Escolar Documentos
Profissional Documentos
Cultura Documentos
de Silva
Plan
Graphical Interpretation
Analytical Linearization (statespace and input-output models)
Model linearization
Real systems are nonlinear
They are represented by nonlinear analytical models
Common techniques (e.g., response analysis, frequency
domain analysis, eigenvalue problem analysis, simulation,
control) use linear models
Method: Linearize each nonlinear term (1st order Taylor
series approximation)
Note: A nonlinear term may be a function of more than one
independent variable.
Graphical Illustration
Nonlinear function f(x) of a single independent variable x
Steps:
1. Express each nonlinear term at an operating point (or, reference condition
2. Select (small) increments of the independent variables from the operating
condition
3. Express Taylor series expansion of nonlinear term about operating point
4. Retain O(1) terms
Note: We need the first-derivatives (local slopes) at operating condition
Example: A nonlinear function f(x,y) of two independent variables x and y
f ( xo , yo )
f ( xo , yo )
f ( x, y ) f ( xo , yo )
x
y with x xo x, y yo y
x
y
or
f ( x , y )
f ( x , y )
f ( x x, y y ) f ( x , y )
x
y
x
y
Note: ( ) or ( )o denotes operating condition
( ) or ( ) denotes a small increment about that condition.
dq 2
f 2 q1 , q 2 , . . ., q n , r1 , r2 , . . ., rm , t
dt
dq n
f n q1 , q 2 , . . ., q n , r1 , r2 , . . ., rm , t
dt
(a)
(b)
(c)
Equilibrium states q 0
Note: At equilibrium (operating point) response (sate vector)
remains steady
Equilibrium states q : Solving n equations f(q, r, t) = 0 for steady
input r
Slightly Excite from Equilibrium State:
If response builds up and deviates further unstable
equilibrium state
If returns to original operating point stable equilibrium
state
If it remains at new state neutral equilibrium state
(a) Stable
(b) Neutral
(c) Unstable
f
f
(q , r , t ) q
(q , r , t ) r
q
r
(q , r )
or x Ax Bu
r
=
u
f df ( x ) x
dx
dx
x x ;
dt
d 2 x
x 2 x ; etc.
dt
Linearization Steps
d ()
1.Set all dt terms to zero, and solve the resulting algebraic
Practical Considerations
(b)
(c)
Saturation
Level
-d
Device Output
Saturation
Level
Linear
Range
(d)
d
Device Output
Device Input
Frictional
(Resistance)
Force
Dynamic
Frictional Force
Device
Input
Device Output
(a)
Device Input
Device Input
(Applied Force)
Illustrative Examples
Example 1
Related Questions
(a) Comment about the model accuracy
(b) State variables: motor speed m , pump-shaft torque Tp, pump speed p . Derive three
state equations nonlinear state model. What are the inputs to the system?
(c) What do parameters 0 and T0 represent with regard to motor behavior? Determine
Tm Tm
Tm
,
and . Verify that first is -ve and other two are +ve. Note: Under normal
m T0
0
operating conditions, we have 0 m 0 (i.e., slip)
(d) Consider steady-state operating point: steady motor speed m . Obtain expressions for
p , T p , and T0 (at operating point), in terms of m and 0
Tm
Tm
Tm
Solution
(a) Backlash and inertia of gear transmission neglected (not accurate in general). Gear efficiency
is assumed constant but usually varies with speed
There is some flexibility in the shaft (coupling) between gear and drive motor
Energy dissipation (in pump load, bearings) is lumped into a single linear viscous-damping
element (In practice nonlinear and distributed)
d p
d m
p = pump rotation
dt
dt
Power = torque speed; Tg = reaction torque on motor shaft from gear; r = gear ratio r m =
output speed of gear Gear efficiency
Tp r m
Tg m
Output Power
Input Power
Tg
Tp
(i)
(iv)
r
J mm Tm bmm Tp
T0 q0 (0 m )
(viii)
(q02 m2 )
(ii)
line frequency); T0
(phase voltage)2
Solution (Contd)
(c) From (viii): When m 0 Tm = T0 T0 = starting torque of motor
From (viii): When Tm = 0 m 0 0 = no-load speed = synchronous speed
Note: Under no-load conditions, no slip
Actual speed of induction motor = speed 0 of rotating magnetic field
Differentiate (viii) wrt T0 , 0 , and m :
Tm q0 (0 m )
1
T0
(q02 m2 )
(at O) (ix)
2
0
(q02 m2 )2
(q02 m2 )2
b
m
(q02 m2 )2
(q02 m2 )2
(at O)
(at O)
(x)
(xi)
p rm (xii);
Tp bp rm
r
0 Tm bm m Tp ; 0 k p (rm p ) ;
0 Tp bp p
(xiii)
T q ( )
Tm bmm r 2 bpm 0 0 2 0 2 m (from (viii))
(q0 m )
m (bm r 2 bp )(q02 m2 )
T0
q0 (0 m )
(xiv)
Note: For the steady operating point, steady inputs T0 and 0 are known. We have 3
equations (xii), (xiii), and (xiv) for the 3 unknowns Tp , m , and p .
