Scribd
Upload
78 views25 pages

Pole Placement and Model Matching

This document discusses pole placement and model matching using feedback control systems. It introduces: 1. Unity feedback and two-degree-of-freedom configurations for pole placement. Compensator equations can be solved using coprime polynomials. 2. Pole placement is possible if the degree of the compensator polynomial is greater than or equal to the degree of the plant polynomial minus one. 3. Model matching transfers can be implemented using proper feedforward and feedback compensators in a two-parameter configuration if the overall transfer function and its inverse are proper and stable.

Uploaded by

Kumara Samy
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
78 views25 pages

Pole Placement and Model Matching

This document discusses pole placement and model matching using feedback control systems. It introduces: 1. Unity feedback and two-degree-of-freedom configurations for pole placement. Compensator equations can be solved using coprime polynomials. 2. Pole placement is possible if the degree of the compensator polynomial is greater than or equal to the degree of the plant polynomial minus one. 3. Model matching transfers can be implemented using proper feedforward and feedback compensators in a two-parameter configuration if the overall transfer function and its inverse are proper and stable.

Uploaded by

Kumara Samy
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 25

9.

Pole placement and model matching


• 9.1 Introduction
• Feedback configuration for pole placement
unity-feedback
• Two-degree-of-freedom configuration
controller-estimator
(plant-input-output-feedback);
two-parameter configuration.
• Compensator equations-classical method
Consider
A(s)D(s)+B(s)N(s) = F(s)
where D(s), N(s), and F(s) are given, and
A(s) and B(s) are unknown to be solved.
• Theorem 9.1 A(s) and B(s) exist for any F(s)
if and only if D(s) and N(s) are coprime.
• If D(s) and N(s) are coprime, there exist
polynomials A(s) and B(s) such that
A ( s ) D ( s ) + B( s ) N ( s ) = 1
• Unity-feedback configuration
• Two-degree-of -freedom configuration
• Any polynomial F(s) satisfies
F( s ) A ( s ) D ( s ) + F( s ) B ( s ) N ( s ) = F( s )
• Thus (the particular solutions)
A(s) = F(s) A (s) and B(s) = F(s) B(s) are solutions.
• The general solutions: Let
Â(s) D(s) + B̂(s) N (s) = 0
• Obviously, Â(s) = − N (s) B̂(s) = D(s) are such solutions.
• Then for any polynomial Q(s)
A(s) = A (s) F(s) + Q(s) Â(s) and B(s) = B(s)F(s) + Q(s)B̂(s)
are general solutions.
• 9.2 Unity-feedback configuration-pole
placement
• Let ĝ(s)=N(s)/D(s) and C(s)=B(s)/A(s).
• Then
pC(s)ĝ(s) pB(s) N (s)
ĝ o (s) = =
1 + C(s)ĝ(s) A(s) D(s) + B(s) N (s)
• The pole-placement problem becomes one
of solving the polynomial equation
A(s)D(s)+B(s)N(s) = F(s)
• Feedback can shift poles but has no effect
on zeros.
• Let deg N(s) < degD(s)=n and Deg B(s) ≤
degA(s)=m.
• Then F(s) has degree at most n+m.
• The solution can be arranged in matrix form as
[A0 B0 A1 B1 … Am Bm]Sm=[F0 F1 F2 … Fn+m]
with ⎡⎢ D0 D1 . Dn −1 0 . 0 ⎤⎥
N0 N1 . N n −1 0 . 0
⎢ ⎥
⎢ 0 D0 . D n −1 Dn . 0 ⎥
⎢ ⎥
Sm := ⎢ 0 N 0 . N n −1 N n . 0 ⎥
⎢ . . . . . . . ⎥
⎢ ⎥
⎢ 0 0 . 0 D0 . Dn ⎥
⎢⎣ 0 0 . 0 N0 . N n ⎥⎦ 2( m +1)×( n + m +1)
• Applying corollary 3.2, the matrix has a
solution for any F(s) if and only if Sm has
full column rank.
• A necessary condition for Sm to have full
column rank is that Sm is square or has more
rows than columns, that is
2(m+1) ≥ n+m+1 or m ≥ n-1
• If m<n-1, then solutions may exist for some
F(s), but not for every F(s): it is not
possible to achieve pole placement.
• Theorem 9.2 Let m ≥ n-1. Then for any
polynomial F(s) of degree (n+m), there
exists a proper compensator C(s)=B(s)/A(s)
of degree m such that
pN (s) B(s) pN (s) B(s)
ĝ o (s) = =
A(s) D(s) + B(s) N (s) F(s)

The compensator can be obtained by


solving 9.13.
• B(s)/A(s) is proper or Am ≠ 0
If N(s)/D(s) is strictly proper. Then Nn = 0,
and the last eq. of 9.13 reduces to
AmDn + BmNn = DnAm = Fn+m
Because F(s) has degree (n+m) desired poles,
we have Fn+m≠0 and, consequently, Am≠0.
• If m=n-1, the compensator is unique.
• The regulation problem (r = 0)
If all poles are selected to have negative real
parts, then for any gain p, the overall system
will achieve regulation.
• The tracking problem: Let r̂ (s) = a/s.
ŷ(s) = ĝ o (s) r̂ (s) = ĝ o (s)a / s
lim y( t ) = lim sŷ(s) = ĝ o (s)a
t →∞ s→ 0

