Master of Science Course in Mechanical Engineering

Politecnico di Milano - Lecco Campus

Control strategies:
Æ feed-back
Æ feed-forward
Æ feed-forward + feed-back

Control and Actuating Devices for Mechanical Systems

Prof. Roberto Corradi

Single-dof vibrating system:

control strategies

fc z
Hypothesis:
f = fd+ fr stable, underdamped single-dof vibrating
system
m
k Objective:
c feeding in this mechanical system with the
control force fc, so as to make the mass m
follow the reference signal zREF in the best
mz + cz + kz = f + f c possible way

f = fd + fr the external forces are the sum of two contributions:

known deterministic forces and unknown random forces

fc control force coming from an actuator, driven according to one of the

following strategies:
Æ feed-back
Æ feed-forward
2
Æ feed-forward + feed-back

1
Feed-back

zr e fcFB + f z
+
controller (PD) mech. system
- +

mz + cz + kz = f + f cFB

f c FB (t ) = k p ( zr (t ) − z (t ) ) + kd ( zr (t ) − z (t ) )

mz + cz + kz = f + f c FB = f (t ) + k p zr (t ) + kd zr (t ) − ( k p z + kd z )

assigned function of time this contribution depends

(this contribution is zero for zr(t)=0) on the system’s state

mz + ( c + kd ) z + ( k + k p ) z = f (t ) + k p zr (t ) + kd zr (t )
Feed-back control:
Æ modifies the system’s parameters (stiffness and damping, for PD)
Æ influences the stability of the controlled system 3

Feed-forward

zr fcFF + f z
inverse dynamics mech. system
+

mz + cz + kz = f + f cFF = f d + f r + f cFF

m, c, k parameters of the real system

mˆ , cˆ, kˆ parameters of the mathematical model

ˆ  + cz
mz ˆ = f (t ) + f (t )
ˆ  + kz d cFF

Once that the mathematical model and the desired response zr(t) are known, the
feed-forward control force is evaluated by solving an inverse dynamic problem:

f c FF (t ) = mz
ˆ r (t ) + cz ˆ (t ) − f (t )
ˆ r (t ) + kz r d

2
 Feed-forward

mz + cz + kz = f d + f r + f cFF

f c FF (t ) = mz
ˆ r (t ) + cz ˆ (t ) − f (t )
ˆ r (t ) + kz r d

⇒ mz + cz + kz = f d + f r + mz
ˆ r + cz ˆ − f
ˆ r + kz r d

z = zr + z zr and z are respectively the reference signal (desired motion)

and the deviation of the actual motion z from the reference itself

zr + (cˆ − c) zr + (kˆ − k ) zr + f r

mz + cz + kz = (mˆ − m) 
If the mathematical model perfectly reproduces the real system’s dynamic
ˆ = m, cˆ = c, kˆ = k ) and no random disturbances act on the system,
behaviour ( m
then the actual system response z(t) is exactly coincident with the desired
one zr(t), i.e. z (t ) = 0
Feed-forward control does not modify the the system’s parameters and does not
influence the stability of the controlled system
5

Feed-forward + feed-back

fcFF + f
zr inverse dynamics
+ + z
mech. system
+ fcFB +
controller (PD)
-

The feed-forward contribution is the main part of the control force

The feed-back contribution acts on the residual error due to random disturbances
or imperfect mathematical model

mz + cz + kz = f d + f r + f cFF + f cFB

f c FF (t ) = mz
ˆ r (t ) + cz ˆ (t ) − f (t )
ˆ r (t ) + kz r d

f c FB (t ) = k p ( zr (t ) − z (t ) ) + kd ( zr (t ) − z (t ) )

⇒ mz + cz + kz = f r + mz
ˆ r + cz ˆ + k ( z − z ) + k ( z − z )
ˆ r + kz r p r d r
6

3
Feed-forward + feed-back

mz + cz + kz = f r + mz
ˆ r + cz ˆ + k ( z − z ) + k ( z − z )
ˆ r + kz r p r d r

z = zr + z

zr + (cˆ − c) zr + (kˆ − k ) zr − k p z − kd z

mz + cz + kz = f r + (mˆ − m) 

zr + (cˆ − c) zr + (kˆ − k ) zr + f r

mz + (c + kd ) z + (k + k p ) z = (mˆ − m) 

If the mathematical model perfectly reproduces the real system’s dynamic

behaviour and no random disturbances act on the system (ideal case), then feed-
back control does not intervene, since z (t ) = 0

Feed-back contribution modifies the system’s parameters (influences stability)

The feed-forward + feed-back control strategy generally allows to obtain the best
system’s performance
7

Example of application: motion control

4
Example of application: motion control

Nc Jθ + 2cR 2θ + 2kR 2θ = f + N c

θ
J, R
1.5

θr
1

0.5

0
0 0.5 1 1.5 2 2.5 3
k c 3

θr 2

0
0 0.5 1 1.5 2 2.5 3
10

θr 0

-10
0 0.5 1 1.5 2 2.5 3
tempo [s]

Example of application: motion control

Nc FEED-FORWARD
θ
J, R
Jθ + 2cR 2θ + 2kR 2θ = f + N cFF

N cFF = Jˆθr + 2cR ˆ 2θ

ˆ 2θr + 2kR r

k c
( Js 2
+ 2cR 2 s + 2kR 2 ) Θ ( s ) = N cFF ( s )