Solution (Contd)
(e) Take increments of state equations (v) (vii):
r
J m m bm m Tp b m 1T0 2 0 (xv); Tp k p (r m p ) (xvi); J p p Tp bp p (xvii)
Note:
T
T
T
Tm m m m T0 m 0 b m 1T0 2 0
m
T0
0
A=
(b bm ) J m
rk p
r ( J m )
0
1 Jp
k p ;
bp J p
0
1 J m
B= 0
0
2 Jm
0
0
; C =
0
0
0 1 k
p
1
0
; D=0
Substitute (xvi) into (xv) to eliminate m ; Substitute (xvii) into the result to eliminate T p
Input-Output Differential Equation:
Jm J p
d 3 p
dt
d 2 p
dt
[k p ( J m
r2J p
) bp (bm b)]
d p
dt
kp(
r 2b p
bm b) p 2rk p 0
(xix)
3rd order differential equation; system is 3rd order; state-space model is also 3rd order.
Solution (Contd)
Observation From (xix):
When 0 is changed by finite step 0 RHS of (xix) will be finite
d 2 p
d 3 p
If as a result, dt 2 or lower derivatives change by a finite step dt 3 should
change by an infinite value (Note: derivative of a step = impulse)
2 ,
dt , and p will not change instantaneously
dt
Note: These three may be considered state variables (which, by definition, cant
change instantaneously, without infinite input)
Only
of 0
d 3 p
dt 3
d 3 p
dt
2 rk p
is J J 0
m p
Electro-Mechanical Coupling
Mechanical damping constant bm comes from bearing
friction and other mechanical sources
Electrical damping constant b comes from the
electromagnetic interactions in the motor
The two occur together (and should be treated
together, in model analysis, simulation, design,
control, etc.).
E.g., whether the response is underdamped or
overdamped depends on the sum bm +b and not the
individual components
Electro-mechanical coupling
Example 2
c =
cutter speed
Related Questions
(a) Use Tm = input; TL = output; m Tk c = state vector. Develop a
complete (nonlinear) state model. What is the system (model) order?
(b) From state model, obtain a single input-output differential equation, with
Tm = input and c = output.
(c) Steady operating conditions: Input and state variables are all constant.
Express operating point values m , Tk , c , and TL in terms of Tm and
model parameters only. Consider cases: Tm > 0 and Tm < 0.
(d) Consider incremental change Tm in motor torque and corresponding
changes , T , , T . Determine a linear state model (A, B, C, D) using
T
x = [ m
Tk
k
T
c ]
as state vector, u =
Tm
as input and y =
TL
as output.
(e) In incremental model, if twist angle of flexible shaft ( m c ) is output
determine a suitable state model. What is the system order then?
(f) In the incremental model, if c is the output, how should the state model
in Part (a) be modified? Then what is the system order?