• Tracking asymptotically:
ĝo(s) must be BIBO stable and
ĝo(0) = 1.
• Thus, we have
N (0) B(0) B N F0
ĝ o (0) = p = p 0 0 =1⇒ p =
F(0) F0 B0 N 0
• This is summarized in the following
Regulation ⇔ ĝo(s) BIBO stable
Tracking step reference input
⇔ ĝo(s) BIBO stable and ĝo(0) = 1.
Tracking ramp reference input
⇔ ĝo(s) BIBO stable and ĝo(0)=1 and
ĝo‘(0)=0
• Robust tracking and disturbance rejection
• Let φ(s) be the least common denominator of
the unstable poles of r̂ (s) and ŵ(s).
• Theorem 9.3 Let ĝ(s)=N(s)/D(s) be strictly
proper and assume that D(s) and N(s) are
coprime. If no root of φ(s) is a zero of ĝ(s),
then there exist a proper compensator such
that the overall system will track r(t) and
reject w(t), both asymptotically and robustly.
• The robust design consists of two steps.
• First find a model 1/φ(s) (internal model
principle) and then design proper B(s)/A(s)
(pole-placement design).
• Embedding internal models
the compensator B(s)/Ã(s)φ(s)⇒B(s)/A(s)
the degree of compensator can be reduced
A(s)D(s) + B(s)N(s) = F(s)
• 9.3 Implementable transfer function
• The model matching problem
given a desired overall transfer function
ĝo(s), find a feeback configuration and
compensators so that the transfer function
from r to y equals to ĝo(s).
Pole-and-zero placement
• Although ĝo(s)=1 (y(t) = r(t)) is the best
overall system, we may not to match it for a
given plant.
• Well posed
If the closed-loop transfer function of every
possible input-output pair is proper.
• If the overall transfer function from r to y is
BIBO stable and if there is no unstable
pole-zero cancellation, then the overall
system is total stable.
• Definition 9.1 Overall transfer function ĝo(s)
is said to be implementable if there exists
no-plant-llinkage configuration and proper
compensators so that the system equals ĝo(s)
and is well posed and totally stable.
• Theorem 9.4 Consider a proper transfer
function ĝ(s). Then ĝo(s) is implementable
if and only if ĝo(s) and
ĝ (s)
t̂ (s) := o
ĝ(s)
are proper and BIBO stable.
• Corollary 9.4 Consider ĝ(s)=N(s)/D(s)
proper. Then ĝo(s)=E(s)/F(s) is
implementable if and only if
1. F(s) is Hurwitz.
2. degF(s)-degE(s)≥degD(s)-degN(s)
3. All zeros of N(s) with zero or positive real
parts are retained in E(s).
• Controller-estimator or plant-input-output-
feedbak configuration

1 C 2 (s)
û(s) = r̂ (s) − ŷ(s)
1 + C1(s) 1 + C1(s)
• Two-parameter configuration

û (s) = C1(s) r̂ (s) − C2 (s) ŷ(s)


• Consider the two-parameter configuration
L(s) M (s )
Let 1
C ( s ) =
A1(s)
(feedforwa rd) C 2 ( s ) =
A 2 (s )
(feedback )

• Assume
L(s ) M (s )
C1(s) = C 2 (s ) =
A (s) A (s )
• The overall transfer function
N (s )
ĝ(s) L(s ) D(s )
ĝ o (s) = C1(s) =
1 + ĝ(s)C2 (s) A(s) 1 + N (s) M (s)
D(s ) A (s )
L( s ) N ( s )
=
A ( s ) D(s ) + M (s ) N ( s )
• In model matching, searching for proper
L(s)/A(s) and M(s)/A(s) meets
E (s ) L(s ) N (s )
ĝ o (s) = =
F(s) A(s) D(s) + M (s) N (s)
• Procedure 9.1
1. Compute
E (s ) N (s ) L( s ) N (s )
ĝ o (s) = =
F (s ) A (s ) D(s ) + M ( s ) N (s )
2. Introduce an arbitrary Hurwitz polynomial
F̂(s) such that the degree of F(s) F̂(s) is 2n-1
or higher.
3. Rewrite ĝo (s) as
E (s) F̂(s) N (s) L( s ) N ( s )
ĝ o (s) = =
F(s) F̂(s) A (s ) D(s ) + M (s ) N (s )
Set L(s) = E (s) F̂(s)
and solve A(s) and M(s) from
A(s) D(s) + M (s) N (s) = F(s) F̂(s)
[A0 M0 A1 M1 … Am Mm]Sm=[F0 … Fn+m]
with
⎡ D0 D1 . D n −1 0 . 0 ⎤
⎢N N1 . N n −1 0 . 0 ⎥
⎢ 0 ⎥
⎢ 0 D0 . D n −1 Dn . 0 ⎥
⎢ ⎥
Sm := ⎢ 0 N 0 . N n −1 N n . 0 ⎥
⎢ . . . . . . . ⎥
⎢ ⎥
⎢ 0 0 . 0 D0 . Dn ⎥
⎢⎣ 0 0 . 0 N0 . N n ⎥⎦

The compensators L(s)/A(s) and M(s)/A(s)


are proper.
• Implementation of two-parameter
compensators

You might also like