A Ph.D. Dissertation Defense

1
Auto-calibration Control Applied To
Electro-Hydraulic Valves

• A Ph.D. Dissertation Defense
• Presented to the Academic Faculty
• By
• PATRICK OP DEN BOSCH
• Committee Members
• Dr. Nader Sadegh (Co-Chair, ME)
• Dr. Wayne Book (Co-Chair, ME)
• Dr. Chris Paredis (ME)
• Dr. Bonnie Heck Ferri (ECE)
• Dr. Roger Yang (HUSCO Intl.)

The George W. Woodruff School of Mechanical
Engineering Georgia Institute of
Technology Atlanta, GA October 30, 2007
2
PRESENTATION OUTLINE

• RESEARCH MOTIVATION
• PROBLEM STATEMENT
• INVERSE MAPPING LEARNING STATE CONTROL
• SIMULATION RESULTS
• APPLICATION TO HYDRAULICS
• EXPERIMENTAL VALIDATION
• CONCLUSION

3
RESEARCH MOTIVATION
Excavator

• CURRENT APPROACH
• Electronic control
• Use of solenoid Valves
• Energy efficient operation
• New electrohydraulic valves
• Conventional hydraulic spool valves are being
  replaced by assemblies of 4 independent valves
  for metering control

Low Pressure
High Pressure
Spool Valve
Spool piece
Spool motion
Kramer (1984), Roberts (1988), Garnjost (1989),
Jansson and Palmberg (1990), Aardema (1999),
Tabor (2005)
Piston
Piston motion
4
RESEARCH MOTIVATION
Backhoes

• CURRENT APPROACH
• Electronic control
• Use of solenoid Valves
• Energy efficient operation
• New electrohydraulic valves
• Conventional hydraulic spool valves are being
  replaced by assemblies of 4 independent valves
  for metering control

Kramer (1984), Roberts (1988), Garnjost (1989),
Jansson and Palmberg (1990), Aardema (1999),
Tabor (2005)
5
RESEARCH MOTIVATION

• ADVANTAGES
• Independent control
• More degrees of freedom
• More efficient operation
• Simple circuit
• Ease in maintenance
• Distributed system
• No need to customize

NASA Ames Flight Simulator

• DISADVANTAGES
• Nonlinear system
• Complex control

Kramer (1984), Roberts (1988), Garnjost (1989),
Jansson and Palmberg (1990), Aardema (1999),
Tabor (2005)
6
RESEARCH MOTIVATION
HUSCOS CONTROL TOPOLOGY
INCOVA LOGIC (VELOCITY BASED CONTROL)
Steady State Mapping (Design)
OPERATOR INPUT Commanded Velocity
INVERSE MAPPING (FIXED LOOK-UP TABLE)
EHPV Opening
COIL CURRENT SERVO (PWM dither)
Inverse Mapping (Control)

• Tabor and Pfaff (2004), Tabor (2004,2005)

HUSCO OPEN LOOP CONTROL FOR EHPVs
7
RESEARCH MOTIVATION
IMPROVED CONTROL TOPOLOGY
INCOVA LOGIC (VELOCITY BASED CONTROL)
Steady State Mapping (Design)
OPERATOR INPUT Commanded Velocity
MAPPING LEARNING CONTROL
EHPV Opening
COIL CURRENT SERVO (PWM dither)
Inverse Mapping (Control)

• Hierarchical control Top Level Controller,
  Mid-Level Controller, and Low Level Controller

HUSCO OPEN LOOP CONTROL FOR EHPVs
8
RESEARCH MOTIVATION

• Theoretical Research Questions
• How well can the systems inverse input-state
  mapping be learned online while trying to achieve
  state tracking control?
• How can the tracking error dynamics and mapping
  errors be driven arbitrarily close to zero with
  an auto-calibration method?
• Experimental Research Questions
• How can the performance of solenoid driven poppet
  valves be improved?
• How well can these calibration mappings be
  learned online?
• How can the learned mappings be used for fault
  detection?

9
RESEARCH MOTIVATION

• Theoretical Objectives
• Development of a general formulation for control
  of nonlinear systems with parametric uncertainty
  and possibly time varying characteristics.
• Development of an auto-calibration method for
  nonlinear systems.
• Analysis of conditions that ensure small state
  tracking errors and mapping errors.
• Experimental Objectives
• Improve flow conductance control of EHPVs.
• Validation of the proposed method.
• Study accuracy of the auto-calibration method.
• Study of online auto-calibration with fault
  diagnostics.

