system identification: black box

36
System Identification: Black Box Robert Grepl Institute of Solid Mechanics, Mechatronics and Biomechanics Faculty of Mechanical Engineering, Brno University of Technology 2010 - 2015

Upload: others

Post on 24-Apr-2022

0 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: System Identification: Black Box

System Identification: Black Box

Robert GreplInstitute of Solid Mechanics, Mechatronics and Biomechanics

Faculty of Mechanical Engineering, Brno University of Technology

2010 - 2015

Page 2: System Identification: Black Box

Black box - agenda

1. Prerequisities: Least squares.

Discrete dynamic system.

2. How to use LS to estimate parameters of DDS? Example: mechanical oscillator

3. Problem: Noise.

www.mechlab.czISMMB, Faculty of Mechanical Engineering, Brno University of Technology

2

Page 3: System Identification: Black Box

www.mechlab.czISMMB, Faculty of Mechanical Engineering, Brno University of Technology

3

Methods: Least Squares 1/2

generally: LS is Linear Optimization method

assume function

(linear)(ordinary) LS requires linear parametrization:example:

consider m measurements... written in matrix form:

if m=n, then Problem: Find parameters of

line if 2 points are given.

Page 4: System Identification: Black Box

www.mechlab.czISMMB, Faculty of Mechanical Engineering, Brno University of Technology

4

Methods: Least Squares 2/2

in real apps, the measurement is not ideal

task is: how to solve this overdetermined problem? (m equations and n unknowns)

answer: minimize criterion

result:

Note: Penrose pseudoinversion

Page 5: System Identification: Black Box

www.mechlab.czISMMB, Faculty of Mechanical Engineering, Brno University of Technology

5

LS example: regression 1D function

fit 1D function

(this task = the fitting of static model linear in parameters)

noise added to measurement

regression matrix:

[E015]

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1-2.5

-2

-1.5

-1

-0.5

0

0.5

1

1.5

2

2.5

t

y

noisy data

original

fit

Page 6: System Identification: Black Box

www.mechlab.czISMMB, Faculty of Mechanical Engineering, Brno University of Technology

6

Recursive Least Squares (RLS) (Plackett's algorithm)

online estimation

gives identical result as OLS

assume model

in matrix form:

Page 7: System Identification: Black Box

Black box - agenda

1. Prerequisities: Least squares.

Discrete dynamic system.

2. How to use LS to estimate parameters of DDS? Example: mechanical oscillator

3. Problem: Noise.

www.mechlab.czISMMB, Faculty of Mechanical Engineering, Brno University of Technology

7

Page 8: System Identification: Black Box

www.mechlab.czISMMB, Faculty of Mechanical Engineering, Brno University of Technology

8

OLS & RLS example: Estimation of mass+spring+damper system 1/3

1) Discretization

Page 9: System Identification: Black Box

www.mechlab.czISMMB, Faculty of Mechanical Engineering, Brno University of Technology

9

OLS & RLS example: Estimation of mass+spring+damper system 2/3

2) Create matrix X and vector Y

Y(k) = y(k);

X(k,:) = [y (k-1), y(k-2), u(k-2)];

{E003}

Page 10: System Identification: Black Box

www.mechlab.czISMMB, Faculty of Mechanical Engineering, Brno University of Technology

10

OLS & RLS example: Estimation of mass+spring+damper system 3/3

3) Solve LS/RLS problem– both LS and RLS provide the same results

– convergence of RLS/precision of LS depends on excitation signal

{E003}0 2 4 6 8 10

-10

0

10

t [s]

Oscilator position

0 2 4 6 8 10-50

0

50

t [s]

inpu

t u

Input excitation force [N]

0 2 4 6 8 10-2

0

2convergence of Recursive LS (RLS)

t [s]

continuous sim

ARX sim

a1

a2

b2

But: there is no noise

considered in our

simulation!

