0.5 0.6


Nor

mal

ized

Tot

al P

ress

ure

Bleed 0% Bleed 50% Bleed 100%

Compressor Pressure Rise versus Normalized Flow Rate for Different Ambient Bleed Openings

Compressor Pressure Rise versus Normalized Flow Rate for Different Ambient Bleed Openings


MITE

0.1

0.15

0.2

0.25

0.3

0.35

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35


Nor

mal

ized

Tot

al P

ress

ure

Bleed 0% Bleed 50% Bleed 100%

Compressor Pressure Rise versus Normalized Flow Rate for Different Recirculation Bleed Openings

Compressor Pressure Rise versus Normalized Flow Rate for Different Recirculation Bleed Openings


MITE


MITE

xc

System Linear controller

Model inversion

Adaptive neural net/fuzzy logic

Model based controller

+

-+

+

x

Adaptive Neuro-fuzzy Controller


MITE


MITE

Adaptive Neuro-fuzzy Controllers

Hybrid control methodology which combines modelinversion with neural nets and fuzzy logic

Parameterization of uncertainty using neural nets andfuzzy logic and adaptation of parameters based onLyapunov stability theory

Rule base adaptation and linear controller gainadaptation to accommodate actuator limits


MITE


MITE

Model based controller

Hybrid controller with fixed linear controller gain

Hybrid controller with variable linear controller gain

Non-dimensional time

Rot

atin

g st

all a

mpl

itud

eResponse to Initial Disturbance with Model

UncertaintyController is based on fifth order compressor mapSimulation is based on third order compressor map


MITE


MITE

Related Work

T700 Engine Fuel Control Using Adaptive Neural networks


MITE


MITE

PI

Network

Neural

FgP ZPPPTNNe ,,,,,,,,0.1 4541341

adu

u

ou u

PREFN

ECU

),:(Feedforward2

TQrr

GovernorRateCompensationand Dynamics

T700Engine

HMU(NonlinearStateFeedback)-

Np

T700 Engine Fuel Controller


MITE


MITE

0 2 4 6 8 10 1290

95

100

105

Time (sec)

pow

er

turb

ine s

peed (

perc

ent)

Inversion ControlSet point

Performance of the T700 Engine Fuel Controller

PowerTurbineSpeed(%)

Time (sec)


MITE


MITE

0 2 4 6 8 10 1290

95

100

105

pow

er

turb

ine s

peed (

perc

ent)

Time (sec)

neural network inversion control

Performance of the T700 Engine Fuel Controller to a Periodic Load Disturbance with and without adaptive neural networks


MITE


MITE

Implemented the observer scheme for on-line identification of stall precursor waves and experimentally evaluated a novel active control scheme based on stall precursors for active surge control in the centrifugal compressor experimental facility at Georgia Tech.

List of Accomplishments (Continued)List of Accomplishments (Continued)


MITE

Centrifugal Compressor SetupCentrifugal Compressor Setup

Control Law

Frequency/Amplitude Observer

Fuel Valve

Control Computer

Throttle Valve

servomotor

servomotor

Throttle and FuelValve Commands

Data AcquisitionComputer

Pressure Measurements

PressureTransducer

Pressures

Self entrainingcombustor

Inlet pressure readout

Control Variables


MITE

Controller EssentialsController Essentials

• Utilizes real time observer that identifies the frequency and amplitude of the most dominant modes of oscillations in the inlet pressure signal

• Sets on-off alarm signal when precursors waves are identified with strong enough amplitude

• Varies the fuel flow rate or other actuators according to the alarm signal


MITE

• Rejecting rather than suppressing stall

• Provides global stability

• Does not require high bandwidth actuator

• Can work with existing fuel injection systems

• Requires very little information about compressor characteristics

Controller Essentials (Cont.)Controller Essentials (Cont.)


MITE

Real-Time Mode ObservationReal-Time Mode ObservationLow Back Pressure, 15 KRPM


MITE

Real-Time Mode ObservationReal-Time Mode ObservationHigh Back Pressure Nearing Surge, 15 KRPM


MITE

Real-Time Mode ObservationReal-Time Mode ObservationUncontrolled Surge, 15 KRPM


MITE



MITE



MITE



MITE



MITE



MITE

Control Using Throttle ActuationControl Using Throttle Actuation


MITE

Control Using Throttle Actuation (Cont.)Control Using Throttle Actuation (Cont.)


MITE

Control Using Fuel Valve ActuationControl Using Fuel Valve Actuation


MITE

Control Using Fuel Valve Actuation (Cont.)Control Using Fuel Valve Actuation (Cont.)


MITE


MITE

Future WorkFuture Work

Further theoretical, simulation (using Dr. Sankar’s CFD models) and experimental evaluations of control actuation schemes (e.g., bleed valve, fuel flowmodulations, etc.) using the centrifugal compressorfacility.

Experimental evaluation of novel controllers.

Further analysis of the effect of inlet parameters on rotating stall in axial compressors.

Documents

School of Aerospace Engineering MITE School of Aerospace Engineering MITE Active and Passive Control of Compressor Flow Instabilities Active and Passive