Systems dynamics and control

ME-3015

Mechanical Engineering
Georgia Institute of Technology

Linearization of nonlinear systems

1
Linear System
Linear systems

u1 (t ) x1 (t )

u2 (t ) x2 (t )

u1 (t ) + u2 (t ) x1 (t ) + x2 (t )

α ⋅ u1 (t ) α ⋅ x1 (t )
2
Linear and Nonlinear System (2)
• Linear, Time-variant systems
2
d x
2
+ (1 − cos 2t ) x = 0
dt
Function of time t

• Nonlinear systems

2
d x dx
2
+ ( x − 1) + x = 0
2

dt dt
NOT linear ! 3
Linear and Nonlinear System (3)
• Linear systems
– The principle of superposition
– Mathematical analysis
– Transfer function analysis/design

• Nonlinear systems
– A nonlinear system may be approximated by a
linear system within limited ranges.

4
Linearization technique
• Taylor series expansion
Nonlinear function z = f (x)

Taylor series expansion about a point (x, z ) z = f (x )

z
z a

x
x 5
Linearization technique (2)
df 1 d2 f 1 d 3
f
f ( x) = f ( x ) + (x − x) + 2
( x − x )2 + 3
( x − x )3 + L
dx x= x 2! dx x= x
3! dx x= x

df
f ( x) ≅ f ( x ) + (x − x)
dx x= x

a
z = z + a( x − x )
z = f (θ ) = sin θ about (0,0)

df
f (θ ) ≅ f (0) + (θ − 0) = 0 + cos θ (θ − 0) = θ
dθ θ =0
6
Linearization (with 2 inputs)
z = f ( x, y ) about z = f (x, y)
p338

⎡ ∂f ∂f ⎤
f ( x, y ) ≅ f ( x , y ) + ⎢ (x − x) + ( y − y )⎥
⎢⎣ ∂x x= x , y = y ∂y x= x , y= y ⎥⎦
⎡ ⎤
1 ⎢∂2 f ∂2 f ∂2 f
+ (x − x) + 2
2
( x − x )( y − y ) + 2 ( y − y)2 ⎥
2! ⎢ ∂x 2 ∂x∂y ∂y ⎥
⎣ x= x , y = y x= x , y= y x= x , y= y ⎦

z = z + a ( x − x ) + b( y − y )
∂f ∂f
a= ,b =
∂x x= x ∂y y= y 7
Example 1: Linearized dynamics
Inclined pendulum

kθ Torsional spring

θ
Joint J = mL2
L

mg Static equilibrium θ
θ
Problem: Obtain natural frequency kθ θ = mgL cos θ

θ : small angle of motion

Linearize about θ
θ : not small
(not about θ=0 !!)
8
Example 1: Linearized dynamics (2)
θ '= θ +θ Displacement measured from the horizontal line

DEOM: Jθ&&′ = mgL cos θ ′ − kθ θ ′

Linearize about static equilibrium θ

Jθ&& = mgL cos(θ + θ ) − kθ (θ + θ ) (θ&& = 0)

Taylor series expansion for cosine

d cos θ
cos(θ + θ ) ≅ cos θ + θ = cos θ − sin θ ⋅θ
dθ θ =θ 9
 Jθ&& = mgL cos(θ + θ ) − kθ (θ + θ )
= mgL(cos θ − θ sin θ ) − kθ θ − kθ θ
= mgL cos θ − mgL sin θ ⋅ θ − kθ θ − kθ θ
= −(mgL sin θ + kθ )θ Static equilibrium

Jθ&& + (mgL sin θ + kθ )θ = 0

kθ = 0, θ = 90 deg
mgL sin θ + kθ
ωn = ωn =
mgL
=
mgL
=
g
J J mL2
L θ
10
Example 2: Linearization (Robot arm)
Robot arm kinematics

(X,Y)
y
L Link length: L

Elbow L

O
L
θ2 θ2
x θ1
shoulder
θ1

Tip position
X = L cos θ1 + L cos(θ1 + θ 2 )
Y = L sin θ1 + L sin(θ1 + θ 2 )

11
Example 2: Linearization (Robot arm) (2)
(ΔX , ΔY )
y
Δθ 2
Arm base posture: θ1 θ2
Δθ1
(e.g. 25 deg, 30 deg)
O θ2
x
θ1
Small angles of motion for two joints:

θ1 : θ1 → θ1 + Δθ1 X : X → X + ΔX
θ 2 : θ 2 → θ 2 + Δθ 2 Y : Y → Y + ΔY

Linearize about θ1 , θ 2

12
Example 2: Linearization (Robot arm) (3)
Linearize about θ1 , θ 2
X + ΔX = L cos(θ1 + Δθ1 ) + L cos(θ1 + Δθ1 + θ 2 + Δθ 2 )
[
= L(cos θ1 − sin θ1 ⋅ Δθ1 ) + L cos(θ1 + θ 2 ) − Δθ1 sin(θ1 + θ 2 ) − Δθ 2 sin(θ1 + θ 2 ) ]
X = L cos θ1 + L cos(θ1 + θ 2 )

Then [
ΔX = L(− sin θ1 ⋅ Δθ1 ) + L − Δθ1 sin(θ1 + θ 2 ) − Δθ 2 sin(θ1 + θ 2 ) ]
[ ]
= − L sin θ1 + sin(θ1 + θ 2 ) Δθ1 − L sin(θ1 + θ 2 )Δθ 2

[ ]
ΔY = L cos θ1 + cos(θ1 + θ 2 ) Δθ1 + L cos(θ1 + θ 2 )Δθ 2

Taylor series expansion for sine

d sin θ
sin(θ + Δθ ) ≅ sin θ + Δθ = sin θ + cos θ ⋅ Δθ
dθ θ =θ 13
Example 2: Linearization (Robot arm) (4)
[ ]
ΔX = − L sin θ1 + sin(θ1 + θ 2 ) Δθ1 − L sin(θ1 + θ 2 )Δθ 2
[
ΔY = L cos θ1 + cos(θ1 + θ )]Δθ
2 1 + L cos(θ1 + θ 2 )Δθ 2

y (X,Y)

ΔX = − LΔθ1 − LΔθ 2
O
45 deg ΔY = LΔθ1
x

θ1 = 0 deg
θ 2 = 90 deg 14
Systems dynamics and control
ME-3015

Mechanical Engineering
Georgia Institute of Technology

Transient response of 1st order and 2nd

order systems

1
Time-domain analysis of dynamic systems

• First-order systems
f mx  bx  f f  v (x )
x
ms  b sX (s)  F (s)
m sX ( s) 1
b  
F ( s) ms  b
• Second-order systems
f
k x
mx  bx  kx  f
m
b 2
Mechanical damper
bxO  k ( xi  xO )
Transfer function: xi xO
b dxO
 xO  xi
xi k dt
k dxO
b  kxO  kxi
xO dt
b
X O ( s) 1
G( s)  
X i ( s) (b / k ) s  1

3
Transient response of 1st order system
xi
X
xi
t
k
xi (t )  X Constant input at t>=0
xO
X
b X i ( s) 
s
1 X
X O ( s)  G( s) X i ( s)  
(b / k ) s  1 s

4
Transient response of 1st order system (cont.)

1 X X X
X O ( s)    
(b / k ) s  1 s s s  ( k / b)

Inverse Laplace transformation

xO (t )  X (1  e( k / b)t )  X (1  et / T )

T  b/k :Time constant

5
 xi,O
input

X
output

slope Within 2% !