Page 11: System Identification: Black Box

Black box - agenda

1. Prerequisities: Least squares.

Discrete dynamic system.

2. How to use LS to estimate parameters of DDS? Example: mechanical oscillator

3. Problem: Noise.

www.mechlab.czISMMB, Faculty of Mechanical Engineering, Brno University of Technology

11

Page 12: System Identification: Black Box

Problem = noise: simple example study – setup 1/4

consider first order system

discretization:

for simplicity, consider NO INPUT

TASK: estimate one parameter a.

simulation data :

www.mechlab.czISMMB, Faculty of Mechanical Engineering, Brno University of Technology

12

Ts

1k kx ax

0 2 4 6 8 100

0.5

1

1.5

2

tx

-1 0 1 2 3

0.5

1

1.5

x(k)

x(k

+1

)

Page 13: System Identification: Black Box

Problem = noise: simple example study – without noise 2/4

Consider ini value x_0 and simulate equation:

On lower Fig. we see the plot x(k) vs. x(k+1).From the equation, it is obvious, that it should be a line.

The slope of this line is a.

With data WITHOUT noise, the estimation of a is precise.

www.mechlab.czISMMB, Faculty of Mechanical Engineering, Brno University of Technology

13

1k kx ax

0 2 4 6 8 100

0.5

1

1.5

2

t

x

-1 0 1 2 3

0.5

1

1.5

x(k)

x(k

+1

)

Page 14: System Identification: Black Box

Problem = noise: simple example study – noise on output 3/4

Consider further the noise on output (e.g. noise of sensor). Equation: (w is random number with normal distribution (Gauss)

We can observe a problem with estimation of parameter a.

www.mechlab.czISMMB, Faculty of Mechanical Engineering, Brno University of Technology

14

1k k

k k

x ax

y x w

0 2 4 6 8 10-2

-1

0

1

2

3

4

5

6

t

x

-1 0 1 2 3 4 5

-1

0

1

2

3

4

5

x(k)

x(k

+1

)

Example:

SI 004

% simulation

x(1) = x0;

for i = 2:length(t)

x(i,1) = a*x(i-1);

y(i,1) = x(i) + randn*0.2;

end

0 2 4 6 8 10-2

-1

0

1

2

3

4

5

6

t

x

-1 0 1 2 3 4 5

-1

0

1

2

3

4

5

x(k)

x(k

+1

)

Page 15: System Identification: Black Box

Problem = noise: simple example study

Next, consider noise „inside“ the systemwhich modifies the state of x(k).Equation:

This noise can be considered as a randomdisturbance (input) influencing the system.

Using the simulation, we can observe, that the estimation ofparameter a is perfect.

www.mechlab.czISMMB, Faculty of Mechanical Engineering, Brno University of Technology

15

1k k

k k

x ax w

y x

Example:

SI 004

% simulation

for i = 2:length(t)

x(i,1) = a*x(i-1) + randn*0.2;

y(i,1) = x(i) ;

end

0 2 4 6 8 10-8

-6

-4

-2

0

2

4

6

8

10

12

t

x

-5 0 5 10

-6

-4

-2

0

2

4

6

8

10

x(k)

x(k

+1

)

0 2 4 6 8 10-8

-6

-4

-2

0

2

4

6

8

10

12

t

x

-5 0 5 10

-6

-4

-2

0

2

4

6

8

10

x(k)

x(k

+1

)

Page 16: System Identification: Black Box

www.mechlab.czISMMB, Faculty of Mechanical Engineering, Brno University of Technology

16

Linear Models for System Identification

q = forward shift operatore.g.: q-1 x(k) = x(k-1)

linear models = relation between y(),u(), and noise v(k) through transfer functions

general form:

Page 17: System Identification: Black Box

www.mechlab.czISMMB, Faculty of Mechanical Engineering, Brno University of Technology

