Dynamic modeling, Analysis, and Control at United Technologies Research Center

1
Dynamic Systems Control Group
2

• Contents
• UTC and UTRC Overview
• United Technologies Corporation Business Units
  and Products
• United Technologies Research Center -
  Organization and Core Capabilities
• Dynamic Systems and Control
• People - group member and university interactions
• Dynamic phenomena at UTC
• Project organization
• Description of selected projects
• Specific features of research done at UTRC

3
UTC and UTRC Overview Products and Organization
4
UNITED TECHNOLOGIES PRODUCTS AND BUSINESS UNITS
Pratt Whitney
Otis
Carrier
5
UNITED TECHNOLOGIES FACT SHEET

• MAJOR BUSINESSES
• Pratt Whitney Aircraft engines, Carrier
  heating and air conditioning systems, Otis
  elevators and escalators, Sikorsky Helicopters,
  Hamilton Sundstrand aerospace systems.
• RANKINGS
• 41st largest U. S. corporation (1998, Fortune
  Magazine), 130th in the world (1998, Fortune
  Magazine, Global 500)

• EMPLOYEES 180,000
• UTC employees, including approximately 105,700
  outside the United states
• REVENUES
• 25.7 BILLION IN 1998,
• SALES TO U. S. GOVERNMENT
• 3.264 billion, or 12.7 of total sales (includes
  sales to NASA)
• RD
• 1.31 billion in company-funded RD in 1998

6
UTRC OUR VALUE TO UTC
To provide technical leadership that increases
the competitiveness of our business units. UTRC
accomplishes this by integrating
technical disciplines and expertise that have
business unit applicability to create technology
for the future needs of the corporation.
7
UTRC MISSION STATEMENT
See It First, Make It Happen
We team with UTCs business units to
foresee technological opportunities and create
solutions that redefine marketplaces, increases
competitiveness, better our society and leave a
legacy of excellence. We aim to be a worldwide,
diverse, and innovative community that is
attractive to top talent and is recognized as a
unique corporate resource. We strive for an
environment of integrity, trust, mutual
respect, fairness and learning in which we can
all grow.
8
UTRC ORGANIZATION
The Office of the Director provides the UTRC with
leadership and strategic direction. A strong
partnership exists between program planning and
execution functions to ensure a clear focus on
impacting the future of the business units.
UTRC Director

• Leadership
• Strategic Direction

• Director,
• Research Programs
• Program Planning

• Director,
• Research Operations
• Program Execution

Office of the Director (UTRC)
9
UTRC SENIOR LEADERSHIP
The senior leaders at UTRC are organized to
support the Research Centers planning and
execution efforts.
Director
Office of the
UTRC Director
Director, Research Operations
Director, Research Programs
10
UTRC CORE CAPABILITIES
Aeromechanical, Chemical Fluid
Systems Acoustics Aerodynamics Heat
Transfer Fluid Dynamics Combustion
Fuels Environmental Science Mechatronic
Systems Dynamic Modeling Analysis Controls
Technology Controls Components Electronics
Technology Advanced Embedded Systems Information,
Computer Communication Technology Advanced
Digital Systems Diagnostic Technology Informistics
Network Technology Systems Software
Materials Structures Engineered
Materials Material Structural
Modeling Materials Characterization Structural
Integrity Surface Engineering Product
Development Manufacturing Product Innovation
Methods Design for X Rapid Product
Realization Nondestructive Evaluation Virtual
Manufacturing Advanced Manufacturing Processes
11
UTRC OUR EMPLOYEES
The Research Center employs close to 800
scientists, engineers, technicians and support
staff worldwide.
1997 DISTRIBUTION
Administration 8
Facilities Support 14
Technical Professionals Technical Support 78
12
UTRC TECHNICAL EMPLOYEES
The Centers engineers and scientists form a
diverse group of technical experts.
Physics 7
Aeronautical 11
B.S. 21
Mechanical 28
Chemical 10
Ph.D. 41
M.S. 38
Materials 8
Computer Science/ Mathematics 12
Engineers - Other 10
Electrical 14
13
UTRC FUNDING SOURCES
Financial support for the Research Centers
operations is provided through corporate,
business unit sponsorship, and through contracts
with industry and government.
Sources of Funds 1998 TOTAL - 107.8 MILLION
14.7 Business Unit Technical Support
29.3 Business Unit Co-Planned Program
31.6
15.8
28.5 Corporate Sponsored Research
13.6
12.6 Business Unit Subcontracts
30.7
16.1
14.9 Direct Contracts
14
UTRC FUNDING USAGE
Selection of technical programs is driven by the
potential to create value for our six business
units. Co-planning of program milestones with the
business units is key to the planning and
selection process.
Business Unit Relevance 1998 TOTAL - 107.8
MILLION
Sikorsky 6
HSD 5
Pratt Whitney 42
UTA 5
Carrier 13
Generic (all Business Units) 19
Otis 10
15
Dynamic Systems and Control People, Products,
Problems, Solutions
16
Dynamic Systems and Control

• Mission
• People and Skills
• University Teaming
• Publications
• Project Organization Products, Problems,
  Solutions
• Selected Project Examples

17
MISSION STATEMENT
We team with UTCs business units to
foresee technological opportunities and create
solutions that redefine marketplaces, increase
competitiveness, leave a legacy of excellence.
We provide world class technical expertise in
the broad areas of dynamic systems and control
including experimental programs, control system
modeling, design, analysis and implementation
and dynamic system analysis and computation.
18
Dynamic Systems and Control People UTRC,
University Partnering, Skills, Publications and
Career Paths
19
People and Program Characteristics