X /T X (1  e4T / T )  0.982 X
t
T 4T
X (1  eT / T )  0.632 X

6
For faster response
xO (t )  X (1  e t /T )

T  b/k :Time constant Need to wait for 4T (sec) for 2% response

Smaller T for faster response Increase k or decrease b

Im
1
Pole of G(s) G( s) 
faster Slower (b / k ) s  1

k
Re p1   ( 1 / T )
O b

7
Ramp response for first-order system
xi xO b

X O ( s) 1
G( s)   (same as the previous ex.)
X i ( s) (b / k ) s  1
xi (t )  rt
xi
r r
X i ( s)  2
t s 8
Ramp response for first-order system (2)
1 r
X O ( s)  G( s) X i ( s)   2
Ts  1 s
1 T T 
 r 2   
s s s  (1 / T ) 

xO (t )  r (t  T  Tet /T )
e(t )  xi (t )  xO (t )  rT (1  Tet /T )

Steady-state error e()  rT

9
xi,O
input
e()  rT
T
output

t very large xO (t )  r (t  T )
10
Transient response of 2nd order systems

f Free vibration with initial conditions

k x
mx  bx  kx  f  0
m
b
x(0)  x0 , x (0)  x0

11
Recall Mass-spring system

k x
m

12
Free vibration
• Pull the mass downward and then
release with arbitrary initial conditions.

mx  kx  0 Equilibrium (x=0)

x
Release! 13
Free vibration (2)
• Use Laplace transformation
mx  kx  0 x(0), x (0)

LDEOM   m[ s X ( s)  sx(0)  x (0)]  kX ( s)  0

2

 m[ s 2 X ( s)  sx0  x0 ]  kX ( s)  0

(ms2  k ) X (s)  msx(0)  mx (0)

mx (0)  msx(0)

Hence X (s) 
ms 2  k
x (0) sx(0)
 2  2
s k /m s k /m 14
Free vibration (3)
m m/k s
X (s)  x (0)  x(0)
k 
s2  k / m
2
 s2  k / m  2

Note that (from Table)

  
L sin k / m t 
k/m
s 
2
 k /m 
2

 
L cos k / m t  s
s 
2
 k /m  2

m k k k
x(t )  x (0) sin t  x(0) cos t n 
k m m m 15
Free vibration with damping

mx  bx  kx  0 x(0)  x0 , x (0)  x0

t
Assume the solution x  Ae x  Aet , x  A2et
Characteristic equation m  b  k  0
2

 b  b  4mk 2

2m
Positive, negative?
Real, Imaginary (complex conjugate)? 16
Free vibration with damping (2)

 b  b  4mk 2

2m
Discriminant

b  4mk  0
2 2 complex conjugate roots

b  4mk  0
2
1 repeated real root b  2 mk
b  4mk  0
2 Critical damping value
2 distinct real roots

Greater Damping coefficient 17

Free vibration with damping (3)
mx  bx  kx  0

x  2n x  n x  0
2

k
Undamped natural frequently n 
m

b Actual damping value

Damping ratio 
2 mk Critical damping value

18
Free vibration with damping (4)

 b  b  4mk 2

2m
Discriminant Damping ratio

b  4mk  0
2 0   1 2 complex conjugate roots Underdamped

b 2  4mk  0  1 1 repeated real root Critically damped

b 2  4mk  0  1 2 distinct real roots Overdamped

Greater Damping coefficient 19

General case
x  2n x  n x  0 2
x(0)  x0 , x (0)  x0

s X (s)  sx(0)  x(0) 2 sX (s)  x(0) 

2
n n
2
X ( s)  0

( s  2 n ) x(0)  x (0)

X ( s) 
s  2 n s  n
2 2

Steps:

(1) Check the damping ratio 0   1  1  1

(2) User Inverse Laplace Transformation

Table 6,7,20, 21 (general case 22) 20
Example 1 Worksheet
Free vibration with initial conditions
k x
mx  bx  kx  0
m x(0)  2, x (0)  0
b
m  1, k  2
Damping ratio

b2
b 2
   0.707  1 Underdamped
2 mk 2 2 1

b 2 2
b2 2  
2 mk 2 2 1
1 Critically damped

b 4
b4  
2 mk 2 2 1
 2 1 Overdamped
21
Example 1 (underdamped)
(ms  b) x0
X ( s)  2
ms  bs  k Output oscillates

2( s  2) Use Table 2-1, #20 & #21

b2 X ( s)  2
s  2s  2
s 1 1
 2  2
( s  1)  1
2
( s  1) 2  1
x(t )  2e t cos t  2e t sin t
 2 2e t sin(t  45O )

22
Example 2 (critically damped)
b2 2
2( s  2 2 ) 2( s  2 2 ) a( s  2 )  b  2( s  2 2 )
X ( s)  2 
s  2 2s  2 (s  2 ) 2 as  ( 2a  b)  2s  4 2
a b
 
s  2 (s  2 )2 a  2, b  2 2
2 2 2
 
s  2 (s  2 )2

Use Table 2-1, #6 & #7

x(t )  2e 2t
 (2 2 )te  2t

23
Example 3 (overdamped)
b4 1, 2  2  2 as  a(2  2 )  bs  b(2  2 )  2s  8
2( s  4) ab  2
X ( s) 
s 2  4s  2 2(a  b)  2 (a  b)  8
a b
 
s  (2  2 ) s  (2  2 )
a  1 2, b  1 2
1 2 1 2
 
s  (2  2 ) s  (2  2 )

Use Table 2-1, #6

x(t )  (1  2 )e( 2 2 )t
 (1  2 )e( 2 2 )t

24
Underdamped oscillation
x  2n x  n x  0 2
Undamped natural frequently
k
n 
x m
Exponential decay envelope b
Damping ratio

2 mk
t
(8-12)

Exponential decay

   1  

 n t
x(t )  e  x(0)  x (0) sin d t  x(0) cos d t 
 
 1   2 d  
 
Damped natural frequency d  n 1   2 d  n 25
MATLAB
Transfer function

1
G(s)  2 >> num=[1]
s  2s  2 >> den=[1 2 2]

0  s2  0  s 1
G( s)  2
s  2s  2 >> sys1=tf(num,den)
>> step(sys1)

26
Logarithmic decrement
x
First positive peak

Displacement at t=0
dependent on
initial conditions x1
x2 xn
x3 t
T t2 t3 tn
t1

T: period of oscillation T  2 / d

t2= t1 +T tn= t1 +(n-1)T

27
Logarithmic decrement (2)
x1 x(t1 ) x(t1 )
  (use 8-12)
x2 x(t2 ) x(t1  T )
Harmonic

  1  

 x(0)  x (0) sin d t  x(0) cos d t 
    2 d  
 1 

Same value at t1, t1+T, t1+2T,… t t +T

x1 e nt1 e nt1 1
 n (t1 T )  nt1 nT  nT  enT
x2 e e e e

x1 e nt1 e nt1 1
 n (t1 ( n1)T )  nt1 n ( n1)T  n ( n1)T  en ( n1)T
xn e e e e
28
Logarithmic decrement (3)

 x1 
x1
 enT or  
ln    ln enT   nT
x2  x2 
 x1 
x1
 en ( n 1)T or  
ln    ln en ( n1)T   n (n  1)T
xn  xn 

Natural logarithm (log base e)