Solution
bcc
Jc
Tk
TL c c c
d
d
(b) Note: c c sgn c ( c c ) (c 2 sgn c ) 2cc sgn c 2 c c
dt
dt
(iv)
d2
( c c ) 2 c c 2c2 sgn(c ) (v) Note: Since sgn(c ) 1 for c 0; 1 for c 0; it is a constant
2
dt
Jm
b
J cc bcc 2c c c 2c c2 sgn(c ) J mc Tm m J cc bcc 2c c c bmc J cc bcc c c c
k
k
Jm Jc
d 3c
d 2c
d 2c
d
dc
d
(
J
b
J
b
)
2
cJ
( J m k J c k bmbc ) c 2bmc c
2cJ m sgn(c )( c ) 2 k (bm bc )c kc c c kTm
m
c
c
m
m
c
3
2
2
dt
dt
dt
dt
dt
dt
Solution (Contd)
(c) Operating Point: 0 Tm bmm Tk (vi); 0 k (m c ) (vii); 0 Tk bcc c c c (viii)
m c ; Tk (bc c c )c ; Tm (bm bc c c )c Signs of m , c , and Tm are identical
Case 1: Tm > 0 c > 0
2
2
Eliminate Tk using (vi) and (viii): 0 Tm bmm bcc cc cc (bm bc )c Tm 0
Positive root: c
(bm bc )2 4cTm
2c
(bm bc )
m
2c
(b b )2 4cT
(bm bc )
m
c
m
From (vi): Tk Tm bm
2c
2c
(b b )2 4cT
(bm bc )
m
c
m
2
2
Case 2: Tm < 0 c < 0 (vi)-(viii): 0 Tm bmc bcc cc cc (bm bc )c Tm 0
(bm bc )2 4cTm
(bm bc )
c
Note: In view of the rotational symmetry of the problem, Case 2 can be derived from the
results of Case 1 (i.e., in the expression for c in case 1, change the sign of Tm and then change
the sign of the entire result.
Solution (Contd)
(d)
Linearize:
J m m = Tm bm m Tk
Tk
= k ( m c )
J c c = Tk bc c 2c c c
Note:
d
dc
( c c )
d
dc
(c 2 sgn c ) 2c sgn c 2 c
with x = [ m Tk c ] and u = Tm
bm J m
A= k
0
1 J m
0
1 Jc
k
;
(bc 2c c ) J c
0
1 J m
0
B=
and y = TL = 2c c c C = 0 0 2c c
and D = [0]
Tk
k
Exactly the same state equations plus this new output equation System order = 3
(e) y m c
(f) y c
d c
c
dt
System order = 4
Example 3
Related Questions
(a) Which system parameters are variable? Explain.
(b) Damping torque Td ( ) at pulley bearing is nonlinear in angular speed of pulley.
T
State vector x [ f v]
f = tension force in cable; v = velocity of car (positive upwards)
Input vector u [Tm ]T
Tm = torque applied by motor to pulley (positive in direction shown in Figure)
Output vector
y = [v]
Obtain a complete, nonlinear, state-space model for the system.
(c) With Tm = input; v = output, convert state model into a nonlinear I/O differential
equation model. What is the system order?
(d) Obtain an equation whose solution steady-state operating speed v of elevator car.
(e) Lienarize the I/O differential-equation model for small changes Tm of input and v of
output, about an operating point.
Note: Tm = steady-state operating-point torque of motor (assumed known).
dTd
Hint: Denote d by b().
(f) Linearize the state model and determine model matrices A, B, C, and D.
Obtain linear I/O differential equation from this state-space model
Verify that it is identical to the result in Part (e)
Solution
(a) r is a variable due to winding/unwinding of cable around pulley; m varies with
occupancy.