• Individual Metrics
• Technical depth - means demonstrated expertise in
  at least one area
• Technical breadth - means the ability to interact
  closely in several areas
• Communication - ability to present results to
  varied audiences
• Organization of Projects
• Business unit problem source
• Multidisciplinary teams for execution
• Intellectual property or competitive advantage as
  deliverables

20
Basic Research Areas of UTRC Interest

• Methods for obtaining reduced order models for
  control of unsteady flow phenomena
• Methods of parameter identification of nonlinear
  dynamical models
• Methods for validation of nonlinear
  physics-based models against experimental data
• Computational tools for complex nonlinear
  dynamical systems
• Methods for on-line optimization of dynamical
  system behavior (e.g., reduce the magnitude of
  oscillations) with adaptive algorithms
• Observers for nonlinear and time-varying systems
• Generation of trajectories obeying state and
  actuator constraints for complex nonlinear
  systems (jet engine control, helicopter control)
• Control strategy for a complex dynamic system
  with redundant actuators of significantly
  different authority operating in the same
  bandwidth upon the multiple objectives of command
  following, disturbance rejection, and stability
  augmentation.
• Methods for optimization of actuator and sensor
  placement for control of complex systems
• Robust real-time model adaptation for a
  multivariable linear control system.

21
Dynamic Systems and Control Group Members
22
Dynamic Systems and Control Group Andrzej
Banaszuk has received Ph.D. in Electrical
Engineering from Warsaw University of Technology
in 1989, and Ph.D. in Mathematics from Georgia
Institute of Technology in 1995. From 1989 to
1997 he has held various research and teaching
positions at Warsaw University of Technology,
Georgia Institute of Technology, University of
Colorado at Boulder, and University of California
at Davis. During that time he performed research
in various areas of control theory including
implicit systems, approximate feedback
linearization of nonlinear systems, trajectory
planning for nonlinear systems, nonlinear
observers, feedback stabilization of periodic
orbits, and control of surge and rotating stall
in jet engines. He is an author or co-author of
about 25 journal papers and numerous conference
papers. Andrzej Banaszuk joined Controls
Technology Group at United Technologies Research
Center in April 1997. His work at UTRC has been
focused on modeling and control of turbomachinery
flutter, rotating stall, combustion instability,
and flow separation. His current research
interest is in reduced order modeling for control
purposes of complex physical phenomena in
turbomachinery, model validation and parameter
identification for nonlinear systems using
experimental data, and control of nonlinear
systems in a neighborhood of non-equilibrium
attractors. In 1998 Andrzej Banaszuk became an
Associate Editor of IEEE Transactions on Control
Systems Technology. Full CV and list of
publication available at http//talon.colorado.edu
/banaszuk. Jim Fuller is a Senior Principal
Engineer in Controls Technology and has 23 years
of experience in modern control system design,
analysis and development, the highlights of which
include development of multivariable, nonlinear
and adaptive control and estimation algorithms
for (1) controlling the flight of the RSRA/X-wing
aircraft, (2) missile guidance, navigation, and
control, (3) aided inertial navigation, (4)
Propfan gas turbine engine, (5) air conditioner
chillers and (6) improving ride and comfort of
elevators. His experience also includes
research into automated nap-of-the-earth
helicopter flight, trajectory generation using
optimal control theory, neural nets, fault
tolerant and robust control algorithm synthesis,
and passive and active ride control
systems. Gonzalo Rey has worked on theoretical
studies of adaptive systems where he has applied
nonlinear dynamical systems analysis tools such
as bifurcation and averaging analysis. His
competencies extend to servo control system
design and control algorithms for aerospace and
industrial motion control applications where he
has acquired a broad experience base. He is
skilled in the areas of robust adaptive control,
linear system parameter identification, linear
control and nonlinear system dynamics. His
recent experience at UTRC includes research in
the areas of active noise control and active
control of flutter in turbo-machinery.
23
Dynamic Systems and Control Group
(continued) Chris Park core competencies
include structural dynamics, linear control
theory, rotor dynamics, non-linear dynamic
modeling, and experimental techniques. He is
also competent in active materials, aerodynamics,
servo control, and active noise control. His
recent experiences at UTRC include active noise
control system development and data analysis for
enclosures, disturbance transmission path
analysis, modeling rotor dynamics for active
control system studies, and development of a real
time active rotor control system for wind tunnel
testing. Clas Jacobson has worked for three
years at UTRC (nine years in academia previously)
in diverse areas of control systems design and
implementation. He has contributed to programs
in active noise control (duct and enclosure),
combustion dynamics and control and compression
system instabilities. His current interests are
mainly in the identification and control of
nonlinear systems for combustion and flow control
applications. Danbing Seto has worked in the
areas of nonlinear adaptive control and control
of complex mechanical systems, where he applied
differential geometric tools to develop control
algorithms for nonlinear systems in a triangular
structure with or without unknown parameters. He
also studied nonlinear vibrational control
theory, from which he derived a mechanical model
for laser cooling. His interdisciplinary
experience include computer-controlled real-time
systems, where he particularly focused on
real-time scheduling, control system upgrade and
software fault tolerance. His recent work at UTRC
has been concentrated on 1) fault tolerance and
2) system identification. The former concerns the
integrated fault management functionality in Otis
elevator control systems with scalability, and