29
Logarithmic decrement (4)
Given x1 , xn , T , obtain the damping ratio.

 x1  T  2 / d (2)
ln     n (n  1)T (1)
 xn  d  n 1   2 (3)

 x1  2 (n  1)
(2)(3)->(1) ln     (n  1)2 / 1   2 
 xn  1  2

30
  x1 
2 (n  1)  1   ln  
2

 xn 
2
  x1  
4  (n  1)  1    ln   
2 2 2 2

  xn  
   x1  
2
   x1  
2

 4 (n  1)   ln       ln   
2 2 2

   xn      xn  

 1  x  
2
  1  x  
2

 2   ln  1   / 4 2   ln  1   
 n  1  xn     n  1  xn   
1  x1 
ln  
n  1  xn 

2
 1  x1  
4  
2
ln   
 n  1  xn  
31
Response time for 2nd order systems (1)
Without initial velocity
 n t
  
x(t )  e  x(0) sin d t  x(0) cos d t 
 1   2


 n t
  
 x(0)e  sin d t  cos d t 
 1   2 

x(0)   
 e  n t 
cos d t  tan 1 
1  2    2 
 1 

a sin   b cos   a 2  b 2 cos(   )

32
Response time for 2nd order systems (2)
Response time: Time in which the response remains within 2% of its initial value

Start from a peak (initial velocity=0)

x(0)
e nt
1  2 A response curve is tangent to
2% the envelope exponentials
t cos   1
x(0)
 e nt
1  2

33
Response time for 2nd order systems (3)
2% response time can be (roughly) estimated by the envelope exponentials.

x(0)
e nt s  0.02 x(0)
1  2
e nt s  0.02 1   2
 nt s   ln(0.02 1   2 )
3.91 
Always greater than 3.91 0 1
Zeta=0, 3.91
Zeta=0.9, 4.74

4
ts 
Very roughly speaking,
 n 34
Step response of 2nd order system
f Free vibration with initial conditions
k x
mx  bx  kx  f
m x(0)  0, x(0)  0
b
m  1, k  2 b  2
Difference from the last time: (1) zero initial conditions (2) external force (input)

f f  2[ N] t  0
2
t 2
F ( s) 
s
35
 F
X ( s)  2
ms  bs  k
a( s 2  2s  2)  bs 2  cs  2
1 2
X (s)   a  b  0,2a  c  0,2a  2
s  2s  2 s
2

 a  1, b  1, c  2
a bs  c
  2
s s  2s  2
1 s2
  2
s s  2s  2
1  d ( s  1) e  ds  d  e  s  2
      d  1, e  1
s  ( s  1)  1 ( s  1)  1 
2 2

1 s 1 1
  
s ( s  1)  1 ( s  1) 2  1
2

x(t )  1  e t cos t  e t sin t General case p.397-398

36
 x

f 2
 1 Static equilibrium point

k 2

37
General case with input and initial
condition
f ~
x  2n x  n x   f
2
x(0)  x0 , x (0)  x0
m
 2

s X (s)  sx(0)  x (0)  2n sX (s)  x(0)  n X (s)  F
~ 2

~
( s  2 n ) x(0)  x (0) F
X ( s)   2
s  2 n s  n s  2 n s  n
2 2 2

Free vibration with initial condition

Transient response by input

38
Example
L
L
3
J  mL2
L
2
 Static equilibrium
(neglect gravity)
Spring constant: k
mass: m
Joint Viscous coefficient: b

L=6, m=2, k=9, b=4

Assume small angle of motion. Obtain

(1) Dynamic equation of motion

(2) Undamped natural frequency
(3) Damping ratio
(4) Consider free vibration with initial conditions:  (0)   0 ,(0)  0
Time to delay from an amplitude of  0 to 0.05 0

Ln(1/0.05)=2.99 39
DEOM   L  L  L  L
J  b    k     0
2 2 3 3
2 2
bL kL
mL2     0
4 9
 b  k
m      0
4 9
2      0
Natural frequency
k 1 2
n     0.707
m 2 2
1 1 2
    0.35  1
2 2 1 2 2 4 40
  x1  2 (n  1)
ln   
 xn  1  2

 1  2
ln    2.99  (n  1)  2.37(n  1)
 0.05  1  2

n  2.26 cycles

2 2  3.14
T  2 / d    9.48
n 1   2 0.707 1  0.352

2.26*9.48=21.42 sec
41
Systems dynamics and control
ME-3015

Mechanical Engineering
Georgia Institute of Technology

Frequency domain analysis

1
Basic input functions u(t) u(t )  1 t  0
• Unit step function 1
t U ( s)  1
s
• Ramp function
u(t) u(t )  rt
r
• Unit impulse function t r
U (s)  2
s

• Sinusoidal function
2
Impulse response
• Unit Impulse function
u(t) 1 /  0t 
u (t )  
1/   0 t  0&t  

1
 t
O U ( s)  1
 0 Delta function  (t )

3
Impulse response (2)
Output

U ( s)  1
System
Y ( s )  G ( s ) 1
G(s)
Unit impulse Input  G(s)
What you observe

y(t )  L1 G(s)U (s)  L1 G(s)

Transfer function only

4
Unit impulse response of 2nd order system
f Free vibration with initial conditions
k x
mx  bx  kx  f
m x(0)  0, x(0)  0
b
m  1, k  2 b  2

f(t)
f(t): unit impulse input F ( s)  1

1
t
O
5
 1
X ( s)  2
ms  bs  k
1
X ( s)  2
s  2s  5
1 2 
  
 2 ( s  1)  4 
2

t
x(t )  0.5e sin 2t

6
Frequency domain analysis
Input
Output
U (s) System

G(s) Y (s)
u(t )  P sin t
Amplitude? Phase?

Frequency response: Steady-state response of a system to a sinusoidal input

For all frequency: 0 to inf

Output-input amplitude ratio
Phase angle between the input and output sinusoids.

7
Sinusoidal transfer function
System y (t )
u(t )  P sin t G(s)
Y (s)
Result: Replaces s with j.

G(s)  G( j )
Y ( j )
GAIN: G ( j )  Amplitude ratio of the output sinusoid
P( j ) to the input sinusoid

G( j)  Y ( j) / U ( j)

Phase shift of the output sinusoid
PHASE: w.r.t. the input sinusoid

8
 u(t)
P
t
O
G( j )
y(t) G( j )

t
O
Same frequency (linear system) 9
Transient response
 1
Example 1 G( s)  Stable 1st order system
s 1
Transfer function: xi xO

b=1, k=1
xi
X O ( s) 1 1
k G( s)   
X i ( s) (b / k ) s  1 s  1
xO
b Lesson #11 or text p389

10
Example 1 (cont.)

u(t )  sin t U ( s)  2
s 2
1
Amplitude=1 G( s) 
s 1
1 
Y ( s)  G ( s)U ( s )   2
s 1 s   2

a bs  c
  2
s 1 s   2

  
a ,b   ,c 
1  2
1  2
1  2 11
Example 1 (cont.2)
0
 t 1
y (t )  e  sin(t  tan  )
1

1  2
1 2
(1) Output: Sinusoidal function Frequency  (same as the input)

(2) Amplitude
1
 G ( j )  Re(G ( j )) 2  Im(G ( j )) 2
1 2
1 1
 
j  1 (1  j )(1  j )
(3) Phase
Im G( j )
G( j )  tan 1
  tan 1 
Re G( j ) 12
Detailed proof: sinusoidal transfer function
K ( s  z1 )(s  z2 ) ( s  zm ) P
G( s)  U (s)  2
( s  p1 )(s  p2 ) ( s  pn ) s 2