10
RESEARCH MOTIVATION

• Benefits
• Alternative method for nonlinear control design
• Better valve control/velocity control
• No individual offline calibration required
• Method can be used as an infield calibration
• Method can be also used for different valve sizes
• Learned mapping more accurately reflects valve
  behavior
• EHPV transients can be corrected
• Valve fault detection can be implemented
• Maintenance scheduling can be implemented from
  monitoring and detecting the deviations from the
  normal pattern of behavior.

11
PRESENTATION OUTLINE

• RESEARCH MOTIVATION
• PROBLEM STATEMENT
• INVERSE MAPPING LEARNING STATE CONTROL
• SIMULATION RESULTS
• APPLICATION TO HYDRAULICS
• EXPERIMENTAL VALIDATION
• CONCLUSION

12
PROBLEM STATEMENT

• Consider a general discrete-time nonlinear
  dynamic plant

13
PROBLEM STATEMENT

• Consider a general discrete-time nonlinear
  dynamic plant

CONTROL PROBLEM
14
PROBLEM STATEMENT

• Consider a general discrete-time nonlinear
  dynamic plant

15
PROBLEM STATEMENT

• Assumptions

16
PROBLEM STATEMENT

• Assumptions

17
PROBLEM STATEMENT

• Assumptions

18
PROBLEM STATEMENT

• Proposition

Similar Results in Levin and Narendra
(1993,1996), Sadegh(1991,2001)
19
PRESENTATION OUTLINE

• RESEARCH MOTIVATION
• PROBLEM STATEMENT
• INVERSE MAPPING LEARNING STATE CONTROL
• SIMULATION RESULTS
• APPLICATION TO HYDRAULICS
• EXPERIMENTAL VALIDATION
• CONCLUSION

20
INVERSE MAPPING LEARNING CTRL
Adaptive Inverse Dynamics Control
Control of plant dynamics and control of plant
noise
Depends on the accuracy of the model
Affected by disturbances, noise, unmodeled
nonlinearities, uncertainties
Widrow (1982, 1987, 1993, 1996)
21
INVERSE MAPPING LEARNING CTRL
Inverse Mapping Control
NN with backpropagation
Driven to reduce the output error
Malinowski (1995)
22
INVERSE MAPPING LEARNING CTRL
Online Direct Learning
Recurrent hybrid NN
Backpropagation training
Uses output error
Uses error predictor
Pham and Oh (1993, 1994)
23
INVERSE MAPPING LEARNING CTRL
Inverse Model Control
Internal Model Control
Recurrent hybrid NN
Direct and indirect learning approach
Backpropagation training
Requires feedback controller
Pham and Yildirim (2000, 2002)
24
INVERSE MAPPING LEARNING CTRL
The plant is linearized about a desired state
trajectory
A Nodal Link Perceptron Network (NLPN) is
employed in the feedforward loop and trained with
feedback state error
The control scheme needs the plant Jacobian and
controllability matrices, obtained offline
Approximations of the Jacobian and
controllability matrices can be used without
loosing closed loop stability
Sadegh (1991,1993,1995)
25
INVERSE MAPPING LEARNING CTRL
NLPN Based Input Matching Control (INMAC)
Direct learning accomplished via
Feedforward control by
26
INVERSE MAPPING LEARNING CTRL
NLPN Based Input Matching Control (INMAC)
Direct learning accomplished via
Feedforward control by
How can this function be approximated/learned?
27
INVERSE MAPPING LEARNING CTRL
NLPN Based Input Matching Control (INMAC)
Direct learning accomplished via
Functional Approximator
Perceptron with single hidden layer
Nodal Link Perceptron Network (NLPN)
Compatible with lookup tables
Local basis function activation
28
INVERSE MAPPING LEARNING CTRL
NLPN Based Input Matching Control (INMAC)
Adaptation
Steepest Descent (SD)
Recursive Least Squares (RLS)
29
INVERSE MAPPING LEARNING CTRL
Composite Input Matching Control (COMPIM)
30
INVERSE MAPPING LEARNING CTRL
Composite Input Matching Control (COMPIM)
31
INVERSE MAPPING LEARNING CTRL
Deadbeat Control and Non-deadbeat Control
Deadbeat Control Law
Non-deadbeat Control Law
Example Linear Time Invariant Plant
Deadbeat
Non-deadbeat
32
INVERSE MAPPING LEARNING CTRL
Composite Input Matching Control (COMPIM)
Stability Analysis
THEOREM 1 Steepest Descent (SD)
(and non-deadbeat)
Control Law
Adaptation
Conditions
Meets PE condition
33
INVERSE MAPPING LEARNING CTRL
Composite Input Matching Control (COMPIM)
Stability Analysis
THEOREM 1 Steepest Descent (SD)
(and non-deadbeat)
Control Law
If
Then
34
INVERSE MAPPING LEARNING CTRL
Composite Input Matching Control (COMPIM)
Stability Analysis
THEOREM 2 Recursive Least Squares (RLS)
(and non-deadbeat)
Control Law
Adaptation
Conditions
Meets PE condition
35
INVERSE MAPPING LEARNING CTRL
Composite Input Matching Control (COMPIM)
Stability Analysis
THEOREM 2 Recursive Least Squares (RLS)
(and non-deadbeat)
Control Law
If
Then
36
INVERSE MAPPING LEARNING CTRL
Composite Input Matching Control (COMPIM) General
Case
Plant
Example
37
INVERSE MAPPING LEARNING CTRL
Composite Input Matching Control (COMPIM) General
Case
Plant
Feedforward
Direct Learning
38
PRESENTATION OUTLINE

