GUIDE:VINOD.B.

PRESENTED BY:MUHAMED SHEREEF P M1,AEI

 Introduction MFAC

CONTENTS

Data driven Control Advantages of MFAC Approach

Dynamic Linearization technique

MFAC System design and stability analysis PPD estimation

CFDL-MFAC
Experimental Investigation Conclusion

References
2

INTRODUCTION
ADAPTIVE CONTROL: In adaptive control the

controller will modify its behavior in response to changes in the dynamics of the process and the character of the disturbances Most adaptive techniques and methodologies typically assume that the structure of the system is known linear and the parameter may be unknown or slow time-varying

 In the case of complex practical systems the structure

of the plant is often difficult to determine and the parameters are hard to identify This is the motivation for Data driven control approach

Data driven control

In this method designing of controller merely using

input and output measurement data of the plant So model of the plant is not required Data driven control methods :Virtual Reference Feed back Tuning(VRFT),Iterative Learning Control(ILC), Model Free Adaptive Control(MFAC)

MFAC
This is proposed for a class of general discrete time

SISO nonlinear systems Instead of identifying a more or less nonlinear model of a plant, an equivalent dynamical linearization model is built along the dynamic operation points of the closed-loop system using a new dynamic linearization technique with a novel concept called pseudo-partial derivative (PPD) The time-varying PPD could be estimated merely using the I/O measurement data of a controlled plant
6

 The dynamic linearization techniques include the

partial form dynamic linearization (PFDL), the compact form dynamic linearization (CFDL), and the full form dynamic linearization (FFDL).

Advantages of MFAC approach

MFAC just depends on the real-time measurement

data of the controlled plant, which implies that we can develop a general controller for a class of the industrial practical processes independently MFAC does not require any external testing signals and any training process , therefore it is a lower cost controller MFAC is simple and easily implemented with small computational burden and has strong robustness

 Under some practical assumptions, the monotonic

convergence and bounded-input bounded-output (BIBO) stability of the CFDL-based MFAC (CFDLMFAC) approach can be guaranteed, which are the highlight features compared with other data-driven control approaches MFAC prototype has been successfully implemented in many practical applications, e.g., chemical industry, linear motor control, injection modelling process, PH value control, and so on
9

DYNAMIC LINEARIZATION TECHNIQUE

The discrete-time SISO nonlinear system to be controlled is given as follows: + 1 = , . . . , , , . . . , (1) where ()and() are the system output and input at time , respectively, and are the unknown orders, and is an unknown nonlinear function. The PFDL of the nonlinear system (1) is based on following assumptions.
10

 A1. The partial derivatives of with respect to

control input , 1 , . . , ( ) are continuous : control input length constant of linearization A2. System (1) is generalized Lipschitz, that is , ( + 1) U(k) for any and U(k) 0 where + 1 = + 1 () U = [ , . . , + 1 ]
11

Theorem1
There exists a parameter vector () , called the PPD

vector, such that system (1) can be transformed into the following equivalent PFDL description: + 1 () U (2)

where = [1 , 2 , . . , ] and ()

12

 Consider the LTI system

1 = 1

(3)

Where 1 =1 + 1 1 + 2 2 + + & 1 =1 + 1 1 + 2 2 + + If ( 1 ) is a stable polynomial, could be expressed as = = 1 (4) 1 where 1 = 0 + 1 + + 1

13

 The model (4) is an approximation of the true system

(3) if the order of is sufficiently large Model (4) can be rewritten in the following form: = 1 5 Comparing (2) & (5) we can see that is time invariant when

14

15

MFAC System design and stability analysis

For the one-step-ahead controller, excessive control

effort is required to bring + 1 to + 1 in one step. + 1 : reference signal

So the following control input index function is used

to design the control law: = + 1 ( + 1) 2 + U() is a weighting constant

(7)

16

 Substituting (2) into (7), differentiating (7) w.r.t.

, and letting it be zero gives = 1

1 1 ()(( +1 ) + + 1 () 2

1 () ()(+1) =2 + 1 () 2

(8)

Where [1 , 2 , , ] is a step-size vector, (0,1]

17

Remarks
constrains the PPD from changing too quickly
is an important parameter for the MFAC system

design. Suitable choice of can guarantee the stability or improve the performance of the control system Since is time varying, we use the modified projection algorithm to estimate

18

PPD estimation

= 1

()(
2

1)

+ 1 > 0 :weighting factor :estimation value of

Using optimal condition

= 0, we have
2

= 1 +

(1)( 1 1 ) + U(1)

(10)

19

 : step-size constant, (0,2)

Combining control law (8) and parameter estimation

(10), the PFDL-MFAC scheme is designed as follows:

(1)( 1 1 ) + U(1)
2

= 1 +

(11) (12)

= 1 if

20

1 1 ()(( + 1 ) = 1 + + 1 () 2

1 () ()(+1) =2 + 1 ()
2

(13)

For a simple SISO nonlinear system, could be set as 1

21

CFDL-MFAC
1 ()=1 1 +
(1)( 1 (1) 1 ) + (1) 2

(14)

1 ()= 1 (1) if 1 ()
= 1 +
1 1 ()(( +1 ) + 1 () 2

15

(16)

22

EXPERIMENTAL INVESTIGATION

23

 Three data-driven control approaches, including

MFAC,VRFT, and IFT, have been implemented on a three-tank system (Fig. 1) to investigate the control performance of the proposed approach The parameters of PFDL-MFAC scheme (11)(13) are = 5, 1 = 2 = = 5 = 1, = = = 1

1 = [1,0,0,0,0] ,

= 105

24

 Root mean square

()2 )/ =1

integral time absolute error

= total sum of squares

=1

()

=
=1

()2

25

MFAC 0.01675 132 333

VRFT 0.01679 144 481

IFT 0.01702 179 1950

26

27

CONCLUSION
The MFAC approach is easy to implement. The control

performance of MFAC is better than that of the other two approaches The MFAC is the online adaptive control approach and does not need any measurement data collection procedure For IFT approach, the control performance is not as good as that of MFAC and VRFT

28

REFERENCES
M. C. Campi, A. Lecchini, and S. M. Savaresi, Virtual Reference Feedback

Tuning (VRFT): A newdirect approach to the design of feedback controllers, in Proc. 39th IEEE Conf. Decision Control, Sydney,Australia, Dec. 2000, pp. 623629. M. C. Campi and S. M. Savaresi, Direct nonlinear control design:The virtual reference feedback tuning (VRFT) approach, IEEE Trans. Autom. Control, vol. 51, no. 1, pp. 1427, 2006. H. Hjalmarsson, M. Gevers, and S. Gunnarsson, Iterative feedback tuning Theory and applications, IEEE Control Syst. Mag., vol. L8,no. 4, pp. 2641, 1998. Z. S. Hou and J. X. Xu, On data-driven control theory: The state of the art and perspective, ACTA Automatica Sinica, vol. 35, no. 6, pp. 650667, 2009.

29