K ( s  z1 )(s  z2 ) ( s  zm ) P
Y ( s)  G( s)U ( s)   2
( s  p1 )(s  p2 ) ( s  pn ) s   2
Partial fractional expansion (1) Stable (real part of poles are negative)
(2) Distinct poles

a a b1 bn
Y ( s)    
s  j s  j s  p1 s  pn

y(t )  ae jt  a e jt  b1e p1t    bne pnt

13
Detailed proof (cont.)
stable

y(t )  ae jt  a e jt  b1e p1t    bne pnt 0

P P P  jG ( j )
a  G( s) 2 ( s  j )   G (  j )   G ( j ) e
s 2 s   j 2j 2j

P P P jG ( j )
a  G( s) 2 ( s  j )  G ( j )  G ( j ) e
s 2 s  j 2j 2j

14
   e  j (t G ( j )) e j (t G ( j )) 
y (t )  G( j ) P  
 2j 2j 

e j (t G ( j ))  cos(t  G( j))  j sin(t  G( j))

 e j (t G ( j ))   cos(t  G( j))  j sin(t  G( j))

y(t )  G( j ) P sin(t  G( j ))

GAIN PHASE

15
Representation of frequency response
Chapter 11 (p608-)
• Bode diagram (Bode plot)
• Nyquist diagram (Nyqiust plot)

16
Bode plot
Sinusoisal Transfer Function G(s)  G( j )
System Output
Input
G(s) y (t )
u(t )  sin t
Y ( j )
GAIN G ( j ) 
U ( j )
PHASE G( j)  Y ( j) / U ( j)
2 Plots: On a logarithmic scale.
(1) Gain vs. Frequency
(2) Phase vs. Frequency 17
Bode plot (cont.)
• Constant gain
• Integral and derivative factor
• 1st order system (1st order lag element )
• 2nd order system

• Multiplication -> addition

• Asymptote

18
Constant gain
G(s)  K  G( j)  K
GAIN: G( j )  K 20 log G( j )  20 log K
PHASE: G( j)  0[deg]
Bode Diagram
7.5

7
Magnitude (dB)

6.5

5.5

K=2 5
1

0.5
Phase (deg)

-0.5

-1
10
0 1
10 19
Frequency (rad/sec)
Integral factor 1
G( s)   G( j ) 
1
s j
1
GAIN: 20 log  20 log [dB]
j
Phase: G( j)  (1 / j)  90[deg]
-20dB/decade
Bode Diagram
5

0
Magnitude (dB)

-5

-10

-15

-20
-89

-89.5
Phase (deg)

-90

-90.5

-91
0 1
20
10 10
Frequency (rad/sec)
Derivative factor G(s)  s  G( j)  j
GAIN: 20 log j  20 log [dB]
Phase: G( j)  ( j)  90[deg]
Bode Diagram
20

15
Magnitude (dB)

10

20dB/decade
0

-5
91

90.5
Phase (deg)

90

89.5

89 21
100 101
Frequency (rad/sec)
1st order system G ( s)  1 G ( j ) 
1
Ts  1 Tj  1
1
GAIN: 20 log G( j )  20 log  20 log T 2 2  1
Tj  1
Phase:  tan 1 T (0 to -90deg)
T  1
Bode Diagram
Asymptote
T  1
0

T=1 -5

-10
Magnitude (dB)

-15
 20 log  2T 2
T  1 -20

 20 log T
-25

-30
-35

Asymptote -40
0
-20dB/decade
 20 log 1  0
Phase (deg)

-45

-90 22
10-2 10-1 100 101 102
Frequency (rad/sec)
2nd order system G(s)  2
n 2
s  2 n s  n
2

n 2
n 2
G ( j )  
( j )  2 n ( j )  n (n   2 )  j (2 n )
2 2 2

1

[1  ( / n ) 2 ]  j (2 / n )
GAIN:
  n 20 log G( j )  1
-40dB/decade
1 
  n 20 log G( j )  20 log  40 log [dB]
( / n ) 2
n
PHASE:
1 1 2 / n
   tan
[1  ( / n ) ]  j (2 / n )
2
1  ( / n ) 2 23
2nd order system (2)
Intersection at   n
  0.2
10 -40dB/decade
0
Magnitude (dB)

-10

-20
0dB/decade
-30

-40
  0  0 deg 0

-45
  n  90 deg
Phase (deg)

-90

-135
    180 deg
-180
-1 0 1
10 10 10

Frequency
n 24
Property of bode diagram (1)
H (s)  1 / G(s) e.g. 1/s and s
G
Magnitude [dB]

GAIN:

0

H 20 log H ( j )  20 log G( j )

H
Phase [deg]
0

PHASE:
G
H ( j)  G( j)
Im

1 1
Re  2 (a  jb )
O a  jb (a  b )
2
25
Property of bode diagram (2)
H (s)  G1 (s)G2 (s)
G1 ( s) G2 ( s)

GAIN: 20 log H ( j )  20 log G1 ( j )G2 ( j )

 20 log G1 ( j )  20 log G2 ( j )

PHASE: H ( j)  G1 ( j)  G2 ( j)

Multiplication -> Addition

26
Property of bode diagram (3)
e.g. Electrical system

Magnitude [dB]
G1
0dB/decade -40dB/decade

0
-20dB/decade
0dB/decade

e.g. Mechanical system 0dB/decade

20dB/decade
Magnitude [dB]

G2 
0

0dB/decade -40dB/decade

20dB/decade
Magnitude [dB]

H
0dB/decade

0

0dB/decade -40dB/decade

-60dB/decade
-20dB/decade 27
Bode plot Example 1
s
Skech a bode diagram of G( s) 
10s  1

1 1
Hint: G( s)  s   s
10s  1 10s  1

G1 ( s) G2 ( s)

28
 Magnitude [dB] 20dB/decade
0


-20dB/decade

0
Phase [deg]
0 45 90

Magnitude [dB]
0dB/decade

0


G2 ( s) 
1
10s  1
T  1 0.1rad/s 20dB/decade
Magnitude [dB]

Phase [deg]
0- 45 90
0dB/decade

0


-20dB/decade
0.1 rad/sec
0
Phase [deg]
0


-90 -45

29
0.1 rad/sec
Bode plot Example 2
1000  s
Sketch a bode diagram of H ( s) 
1 s

1000  s 1
Hint: H ( s)   (1000  s)
1 s 1 s
H1 ( s ) H 2 ( s)

30
Magnitude [dB] H1 (s)  (10000  s)

0dB/decade 20dB/decade
80


0
0dB/decade
90

Magnitude [dB]

Phase [deg]

80
-20dB/decade 
45
0

0dB/decade (=20-20 dB/dec)

1 10000 rad/sec
H 2 (s)  1 rad/sec 10000 rad/sec
1 s 0

Phase [deg]
0

Magnitude [dB]

-90 -45
0dB/decade
0


-20dB/decade

0
Phase [deg]


-90 -45

31
1 rad/sec
Nyquist Plots (Vector Locus)
• Plot of G( j )  (for all  ) 0    
on Real-imaginary plane

Im
Text p632
G( j )
G( j )