(b) State Equations: Apply Newtons 2nd law to two inertia elements and Hookes law
to spring element: J Tm rf Td ( ) (i); f k (r v) (ii); mv f mg (iii)
Output y =v
(c) Eliminate f by substituting (iii) into (i) and (ii):
J Tm r m(v g ) Td ( ) (iv); mv k (r v) (v)
1 m
1 m
From (v): r ( k v v) r ( k v v)
Substitute these into (iv), to eliminate :
J m
1 m
( v v) Tm r m(v g ) Td ( ( v v))
r k
r k
(vi)
r
mg
T
(
) 0 Solve steady-state operating speed
Steady-state Equation: m
d
r
Solution (Contd)
J m
(e)Linearize (vi): r ( k v v) Tm r mv b( ) (vii)
From (v): mv k (r v)
J m
b( )m
J
b( )
v
v ( rm)v
v Tm
r k
rk
r
r
J m
1 m
( v v) Tm r mv b( ) ( v v)
r k
r k
rk
A =
0
r
J
0
1
m
dTm ( )
d
(viii)
where b( )
k
;
0
f k (r v) (x); mv f (xi)
f v]T
1 J
0
B= ; C = 0 0 1 ;
0
D=0
Example 4
Vertical
Direction
Y
Spacecraft
Mass m
(X, Y)
f(t)
X
Horizontal on Earths
Surface
Example (Contd)
A rocket-propelled spacecraft of mass m is fired vertically up (in the Ydirection) from earths surface (see Figure).
Vertical distance of spacecraft centroid (altitude) = y
(measured from earths surface)
Upward thrust force of rocket = f(t)
R
mg
R y
Gravitational pull on spacecraft =
dy
.
dt
Related Questions
(a) Derive the input-output differential equation for the system.
f = input; y = output.
(b) Spacecraft accelerates to altitude yo and maintains a constant speed vo
(still moving in the same vertical (Y) direction)
Determine an expression for rocket force needed for this
constant-speed motion.
Express answer using yo, vo, time t, and system parameters m, g, R, r, k.
Show that this force decreases as the spacecraft ascends.
(c) Linearize the I-O model (Part (a)) about steady operating condition
y and y in position and speed
(part (b)), for small variations
of spacecraft, due to a force disturbance
f (t ) .
(e) Linearize state-space model (d) about steady conditions (b) for small
variations
disturbance
and
f (t ) .
(f) From linear state model (e) derive the linear input-output model and
show that the result is identical to what you obtained in Part (c).
Solution
2
R
y r
my
mg
f (t ) (i)
k y ye
(a) Newtons 2nd law in Y-direction:
R y
k
v
v
e
f s (t )
o
o
2
Substitute (ii)-(iv) in (i):
vot yo
1 R
vo > 0
f s (t )
mg
vot yo
1 R
kvo2e(vot yo )
Solution (Contd)
(c) Derivatives needed for linearization (O(1) Taylor series terms):
y y y r
d
1
2
1
y r
y r
y
;
y
ye
e
;
y
ye
2
y
e
2
3
dy
R
y
r
y
y
y
1
R
R
my
2mg
1
k
y ye y r y 2k y e y r y f (t )
3
R
r
y
1
(v)
(Note: Steady terms cancel out because they satisfy the steady operating condition)
Note: y is given by (iv) and is time-varying.
2mg
1
k 2 y r
vo e
y 2kvo e y r y f (t )
3
R
r
y
1
Steady state : y y vo 0 my
Note: Unstable system.
(d) State vector x x1
x2 y
x1 x2
x2
Output equation :
(vi)
g
x1
1
R
y1 x1
k
1
x2 x2e x1 r f (t )
m
m
(vii)
(v)
Solution (Contd)
(e) To linearize, use derivatives (local lopes) as before:
x2 x2 x1
d
1
2
1
x1 r
;
x
x
e
e
2 2
dx1 x1 2
R x1 3 x1
r
1
1
R
R
x2 x2e x1 r 2 x2 e x1
x2
Output equation:
2g
x
R 1 1
R
x1
k 2 ( x1 ) r
k
1
vo e
x1 2 voe x1 r x2 f (t )
mr
m
m
(ix)
y1 x1 (x)
x1
2mg
x
R 1 1
R
x1
k 2 ( x1 ) r
k
1
vo e
x1 2 voe x1 r x1 f (t )
mr
m
m
Graphical Solution to
Linearization:
Use of Experimental Operating
Curves
Graphical Linearization
(Two-variable Case)
1. Keep one variable constant at some operating condition and
vary the other variable in steps
2. Plot the data curve
3. Repeat steps 1 and 2 by changing the operating condition of
the first variable by a small step. Obtain a series of
characteristic curves in this manner
4. Linearization: For a specific operating condition, find the
slope of the curve passing through it (i.e., with the first
variable kept constant); Also find the vertical increment of two
adjacent curves enclosing the operating point (i.e., with the
second variable is kept constant)
5. Slopes with respect to the two independent variables are
obtained from the two values in Step 4.
