ANALOG AND LABVIEW- BASED CONTROL OF A MAGLEV … IMECE2005_paper81600_sli… · ANALOG AND...
Transcript of ANALOG AND LABVIEW- BASED CONTROL OF A MAGLEV … IMECE2005_paper81600_sli… · ANALOG AND...
ANALOG AND LABVIEW-BASED CONTROL OF A MAGLEV SYSTEM WITH
NI-ELVISRISHABH SINHACooling Design CenterLarge Power Systems DivisionCaterpillar Inc.Mossville, IL 61552 [email protected]
MARK NAGURKADepartment of Mechanical &
Industrial EngineeringMarquette University
Milwaukee, WI 53201 [email protected]
IMECE2005-816002005 ASME International Mechanical Engineering Congress and Exposition
November 5-11, 2005, Orlando, Florida USA
Sinha & NagurkaIMECE2005-81600 2
Marquette UniversityMarquette University
View from 5th floor of Olin Engineering Bldg, Oct 27, 2005
Sinha & NagurkaIMECE2005-81600 3
Outline of PresentationOutline of Presentationo Objectiveso Backgroundo Maglev Testbedo Control Strategieso Simulation Studieso Experimental Studieso Summaryo Future Work
Sinha & NagurkaIMECE2005-81600 4
ObjectivesObjectivesUsing a low cost maglev system for
mechatronics education:
o Develop a linearized modelo Investigate classical linear controllerso Implement controllers using analog
circuits on NI-ELVISo “Implement” controllers using LabVIEWo Compare performance of controllers
Sinha & NagurkaIMECE2005-81600 5
Outline of PresentationOutline of Presentationo Objectiveso Backgroundo Maglev Testbedo Control Strategieso Simulation Studieso Experimental Studieso Summaryo Future Work
Sinha & NagurkaIMECE2005-81600 6
What is Maglev?What is Maglev?o Maglev = Magnetic Levitation
o Levitate objects by electro-magnetic forces to cancel effect of gravity
o Established technology : o high-speed maglev vehicleso maglev bearingso vibration isolation systems, etc.
Sinha & NagurkaIMECE2005-81600 7
ExamplesExamples
Maglev train (China)(speeds of 430 km/h) Maglev Bearing
Levitron
Sinha & NagurkaIMECE2005-81600 8
Control StrategiesControl Strategieso Linear Control
o Classical PIDo PID with gain schedulingo Phase-lead, Phase-lago LQR, LQE, LQG, H∞, µ-synthesis
o Nonlinear Controlo On-offo PWMo Fuzzy-logico Neural-network controlo Feedback linearizationo Adaptive controlo Backstepping theory
Sinha & NagurkaIMECE2005-81600 9
Outline of PresentationOutline of Presentationo Objectiveso Backgroundo Maglev Testbedo Control Strategieso Simulation Studieso Experimental Studieso Summaryo Future Work
Sinha & NagurkaIMECE2005-81600 10
Physical TestbedPhysical Testbed
Sinha & NagurkaIMECE2005-81600 11
Testbed PictorialTestbed Pictorial
Sinha & NagurkaIMECE2005-81600 12
o Infrared emitter and detector pair
o Acts as a variable resistoro Levitated object blocks
path, changes light intensity
o Linear behavior in operating region
Infrared (IR) SensorInfrared (IR) Sensor
sv xβ=
Displacement
Sen
sor O
uput
svxβ
=
IR Emitter IR DetectorLevitatedObject
Sinha & NagurkaIMECE2005-81600 13
Sensor and Controls LogicSensor and Controls Logico The emitter generates constant
light intensity.o The detector signal is amplified
and compared with a reference voltage.
o Difference of signals used to adjust current to electromagnet.o If the levitated object is too close to
electromagnet (detected IR signal too small), the current is reduced.
o If the levitated object is too far (detected IR signal too large), the current is increased.