• RESEARCH MOTIVATION
• PROBLEM STATEMENT
• INVERSE MAPPING LEARNING STATE CONTROL
• SIMULATION RESULTS
• APPLICATION TO HYDRAULICS
• EXPERIMENTAL VALIDATION
• CONCLUSION

39
SIMULATION RESULTS
FIRST ORDER LINEAR PLANT
Sampling Time
Plant
Parameters
40
SIMULATION RESULTS
FIRST ORDER LINEAR PLANT
Sampling Time
Plant
Parameters
41
SIMULATION RESULTS
FIRST ORDER NONLINEAR PLANT
Plant
Initial Mapping
42
SIMULATION RESULTS
FIRST ORDER NONLINEAR PLANT
RLS
SD
43
SIMULATION RESULTS
SECOND ORDER NONLINEAR PLANT
States
Inputs
Current sent to S1
Current sent to S2
44
SIMULATION RESULTS
SECOND ORDER NONLINEAR PLANT
State 1
45
SIMULATION RESULTS
SECOND ORDER NONLINEAR PLANT
State 2
46
PRESENTATION OUTLINE

• RESEARCH MOTIVATION
• PROBLEM STATEMENT
• INVERSE MAPPING LEARNING STATE CONTROL
• SIMULATION RESULTS
• APPLICATION TO HYDRAULICS
• EXPERIMENTAL VALIDATION
• CONCLUSION

47
APPLICATION TO HYDRAULICS
ELECTRO-HYDRAULIC POPPET VALVE (EHPV)
Coil Cap
Adjustment Screw

• Poppet type valve
• Pilot driven
• Solenoid activated
• Internal pressure compensation
• Virtually zero leakage
• Bidirectional
• Low hysteresis
• Low gain initial metering
• PWM current input

Modulating Spring
Input Current
Coil
Armature
Pilot Pin
Control Chamber
Armature Bias Spring
U.S. Patents (6,328,275) (6,745,992)
Pressure Compensating Spring
Main Poppet
Forward (Side) Flow
Reverse (Nose) Flow
48
APPLICATION TO HYDRAULICS
VALVE OPENING

• VALVE CHARACTERIZATION
• Flow Conductance

49
APPLICATION TO HYDRAULICS

• FORWARD MAPPING
• REVERSE MAPPING

Side to nose
Forward Kv at different input currents A
Nose to side
Reverse Kv at different input currents A
50
APPLICATION TO HYDRAULICS

• SIMPLIFIED EHPV MODEL

Forward Kv at different input currents A
Forward Kv
Reverse Kv at different input currents A
51
APPLICATION TO HYDRAULICS

• SIMPLIFIED EHPV MODEL

Forward Kv at different input currents A
Reverse Kv
Reverse Kv at different input currents A
52
PRESENTATION OUTLINE

• RESEARCH MOTIVATION
• PROBLEM STATEMENT
• INVERSE MAPPING LEARNING STATE CONTROL
• SIMULATION RESULTS
• APPLICATION TO HYDRAULICS
• EXPERIMENTAL VALIDATION
• CONCLUSION