Graphical Linearization
(Two-variable Case)
Response
y
y y ( x1, x2 )
x2+x2
y
y
y x1
x2
x1
x2
y
b
y
x1
x1
x2
or y bx1 kx2
y
x2
x2
1
x1 Variable
x2 constant
y
x2
x1 constant
y
x2
x2 b x1 k x2
x1
DC Motor Example
(Control Circuit and Mechanical Loading)
(a) Equivalent circuit of a dc motor (separately excited); (b) Armature mechanical loading diagram
Torque-Speed Curves: (a) Shunt-Wound; (b)Series-Wound; (c) Compound-Wound; (d) General Case
Note: In each curve, excitation voltage vc is maintained constant
vc constant
Note: Mechanical damping included in this b depends on mechanical damping present in the
measured torque during test (primarily bearing friction)
Draw a vertical line through operating point O Torque intercept Tm between two curves
Note: Vertical line is a constant speed line
kv
Voltage Gain:
T
m
Tm
m
Tm
vc
m
vc
m constant
Tm
vc
Tm
vc
Note: Torque needed to drive rotor inertia is not included in this equation
(because steady-state curves are used in determining parameters)
Inertia term is explicitly present in Mechanical Equation of motor rotor (See Figure (b)):
dm
d m
Jm
Tm TL Linearized (incremental) model: J m
Tm TL b m kv vc TL
dt
dt
Note: Tm and b include overall damping (both electrical and mechanical) of motor.
Linearization Steps
1. Keep one variable constant at some operating condition and
vary the other variable in steps
2. Repeat for an increment of the variable that was kept constant
3. Plot the successive operating curves obtained in this manner
4. Determine the local slopes at an operating point
5. The rest is the same as with analytical linearization
Example
Water Out
Jt
Fuel Rate
bt
Tk
Jp
bp
For Shaft :
Torional stiffness k 20.0 Nm/rad
Torque in the shaft =Tk
Gas Turbine
Centrifugal Pump
Torque T (N.m)
Water In
Fuel Input
Rate
Q (gal/s)
160
10
140 9
120
100
8
7
6
5
4
80
60
Steady-state
operating curves
40
20
200
400
600
800
Speed (rpm)
Example (Contd)
Related Questions:
T , Q
Tm
Tk
Jp
Jt
Tk
Shaft
bt
Gas Turbine Rotor
Tp
p
bp p
Pump Rotor
Example (Contd)
State-space Model (Linearized)
x [, k , p ]T ; u [q]
A
k
C (0 0 1);
1/ J t
0
1/ J p
D (0)
k ;
bp / J p
0
kq / J t
B 0
0
Example (Contd)
Local Slopes (from Experimental Curves)
b = slope at operating point on the Q 8 gal/s curve
(130 80)(Nm)
50
1.326 Nm/rad/s
2
2
(560 200)
(rad/s) 360
60
60
k q = increment of T over the change in Q from7 to 9 gal/s with kept constant at 400
rpm
(110 92)(Nm)
9.0 Nm/gal/s
(9 7)(gal/s)
18.26 s -1
Jt
0.1
1
10.0 (kg.m 2 )-1
Jt
1
1
20.0 (kg.m 2 ) 1
J p 0.05
bp
Jp
kq
Jt
3.0
60.0 s-1
0.05
9.0 (N.m/gal/s)
90.0 gal/s3
2
0.1 (kg.m )
Torque T (N.m)
Fuel Input
Rate
Q (gal/s)
160
10
140 9
120
8
7
6
100 5
4
80
60
40
20
200
400
600
800
Speed (rpm)
Example (Contd)
State-space Model (Linearized)
x [, k , p ]T ; u [q]
0
18.26 10.0
A 20.0
0
20.0 ;
0
20.0 60.0
C (0 0 1);
D (0)
90.0
B 0
0