You are on page 1of 15

Automatic Control

Mode analysis
Mode analysis – Introduction

• Consider an LTI system in state space form:

x  t   Ax  t   Bu  t 
y  t   Cx  t   Du  t 

• Let 𝜆 𝐴 be the eigenvalues of 𝐴:

 ( A)  1 , 2 ,..., n x

• These eigenvalues are fundamental: they determine
- the stability
- the dynamic properties
- the free and forced evolution
of a system.
Mode analysis – Natural modes

• Suppose that the eigenvalues of 𝐴 are the following:


( A )  1 , 2 , 3 , 4 , 5 , 1 , 2 , 3 real ,  4 , 5 compl. conj.
 3   2  1 ,  5  *4  1   2

• We can associate partial fractions to the eigenvalues as follows:

eigenvalue partial fraction


r1
real, mult. 1 : 1
s  1
 r2 where
s  
 ri  C
real, mult. 2 :  2   3
2

 r3 ri  0
 s   2 2
r4 r4*
compl. conj. :  4  *5 
s   4 s  *4
Mode analysis – Natural modes

• Apply to each partial fraction the inverse Laplace transform:

eigenvalue partial fraction mode


r1
real, mult.1 : 1 e 1t
s  1
 r2
s   e 2 t

real, mult. 2 : 2  3
2

 r3 te 2t
 s  2 2
r4*
e Re  λ4 t cosIm λ4 t   
r4
compl. conj. : 4  *5 
s  4 s  *4

Definition. The natural modes of the system are the inverse Laplace
transforms of the partial fractions associated to the eigenvalues.
Mode analysis – Comments

• Relation eigenvalues  modes:

real eigenvalue with mult. 1  1 mode


real eigenvalue with mult. k  k modes
couple of c.c. eigevalues with mult. 1  1 mode
couple of c.c. eigevalues with mult. k  k modes

• The modes are functions of time of the following types:

  t  , e pt , e t cos t    , t m , t me pt , t me t cos t   


Mode analysis – Classification I

• Convergent modes: p  0,   0, m  0
e pt , e t cos t    , t me pt , t me t cos t   

• Stationary modes:

  t  , cos t   

• Divergent modes: p  0,   0, m  0; p  0,   0, m  0

e pt , e t cos t    , t m , t me pt , t me t cos t   


Mode analysis – Classification I

ept, p<0 tept, p<0 A eσt cos(wt+φ), σ<0


• Convergent
modes:

Linear combination of
ε(t) A cos(wt+φ) 2 stationary modes
• Stationary
modes:

ept, p>0 t, t2 A eσt cos(wt+φ), σ>0


• Divergent
modes:
Mode analysis – Classification II

• Real modes (associated to real eigenvalues):

ept, ε(t), 𝑡, 𝑡 2 , 𝑡 ept, 𝑡 2 ept...

• Complex modes (associated to complex eigenvalues):

𝐴cos 𝑤𝑡 + 𝜑
𝐴𝑒𝜎𝑡 cos 𝑤𝑡 + 𝜑 oscillatory behavior
𝐴𝑡𝑒𝜎𝑡 cos 𝑤𝑡 + 𝜑
...
Mode analysis – Classification II

ept, p<0 ε(t) ept, p>0

• Real
modes:

A eσt cos(wt+φ), σ<0 A cos(wt+φ) A eσt cos(wt+φ), σ>0


• Complex
modes:
Mode analysis – Modes of a transfer function

• Consider a transfer function 𝐺(𝑠) and its partial fraction expansion:


R1 R2  R4 R4* 
G( s)        *
 
s  p1 s  p2  s  p4 s  p4 

Definition. The modes of a transfer function 𝐺(𝑠) are the inverse Laplace
transforms of its partial fractions.

• Analogously:

Definition. The modes of a signal 𝑌(𝑠) are the inverse Laplace transforms
of its partial fractions.

• If 𝑌 𝑠 = 𝐺 𝑠 𝑈(𝑠), then 𝑌 𝑠 = 𝑀𝐺 𝑠 + 𝑀𝑈 𝑠 , where


- 𝑀𝐺 𝑠 : partial fractions of 𝐺 𝑠 ,
- 𝑀𝑈 𝑠 : partial fractions of 𝑈 𝑠 .
The modes of 𝑦(𝑡) include the modes of both 𝐺 𝑠 and 𝑢(𝑡).
Mode analysis – Modes of a transfer function

• Consider an LTI system described by its transfer function:

Y ( s )  G ( s )U ( s ), G ( s )  C sI  A B  D
1

• If the system is completely controllable and observable (CCO)

 { p1 ,..., pn }  {1 ,..., nx }


 the natural modes coincide with the modes of 𝐺(𝑠).

• If the system is not CCO, the modes of 𝐺(𝑠) are a subset of the
natural modes of the system.
Typical first order system

1
• Transfer function: G s 
1 s
 : time constant of the system.

t
1 1 
• Step response: Y s   y t   1  e 
1 s s

 > 0 : stable pole


(convergent mode)

 < 0 : unstable pole


(divergent mode)
Typical second order system

1
• Transfer function: G s 
 1
1 2 s  2 s2
n n
𝜁 : damping.
 n  0 : natural frequency.

Poles: 𝑝 = 𝜔𝑛 −𝜁 ± 𝑗 1 − 𝜁 2

𝜁<0 → unstable poles (divergent modes)


𝜁=0 → compl. conj. poles (stationary oscillatory mode)
𝜁≥1 → real as. stable poles (convergent modes)
0 < 𝜁 <1 → stable compl. conj. poles (convergent mode)
Typical second order system

• Step response:

e nt
y t   1  sin  n 1   2 t  arccos   
1  2  

y(t)

ωn t
Typical second order system

• Step response for different values of n:

s=tf('s');

zeta=0.4;
wn=1;
G1=1/(1+2*zeta/wn*s+s^2/wn^2);

step(G1,15), hold on

wn=5;
G2=1/(1+2*zeta/wn*s+s^2/wn^2);

step(G2,15)
ωn=1,5

You might also like