17

Linear Models for System Identification

AR – autoregressive model

MA – moving average

example: smoothing

Page 18: System Identification: Black Box

www.mechlab.czISMMB, Faculty of Mechanical Engineering, Brno University of Technology

18

Linear Models for System Identification

ARMA

ARX

ARMAX

OE

BJExample: system completely deterministic,

noise added only to output (e.g. position

sensor = potentiometer)

Very popular – linear in parameters

OLS

Page 19: System Identification: Black Box

Black-box identification- the use of a tool System Identification/GUI

Robert GreplInstitute of Solid Mechanics, Mechatronics and Biomechanics

Faculty of Mechanical Engineering, Brno University of Technology

2010 - 2015

Page 20: System Identification: Black Box

System Identification Toolbox - GUI

System Identification Toolbox very (very) complex tool, author L. Ljung

you can work using

– command line

– GUI

how to run it?: ident

www.mechlab.czISMMB, Faculty of Mechanical Engineering, Brno University of Technology

20

Page 21: System Identification: Black Box

System Identification Toolbox – GUI: Import data

www.mechlab.czISMMB, Faculty of Mechanical Engineering, Brno University of Technology

21

Page 22: System Identification: Black Box

System Identification Toolbox – GUI: Preprocessing

www.mechlab.czISMMB, Faculty of Mechanical Engineering, Brno University of Technology

22

preprocessing of the data check data

select estimationand verification data

Check

imported data

divide

Estimation and

Verification

data

Page 23: System Identification: Black Box

System Identification Toolbox – GUI: Select data

Select data for Estimation and for Verification

www.mechlab.czISMMB, Faculty of Mechanical Engineering, Brno University of Technology

23

Data for

estimation

Data for

verification

Page 24: System Identification: Black Box

System Identification Toolbox – GUI:

Estimate order of the system

You can estimateseveral ARX models atonce.

Think about noise...

www.mechlab.czISMMB, Faculty of Mechanical Engineering, Brno University of Technology

24

Example:

ARX

Example:

OE

Page 25: System Identification: Black Box

Exercise

S005

www.mechlab.czISMMB, Faculty of Mechanical Engineering, Brno University of Technology

25

Page 26: System Identification: Black Box

Dodatky

• Offline vs. online odhad parametrů

• IV (Instrumental variable) method

• Nelineární nejmenší čtverce

www.mechlab.czISMMB, Faculty of Mechanical Engineering, Brno University of Technology

34

Page 27: System Identification: Black Box

www.mechlab.czISMMB, Faculty of Mechanical Engineering, Brno University of Technology

35

Offline vs. online odhad parametrů

batch estimation– measurement in time (0,T)

– data processing

– parameter estimation using all data

online estimation– parameter estimation during system operation

– recursive algorithm

– advantages:

» memory requirements (often) reduced – OLS vs. RLS

» the slow changes of parameters can be captured

Page 28: System Identification: Black Box

www.mechlab.czISMMB, Faculty of Mechanical Engineering, Brno University of Technology

36

Recursive Least Squares (RLS) (Plackett's algorithm)

online estimation

gives identical result as OLS

assume model

in matrix form:

Page 29: System Identification: Black Box

www.mechlab.czISMMB, Faculty of Mechanical Engineering, Brno University of Technology

37

OLS & RLS example: Estimation of mass+spring+damper system 3/3

Solve LS/RLS problem– both LS and RLS provide the same results

– convergence of RLS/precision of LS depends on excitation signal

{E003}0 2 4 6 8 10

-10

0

10

t [s]

Oscilator position

0 2 4 6 8 10-50

0

50

t [s]

inpu

t u

Input excitation force [N]

0 2 4 6 8 10-2

0

2convergence of Recursive LS (RLS)

t [s]

continuous sim

ARX sim

a1

a2

b2

No noise considered in our

simulation!