G( j )
Re
O

MATLAB: nyquist(sys) NOTE (default):     32

Nyquist plot of 1st order system
1 1
G(s)  G ( j ) 
Ts  1 Tj  1
Im

  0.5 1
Re
O
 0
-0.5
  1/ T

33
sample19_3.m
Nyquist Diagram Nyquist Diagram
10 1

8 0.8
1 1
6
G( s)  0.6
G( s) 
4
s 0.4
s 1

Imaginary Axis
Imaginary Axis

2 0.2

0 0

-2 -0.2

-4 -0.4

-6 -0.6

-8 -0.8

-10 -1
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
Real Axis Real Axis

1
G(s) 
s 2  2s  1
3
Nyquist Diagram
1
Nyquist Diagram

0.8
2

  0.2   1.2
0.6

Imaginary Axis 0.4

Imaginary Axis

1
0.2

0 0

-0.2
-1
-0.4

-0.6
-2
-0.8
34
-3 -1
-1.5 -1 -0.5 0 0.5 1 1.5 2 -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
Real Axis Real Axis
Vibration isolation
• Automobile suspension system
• Transimissibility: a measure of the reduction
of a transmitted force/motion afforded by an
isolator
m x Vehicle displacement

k b |x| versus |p|

p
Road displacement 35
 suspension Output
Input
G(s)
p(t )  P sin t x(t )
P: constant
X ( j )
Transmissibilily (GAIN) G ( j ) 
P( j )

mx  b( x  p )  k ( x  p)

(ms  bs  k ) X ( s)  (bs  k ) P( s)
2

X ( s) bs  k
 2
P( s) ms  bs  k
36
Sinusoidal transfer function
X ( j )
G ( j )  
P ( j )
bj  k k  jb

m( j )  bj  k  m 2  k  jb
2

Transmissibilily (GAIN)

k 2  b 2 2
G( j ) 
(m  k )  b 
2 2 2 2

37
%MATLAB code
m=500;
k=200;
b=100;
G=tf([b k],[m b k])
bode(G); %bode plot
grid

omega=logspace(-1,2); b ?
TR=sqrt(k^2+b^2*omega.^2)./sqrt((-m*omega.^2+k).^2+b^2*omega.^2);
figure(2)
loglog(omega,TR)
grid

Bode Bode Diagram TR

20 1
10
10

0
Magnitude (dB)

-10

0
-20 10

-30

-40

-50
-1
-60 10
0

-45
Phase (deg)

-2
-90 10

-135

-180 -3
-1 0 1 2
10
-1 0 1 2
10 10 10 10 10 10 10 10
Frequency (rad/sec)
38
Free vibration analysis: Summary
• Goal : To identify natural frequencies and
corresponding modes

Natural frequencies can be found by…

• Eigenvalue analysis of DEOM

• Pole(s) of transfer function (if SISO)
• Eigenvalues of State Matrix A
(Eigenvectors  Modes)

39
Review: Free vibration of 1 DOF Spring-
mass system
x  n 2 x  0
k
Natural frequency n  [rad/sec]
k m
TF from (fictitious) input f to x
m X ( s) 1 k
G( s)   2 p1, 2 j
F ( s) ms  k m
x(t)

t
O
40
Free vibrations in multiple-degrees-of-
freedom systems
Two DOF system Free vibration=No input

x1 k2 x2 k
k1 3
m1 m2

DEOM m1x1  k1 x1  k2 ( x1  x2 )

m2 x2  k3 x2  k2 ( x2  x1 )

41
Vibration of linear mechanical systems
• Linear sum of the vibration modes
(principle of superposition)

Mode 1

+
Mode 2

In general, amplitudes and phases are dependent on initial conditions

42
Solution in the time domain
m1x1  k1 x1  k2 ( x1  x2 ) For simplicity

m2 x2  k3 x2  k2 ( x2  x1 ) m1  m2  m, k1  k2  k3  k

Assume x1  A sin t x1   A 2 sin t

Harmonic
solution x2  B sin t x2   B 2 sin t
(mA 2  2kA  kB) sin t  0

Since
(mB  2kB  kA) sin t  0
2

sin t  0
(mA 2  2kA  kB)  0  m 2  2k k   A
or     0
(mB 2  2kB  kA)  0  k  m  2k   B 
2

The equation needs to have nonzero (nontrivial) solution

 m 2  2k k 
det  0 43
 k  m  2k 
2
  m 2  2k k
 (m 2  2k ) 2  k 2  0
k  m 2  2k

 2 k  2 3k  k 3k
        0 1  , 2 
 m  m m m

 k 
  m  2k k   A1  A1
k
1  
m
k    0 1
m  k  m  2k   B1  B1
 m 

 3k 
  m  2k k   A2 
3k m A2
2   3k    0  1
m  k  m  2k   B2  B2
 m 

44
Laplace domain analysis (for natural frequency)
m1x1  k1 x1  k2 ( x1  x2 )
m2 x2  k3 x2  k2 ( x2  x1 )
m[ s 2 X 1 ( s)  sx1 (0)  x1 (0)]  2kX1 ( s)  kX 2 ( s)  0
m[ s 2 X 2 ( s)  sx2 (0)  x2 (0)]  2kX 2 ( s)  kX1 ( s)  0

Equation of X1 in the s-domain

 2k  k
[ sx1 (0)  x1 (0)] s 2    [ sx2 (0)  x2 (0)]
X 1 (s)   m m
4k 2 3k 2
s 
4
s  2
m m

k 3k
p1, 2 j , p3, 4   j 45
m m
State-space approach We have u for descriptive purpose.
u does not affect on natural frequencies.
u
x1 k x2 k
k
m m

mx1  kx1  k ( x1  x2 )  u
mx2  kx2  k ( x2  x1 )
x  z1 z 4   x1 x 2 
T
State vector z2 z3 x1 x2

 z1   2k
0 1 0 0 z  z1 
 1  0 
 z   
k
0 0  z   1  z 
 2   m m  2    m u x2  0 0 1 0 2 
 z3   0 0 0 1   z3   0   z3 
   k 0 
2k    
 
 z4   m m
0  z 4   0 
  z4  46
cont: Numerical eigenvalue analysis on A

 0 1 0 0 Numerical solution m=1, k=100

 2k k 
 m 0
m
0
A
0 0 0 1 1 
k
 10, 2 
3k
 17.32
 k 2k 
 0  0 m m
 m m 
A=
MATLAB eig command:
0 1 0 0
-200 0 100 0
0 0 0 1
100 0 -200 0

>> eig(A)
ans =

0 +17.32050807568879i
0 -17.32050807568879i
0 +10.00000000000000i
0 -10.00000000000000i 47
Systems dynamics and control
ME-3015

Mechanical Engineering
Georgia Institute of Technology

Time domain analysis and design of

control systems

1
Automatic control
• Open-loop control
– The output of the system has no effect on the
control action.
e.g. toaster (does not measure temperature, nor
color of bread)
• Closed-loop control (feedback control)
– A system that maintains a prescribed
relationship between the output and reference
input by comparing them.
e.g. Cruise control (measure speed), air conditioner
(measure temperature)
2
Automatic control (2)
Goal: reference input=output
Plant: a system to be controlled

Controller (compensator) : takes a reference input and calculates

control input to the plant, e.g., microprocessor

Open-loop control

Reference Output
input input
Controller Plant

Disturbance

Open-loop control: determines the reference input based on a priori knowledge

3
(model) of the plant
Automatic control (3) Closed-loop control
Error detector
Robot
Microprocessor
Reference DC motor Link mechanism
input Torque Arm angle
Controller Actuator Plant

ERROR

Sensor
Camera
Rotary encoder

Disturbance (unwanted noise, friction, etc)

Reference Output
input + ERROR
Controller Plant
- input
4
 D(s)

C(s)
+
R(s) K(s) P(s)
- input

Unity feedback (special case where K(s)=1)