the latter involves development of
tools/methodologies for model validation of
nonlinear systems as well as modeling jet engines
using the state-of-the-art identification tools.
His long-term technical goal at UTRC is to
investigate estimation theory applied in
integrated control systems, which unifies the
research areas of model identification and state
estimation together with control design.
24
Dynamic Systems and Control Group
(continued) Alexander Khibnik has a background
in analysis of nonlinear dynamical systems with
an emphasis on analytical and numerical issues in
bifurcation theory. He joined UTRC in 1997,
after spending more than 20 years in academia.
His experience with systems ranging from ecology
to neurobiology to nonlinear physics is focused
on the development and application of numerical
tools for the analysis of their qualitative
nonlinear dynamical behavior. His competencies
extend to self-excited oscillations, coupled
oscillators, resonance, fast-slow systems,
continuation techniques, integrated with software
and computer tool development. His recent
experience at UTRC has focused on the analysis of
compressor and combustion dynamics with an
emphasis on modeling nonlinear dynamics from
data. He is currently leading a team in flow
control area which studies low dimensional
dynamics of separation in diffuser flows and its
utilization for model-based control of
separation. Satish
Narayanan comes from an experimental fluid
mechanics background and has applied the
nonlinear dynamical systems approach to extract
low-dimensional models of complex fluid flow
phenomena. In doing so he has developed active
nonlinear flow control strategies for turbulent
flows of wide technological relevance such as
jets and shear layers. His areas of expertise
include nonlinear dynamics, reduced-order
modeling, flow control, experimental fluid
mechanics, turbulence, vortex dynamics and
hydrodynamic stability. His current projects in
UTRC involve dynamical modeling and active
control of flow separation phenomena
(experimental and numerical), the development and
the implementation of a phased array a new jet
noise source localization technique, and the
testing of new active control methods for jet
noise reduction. Richard Murray is an expert
in the area of dynamical systems and nonlinear
control, with applications to motion and flow
control. His past work includes studies in
geometric mechanics for Lagrangian systems with
symmetries and nonholonomic constraints,
real-time trajectory generation for motion
control systems using differential flatness, and
active control of compression, combustion, and
cavity flow instabilities. Murray and his
research group at Caltech have designed, built
and operating a variety of experiments, including
a thrust vectored flight control experiment, an
axial flow compression system facility, and a
cavity flow instability experiment. At UTRC,
Murray is an active participant in programs
relating to flow control, combustion dynamics and
control, modeling and analysis, and smart
products.
25
Dynamic Systems and Control Group
(continued) Leena Singh has intensive
experience in methods of motion control and
trajectory generation of Lagrangian systems,
specifically, articulated multi-link manipulators
such as robot arms and hands. Key competencies
and areas of interest are modern control theory,
optimal control, passivity-based control,
attitude control and exact, analytical algorithms
for online trajectory generation in
constraint-based spaces. She also has experience
in modeling the spatial kinematics and dynamics
of mechanical systems. At UTRC (since July 1997)
she has worked on projects in the areas of
kinematic modeling and control, and estimator
design. Bernd R. Noack has a fluid dynamics
background. He has joined UTRC in December 1998
after 6 years in research institutes and
academia. He has worked in the areas of wake
flow, several open and confined flows, turbulence
of superfluid helium, brain activity and
time-signal analysis. He has experience with
phenomenological modeling, Navier-Stokes
simulation, Galerkin methods, linear and
nonlinear stability analysis, Floquet theory,
nonlinear dynamics, low-dimensional modeling,
mean-field theories, center-manifold methods,
harmonic balances, turbulence modeling and
control. Particular UTRC applications include
modeling and control of flow separation and
mixing enhancement. Mike Dorobantu is
interested in the efficient computations of
numerical solution to PDEs. In academia he
focused on flow problems, such a flow through
porous media, using multi-scale techniques, the
application of wavelet-based preconditioning and
homogenization, multi-grid preconditioning, and
streamline diffusion stabilization methods. At
UTRC he is developing classification algorithms
based time-frequency analysis and multi-phase
non-newtonian fluids mixing models. He is also
involved in convergence acceleration and
extracting spectral information from time-domain
simulations of flow problems and deriving
data-driven reduced order models for transient
flows.
26
Dynamic Systems and Control External
Collaborations
27
Modeling for Control of Mixing

• Academic contacts Professors Igor Mezic,
  University of California at Santa Barbara,
  Professor Luca Cortelezzi, McGill University
• UTRC Contacts Dr. Bernd Noack, Dr Andrzej
  Banaszuk
• Project Goalcreate a low order model and derive
  model-based control laws for mixing enhancement.
• Approach vortex methods for modeling flow
  dynamics and dynamical system methods for control
  law derivation are investigated.
• Applications modeling for control of combustion
  phenomena.
• Status research in progress.
• Publications Conference and journal submissions
  expected by mid 1999.

28
Model Validation for Nonlinear Systems

• Academic contacts Professors Igor Mezic and Roy
  Smith, University of California at Santa Barbara
  
• UTRC contact Dr. Andrzej Banaszuk
• Project goal create new methods for validation
  of nonlinear models with non-equilibrium behavior
  and stochastic disturbances against experimental
  data.
• Approach methods from ergodic theory for
  comparison of behavior of dynamical systems and
  extensions of classical linear model validation
  concepts are investigated.
• Applications modeling for control of combustion
  instability, flow separation, and rotating stall.
  