53
EXPERIENTAL VALIDATION

• HYDRAULIC TEST-BED

CAN bus interface
Balluff position/velocity transducer
XPC-Target (SIMULINK)
Pressure Control
Flow Control
54
EXPERIENTAL VALIDATION

• SUPPLY PRESSURE CONTROL

First order nonlinear plant
Coordinate Transformation
55
EXPERIENTAL VALIDATION

• SUPPLY PRESSURE CONTROL

Desired Flow Conductance Kv
Pump Flow Characteristics
56
EXPERIENTAL VALIDATION

• SUPPLY PRESSURE CONTROL Generic Initial mapping
  

Flow Conductance Kv
Supply Pressure PS
57
EXPERIENTAL VALIDATION

• SUPPLY PRESSURE CONTROL Calibrated Initial
  mapping

Flow Conductance Kv
Supply Pressure PS
58
EXPERIENTAL VALIDATION

• SUPPLY PRESSURE CONTROL SD COMPIM with Generic
  Initial mapping

Flow Conductance Kv
Supply Pressure PS
59
EXPERIENTAL VALIDATION

• SUPPLY PRESSURE CONTROL RLS COMPIM with Generic
  Initial map

Flow Conductance Kv
Supply Pressure PS
60
EXPERIENTAL VALIDATION

• SUPPLY PRESSURE CONTROL

SD Flow Conductance Kv
RLS Flow Conductance Kv
61
EXPERIENTAL VALIDATION
FLOW CONTROL

• Modeling of valves flow mapping
• Online approach without removal from overall
  system
• Combination of model based approach,
  identification, and NN approximation
• Comparison among automated modeling, offline
  calibration, and manufacturers calibration

Liu and Yao (2005)
62
EXPERIENTAL VALIDATION

• FLOW CONTROL

Control Topology
INCOVA LOGIC (VELOCITY BASED CONTROL)
OPERATOR INPUT Commanded Velocity
INVERSE MAPPING (FIXED LOOK-UP TABLE)
INVERSE MAPPING (ADAPTIVE LOOK-UP TABLE)
EHPV Opening
COIL CURRENT SERVO (PWM dither)
63
EXPERIENTAL VALIDATION

• FLOW CONTROL

Flow Conductance Kv
Piston Position/Velocity
64
EXPERIENTAL VALIDATION

• FLOW CONTROL

Flow Conductance Kv
Piston Position/Velocity
65
EXPERIENTAL VALIDATION

• FLOW CONTROL

Flow Conductance Kv
Piston Position/Velocity
66
EXPERIENTAL VALIDATION

• FLOW CONTROL

Flow Conductance Kv
Piston Position/Velocity
67
EXPERIENTAL VALIDATION

• HEALTH MONITORING

Flow Conductance Bounds
Control Topology
68
EXPERIENTAL VALIDATION

• HEALTH MONITORING

Health Indicator Logic
69
EXPERIENTAL VALIDATION

• HEALTH MONITORING

70
EXPERIENTAL VALIDATION

• HEALTH MONITORING

71
PRESENTATION OUTLINE

• RESEARCH MOTIVATION
• PROBLEM STATEMENT
• INVERSE MAPPING LEARNING STATE CONTROL
• SIMULATION RESULTS
• APPLICATION TO HYDRAULICS
• EXPERIMENTAL VALIDATION
• CONCLUSION

72
CONCLUSIONS

• RESEARCH CONTRIBUTIONS
• Deadbeat/non-deadbeat control method based on
  input matching with composite adaptation
• Rigorous closed-loop stability analyses for the
  above controllers using steepest descent and
  recursive least squares methods
• A procedure to handle arbitrary state and input
  delays
• A model of the EHPV
• Intelligent control technology for the EHPV

• RESEARCH IMPACT
• An alternative discrete-time control design based
  on an auto-calibration scheme for nonlinear
  systems
• Improvement of hydraulic controls using solenoid
  driven valves based on calibration routines
• Intelligent control technology for the hydraulic
  industry
• Easily extended to other engineering applications

73
CONCLUSIONS

• FUTURE RESEARCH
• Extend these results for output control
• Consider/develop other schemes that suffers less
  from the curse of dimensionality
• Relax the PE condition
• Apply this scheme to other hydraulic component
  with higher order dynamics
• Apply this control method to other metering modes
  along with multi-function cases and mode
  switching

THANK YOU FOR YOUR ATTENTION