Professional Documents
Culture Documents
I.D. Landau
Laboratoire d Automatique de Grenoble, (INPG/CNRS), France
April 2002
1 I.D. Landau : Identification in closed loop/A powerful design tool
Outline - Introduction - An example (flexible transmission) - Objectives of identification in closed loop - Basic Schemes - The CLOE Algorithms (closed loop output error) - Properties of the algorithms - Properties of the estimated models - Validation of models identified in closed loop - Iterative identification in c. l. and controller re-design - Experimental results ( flexible transmission) - Filtered open loop identification algorithms - Conclusions
2 I.D. Landau : Identification in closed loop/A powerful design tool
IDENTIFICATION
Ref. CONTROLLER
+ +
PLANT
1) Identification of the dynamic model 2) Performance and robustness specifications 3) Compatible controller design method 4) Controller implementation 5) Real-time controller validation (and on site re-tuning) 6) Controller maintenance (same as 5)
u Plant
+ + y
y + -
Controller
Design system (simulation)
Model
Comparison of achieved and desired performances A useful interpretation : Check to what extent the model used for design allows achievement of: - desired nominal performances - desired robustness specs.(sensitivity functions)
4 I.D. Landau : Identification in closed loop/A powerful design tool
May provide better design models Cannot be dissociated from the controller and robustness issues
6 I.D. Landau : Identification in closed loop/A powerful design tool
m
d.c. motor
axis position
Controller
A u(t) + D C +
R-S-T controller
ref
fs = 20Hz
0% load
Benefits of identification in closed loop (1) controller design using the open loop identified model output
output
cl. loop syst. pole computed pole
real system
real system
control control
The pattern of identified closed loop poles is different from the pattern of computed closed loop poles
I.D. Landau : Identification in closed loop/A powerful design tool
Benefits of identification in closed loop (2) controller computed using the closed loop identified model
output output
real system
real system
control control
The computed and the identified closed loop poles are very close
10 I.D. Landau : Identification in closed loop/A powerful design tool
+ r T
q-d B A Plant
Sup
ru
AS P
AR P + u +
w q-d B + A + y
Syp
Objective : development of algorithms which: - take advantage of the improved input spectrum - are insensitive to noise in closed loop operation
11 I.D. Landau : Identification in closed loop/A powerful design tool
Objective of the Identification in Closed Loop (identification for control) Find the plant model which minimizes the discrepancy between the real closed loop system and the simulated closed loop system. r
Controller
Plant
Controller
+ y + + y
noise
CLOE Algorithms
Model
Simulated System
CL
1/S
Closed Loop Output Error Identification Algorithms (CLOE) Excitation added to reference signal
+ r T R + R + u 1/S q-d B A Model yr 1/S u w q-d B + A + Plant y +
CL
ru
Plant + w q-d B A + + T R
+ 1/S +
CL
1/S + u
+ q-d B A R y
u = -R y + T r S S
u = -R y + T r S S
u = - R y+ ru S
u = - R y +ru S
Notations
r 1
v(t) u
p(t)
G
R (q 1 ) K(q ) = S(q 1 )
1
q d B(q 1 ) G (q ) = A(q 1 )
Sensitivity functions :
Syp (z 1 ) = 1 K G KG ; Sup ( z 1 ) = ; Syv (z 1 ) = ; Syr ( z 1 ) = 1 + KG 1 + KG 1 + KG 1 + KG
P(z 1 ) = A(z 1 )S(z 1 ) + z d B(z 1 )R (z 1 )
True closed loop system :(K,G), P, Sxy ), P xy ,S Nominal simulated(estimated) closed loop : (K, G
Closed Loop Output Error Identification Algorithms (CLOE) Excitation added to reference signal
u
r
+ y + +
G
K
u
y
+
CL
CL
P.A.A.
K
P.A.A.
CLOE Adjustable predictor (closed loop) y (t + 1) = T (t ) (t ) y (t + 1) = T (t + 1) (t ) R u (t ) = y (t ) + ru S T (t ) = a 1 (t ),...b n (t ) 1 (t ),...a n (t ), b T (t ) = [ y (t ),.. y (t n A + 1), u (t d ),..u (t d n B )] The Parameter Adaptation Algorithm
0 T (t ) (t ) = y(t + 1) y CL (t + 1) = y(t + 1) 0 (t + 1) 0 (t + 1) = (t ) + F (t + 1) (t ) CL (t + 1)
AF-CLOE
CLOE Properties Case 1: The plant model is in the model set (i.e. the estimated model has the good order) Asymptotic unbiased estimates in the presence of noise subject to a (mild) sufficient passivity condition CLOE F-CLOE AF-CLOE
S / P /2 / P/2 P
none
Case 2: The plant model is not in the model set (ex.: the estimated model has a lower order) See the next slides
20 I.D. Landau : Identification in closed loop/A powerful design tool
= arg min
S yp
S yp r ( ) + p ( )]d [G G
will minimize the 2 norm between the true sensitivity -G function and the sensitivity function of the closed loop estimator when r(t) is a white noise (PRBS) -Plant -model error heavily weighted by the sensitivity functions -The noise does not affect the asymptotic estimation
21 I.D. Landau : Identification in closed loop/A powerful design tool
= arg min S yp
S up r () + p ( )]d [G G
will minimize the 2 norm between the true sensitivity -G function and the sensitivity function of the closed loop estimator when r(t) is a white noise (PRBS) -Plant -model error heavily weighted by the sensitivity functions -The noise does not affect the asymptotic estimation
22 I.D. Landau : Identification in closed loop/A powerful design tool
Identification in closed loop - Some remarks - The quality of the identified model is enhanced in the critical frequency regions for control (compare with open loop id.) = arg min CLOE
* 2
S yp
S yp r ( ) + p ( )]d [G G
OLOE
* = arg min
r ( ) + p ( )]d [G G
-Identification in closed loop can be used for model reduction. The approximation will be good in the critical frequency regions for control.