• Status research in progress.
• Publications conference and journal submissions
  expected by mid 1999.

29
Control Theory for Systems with
Non-equilibrium Attractors

• Academic Contact Professor John Hauser,
  University of Colorado at Boulder
• UTRC Contact Dr. Andrzej Banaszuk
• Project Goal create methods and tools for
  control of models with non-equilibrium
  attractors, like periodic orbits. Typical goal is
  to achieve acceptable performance with limited
  actuator authority in the cases when
  stabilization of an equilibrium is not achievable
  or undesirable.
• Approach dynamical system topological and
  Lyapunov function methods
• Applications control of combustion instability,
  flow separation, and rotating stall.
• Status preliminary results for shrinking of
  planar periodic orbits with saturated actuators
  available. Extensions to non-planar periodic
  orbits and to other type of attractors expected.
• Publication Control of planar periodic orbits,
  accepted for 1999 CDC. Journal submission in
  progress.

30
Performance and Stability Analysis of
Extremum Seeking Methods

• Academic contacts Professor Miroslav Krstic,
  University of California at San Diego, Mario
  Rotea, Purdue.
• UTRC contact Dr. Andrzej Banaszuk
• Project goal create methods and tools for
  performance and stability analysis for
  extremum-seeking algorithms.
• Approach combination of methods from linear,
  nonlinear, and adaptive control
• Applications adaptive control of combustion
  instability and flow separation
• Status work in progress.
• Publication conference and journal submission
  expected by late 1999.

31
Development of Parametric Analysis Techniques for
Large Scale Systems

• Academic Contact Dr. Kurt Lust, Cornell
  University Katholic University of Leuven
  (http//www.cs.kuleuven.ac.be/kurt)
• UTRC Contact Dr. Alexander I. Khibnik
• Project Goals development of tools for
  parametric analysis that utilize existing CFD
  time simulation codes to compute and analyze
  steady-state solutions of large-scale models.
• Approach acceleration of iterative methods (RPM,
  GMRES), effective spectral computations (Arnoldi,
  Jacobi-Davidson), continuation techniques
• Applications large-scale models in fluid flows,
  combustion, acoustics, aeromechanics.
• Status work in progress.
• Publication conference and submission expected
  by late 1999.

32
Selected Recent Publications

• System Identification for Limit Cycling Systems
  A Case Study for Combustion Instabilities, R. M.
  Murray, C. A. Jacobson, R. Casas, A.I Khibnik,
  C.R. Johnson Jr., R. Bitmead, A.A. Peracchio,
  W.M. Proscia, 1998 American Control Conference
• Self-Tuning Control of a Nonlinear Model of
  Combustion Instabilities, M. Krstic, A.
  Krupadanam, C.A. Jacobson, 1997 IEEE Conference
  on Control Applications
• Active Control of Combustion Instability in a
  Liquid Fueled Low NOx Combustor, J. M. Cohen, N.
  M. Rey, C. A. Jacobson, T. J. Anderson, 1998 ASME
  Turbo Expo.
• Linear and Nonlinear Analysis of Controlled
  Combustion Processes. Part I Linear Analysis.
  Part II Nonlinear Analysis,A. Banaszuk, C.A.
  Jacobson, A.I. Khibnik, and P.G. Mehta, 1999 CCA,
  August 1999, Hawaii.
• - Active Control of Combustion Instability in a
  Liquid-Fueled Sector Combustor,J.R. Hibshman,
  J.M. Cohen, A. Banaszuk, T.J. Anderson, and H.A.
  Alholm, 1999 ASME Turbo Expo, 1999, Indianapolis.
  

33
Selected Recent Publications (continued)

• - Adaptive detection of instabilities and
  nonlinear analysis of a reduced-order model for
  flutter and rotating stall in turbomachinery,
• G.S. Copeland, I.G. Kevrekidis, R. Rico-Martinez,
  1999 CCA, Hawaii.
• A Backstepping Controller for a Nonlinear
  Partial Differential Equation Model of
  Compression System Instabilities, A. Banaszuk,
  H.A. Hauksson, and I. Mezic, SIAM Journal of
  Control and Optimization , 1999, to appear.
• - Design of Controllers for MG3 Compressor Models
  with General Characteristics Using Graph
  Backstepping, A. Banaszuk and A.J. Krener,
  Automatica , 35 (8) 1999, 1343 -1368.
• - On control of planar periodic orbits
• A. Banaszuk and J. Hauser, 1999 CDC, December
  1999, Phoenix.
• - Analysis of low dimensional dynamics of flow
  separation.
• Khibnik, A.I, Narayanan, S., Jacobson, C.A. and
  Lust, K. Submitted to Notes in Computational
  Fluid Dynamics (Proceedings of Ercoftac and
  Euromech Colloqium 383 "Continuation Methods in
  Fluid Dynamics", Aussois, France, 6-9 September
  1998).
• - Low-dimensional model for active control of
  flow separation.
• Narayanan, S., Khibnik, A.I. Jacobson, C.A.,
  Kevrekidis, Y., Rico-Martinez, R. and Lust, K,
  CCA '99 (Hawai, August 1999).
• - Control of laminar mixing enhancement in a
  recirculation region,
• B.R. Noack, A. Banaszuk, and I. Mezic, to be
  submitted to Physica D, 1999.