IR Emitter IR Detector
Sinha & NagurkaIMECE2005-81600 14
ActuatorActuatoro Actuator is electro-
magnetic coil with steel bolt core
o Electromagnet obtained commercially
o FEA performed to check for magnetic saturation
o Power supply:18 V / 3 A
Sinha & NagurkaIMECE2005-81600 15
Maglev Temperature Expt.Maglev Temperature Expt.
Mechatronics Laboratory, Marquette University
Thermistor mounted to coil used to
measure temperature
during levitation
Sinha & NagurkaIMECE2005-81600 16
( ) 1o oL xL x Lx
= +
Nonlinear ModelNonlinear Model( ) ( ) ( )2
, ,2
i t dL xf x i t
dx= −
( ) ( )( )
( )( )
2 2
, ,2o o i t i tL xf x i t C
x t x t⎛ ⎞ ⎛ ⎞
= =⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟⎝ ⎠ ⎝ ⎠
( ) ( )2
2 , ,d x t
m mg f x i tdt
= −
( ) ( )( )
22
2
d x t i tm mg C
dt x t⎛ ⎞
= − ⎜ ⎟⎜ ⎟⎝ ⎠
Sinha & NagurkaIMECE2005-81600 17
Linearized ModelLinearized Model
( ) ( ) ( )2 2
2 3
2 2, , ...o o o
o o o
i Ci Cif x i t C i t x tx x x
⎛ ⎞ ⎛ ⎞ ⎛ ⎞= + − +⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎝ ⎠ ⎝ ⎠ ⎝ ⎠
( ) 1, , ...of x i t f f= + +
( )2
, oo
o
if x i C mgx
⎛ ⎞= =⎜ ⎟
⎝ ⎠
( ) ( ) ( )2
1 2 3
2 2, , o o
o o
Ci Cif x i t i t x tx x
⎛ ⎞ ⎛ ⎞= −⎜ ⎟ ⎜ ⎟⎝ ⎠ ⎝ ⎠
( )2
2
d x tm mg f
dt= −
( ) ( ) ( )2 2
2 2 3
2 2o oo
o o
d x t Ci Cim mg f i t x tdt x x
⎛ ⎞ ⎛ ⎞= − − +⎜ ⎟ ⎜ ⎟
⎝ ⎠ ⎝ ⎠
Sinha & NagurkaIMECE2005-81600 18
Linearized Model (cont’d)Linearized Model (cont’d)2
oo
o
if C mgx
⎛ ⎞= =⎜ ⎟
⎝ ⎠
( ) ( ) ( )2 2
2 2 3
2 2o o
o o
d x t Ci Cim i t x tdt x x
⎛ ⎞ ⎛ ⎞= − +⎜ ⎟ ⎜ ⎟
⎝ ⎠ ⎝ ⎠
( ) ( ) ( )2
22 3
2 2o o
o o
Ci Cims X s I s X sx x
⎛ ⎞ ⎛ ⎞= − +⎜ ⎟ ⎜ ⎟
⎝ ⎠ ⎝ ⎠
( ) ( )( )
2 2
2 22 2
3 3
2 2
2 2
o o
o o
o o
o o
Ci Cix mxX s
G sI s Ci Cims s
x mx
⎛ ⎞ ⎛ ⎞− −⎜ ⎟ ⎜ ⎟⎝ ⎠ ⎝ ⎠= = =⎛ ⎞ ⎛ ⎞
− −⎜ ⎟ ⎜ ⎟⎝ ⎠ ⎝ ⎠
Sinha & NagurkaIMECE2005-81600 19
Block DiagramBlock Diagram
( ) ( )( )
2 2
2 22 2
3 3
2 2
2 2
o o
o o
o o
o o
Ci Cix mxX s
G sI s Ci Cis m s
x mx
⎛ ⎞ ⎛ ⎞− −⎜ ⎟ ⎜ ⎟⎝ ⎠ ⎝ ⎠= = =⎛ ⎞ ⎛ ⎞
− −⎜ ⎟ ⎜ ⎟⎝ ⎠ ⎝ ⎠
Sinha & NagurkaIMECE2005-81600 20
Alternate Plant Transfer FunctionAlternate Plant Transfer Functiono The electromagnet
represented as a series combination of a resistor and inductor
o Consider sensor output as the system output
o Consider voltage to the electromagnet as the system input
( ) ( ) ( ) ( ), ,di t
v x i t Ri t L xdt
= +
( ) ( )( )
2
22
3
2
2
o
oso
o
o
CimLxV s
G sV s CiRs s
L mx
β⎛ ⎞−⎜ ⎟⎝ ⎠= =
⎛ ⎞⎛ ⎞+ −⎜ ⎟⎜ ⎟⎝ ⎠⎝ ⎠
Sinha & NagurkaIMECE2005-81600 21
Block Diagram IIBlock Diagram II
( ) ( )( )
2
22
3
2
2
o
oso
o
o
CimLxV s
G sV s CiRs s
L mx
β⎛ ⎞−⎜ ⎟⎝ ⎠= =
⎛ ⎞⎛ ⎞+ −⎜ ⎟⎜ ⎟⎝ ⎠⎝ ⎠
Sinha & NagurkaIMECE2005-81600 22
Open-loop AnalysisOpen-loop Analysis
( ) ( )( ) ( )824000
232 49.5 49.5oG ss s s
=+ + −
The root-locus has a pole in the right-half plane; the system is
unstable
Unstable
Sinha & NagurkaIMECE2005-81600 23
Outline of PresentationOutline of Presentationo Objectiveso Backgroundo Maglev Testbedo Control Strategieso Simulation Studieso Experimental Studieso Summaryo Future Work
