21.07.2013 Views

Fractional System Toolbox for Matlab - mechatronics

Fractional System Toolbox for Matlab - mechatronics

Fractional System Toolbox for Matlab - mechatronics

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

BORDEAUX 1 UNIVERSITY - ENSEIRB Engineering School UMR 5131 CNRS<br />

CONTROL engineering, CIM, SIGNAL&IMAGE processing<br />

41st IEEE Conference on Decision and Control<br />

Tutorial Workshop<br />

"<strong>Fractional</strong> Calculus Applications in Automatic Control and Robotics"<br />

December 9, 2002, Las Vegas, Nevada, USA<br />

<strong>Fractional</strong> systems toolbox <strong>for</strong> <strong>Matlab</strong>: <strong>Matlab</strong><br />

applications in <strong>System</strong> Identification<br />

Dr Olivier COIS, Dr Pierre MELCHIOR, Dr Patrick LANUSSE,<br />

Frédéric DANCLA and Pr Alain Oustaloup<br />

LAP - UMR 5131 CNRS<br />

Université Bordeaux 1 - ENSEIRB<br />

351 cours de la Libération - F33405 TALENCE Cedex - FRANCE<br />

Tél. : +33 (0)5 56 84 66 07 - Fax : +33 (0)5 56 84 66 44<br />

Email : melchior@lap.u-bordeaux.fr - URL : http://www.lap.u-bordeaux.fr


A new CAD toolbox: toolbox:<br />

"The CRONE toolbox: toolbox<br />

<strong>Fractional</strong> <strong>System</strong>s <strong>Toolbox</strong>"<br />

<strong>Toolbox</strong><br />

• Applications<br />

• Concept<br />

The original theoretical mathematical concepts developed in<br />

our laboratory are based on fractional or non integer<br />

differentiation.<br />

CRONE control (French abbreviation <strong>for</strong> Non Integer Order<br />

Robust Control) and <strong>System</strong> Identification by fractional model<br />

are typical applications.<br />

•Aims<br />

•Facilitate transfer of concepts to :<br />

- engineers, researchers in mathematics and engineering sciences,<br />

- universities,<br />

- industrials.<br />

•Make our research results widely available and develop international relations<br />

• Identification & Robust control (electromechanical, thermal, pneumatic,<br />

hydraulic systems, …)<br />

• Electronic (filter synthesis),<br />

• Path planning : Path description (cutting table) & Path tracking<br />

• Signal and image processing, ...