34
UTRC Technical Career Path Increasing Program
Responsibility
Expert
Program manager (responsibility for technical
direction and resourcing)
Principal investigator (responsibility for
technical direction)
Individual contributor in single technical area
35
UTRC Career Paths Cover Technical and Management
Line Managers
Program Managers
51
Technical Council
50
49
Program Management Track
48
46
Common Competencies
36
Dynamic Systems and Control Projects Organization,
Content, Solutions
37
UTRC research in dynamic modeling and control

• UTC Business unit relevance drives the research
• research always tied to a product need
• emphasis on potential benefits to business units
  in either product or process
• full scale experimental rigs validate modeling
  and control concepts
• Ability to communicate with people of different
  background (coworkers, management, engineers in
  business units) is essential.
• Breadth of programs is typical
• Evaluation of problem
• Modeling at multiple time and spatial scales
• Control concepts evaluated to influence dynamics
• Proof of concept on full scale hardware

38
Dynamics and Control Approach
Phenomena Characterization

• Business case
• Risk assessment
• Product plan

Business Unit Need
Modeling

• Actuation limits
• Scaling laws

Control

• Control design
• Actuation system
• Fundamental limits

Demonstration
Product or Process Improvement
39
UTC cares for dynamic modeling, analysis, and
control because dynamics (usually undesirable)
affect UTC products.
Customer requirements
Product
Problem
Undesirable dynamics
Solution
Understand dynamics
Change dynamics
40
Problems undesirable dynamics affects UTC
products

• Carrier
• Noise (ducts, compressors, combustors)
• Compressor surge and stall
• Otis
• Elevator/cable dynamics
• Noise
• Power electronics dynamics
• Electric drives dynamics

• Pratt Whitney
• Compressor stall and surge
• Fan flutter, turbine buffeting
• Compressor stator vortex shedding
• Blade cracks propagation
• Turbine blades temperature transients
• Diffuser/duct flow separation
• Inlet flow distortion
• Jet noise
• Combustor instability
• Sikorsky
• Structure noise and vibration
• Blades/structure interaction with air flow

Generic 1. Noise and vibrations 2. Flow
separation - efficiency loss 3. Flow/structure
interaction - structural damage
41
Path to solutions Understanding Dynamics
Basic understanding of physics
Sensor selection
Actuator selection
Experimental data

• Physics-based modeling
• Construction of dynamical system model
• Identification of model parameters
• Validation of models against data
• Model reduction (Galerkin, POD, )

• Data-based modeling
• Construction of dynamical system model
• - linear frequency response
• - nonlinear embedding, neural nets, ...

• Study dynamical system properties (attractors,
  stability, bifurcations,...)
• Link model parameters to design parameters
• Identify sensor/actuator selection for active
  control

42
Short/mid term solutions change dynamics (fix
the problem)
Issues, tradeoffs
Options

• Redesign product to avoid the undesired behavior
• Modify dynamics by passive fixes
• Modify dynamics by active control

• Can be impossible (product has to be shipped in 6
  months ...)
• Can be expensive, difficult
• What if the control system fails ...

Long term solutions design dynamics (prevent
the problem)

• Incorporate dynamical system models early at the
  design process to avoid the undesired behavior
• Use the dynamical models to build the system with
  embedded sensors and actuators for active control
• Educate design engineers about dynamics

43
Dynamic Systems and Control Example Combustion
Instabilities
44
Performance Limitations in Aircraft Engines

• Flutter and high cycle fatigue
• Aeromechanical instability
• Active Control a possibility
• Combustion instabilities
• Large oscillations cannot be tolerated
• Active control demonstrated at UTRC
• Jet noise and shear layer instabilities
• Government regulations driving new ideas

• Inlet separation
• Separation of flow from surface
• Possible use of flow control to modify
• Distortion
• Major cause of compressor disturbances
• Rotating stall and surge
• Control using BV, AI, IGVs demonstrated
• Increase pressure ratio Þ reduce stages

45
Combustion Dynamics Control Programs

• PW/UTRC Joint Planned Programs
• Combustion Dynamic Modeling
• Active Instability Control (AIC)
• DARPA AIC - Liquid Fuel
• NASA Direct Injection Aeroengine AIC

46
Combustion Dynamics Control Capabilities

• Experimental
• BFSC, ASDC
• High Pressure SNR
• Sector Rig
• Engine

• Modeling
• Unsteady CFD
• Euler Code
• Lumped/Linear Acoustics
• Reduced Order Heat Release

• Dynamic Analysis Control
• Model Analysis
• (Stability, Amplitude)
• System Identification
• Control Analysis Design
• (Adaptive, Robustness)
• Control Implementation

• Sensing Actuation
• Pressure
• PMT
• 2D Flame Imaging
• Fuel Valves
• Solenoid
• PZT
• MOOG DDV other
• Acoustic Forcing Bleed Valve

47
Combustion Dynamics Control Team

• Modeling
• A. Peracchio
• G. Hendricks
• D. Choi
• A. Khibnik
• B. Wake

• Sensing Actuation
• T. Anderson
• N. Rey
• J. Haley
• MOOG

• Product Integration
• T. Rosfjord
• W. Proscia
• J. McVey
• W. Sowa
• J. Lovett (PW)
• S. Syed (PW)