Sinha & NagurkaIMECE2005-81600 24
Control Block DiagramControl Block Diagram
o Error signal calculated as difference of setpoint and sensor output
o Error signal is input to controllero Bias is added to controller output, amplified,
and input to “maglev plant”o Controlled output is actual output
Sinha & NagurkaIMECE2005-81600 25
Control StrategiesControl Strategieso Proportional Control
o Controller output is proportional to error
o Lead-Lag Controlo Adds a zero and a pole to the
open-loop system
( )P PG s K=
( )P PG s K=
( ) ( ) ( )Leads zG s K K s z ps p+
= + →∞+
( )/Lead Lags zG s Ks p+
=+
00
Lead for z pLag for p z
< << <
Sinha & NagurkaIMECE2005-81600 26
Control StrategiesControl Strategieso PD Control
o Adds derivative term
o PID Controlo Integral term makes steady-state
error zero
( )PD P DG s K K s= +
( ) IPID P D
KG s K K ss
= + +
Sinha & NagurkaIMECE2005-81600 27
Outline of PresentationOutline of Presentationo Objectiveso Backgroundo Maglev Testbedo Control Strategieso Simulation Studieso Experimental Studieso Summaryo Future Work
Sinha & NagurkaIMECE2005-81600 28
Proportional ControllerProportional ControllerProportional Controller acts
as a gain; cannot move poles into left-half plane
Proportional Controller by itself cannot stabilize
open-loop unstable system
Unstable
Sinha & NagurkaIMECE2005-81600 29
Lag CompensatorLag CompensatorLag Compensator
cannot move poles into left-half plane
Lag Compensator by itself cannot stabilize
an open-loop unstable system
Unstable
Sinha & NagurkaIMECE2005-81600 30
Lead CompensatorLead CompensatorLead Compensator
moves poles into left-half plane
Lead Compensator can stabilize open-
loop unstable system
Gain Margin: 13.2 dBPhase Margin: 15.8 deg
Stable
Sinha & NagurkaIMECE2005-81600 31
Lead Compensator ResponseLead Compensator ResponseStep ResponseSettling Time: 0.157sSteady-State Error: 0.24
Impulse ResponseSettling Time: 0.164s
(Zero at -100 rad/s, Pole at -1000 rad/s)
Sinha & NagurkaIMECE2005-81600 32
PD ControllerPD ControllerPD Controller
moves poles into left-half plane
PD Controller can stabilize open-loop
unstable systemGain Margin: 2.61 dB
Phase Margin: 7.28 deg
Stable
Sinha & NagurkaIMECE2005-81600 33
PI ControllerPI Controller
PI Controller cannot stabilize
open-loop unstable system
PI Controller cannot move poles into left-half plane
Unstable
Sinha & NagurkaIMECE2005-81600 34
PID ControllerPID ControllerPID Controller
moves poles into left-half plane
PID Controller can stabilize open-loop
unstable systemGain Margin: 3.42 dB
Phase Margin: 8.94 deg
Stable
Sinha & NagurkaIMECE2005-81600 35
PID Controller ResponsePID Controller ResponseStep ResponseSettling Time: 0.354sSteady-State Error: 0
Impulse ResponseSettling Time: 0.172s
(KP=0.68, KI=3.12, KD=0.0141)
Sinha & NagurkaIMECE2005-81600 36
Outline of PresentationOutline of Presentationo Objectiveso Backgroundo Maglev Testbedo Control Strategieso Simulation Studieso Experimental Studieso Summaryo Future Work