C(s)
+
R(s) P(s)
- input

5
General requirements
• Must be stable (how to check stability)
• Must be reasonably fast (how to improve
performance)
• Small or no steady-state error (controller type)

elevator

Steady-state error

weight 6
Automatic controllers
•Two-position or ON-OFF
•Proportional
•Integral
•Proportional-plus-Integral (PI)
•Proportional-plus-Derivative (PI)
•Proportional-plus-Derivative-plus-Integral
(PID)
But not limited to these types.
7
Proportional control of 1st order system
J

Input torque

 bearing

No viscous friction at bearings:

bearing

Goal: Velocity (speed) control:   r, t  

J   , (0)  0
1
( s)    P( s)
Js 8
Proportional control of 1st order system (2)
Basic control storategy

 (r  0)  ( r )  ( r )
  
Faster
OK Slower
Apply negative torque Apply positive torque
=decrease angular velocity Do nothing
=increase angular velocity

Proportional control:
(proportional to the error)
  k pe

e( r   )
O
9
Proportional control of 1st order system (3)
K(s) input P(s)
e 1
r
+
kp 
-  Js

P( s ) K ( s )
Pole of closed-loop system: 1  P( s) K ( s)
kp
P( s) K ( s)
 Js  k p
1  P( s) K ( s) k p Js  k p
1
Characteristic equation:
Js
1  P( s ) K ( s )  0 Pole: p  k p / J 10
Stability and performance
Stability Pole: p  k p / J

Stable(=real part negative) kp  0 OK!

Performance
r(t)=r (>0) (constant)
kp k pr
r s( s) 
( s )  Js  k p
Js  k p s
k pr k pr
lim  (t )  lim s(s)  r
Final value theorem:
t  s 0 Js  k p kp
s 0

 (t )  r OK! 11
Example 1
1
P( s ) 
s( s  3)

K ( s)  2 Proportional control

(1) Draw block diagram for the closed loop system

(2) Obtain closed-loop TF (reference to output x) and check the stability
(3) Obtain x(t) for a unit step input. Plot x(t) versus time.

12
Example 1 solution
K(s) input P(s)
+ 1
r 2 x
- s ( s  3)

Open-loop TF
1
2
P( s) K ( s ) s( s  3) 2 2
  2 
1  P( s) K ( s) 1  2  1 s  3s  2 ( s  1)( s  2)
s( s  3)
1+Open-loop TF

p1, 2  1,2 (stable)

13
Example 1 solution (2)
2 1
X ( s)  
( s  1)( s  2) s
1 2 1
  
s s 1 s  2

x(t )  1  2et  e2t

x

t
O 14
Closed loop control (P displacement)
u (motor)

m m m
b
x

Goal1 (x=-1) Initial position (x=0) Goal 2 (x=1)

u P(s) x
1
P( s )  2
ms  bs
Design feedback controller K(s): x  r, t  
15
 Virtual spring (controller mimics this)

Basic idea Neutral position (e=0)

k

m
b
e

O (r=x)

Spring-mass-damper system: e  0( x  r ), t  

u  ke
Proportional

e K ( s)  k u 16
Closed loop control (displacement) (3)
K(s) input P(s)
+ 1
R k X
- ms 2  bs

Spring-mass-damper system?? P( s ) K ( s )
1  P( s) K ( s)
x

r x  r, t  
Underdamped, critically-damped, overdamped
initial
t
17
Closed loop stability (displacement, P)
1
k 2
P( s ) K ( s)
 ms  bs  2
k
1  P( s ) K ( s) 1  k 1 ms  (k  bs)
ms  bs
2
Stability check

Poles of the closed loop system: Im

 b  b 2  4mk
p1 , p2  Re

2
System is stable = real part negative k 0 (use normal spring)

18
Transient response
TF (zero initial )
k k r
X ( s)  2 R( s )  2 
ms  (k  bs) ms  (k  bs) s
Step input r (constant)

k/m n 2
r
X (s)  2 R( s )  2 
s  (k / m  b / ms) s  2n s  n s
2

b
n  k / m ,  
2 km
19
Transient response (2)

n 2
r
X (s)  2 
s  2 n s  n s
2

1  s   n 
 r   2 n
 2 
2 
 s s  2 n s  n s  2 n s  n 
2

20
Transient response (3)
Transient response of spring-mass-damper

 1 
x(t )  r 1  e nt sin(n 1   2  t   ) 
   2 
 1 
1  2
(10-14, Table #23)
  tan 1


x

t
O
21
What If P control only…
Neutral position (e=0)
k

m The system does not have a damper.

n 2 r O (r=x)
X (s)  2 
s  n s
2
r

 1 n 2 
 r   2 
2 
 s s  n 
O
1 s 
 r   2 
2 
 s s  2 n s  n  Need derivative control (add damping)

x(t )  r  r cos nt 22

ON-OFF (two position) control
The simplest control strategy
e=r-x>0 e=r-x<0
u=F(const) u=-F(const)
m m m
x

reference r

Two position or On-Off control

u
F
e
-F
23
Closed loop control (displacement) (3)
input
u
+ F 1
R e X
- -F ms 2

r oscillation around r

initial
t

Good to approach r, but… x  r, t   24

Closed loop control (displacement)
u (motor)

m m m
x

Goal1 (x=-1) Initial position (x=0) Goal 2 (x=1)

u P(s) x
1
P( s)  2
ms
Design feedback controller K(s): x  r, t  
25
 Virtual spring and damper (controller mimics these)

Position PD Neutral position (e=0)

k

m
b
e

O (r=x)

Spring-mass-damper system: e  0( x  r ), t  

u  ke  be
Proportional Derivative PD control

e K (s)  k  bs u 26
Position PD (2)
K(s) input P(s)
+ 1
R k  bs 2 X
- ms

Spring-mass-damper system?? P( s ) K ( s )
1  P( s) K ( s)
x

r x  r, t  
Underdamped, critically-damped, overdamped
initial
t
27
Position PD (3)
1
(k  bs) 2
P( s ) K ( s) ms (k  bs)
  2
1  P( s) K ( s) 1  (k  bs) 1 ms  (k  bs)
2
ms Stability check

Poles of the closed loop system: Im

 b  b 2  4mk
p1 , p2  Re

2
System is stable = real part negative k, b  0 (use normal spring & damper)

28
Position PD (4)
TF (zero initial )
(k  bs) (k  bs) r
X ( s)  2 R( s )  2 
ms  (k  bs) ms  (k  bs) s
Step input r (constant)

(k / m  bs / m) n  2n s r
2
X (s)  2 R( s )  2 
s  (k / m  b / ms) s  2n s  n s
2

b
n  k / m ,  
2 km
29
Position PD (5)

n  2 n s r
2
X (s)  2 
s  2 n s  n s
2

1 s 
 r   2 
2 
 s s  2 n s  n 

30
Position PD (6)
Transient response of spring-mass-damper

 1 
x(t )  r 1  e nt sin(n 1   2  t   ) 
   2 
 1 
1  2
  tan 1 (Table #23)

x

t
O
31
Transient response specifications
k
Recall last time m
b
e
O (r=x)

1
P( s )  2 K ( s)  k (P)
ms  bs

K(s) input P(s)

+ 1
R k X
- ms 2  bs
32
Transient response specifications (2)
b
n  k / m ,  
2 km

r
x(t )  r  e nt sin(n 1   2  t   )
1  2

1  2
  tan 1 (Table #23)