• Experimental
• J. Cohen
• D. Kendrick
• H. Alholm
• R. Decker

• Control
• C. Jacobson
• A. Banaszuk
• Y. Zhang
• G. Rey
• R. Murray
• R. Bitmead
• M. Krstic

48
Example industrial combustor design
Customer requirements low emission level
lean mixture
Product
Undesirable dynamics lean mixture
violent pressure oscillations high cycle
fatigue, combustor destruction
Problem
Solution
Understand dynamics reduced order physics-based
model model analysis predicts limit cycle model
parameters linked to design parameters model
allows to identify effective actuation mechanism
Change dynamics. Options 1. Redesign
combustor 2. Use passive devices to reduce
oscillations 3. Use active control to reduce
oscillations
49
Combustion Instabilities Will Occur
Combustion Instabilities Limit Minimum Achievable
NOx Emissions

• Goals
• NOx/CO limits
• RMS pressure limits
• Wide range of operating conditions
• 50 - 100 power
• -40 to 120 F ambient temp.
• Instabilities inevitable
• combustion delay
• convective delay
• Passive design solution may be possible
• AIC can enable product

Product Need
Stability boundary defined as maximum allowable
pressure fluctuation level
50
Combustors Experience Instabilities
Data obtained in single nozzle rig environment
showing abrupt growth of oscillations as
equivalence ratio is leaned out to obtain
emissions benefit
51
Combustion Dynamics
Active Control

• Development of dynamic models
• Improved acoustic models 1D 3D
• Improved flame models
• Atomization mixing models
• Development of prediction and analysis tools
• Predict stability boundaries reliably and early
  in the development process
• Development of design test protocols
• Extract data from component tests
• Integrate physical understanding into design
  process

• Engine-ready sensing and actuation
• Modeling enables requirements specifications for
  vendors
• Modeling enables scaling effects to be understood
  
• Robust algorithms architectures
• Modeling enables development of self-tuning
  algorithms for hands-off operation over long
  periods
• Modeling enables integrated diagnostics
  prognostics
• Control at finer spatial scales
• Fuel/air ratio control for pattern factor
• Mixing control of higher power, lower emissions

Process Standard Work
Product Improved Engines
52
Observed Unacceptable Time Response Behavior
Model description capturing system dynamics
Alter system dynamics to obtain acceptable
behavior
System Level Model Showing Feedback Coupling
Evaluation of Design Options

• Evaluation of model sensitivities
• Development of experimental protocols and model
  calibration
• Evaluation of paths to mitigate undesirable
  behavior

Effects of Parameter Variation on Stability
Boundary
Enabling effective use of dynamic model
Parametric analysis of system model
53
System Level Model
Air Feed System
Combustion Response
Combustor Acoustics
Fuel Feed System
System Level Model Captures
Purpose of System Level Model
- Key components and interactions
System modifications - Preliminary design (
scaling) - Design optimization Active
Control - Actuator authority - Control
algorithm development
- Experimentally obtained information
- Lessons learned in transferable code
System Level Model Analysis
- Linear stability boundaries
- Amplitude prediction
- Closed-loop control performance
54
Thermoacoustic Modeling and AnalysisLean
Premixed Combustion Instability Mechanism

• Thermoacoustic instability - feedback
  interconnection of acoustic and heat release
  component subsystems - instability of feedback
  system is mechanism of pressure oscillations
• Acoustic resonance sets the frequency of
  oscillation
• Heat release rate dependent on
• Instantaneous equivalence ratio
• Instantaneous flame surface area
• Linear dynamics define system stability
• Nonlinear effects determine limit cycle amplitude
• Acoustic damping
• Heat release

Fluctuating pressure driven by unsteady heat
release
Fluctuating heat release driven by unsteady
velocity
55
Analysis shows model captures phenomenon
Frequency varies with delay
Amplitude vs
Frequency varies with equivalence ratio
Phenomena Increasing oscillations with decreasing
mean equivalence ratio

• Mechanism
• System level model capturing phenomenon - 6th
  order nonlinear delay differential equation
• Key parameters are acoustic damping and mean
  equivalence ratio (heat release time delay is a
  function of mean equivalence ratio_

• Analysis
• Linear stability boundaries
• Amplitudes of oscillation and character of loss
  of stability (bifurcation)

56
Combustion Dynamics Control Model Calibration
and Use in Evaluation of System Modifications
Analysis allows calibration of model from data
to enable quantitative studies

• Evaluation of Mitigation Strategies
• Evaluate passive design changes (resonators) for
  size, placement, prediction of performance
• Evaluate active control for actuation
  requirements (bandwidth) and prediction of
  performance