Sinha & NagurkaIMECE2005-81600 37
Maglev Experimental StationMaglev Experimental Station
Mechatronics Laboratory, Marquette University
Sinha & NagurkaIMECE2005-81600 38
NI-ELVISNI-ELVIS
Mechatronics Laboratory, Marquette University, Oct 27, 2005
National Instruments’Educational Laboratory Virtual Instrumentation Suite
Sinha & NagurkaIMECE2005-81600 39
NI-ELVIS Prototyping BoardNI-ELVIS Prototyping Board
Mechatronics Laboratory, Marquette University, Oct 27, 2005
National Instruments’Educational Laboratory Virtual Instrumentation Suite
Sinha & NagurkaIMECE2005-81600 40
Analog Control with NI-ELVISAnalog Control with NI-ELVIS
Mechatronics Laboratory, Marquette University, Oct 27, 2005
Sinha & NagurkaIMECE2005-81600 41
Lead Compensator Analog CircuitLead Compensator Analog Circuit
( ) 5 1
5 6
5 6 1
1
D
sensor
sR CVG s R RV s
R R C
+= =
++
Sinha & NagurkaIMECE2005-81600 42
Lead Compensator Analog ImplementationLead Compensator Analog Implementation
R11
1k
0
Coil
18V
Q22N3055
RS 276-145
0
IR PhotoTransistor
IRDiode
0
R810k
R2
33k
R1150 Ohm 25W
15V 15V
R10
15k
RS 276-143
0
D1
1N5400
-
+
U1
LM7413
26
7 14 5
15V
-15V
C1
0.1u
R5
100k
R6
11k
0
Sinha & NagurkaIMECE2005-81600 43
PD Analog CircuitPD Analog Circuit
SUMMER
-
+
U4
26
3
DERIVATIVE
INVERTER
Verror
RP1
00
RD
-
+
U1
26
3
RP2
R5
PROPORTIONAL
R1
R4
R2
0
Vcontrol0
CD-
+
U2
26
3
-
+
U3
26
3
R3
( ) ( )( ) ( )2
1control
errorD
P
V s RPG s RD CD sV s RP
KK
⎛ ⎞= = + ×⎜ ⎟⎝ ⎠
Sinha & NagurkaIMECE2005-81600 44
Sensor Output
Setpoint
Proportional
Derivative
Bias
SummationController
Output
Error
DifferenceSensor
PD Analog Circuit ImplementationPD Analog Circuit Implementation
KP=0.68, KD=0.0141
Sinha & NagurkaIMECE2005-81600 45
PID Analog CircuitPID Analog CircuitSUMMER
CI
-
+
U5
26
3
CD
INVERTER
Verror
INTEGRAL
RP1
0
R3
0
0
-
+
U1
26
3
RP2
R5
PROPORTIONAL
R1
R4
R2
0
Vcontrol
RI
0
-
+
U2
26
3
-
+
U4
26
3
RD
DERIVATIVE
-
+
U3
26
3
R4
( ) ( )( ) ( )2 1 1
1control
errorD
P I
V s RPG s RD CD sV s RP RI CI s
KK K
⎛ ⎞ ⎛ ⎞= = + + ×⎜ ⎟ ⎜ ⎟×⎝ ⎠ ⎝ ⎠
Sinha & NagurkaIMECE2005-81600 46
PID Analog Circuit ImplementationPID Analog Circuit Implementation
Sensor Output
Setpoint
Setpoint for Antiwindup
Proportional
Integral
Derivative
Bias
Summation Controller Output
Error
Difference
Difference
KP=0.68, KI=3.12, KD=0.0141
Sinha & NagurkaIMECE2005-81600 47
Analog ImplementationAnalog Implementation
Power Transistor
Setpoint for Anti-windup
Setpoint
Bias
Power Resistor
Mechatronics Laboratory, Marquette University
Sinha & NagurkaIMECE2005-81600 48
LabVIEW-based ControlLabVIEW-based Control
IRDiode
RS 276-145
RS 276-143IR PhotoTransistor
0
R1150 Ohm 25W
0
DA
C 0NI-ELVIS
0
18V
Coil
R233k
Q22N3055
R312k
15V
R33
180
15V D4
1N5400
AC
H 2
0
LabVIEW - Maglev.VI
Sinha & NagurkaIMECE2005-81600 49
LabVIEW-based ControlLabVIEW-based Control
Power Transistor
Power Resistor
Analog Output