Page 30: System Identification: Black Box

www.mechlab.czISMMB, Faculty of Mechanical Engineering, Brno University of Technology

38

OLS & RLS example: Estimation of mass+spring+damper system with noise

Assume noise added to measurement (e.g. if potentiometer used)

noise_stdvar = 0.1;

(later, we name this noise model “Output Error” = OE model)

{E003} 5.5 6 6.5 7 7.5 8

-6-4-202

t [s]

Oscilator position

0 2 4 6 8 10-50

0

50

t [s]

inpu

t u

Input excitation force [N]

0 2 4 6 8 10

-1

0

1

2

convergence of Recursive LS (RLS)

t [s]

continuous sim

ARX sim

a1

a2

b2

Page 31: System Identification: Black Box

www.mechlab.czISMMB, Faculty of Mechanical Engineering, Brno University of Technology

39

Estimation of OE: Instrumental Variable (IV) Method

solution for consistency problem

1. Estimate ARX

2. Simulate model with estimated parameters

3. Create instrumental matrix Z

4. Estimate

repeat steps 2-4 (usually upto 3-4 steps) [E009]

Page 32: System Identification: Black Box

www.mechlab.czISMMB, Faculty of Mechanical Engineering, Brno University of Technology

40

Nonlinear Least Squares 1/2

motivation: need to find parameters for functions which are nonlinear in parameters

assume vector field

goal: minimize criterion

Taylor series

– gradient:

Page 33: System Identification: Black Box

www.mechlab.czISMMB, Faculty of Mechanical Engineering, Brno University of Technology

41

Nonlinear Least Squares 2/2

all NLS methods are iterative

methods for computation of h:– Steepest descent

– Newton’s method

– Gauss-Newton

– Levenberg-Marquardt

Page 34: System Identification: Black Box

www.mechlab.czISMMB, Faculty of Mechanical Engineering, Brno University of Technology

42

Example NLS: Solution of overdetermined system

Assume vector field

clearly, the system is overdetermined

x = 1 / 0.4115

Task: Find solution which has minimal mean square error.

[E002]0 2 4 6 8 10 12 14 16 18 20

0

0.2

0.4

0.6

0.8

1

convergence of iteration method

iterations

valu

es o

f x

0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.30

0.2

0.4

0.6

0.8

variable x

Plot of square error

LM lambda = 2

GN alpha = 0.5

solution of individual equations

f12

f22

f12 + f

22

Page 35: System Identification: Black Box

www.mechlab.czISMMB, Faculty of Mechanical Engineering, Brno University of Technology

43

References

[Iserman 2011] Isermann, R., Muenchhof, M.: Identification of Dynamics Systems, Springer 2011

[Nelles, 2001] O. Nelles: Nonlinear system identification: from classical approaches to neural networks and fuzzy models, Springer-Verlag, 2001google books

[Noskievič, 1999] P. Noskievič: Modelování a identifikace systémů, Montanex, 1999

[Ljung 2008] L. Ljung: Perspectives on System Identification, Seoul 2008, http://www.control.isy.liu.se/~ljung/

[Denery 2005] T. Denery: Modeling Mechanical, Electric and Hydraulic Systems in Simulink, Matlab webinars, 2005

[Blaha] doc. Ing. Petr Blaha, PhD., FEKT, CAKhttp://sites.google.com/site/modelovaniaidentifikace/

[Gander, 1980] Gander, W.: Algorithms for QR-Decomposition, research report, 1980

http://www.emeraldinsight.com/journals.htm?articleid=1943785&show=html

http://courses.ae.utexas.edu/ase463q/design_pages/fall02/AWE_web/secondorder.htm

http://www.ni.com/white-paper/4028/en/

Page 36: System Identification: Black Box

Kontakt: www.mechlab.czÚstav mechaniky těles, mechatroniky a biomechaniky

Fakulta strojního inženýrstvíVysoké učení technické v Brnědoc. Ing. Robert Grepl, Ph.D.

[email protected]