• Calibration
• System level model captures experimental data
  quantitatively

Data Analysis Key parameters extracted from
experiment (forced response tests) - trend in
equivalence ratio (time delay) drives dynamical
behavior
57
Dynamic Systems and Control Examples
58
Dynamic Systems and Control Flight Systems
59
Dynamics and Control Program Constrained
Multivariable Control
60
Problem Many UTC products are multi-input,
multi-output systems but multivariable control
theory is not useful for designing their control
systems difficult ad hoc designs
Reason Balancing performance against product
cost and weight results in products operating
near many physical constraints Popular control
synthesis methods do not include constraints in
their formulations
Approach Develop a multivariable control
synthesis method that explicitly recognizes
constraints in its formulation
61
The demand for efficiency pushes engine operation
to the physical limits controller must must
meet many constraints
Thermal efficiency increases with burner
temperature - nominal temperatures near melting
point of parts - temperature overshoots rapidly
degrade turbine life
Propulsive Efficiency requires larger fans -
structural constraints on fan speed
Fan and compressor efficiency best near stall,
surge, and flutter boundaries - operating
constraints to avoid instabilities
62
Automatic Flight Control Systems Must Control 4
Degrees of Freedom While Meeting Many Constraints
Dont stall rotor
Dont strike fuselage
Actuator stroke and rate limits
Pitch and roll limits
Limited engine power and responsiveness
63
Approach Optimization-Based Control Algorithm
Commands PLA
Performance Index weights, timing parameters
Constraints actuator limits actuator rate
limits max T4 max,min N1,N2 compressor op lines
Onboard, Constrained Actuator Time
History Optimization
Actuator Commands
Onboard Model/Estimation continuously
estimating state and uncertain parameters
Sensor Signals
Focus of control logic designer
64
UTRC Team
Dynamic Systems and Controls Jim Fuller,
Leena Singh, Danbing Seto
Informistics ( numerical algorithms )
Martin Appel
Software Technology William Weiss
65
Dynamics and Control Program Virtual Alignment
Via Misalignment Estimation
66
The Commanche Alignment Algorithm Transforms
Target Location and Velocity from EOTAS to INS
and Weapon Coordinates
EOTAS Electro-optic Target Acquisition System
NQIS Inertial Navigation System
Gun
The Alignment Algorithm Compensates for Bending,
Installation Misalignments, and Sensor Errors
for More Accurate Fire-control Goal maintain
alignment during even aggressive maneuvers
67
The Team
Key UTRC Team Jim Fuller and Leena Singh
This activity is part of the Comanche development
including Sikorsky Aircraft, Comanche
weapon system integrators Martin Marietta,
Electro-optic Target Acquisition System
Litton, NQIS and rate gyros General
Electric, gun system
68
Target search or tracking
EOTAS rate gyro frame
NQIS rate gyro frame
Maneuver stresses
Nominal rotation of seeker base wrt NQIS
platform
Tilting of gyro mount wrt EOTAS platform
Rotation thru EOTAS gimbals to seeker base
Rotation of seeker base wrt its
nominal orientation due to fuselage bending
Tilting of gyro mount wrt NQIS platform
NQIS Mis-Mount Estimator
Seeker Mis-Mount Estimator
Gimbal angle resolvers
Aircraft Blueprints
Bending Estimator
SG C SE
SE C SB0
SB0 C SB
SB C NB
NB C NG
Target position in seeker coordinates
Target position in aircraft coordinates
Red and Blue are part of the Kalman Filter Based
Alignment Algorithm
69
Rationale and Approach
The inertial navigation, target acquisition, and
gun systems each have a triad of rate gyros to
support their operation
The difference between the angular rates of two
components - is primarily a measure of
bending rate - secondarily a measure of mount
errors, gyro biases, and bending - can
be put in form of Kalman filter measurement
equation
Solution Estimate the misalignment terms via a
Kalman filter
70
Complicating Factors
For any given rotation, -only the
2-DOF of bending orthogonal to the rotation
are observable -degree of
observability is proportional to rotation rate
magnitude
Rate gyros have slowly varying random biases
- integral of bending rate measurement has large
low frequency errors
Need time varying filter gains, but covariance
propagation requires too much computation
Solution Quasi-steady state Kalman filter
71
Bending Estimator Formulated as Kalman Filter
IwSGSG
y1
SBCSG


3 EOTAS rate gyros
Direction Cosine
Rate/Position estimator
qS

-
Gimbal Kinematics
3 measurements
5 EOTAS gimbal resolvers
SGwSBSB
SB0CNG
IwNGNG
y2
3 NQIS rate gyros
Predicted measurement
H
F
Z-1
-
Estimate of fuselage bending between EOTAS base
and NQIS base


K

15 states
Kalman Gain
Quasi-steady Kalman gains are scheduled analytical
ly via an invented time varying transform
72
Dynamic Systems and Control Example Control of
Separated Flows
73
Subsonic Engine Flow Control

• External Cowl
• Drag reduction

• Fan Nozzle
• Area control

• Subsonic Diffuser
• Separation control

• Jet Noise
• Community noise

• Fan and Compressor
• Separation control (fewer stages)
• Clearance (margins, performance)
• Noise

• Inlet Lip
• Separation control

• Combustion Mixing
• Dynamic mixing enhancement

74
Enabling process
Dynamic signature of separation
Diffuser rig
- subscale experiment - parametric studies -
testbed for dynamic- model-based control -
testbed for CFD-enabled model and control design
Low order dynamic models (Galerkin, black box,
phenom.)
Model-based control (experiment, CFD)
Sikorsky, PW
DEMO
- impact UTC products - implement and evaluate
dynamic-model-based control design on real life
applications
75
Methods and Issues
Fluid dynamics - diffuser geometry - boundary
conditions - boundary layer - shear layer - onset
of separation - flow transitions - hydrodynamic
instabilities - large-scale structures and
their temporal dynamics - turbulence and mixing -
mechanics of actuation and affects on flow
structures
Dynamic modeling - phenomenological models -
dimensional analysis - simplified NSE (integral
eqs., parabolized eqs., self-similar
solutions) - vortex methods - flow simulations
(DNS, turbulence modeling) - POD methodology -
Galerkin/POD models (analytical,
solver-based) - black-box models (ANN) -
CFD-based - based on experimental data - model
analysis (ROM) - model analysis (tied to CFD
models)
Control of separation - control strategy -
model-based control - actuators (local,
distributed) - cost functional and actuation
authority - observers (POD-based, NLPC
based) - optimization of control parameters
- design optimization
Team Satish Narayanan, Bernd Noack, Alexander
Khibnik, Andrzej Banaszuk University connections
Princeton, Cornell, U.Houston, KIAM (Moscow),
McGill, UCSD, UCSB, Max-Planck Inst., KU Leuven
76