P518, Figure 10-24 : Hand out

Generally we prefer 0.4    0.8 33

Transient response specifications (3)

x Peak time x(t p )  x()

tp Mp 
x ( )
Maximum over shoot [%]
Mp

 0.02r
ts
Settling time

t
tr

Rise time (0 to 100%, 10 to 90%) 34

Peak time, overshoot, settling time
p.518 (10-15)

    
x(t )  r 1  e nt  sin(n 1   2  t )  cos(n 1   2  t )  
  1  2 
  
Peak time: dx n
r e nt sin(n 1   2  t )  0
dt 1  2

tp 
n 1   2

Maximum overshoot tp 
n 1   2

x(t )  r     /
Mp   e  / 1 2  sin( )  cos( )   e 1 2

r  1  2 
  35
Pole-plot: 2nd order system
n 2
Closed loop TF: G(s)  2
s  2 n s  n
2

1
Poles: Characteristic equation: 1  P( s ) K ( s )  2 k 0
ms  bs
Or the denominator of the closed-loop TF:


a  (n ) 2  n 1   2  
2 p1 , p2  n  jn 1   2
n
Im
jn 1   2

Re
  n

 jn 1   2
36
Pole-plot: 2nd order system (2)
Same overshoot (MP)

 / 1 2
Envelope the same   n
Mp  e
Im

Frequency the same

 tan   / 1   2
d  n 1   2
Re

37
Proportional control of 1st order system
J

Input torque

 bearing

Viscous friction at bearings:

Torque is proportional to angular velocity
bearing

Goal: Velocity (speed) control:   r, t  

J  b   ,  (0)  0
1
( s)    P( s)
Js  b P(s) 38
Proportional control of 1st order system (2)
  k pe K ( s)  k p
K(s) input P(s)
e 1
r
+
kp 
-  Js  b

Pole of closed-loop system:

kp
P( s) K ( s) Js  b  kp

1  P( s) K ( s) kp Js  b  k p
1
Js  b
39
Proportional control of 1st order system (3)
Step response r(t)=r (>0) (constant)
Final value theorem:
k pr k pr
lim  (t )  lim s(s)  r
t  s 0
Js  b  k p kp  b
s 0
kp
 (t )  r r
kp  b

r b
Offset (steady-state error) r
kp  b

initial  (t )
t

Larger gain kp , smaller offset 40

Integral control
• Eliminate offset (steady-state error)
• Use Integral of the error
t
r
 e(t )dt
0
Increases if steady-stead
error exists

initial  (t )
t

K ( s)  k p P (proportional) control

ki
K (s)  I (integral) control
s
ki
K (s)  k p  PI (proportional-integral) control
k s
K ( s)  k p  i
 kv s PID (proportional-integral-derivative) control 41
s
Use of Integral control
Step response r(t)=r (>0) (constant)
K(s) input P(s)
e 1
r
+
ki / s 
-  Js  b

ki 1
P( s) K ( s)
 s Js  b  2
ki
1  P ( s ) K ( s ) 1  ki 1 Js  bs  ki
s Js  b
Final value theorem:
ki r
lim  (t )  lim s( s)  lim s 2 r
t  s 0 s 0 Js  bs  ki s
 (t )  r (no steady-state error!!) 42
PID control
• Proportional-Integral-derivative control
t
u  k p e  ki  edt  kv e
0

ki
HW7 K ( s)  k p   kv s
Parameter turning (p566)
s
Ziegler-Nichols rules

Approach reference Add damping

Compensate for steady-state error

43
Example 2 (PID control)
1
P( s )  2
s s2
1
K1 ( s)  1   s PID control
s
K 2 ( s)  1  s PD control

Check the steady-state error for K1 and K2 for unit-step response.

Use final value theorem.

44
Example 2 (PID control) Solution
Closed-loop TF

PID control  s2  s 1 1

  2
P( s) K ( s)  s  s  s  2 s 2
 s 1
  2
1  P( s) K ( s)  s  s 1
2
1 ( s  s  1)  s( s 2  s  2)
1    2
 s s s2
Final value theorem:
s2  s 1
lim x(t )  lim sG( s)  lim s 2 1
t  s 0 s 0 ( s  s  1)  s ( s 2  s  2)

PD control s  1 2 1
s 1
P( s ) K ( s )
 s  s  2 
1  P( s) K ( s) 1  s  1 1 ( s  1)  ( s 2  s  2)
s2  s  2
s 1 1
lim x(t )  lim sG( s)  lim s  1/ 3  1 45
t  s 0 s 0 ( s  1)  ( s 2  s  2) s
System types (to check the steady-state error)
(1) Plant only
Unity feedback (special case where K(s)=1) (2) Controller + Plant

+
R(s) G(s) C(s)
-

K ( s  z1 )(s  z2 ) ( s  zm )
G( s)  N
s ( s  p1 )(s  p2 ) ( s  pn )

Number of poles at the origin.

Type 0 system if N=0.

Type 1 system if N=1.
Type 2 system if N=2….
46
System types (2) K p  lim G ( s )
s 0

K v  lim sG( s )
s 0

K a  lim s 2G ( s )
s 0
Steady-state error

step input Ramp input Acceleration input

r (t )  1 r (t )  t r (t )  t 2 / 2

Type 0 system 1  
1 K p
1
Type 1 system 0 
Kv
1
Type 2 system 0 0
Ka
47
Example 3 G(s)
P(s)
e 1
r
+
K(s) 
- Js  b

P control: Steady-state error for unit-step response

1
G( s)  k p Type 0
Js  b
I or PI control: No steady-state error for unit-step response
ki 1
G( s)  Type 1
s Js  b
ki  k p s 1
G( s)  Type 1 48
s Js  b
Example 3 (2)
PI control:
The table predicts Type 1 system will have steady-state error for a unit ramp response.
ki  k p s 1
G( s) 
s Js  b

ki  k p s
1
G( s) s Js  b ki  k p s
H ( s)   
1  G( s) ki  k p s 1 s( Js  b)  (ki  k p s)
1
s Js  b
ki  k p s 1 ki
K v  lim sH ( s)  lim s 
s 0 s 0 s Js  b b
1 b
Steady-state for a unit-ramp input:  49
K v ki
Unit ramp response

lim e(t )  lim r (t )   (t )  lim s[ R( s )  ( s )]

1
R( s)  2
s t  t  s 0

1 ki  k p s 1
 lim s  2  2
s 0
 s s ( Js  b)  (ki  k p s ) s 

1 ki  k p s 1
 lim   

s 0 s s ( Js  b)  (ki  k p s ) s 
 s ( Js  b)  (ki  k p s )  ki  k p s 
 lim  
s 0
 s ( Js  b)  s (ki  k p s )
2

t
 ( Js  b)  b
 lim   
s 0 s ( Js  b)  ( k  k s )
 i p  ki
50
Summary: Steady-state error
compensation
(1) Increase control gains
(2) Add integrator(s) (i.e., change system type)

51
DC motor modeling
Ra La=0
J b


ea(t) ia eb(t)

Constant air gap flux

(constant field current)
  K  ia K: motor-torque constant Not negligible but too small
to expect sufficient damping.
eb  Kb  Back EMF voltage
PD control
( s) K
 G( s)  (10-27)
Ea ( s ) s( sJ  b) 52
Position control systems with velocity
feedback
• How to obtain velocity (angular speed)
signal?
– Derivative of angle  w.r.t. time
– Tachometers (angular speed sensor)
• Back EMF
• Better S/N ratio

eh (t )  K h 

53
 P(s)

+ + K 1
r 
- - Js  b s

Kh

P(s)