1. <strong>System</strong> Identification<br />

Content<br />

A new tool <strong>for</strong> mathematical models of physical systems<br />

Principles of parameter estimation methods<br />

<strong>Matlab</strong> <strong>Toolbox</strong><br />

Application example


A new tool <strong>for</strong> mathematical models<br />

of physical systems<br />

What is exactly a fractional derivative ?<br />

⎛<br />

⎜<br />

⎝<br />

n<br />

d ⎞<br />

⎟ f<br />

dt⎠<br />

() t<br />

d n<br />

⎛ ⎞<br />

⎜ ⎟<br />

⎝ dt ⎠<br />

f<br />

() t<br />

Mathematical viewpoint<br />

Time domain Operational domain<br />

• Riemann Liouville definition<br />

∆<br />

=<br />

Γ<br />

( m − n)<br />

• Complicated <strong>for</strong>mula<br />

1<br />

m<br />

⎛ d ⎞ ⎛ t<br />

⎜ ⎟ ⎜<br />

⎝ dt⎠<br />

⎜∫0<br />

−<br />

⎝<br />

Laplace trans<strong>for</strong>m<br />

( )<br />

( ) ( ) ⎟ f τ ⎞<br />

dτ<br />

−<br />

1−<br />

m n<br />

t τ<br />

• Global operator: takes into account<br />

all the values of function f(t)<br />

⎠<br />

Magnitude (dB)<br />

Phase (deg)<br />

-50<br />

s F()<br />

s<br />

n<br />

• Expression simpler than<br />

in the time domain<br />

50<br />

0<br />

200<br />

10 -1<br />

0<br />

-200<br />

10 -1<br />

n=-1.5<br />

n=-1<br />

n=-0.5<br />

n=0<br />

n=0.5<br />

n=1<br />

n=1.5<br />

Bode diagrams<br />

10 0<br />

Frequency (rad/s)<br />

10 0<br />

Frequency (rad/s)<br />

10 1<br />

10 1


A new tool <strong>for</strong> mathematical models<br />

of physical systems<br />

<strong>Fractional</strong> differential equation<br />

n<br />

⎛ d ⎞<br />

a0⎜<br />

⎟<br />

⎝ dt<br />

⎠<br />

a0<br />

y<br />

n<br />

⎛ d ⎞<br />

1⎜<br />

⎟<br />

⎝ dt<br />

⎠<br />

a1<br />

n<br />

⎛ d ⎞<br />

L⎜<br />

⎟<br />

⎝ dt<br />

⎠<br />

aL<br />

n<br />

⎛ d ⎞<br />

0⎜<br />

⎟<br />

⎝ dt<br />

⎠<br />

n<br />

⎛ d ⎞<br />

1⎜<br />

⎟<br />

⎝ dt<br />

⎠<br />

n<br />

⎛ d ⎞<br />

M ⎜ ⎟<br />

⎝ dt<br />

⎠<br />

() t + a y()<br />

t + ... + a y()<br />

t = b u()<br />

t + b u()<br />

t + ... + b u()<br />

t<br />

<strong>Fractional</strong> state space representation<br />

⎧<br />

⎛ ⎞<br />

⎪⎜<br />

⎟<br />

⎨⎝<br />

dt<br />

⎠<br />

⎪<br />

⎩y<br />

d n<br />

x<br />

() t = A x()<br />

t + B u()<br />

t<br />

() t = C x()<br />

t + D u()<br />

t<br />

Mathematical viewpoint<br />

b0<br />

b1<br />

•Modal decomposition<br />

•Output analytical expression<br />

•Stability properties<br />

bM


A new tool <strong>for</strong> mathematical models<br />

of physical systems<br />

Thermal system studies<br />

Semi-infinite medium<br />

• Diffusive equation<br />

→ x = 0<br />

⎛ d ⎞<br />

⎜ ⎟<br />

⎝ dt<br />

⎠<br />

0.<br />

5<br />

→ x > 0 : series expansion<br />

⎛<br />

⎜<br />

⎝<br />

d<br />

dt<br />

⎞<br />

⎟<br />

⎠<br />

0<br />

. 5<br />

T<br />

T<br />

∂T<br />

∂t<br />

1<br />

λ ρ C<br />

( 0,<br />

t)<br />

= φ()<br />

t<br />

1<br />

λ ρ C<br />

∑ ∞<br />

n=<br />

0<br />

( x,<br />

t)<br />

=<br />

c φ()<br />

t<br />

p<br />

( x,<br />

t)<br />

= c ( x,<br />

t)<br />

n<br />

∂<br />

2<br />

∂x<br />

T<br />

p<br />

2<br />

⎛ d ⎞<br />

⎜ ⎟<br />

⎝ dt<br />

⎠<br />

n<br />

2<br />

Physical viewpoint<br />

φ(t)<br />

0<br />

T(x,t)<br />

Semi-infinite medium<br />

x


A new tool <strong>for</strong> mathematical models<br />

of physical systems<br />

T<br />

Other geometries : fractional differential equation<br />

0.<br />

5<br />

1<br />

0.<br />

5M<br />

0.<br />

5 N<br />

() t + a D T () t + a D T () t + ... + a D T () t = b φ()<br />

t + b D φ()<br />

t + ... + b D φ()<br />

t<br />

1<br />

2<br />

M<br />

Physical viewpoint<br />

Studies of various geometries (cylindrical, spherical …) show that fractional<br />

differential equations, with differentiation orders being multiples of 0.5, permit better<br />

approximations of analytical solutions than classical differential equations<br />

The result can be extended to all the physical systems governed by<br />

- the diffusive partial differential equation (electrochemical systems,<br />

electromagnetic systems)<br />

- the Telegraphist partial differential equation (long transmission cables…)<br />