• Flow separation
• Motivation objectives
• predict dynamics of separated flows
• understand physics/dynamics of separation
  (low-D ?)
• develop dynamical models capturing essential
  dynamics
• enhance performance of devices involving flow
  separation
• design demonstrate model/physics-based flow
  control strategies
• active control stall in high-angle-of-attack
  airfoils,
• engine/axial fan inlet flows, thrust vectoring
• Approach (flow separation in 2D diffuser)
• Numerical (2D CFD low Re, exact)
  spatiotemporal flow fields
• Dynamical analysis modeling identify dominant
  modes, low-D
• extract (parameter-dependent) dynamical models
• parametric/bifurcation analysis of models

77
2D DNS results (Rew1 30,000) N/w14 first
transition 6o 2q6o
2q6.5o
2q8o
Snapshots of kinetic energy fields
Longitudinal velocity traces (centerline of
expansion exit)

• Appearance of low freq. oscillations
• Onset of asymmetry

78
Empirical eigenfunctions

• Spatial patterns and temporal dynamics computed
  using POD (Karhunen-Loève) method
• POD modes used for Galerkin projection of
  governing equations
• POD coefficients used for training black-box
  models

• How POD is done?
• method of snapshots (equivalent to SVD)
• data mapped to standard rectangular domain (grid
  same for different angles)
• data symmetrized by adding mirror image
• data sparsed and mean subtracted (definition of
  mean?)
• scalar fields weighted stacked together
  (scaling? choice of fields?)
• data for several angles stacked to form
  representative data set to span fields for
  parameter range of interest (equal
  representation?)

79
Flow reconstruction
fi(x)
2q8.5o
uN(x,t) Si ai(t) fi(x) i1, ., N

• POD modes computed for ensemble of KE fields
  spanning 5.4o
• 1,5 20 modes capture 47,76 95 energy

Cumulative POD energy spectrum
Notes - 1 mode captures location of large
structures - 5 modes capture asymmetry - 10
modes start capturing small scale details - 20
modes add very little to the picture of 10 modes
80
Galerkin solver-based model

• Idea Use CFD as a time-stepper and build a
  projection layer around it.
• Takes care of geometry automatically
• Parametric/bifurcation analysis feasible

CFD time-stepper
CFD/ Galerkin model
U - POD modes

• Neural Network model
• NN model to predict interpolate system dynamics
• NN model trained on limited temporal data set
  (POD coeffs.)
• goal trace attractors (long term predictions) as
  parameters vary
• Discrete network two-hidden-layer network for
  discrete time DS identification
• fitted function X(n1) F( X(n), X(n-1),
  ... P) X - state variable, P - parameter

81
Bifurcation scenario
Asymmetric chaotic regime
Unknown
Asymmetric invariant torus
Asymmetric limit cycle/ chaotic regime
Asymmetric limit cycle
Secondary Hopf
Asymmetric equilibrium
Hopf
Symmetry breaking
Symmetric equilibrium
Symmetric limit cycle
0o
10o
2q
Multistability
82
Dynamic Systems and Control Example Enclosure
Noise Control
83
UTC Products Require Quiet Interiors

• Similarities
• Mechanisms exterior excitation, structureborne
  and airborne paths point and distributed sources
• Content broadband and tonal, low to high
  frequencies
• Complicated subsystem coupling
• Goal reduce cabin noise using active control

84
Product Requirements Drive Program Content
Division Product Requirements Helicopter Cabin
Noise Automotive Interior Noise Elevator
Interior Noise Equipment Room Noise Commuter
Aircraft Noise
Research Program Content
Requirements
Technology
Integrated System Design
85
Typical Helicopter Spectrum
86
Number of Acoustic Modes
600
500
400
Number of modes
300
200
100
0
0
200
400
600
800
Frequency

• For an acoustic space 5x6x9
• Global control with speakers and microphones is
  not feasible

87
Gear-Mesh Noise Control Architecture
Transmission
Sensors
Controller
Actuators
Noise
source
88
Control System Schematic
Rotor Speed Reference Signal
Adaptation
h
h
Sensors
u
z
y
Controller
S
x
x
Sample
Harmonic Estimator (Demodulate)
Remodulate
Plant T
i,j
89
Performance

• Simultaneous performance at three tones
• Optimized actuation configuration with minimum
  degrees of freedom

Fundamental, f1
Fundamental, f2
Harmonic 2f1
100
100
100
90
90
90
80
80
80
Performance (dB)
70
70
70
60
60
60
50
50
50
5
10
15
20
25
30
5
10
15
20
25
30
5
10
15
20
25
30
Microphone number
Microphone number
Microphone number
Overall
90
Adaptation Performance

• Vary frequency by /- 1 (10 seconds for full
  cycle)
• Adaptation maintains performance