+ K 1
r 1 Kh s 
- Js  b s

54
PD control (of theta) : Same diagram but implementation is different.
Stability and performanceK
+ K Js  b 
K
Js  b 1 Kh
K Js  (b  KK h )
-
Js  b

Kh K
Js  (b  KK h )

+ K
r
s( Js  (b  KK h )) 
-
K
s( Js  (b  KK h )) K
 2
1
K Js  (b  KK h ) s  K
s( Js  (b  KK h )) 55
Stability and performance (2)
Characteristics equation: Js 2  (b  KK h ) s  K
0    1.0  s 2  2 n s  n  0
2
Im
K b  KK h
n  , 
Re J 2 JK

x
r x  r, t  
Underdamped, critically-damped, overdamped
initial
t

Desirable 0.4    0.8

56
Systems dynamics and control
ME-3015

Mechanical Engineering
Georgia Institute of Technology

Stability analysis

1
Stability
Closed loop poles Im

stable
Real part(s) negative
Re
O

Left-half s-plane

2 1 s2
( s  1)( s  2) 2 ( s  5) s 1 ( s  1)( s  100)

General (higher order, i.e., greater than 2) case? 2

Routh-Hurwitz stability criterion
• In textbook, Routh’s method only
The denominator of closed loop TF:
n 1
b s    b1s  b0
G (s)  n 1
n 1
s  a s    a1s  a0
n
n 1
Or characteristic equation

s n  an 1s n 1    a1s  a0  0

We want to check if then the system is stable.

Re( pi )  0, i 3
Routh’s stability table
n 1
Characteristic equation: s  an 1s
n
   a1s  a0  0
sn R11  1 R12  an  2 R13  an  4 
s n 1 R21  an 1 R22  an 3 R23  an 5 
an 1an  2  an 3 an 1an  4  an 5 an 1an 6  an 7
s n2 R31  R32  R33  
an 1 an 1 an 1
R31 R22  R21 R32 R31 R23  R21 R33 R31 R24  R21 R34
s n 3 R41  R42  R43 
R31 R31 R31

s1 Rn1 0 0
s0 R( n 1)1 0 0

The system is stable if and only if:

ai  0 (i  0,1,2,  , n  1)
Ri1  0 (i  3, , n  1) 4
Example 1
• Check the stability of the following closed-loop transfer
function. Use Routh’s method.
1
r 
s 5  3s 4  2s 3  s 2  3s  2

s5 1 2 3

s4 3 1 2
3 2  3  1 1 5 3  3  1 2 7
s R31   R31   0
3 3 3 3
s 2 (5 / 3) 1  (7 / 3)  3 16 (5 / 3)  2  0  3
R41   R32  2 0
5/3 5 (5 / 3)
(16 / 5)  (7 / 3)  (5 / 3)  2 27
s1 R51   0
 16 / 3 8
s0 2 5
Example 2
• Check the stability of the following closed-loop transfer
function. Use Routh’s method.
1
r 
s 4  2s 3  3s 2  4s  5

s4 1 3 5
3
s 2 4 0
2 2  3  1 4 2  5  1 0
s R31  1 R31  5 0
2 2
4 1  2  5
s1 R41 
1
 6 0 0
(6)  5  1 0
s0 R51 
6
5

6
Example 3
• Determine k such that the closed-loop system is stable.
Use Routh’s method.
+ k 1
r x
- s2 s ( s  1)

1  PK  0  s 3  3s 2  2s  k  0
s3 1 2
2
s 3 k
2  3  1 k
1 R31   (6  k ) / 3 0
s 3
(6  k ) / 3  k  0
s0 R41  k 0
(6  k ) / 3
(6  k ) / 3  0 and k  0  0  k  6 7
Example 3: Pole-zero Map sample22_1.m on T-square

k=2 (stable) k=10 (unstable)

Step Response Step Response
1.4 8

1.2 6

4
1

Amplitude
Amplitude

0.8
0
0.6
-2

0.4
-4

0.2 -6

0 -8
0 5 10 15 20 25 30 35 40 45 0 5 10 15
Time (sec) Time (sec)

Pole-Zero Map Pole-Zero Map

1 2

0.8
1.5
0.6
1
0.4

Imaginary Axis
Imaginary Axis

0.5
0.2

0 0

-0.2
-0.5

-0.4
-1
-0.6
-1.5
-0.8

-1 -2
8
-3 -2.5 -2 -1.5 -1 -0.5 0 -3.5 -3 -2.5 -2 -1.5 -1 -0.5 0 0.5
Real Axis Real Axis
Hurwitz’s method
Characteristic equation: s n  an 1s n 1    a1s  a0  0
principal minor
an 1 an 3 an 5  0 
 1 an  4  0 
an 1 an 3
 an  2 H2 
1 an  2
 0 an 1 an 3 0
H   an 1 an 3 an 5
 0 1 an  2 0
H3  1 an  2 an  4
     
  0 an 1 an 3
 0 0  a0 

The system is stable if and only if:
ai  0 (i  0,1,2,  , n  1)
H i  0 (i  2,  , n  1) 9
Routh’s method and Hurwitz’s method are essentially the same.

H i 1
Ri1   0 (i  3,  , n)
H i 2

10
Example 4
• Determine k such that the closed-loop system is stable.
Use Horwitz’s method.
+ k 1
r x
- s2 s ( s  1)

1  PK  0  s 3  3s 2  2s  k  0
3 k 0  3 k

H  1 2 0   H2   6k  0
1 2
0 3 k 
Same result

(6  k )  0 and k  0  0  k  6 11
Example 3: Pole-zero Map
k=2 (stable) k=10 (unstable)
Step Response Step Response
1.4 8

1.2 6

4
1

Amplitude
Amplitude

0.8
0
0.6
-2

0.4
-4

0.2 -6

0 -8
0 5 10 15 20 25 30 35 40 45 0 5 10 15
Time (sec) Time (sec)

Pole-Zero Map Pole-Zero Map

1 2

0.8
1.5
0.6
1
0.4

Imaginary Axis
Imaginary Axis

0.5
0.2 -3.3
0 0

-0.2
0 -0.5 -2 -1 0
-0.4
-1
-0.6
-1.5
-0.8

-1 -2
12
-3 -2.5 -2 -1.5 -1 -0.5 0 -3.5 -3 -2.5 -2 -1.5 -1 -0.5 0 0.5
Real Axis Real Axis
Worksheet Stability analysis 1
• Check the stability of the following closed-loop transfer
function. Use Routh’s method.
1
r 
s 4  2s 3  s 2  s  2

4
s
3
s
2
s
s1
s0
13
Worksheet Stability analysis 1 (solution)
• Check the stability of the following closed-loop transfer
function. Use Routh’s method.
1
r 
s 4  2s 3  s 2  s  2

s 4 1 1 2
3
s 2 1 0
2 1 2  1 1 1 2  2  1 0
s R31   R31  2 0
2 2 2
1 / 2 1  2  2 7
s1 R41 
1/ 2

2
0 0
(7 / 2)  2  1 0
s0 R51 
7/2
2

UNSTABLE 14
Worksheet Stability analysis 2
• Check the stability of the following closed-loop transfer
function. Use Horwitz’s method.
1
r 
s 3  2s 2  s  1

 

H   H2 

 

15
Worksheet Stability analysis2 (solution)
• Check the stability of the following closed-loop transfer
function. Use Horwitz’s method.
1
r 
s 3  2s 2  s  1

an 1 an 3 0   2 1 0
H   1 an  2 0   1 1 0
 0 an 1 an 3  0 2 1

2 1
H2   2 1  1  0
1 1
STABLE 16