Preliminary conclusion: fractional differentiation is an efficient tool<br />

<strong>for</strong> modeling some real physical systems<br />

0<br />

1<br />

0.<br />

5<br />

N


1.5<br />

1<br />

0.5<br />

0<br />

y<br />

n<br />

⎛ d ⎞<br />

1⎜<br />

⎟<br />

⎝ dt<br />

⎠<br />

a1<br />

<strong>System</strong> Identification<br />

Goal: establish a mathematical model capable of reproducing the system's<br />

physical behavior as faithfully as possible from a series of observations<br />

u ( t)<br />

y(<br />

t)<br />

Real physical <strong>System</strong><br />

0 20 40 60 80<br />

Parameter estimation method<br />

n<br />

⎛ d ⎞<br />

L⎜<br />

⎟<br />

⎝ dt<br />

⎠<br />

aL<br />

n<br />

⎛ d ⎞<br />

0⎜<br />

⎟<br />

⎝ dt<br />

⎠<br />

n<br />

⎛ d ⎞<br />

1⎜<br />

⎟<br />

⎝ dt<br />

⎠<br />

n<br />

⎛ d ⎞<br />

M ⎜ ⎟<br />

⎝ dt<br />

⎠<br />

() t + a y()<br />

t + ... + a y()<br />

t = b u()<br />

t + b u()<br />

t + ... + b u()<br />

t<br />

b0<br />

b1<br />

3<br />

2<br />

1<br />

0<br />

-1<br />

0 20 40 60 80<br />

bM


<strong>System</strong> Identification:<br />

parameter estimation methods<br />

<strong>for</strong> fractional models<br />

y<br />

n<br />

⎛ d ⎞<br />

1⎜<br />

⎟<br />

⎝ dt<br />

⎠<br />

a1<br />

linearity with respect to coefficients<br />

non-linearity with respect to differentiation orders<br />

2 types of estimation methods<br />

Only coefficients are estimated<br />

orders are fixed by the user<br />

Equation error method<br />

n<br />

⎛ d ⎞<br />

L⎜<br />

⎟<br />

⎝ dt<br />

⎠<br />

linear optimisation techniques<br />

aL<br />

Least Square, Instrumental Variable...<br />

<strong>Fractional</strong> models<br />

n<br />

⎛ d ⎞<br />

0⎜<br />

⎟<br />

⎝ dt<br />

⎠<br />

n<br />

⎛ d ⎞<br />

M ⎜ ⎟<br />

⎝ dt<br />

⎠<br />

() t + a y()<br />

t + ... + a y()<br />

t = b u()<br />

t + ... + b u()<br />

t<br />

b0<br />

Both orders and coefficients<br />

are estimated<br />

Output error method<br />

non-linear optimisation techniques<br />

Gradient, Marquardt ...<br />

bM


<strong>Matlab</strong> <strong>Toolbox</strong>


Application example<br />

Application Field: Machining by turning<br />

The efficency of a machining depends<br />

on the heat flux through the tool<br />

It cannot be measured directly<br />

during machining<br />

Tools Temperature, Flux ?<br />

How to estimate the heat flux through the tool during machining


2 stages<br />

Strategy<br />

Heat surface<br />

Thermocouple<br />

T<br />

Insert tool<br />

Sensor<br />

Tool holder<br />

1. Be<strong>for</strong>e machining: Identify the thermal dynamic behavior between<br />

- heat flux through the tool<br />

- temperature close to the tip of the tool<br />

2. During machining: Estimate the heat flux through the tool using:<br />

- the identified thermal model<br />

- temperature measurments close to the tip of the tool<br />

q


T<br />

Identification of the thermal dynamic behavior of the tool<br />

Heat resistor<br />

Model structure choice<br />

thermocouple embeded in the tool<br />

Heat surface<br />

q<br />

Thermocouple<br />

T<br />

Thermal system <strong>Fractional</strong> model<br />

0.<br />

5<br />

1<br />

0.<br />

5M<br />

Insert tool<br />

Sensor<br />

Tool holder<br />

0.<br />

5 N<br />

() t + a D T () t + a D T () t + ... + a D T () t = b φ()<br />

t + b D φ()<br />

t + ... + b D φ()<br />

t<br />

1<br />

2<br />

M<br />

0<br />

1<br />

0.<br />

5<br />

N


T<br />

Power supply<br />

1.5<br />

1<br />

0.5<br />

Identification of the thermal dynamic behavior of the tool<br />

0<br />

0 20 40 60 80<br />

Thermocouple<br />

3<br />

2<br />

1<br />

0<br />

-1<br />

0 20 40 60 80<br />

Heat flux Temperature<br />

0.<br />

5<br />

1<br />

0.<br />

5M<br />

0.<br />

5 N<br />

() t + a D T () t + a D T () t + ... + a D T () t = b φ()<br />

t + b D φ()<br />

t + ... + b D φ()<br />

t<br />

1<br />

2<br />

Parameter estimation method<br />

M<br />

0<br />

1<br />

0.<br />

5<br />

N


Identification of the thermal dynamic behavior of the tool<br />

Data acquisistion : estimation and validation data (Ts=0.1s)


Identification of the thermal dynamic behavior of the tool<br />

Parameter estimation : 5 parameters<br />

0.<br />

5<br />

1<br />

() t + 10. 90 D T () t −1.<br />

04 D T () t + 39.<br />

81D<br />

T () t = 6.<br />

22φ()<br />

t −1.<br />

65D<br />

() t<br />

T φ<br />

1.<br />

5<br />

0.<br />

5


Identification du comportement thermique de l'outil<br />

Model validation


Conclusion (Syst ( Syst. . Ident.) Ident.)<br />

A further class of mathematical models, called fractional models, has been<br />

developed and analyzed (stability theorem, output analytical expression…)<br />

Various estimation methods of fractional models have been developed and<br />

applied to the identification of real systems<br />

A <strong>Matlab</strong> <strong>Toolbox</strong> has been developed<br />

<strong>Fractional</strong> differentiation is shown to be an efficient tool <strong>for</strong><br />

modeling certain real systems, using few parameters<br />

Application to the identification of electro-chemical systems (batteries)

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!