23 I.D. Landau : Identification in closed loop/A powerful design tool
+ -
B + R/S
q-1A* +
Validation of Models Identified in Closed Loop Controller dependent validation ! 1) Statistical Model Validation 2) Pole Closeness Validation 3) Time Domain Validation
2.17 RN (i ) ; i 1 N
normalized crosscorelation number of data 1 N 2.17 N 97%
Uncorrelation Test
N = 256 RN (i ) 0.136
pratical value :
RN (i ) 0.15
26
Identification in Closed Loop Pole Closeness Validation Closeness tests of achieved and desired (computed) poles and/or sensitivity functions by identification of the closed loop
+ 1 S R q-dB P P.A.A. u q-dB A
PLANT
y +
Rem: -use of open loop identification algorithms -same signals as those used for the identification of the plant model in closed loop operation
I.D. Landau : Identification in closed loop/A powerful design tool
27
Identification in Closed Loop Time Domain Validation Comparison of achieved and simulated performance in the time domain r Controller u
noise
Plant
+ y + +
y
Controller
Design system (simulation)
Model
Rem: - not enough accuracy in many cases - difficult interpretation of the results in some cases
28 I.D. Landau : Identification in closed loop/A powerful design tool
OPEN LOOP BASED CONTROLLER PLANT IDENTIFICATION IN CLOSED LOOP (OLBC) IDENTIFICATION OF THE CLOSED LOOP COMPARATIVE CLOSED LOOP VALIDATION BEST PLANT MODEL
CLOSED LOOP BASED CONTROLLER (CLBC) 29 I.D. Landau : Identification in closed loop/A powerful design tool
w y + + + y -
Plant
q-d B/A
CL
Model
Step 1 : Identification in Closed Loop -Keep controller constant -Identify a new model such that CL Step 2 : Controller Re Design - Compute a new controller such that CL Repeat 1, 2, 1, 2, 1, 2,
30 I.D. Landau : Identification in closed loop/A powerful design tool
Experimental Results
axis position
d.c. motor
Controller
A u(t) + D C +
R-S-T controller
ref
CL model
OL model
Flexible Transmission
Model identified in open loop Model identified in closed loop The model identified in closed loop provides computed poles closer to the real poles than the model identified open loop
34 I.D. Landau : Identification in closed loop/A powerful design tool
0.8
0.8
position (V)
0.4
0.6
0.6
0.4
0.2
0.2
0 29
Model identified in open loop Model identified in closed loop The simulation using the model identified in C.L. is closer to the real response than the simulation using the O.L.identified model
35 I.D. Landau : Identification in closed loop/A powerful design tool
Controller Re-design Based on the Model Identified in Closed Loop (on-site controller re-tuning)
1 1
0.8
0.8
position (V)
0.4
0.6
0.6
0.4
0.2
0.2
0 29
0 29 30 31 32 33 temps (s) 34 35 36
Re-designed controller (CLBC) Initial controller (OLBC) The CLBC controller provides performance which is closer to the designed performance than that provided by the OLBC controller
36 I.D. Landau : Identification in closed loop/A powerful design tool
Adjustable Model
-Biased estimates -Require (theoretically) time varying filters -FOL alg. can be seen as approximations of CLOE alg. -Are used instandard indirect adaptive control
37 I.D. Landau : Identification in closed loop/A powerful design tool
Concluding Remarks - Methods are available for efficient identification in closed loop - CLOE algorithms provide unbiased parameter estimates - CLOE provides control oriented reduced order models (precision enhanced in the critical frequency regions for control) -The knowledge of the controller is necessary -In many cases the models identified in closed loop allow to improve the closed loop performance -For controller re-tuning, opening the loop is no more necessary -Identification in closed loop can be used for model reduction -By duality arguments one can use the algorithms for controller reduction - Successful use in practice - A MATLAB Toolbox is available (CLID) -A stand alone software is available (WinPIM/Adaptech)
38 I.D. Landau : Identification in closed loop/A powerful design tool
Personal References
Landau I.D., Karimi A., (1997) : Recursive algorithms for identification in closed -loop a unified approach and evaluation , Automatica, vol. 33, no. 8, pp. 1499-1523. Landau I.D., Lozano R., MSaad M., (1997) : Adaptive Control, Springer, London,U.K. Landau I.D., (2001) : Identification in closed loop : a powerful design tool (better models, simple controllers) , Control Engineering Practice , vol. 9, no. 1, pp. 51- 65. Landau I.D., (2002) Identification et Commande des Systmes, 3rd edition, Hermes, Paris (June)