Mechatronics Laboratory, Marquette University
Sinha & NagurkaIMECE2005-81600 50
LabVIEW VI Front PanelLabVIEW VI Front Panel
Sinha & NagurkaIMECE2005-81600 51
LabVIEW VI ProgramLabVIEW VI Program
Sinha & NagurkaIMECE2005-81600 52
Levitation of AA BatteryLevitation of AA Battery
Mechatronics Laboratory, Marquette University, Oct 27, 2005
Sinha & NagurkaIMECE2005-81600 53
Levitation of Thin RingLevitation of Thin Ring
Mechatronics Laboratory, Marquette University, Oct 27, 2005
Sinha & NagurkaIMECE2005-81600 54
Levitation of Nested WasherLevitation of Nested Washer
Mechatronics Laboratory, Marquette University
Sinha & NagurkaIMECE2005-81600 55
Levitation of 9V BatteryLevitation of 9V Battery
Mechatronics Laboratory, Marquette University, Oct 27, 2005
Sinha & NagurkaIMECE2005-81600 56
Levitation of Various ObjectsLevitation of Various Objects
Mechatronics Laboratory, Marquette University
Sinha & NagurkaIMECE2005-81600 57
Levitation of Various ObjectsLevitation of Various Objects
Mechatronics Laboratory, Marquette University
Sinha & NagurkaIMECE2005-81600 58
Levitation of Various ObjectsLevitation of Various Objects
Mechatronics Laboratory, Marquette University
Sinha & NagurkaIMECE2005-81600 59
Current ComparisonsCurrent Comparisons
N/AN/A2.27N/AN/A639.48%50.95Large Bolt
N/AN/A2.19N/AN/A587.37%47.369-Volt Battery
1.75N/A1.791.78N/A542.53%44.27Splined Ring
1.261.371.341.361.27240.93%23.49Thin Ring
1.221.301.241.291.16174.17%18.89AA Battery
1.721.721.601.681.59106.53%14.23Bolt 2
1.371.391.321.361.2954.43%10.64Bolt
1.461.491.471.491.3429.75%8.94Nested Washer
LV - PIDLV - PDA - PIDA - PDA -Lead
CURRENT (A)% VariationFrom 6.8gMASS (g)OBJECT
Sinha & NagurkaIMECE2005-81600 60
Addition of Second CoilAddition of Second Coil
Mechatronics Laboratory, Marquette University
Sinha & NagurkaIMECE2005-81600 61
StiffnessStiffness
Sinha & NagurkaIMECE2005-81600 62
PD Controller ResponsePD Controller Response
Sinha & NagurkaIMECE2005-81600 63
Outline of PresentationOutline of Presentationo Objectiveso Backgroundo Maglev Testbedo Control Strategieso Simulation Studieso Experimental Studieso Summaryo Future Work
Sinha & NagurkaIMECE2005-81600 64
SummarySummaryo Developed nonlinear and linearized modelo Conducted simulation studies to compare
controllers and select gains o Implemented PID-type controllers using analog
circuits with NI-ELVIS boardo Developed software-based controllers using
LabVIEW o Compared "hardware” vs. "software" based
controllers o Added second coil for determining stiffness and
system responseo Tested practical differentiator and anti-windup
circuits
Sinha & NagurkaIMECE2005-81600 65
MovieMovie
Sinha & NagurkaIMECE2005-81600 66
ClosingClosingo Applied classical linear control
strategies to a linearized model of a nonlinear system
o Implemented analog controllers using NI-ELVIS
o Developed LabVIEW controllerso Can dynamically change controller
type and parameters and observe effect on system in real-time
Sinha & NagurkaIMECE2005-81600 67
Outline of PresentationOutline of Presentationo Objectiveso Backgroundo Maglev Testbedo Control Strategieso Simulation Resultso Experimental Resultso Summaryo Future Work
Sinha & NagurkaIMECE2005-81600 68
Future WorkFuture Worko Implement nonlinear and optimal control
strategies in LabVIEW & compare performance with linear control strategies
o Develop a detailed FEA model to optimize the shape of the electromagnet for maximum field strength
o Design an array of sensors to translate levitated object vertically.
o Add horizontal electromagnets for two degree-of-freedom system.
Sinha & NagurkaIMECE2005-81600 69
Future Work (cont’d)Future Work (cont’d)
o Add velocity sensor for controlled damping
o Develop more advanced modelo Include gyro-dynamics
o Include skin friction using a CFD model
o Test remote access to VI through internet
o Create a virtual reality model using Virtual Reality Toolbox o Visualization of simulation without actual
implementation
The EndThe End
Questions ...Questions ...
Sinha & NagurkaIMECE2005-81600 72
Movie of Levitated BallMovie of Levitated Ball
Sinha & NagurkaIMECE2005-81600 73
Backup Slides
Sinha & NagurkaIMECE2005-81600 74
Magnetic Field FEA ResultsMagnetic Field FEA Results
Using FEMM package
Sinha & NagurkaIMECE2005-81600 75
Practical DifferentiatorPractical DifferentiatorCN
-
+3
26
0
CDVin
RD
Vout
o A pole needs to be added to differentiator circuit for “roll-off” of high frequency response.
o Without additional pole, differentiator circuit just amplifies any high frequency noise signals.
o Need to filter high frequency signals from signal going to the differentiator.
o Capacitor, CN, added in the differentiator circuit
( )( )
( )( ) 1
out
in
V s RD sV s RD CN s
=× +
Sinha & NagurkaIMECE2005-81600 76
Anti-windup CircuitAnti-windup Circuito The integral circuit
integrates the error with respect to time
o If error is large & constant, e.g., when the object is not present, integral action will continuously increase until saturation
o Need an “anti-windup” circuit
Sinha & NagurkaIMECE2005-81600 77
System IdentificationSystem Identification
Transistor constantβF
V/mSensor gainβ
HCoil inductanceL
ΩCoil resistanceR
N-m2/A2Force constantC
AEquilibrium currentio
VVoltage supplied to electromagnetVin
mEquilibrium distancexo
kgMass of the objectm
SI UNITSDESCRIPTIONSYMBOL
Sinha & NagurkaIMECE2005-81600 78
SISO Design ToolSISO Design Tool
Sinha & NagurkaIMECE2005-81600 79
PD Controller ResponsePD Controller ResponseStep Response
Settling Time: 0.157sSteady-State Error: 2.15
Impulse ResponseSettling Time: 0.562s
(KP=0.68, KD=0.0141)
Sinha & NagurkaIMECE2005-81600 80
PID Controller ResponsePID Controller Response
Sinha & NagurkaIMECE2005-81600 81
Margin Plots for PD ControllerMargin Plots for PD Controller
Sinha & NagurkaIMECE2005-81600 82
Margin Plots for PID ControllerMargin Plots for PID Controller