N cFF ( s ) = Js (
ˆ 2 + 2cR ˆ 2 Θ (s)
ˆ 2 s + 2kR r )
NcFF f
Θr + 1 Θ
ˆ 2 + 2cR
Js ˆ 2
ˆ 2 s + 2kR
+ Js 2 + 2cR 2 s + 2kR 2

Θ(s)
=1 for perfect mathematical model ⇒ Θ ( s ) = Θr ( s )
Θr ( s ) 10

5
 Example of application: motion control

Nc FEED-BACK
θ
J, R
Jθ + 2cR 2θ + 2kR 2θ = f + N cFB
N cFB = k p (θ r − θ ) + kd θr − θ ( )
k c
( Js 2
+ 2cR 2 s + 2kR 2 ) Θ ( s ) = N cFB ( s )

N cFB ( s ) = k p (1 + Td s ) ( Θr ( s ) − Θ ( s ) )

Θr NcFB f Θ
+ +
k p (1 + Td s )
1
Js + 2cR 2 s + 2kR 2
2

- +

Θ(s) k p (1 + Td s )
L ( s) = =
Θr ( s ) Js + ( 2rR + k pTd ) s + ( 2kR 2 + k p )
2 2 11

Example of application: motion control

θ Nc FEED-FORWARD + FEED-BACK
J, R Jθ + 2cR 2θ + 2kR 2θ = f + N cFF + N cFB
N cFB = k p (θ r − θ ) + kd θr − θ ( )
N = Jˆθ + 2cR ˆ 2θ + 2kRˆ 2θ
cFF r r r
k c
( Js 2
+ 2cR 2 s + 2kR 2 ) Θ ( s ) = N cFF ( s ) + N cFB ( s )
N cFB ( s ) = k p (1 + Td s ) ( Θr ( s ) − Θ ( s ) )
N ( s ) = Js
cFF (
ˆ 2 + 2cR
ˆ 2 s + 2kR ˆ 2 Θ (s) ) r

NcFF
+ f
Θr ˆ 2 + 2cR
Js ˆ 2
ˆ 2 s + 2kR
+ + 1
Θ
+ NcFB + Js + 2cR 2 s + 2kR 2
2

k p (1 + Td s )
-
12

6
 Example of application: motion control

FEED-FORWARD + FEED-BACK
NcFF
Θr Gˆ ( s ) −1 + Θ
G( s)
+ NcFB
k p (1 + Td s ) +
-

Θ ( s ) = ( N cFF ( s ) + N cFB ( s ) ) G ( s )

N cFF ( s ) = Gˆ ( s ) Θr ( s )
−1
NcFB ( s) = kp (1+ Td s) ( Θr ( s) −Θ( s) )

Θ ( s ) = Θr ( s ) Gˆ ( s ) G ( s ) + k p (1 + Td s ) ( Θr ( s ) − Θ ( s ) ) G ( s )
−1

Θ ( s ) Gˆ ( s ) G ( s ) + k p (1 + Td s ) G ( s )
−1
Θ(s)
= =1 for perfect
Θr ( s ) 1 + k p (1 + Td s ) G ( s ) Θr ( s ) mathematical model
13

Example of application: motion control

1.5

θr 1

0.5

0
0 0.5 1 1.5 2 2.5 3
3

θr 2

0
0 0.5 1 1.5 2 2.5 3
10

θr 0

-10
0 0.5 1 1.5 2 2.5 3
tempo [s]

14

7
Example of application: motion control

FEED-FORWARD

Jˆ = J , cˆ = c, kˆ = k

1.4
Response
Reference
1.2

0.8
[rad]

0.6

0.4

0.2

0
0 0.5 1 1.5 2 2.5 3
[s]

15

Example of application: motion control

FEED-BACK (PD)

Response
Reference
1.2

0.8
[rad]

0.6

0.4

0.2

0
0 0.5 1 1.5 2 2.5 3
[s]

16

8
Example of application: motion control

FEED-FORWARD + FEED-BACK (PD)

Jˆ = J , cˆ = c, kˆ = k

1.4
Response
Reference
1.2

0.8
[rad]

0.6

0.4

0.2

0
0 0.5 1 1.5 2 2.5 3
[s]

17

Example of application: motion control

FEED-FORWARD
k
Jˆ = J , cˆ = c, kˆ = ⇒ underestimated stiffness
2

Response
Reference
1.2

0.8
[rad]

0.6

0.4

0.2

0
0 0.5 1 1.5 2 2.5 3
[s]

18

9
 Example of application: motion control

FEED-FORWARD + FEED-BACK (PD)

k
Jˆ = J , cˆ = c, kˆ = ⇒ underestimated stiffness
2

Response
Reference
1.2

0.8
[rad]

0.6

0.4

0.2

0
0 0.5 1 1.5 2 2.5 3
[s]

19

Example of application: motion control

k
Jˆ = J , cˆ = c, kˆ = ⇒ underestimated stiffness
2
Response FB (PID)
Reference
1.2
Response FF+FB (PID)

1 PID controller

0.8
[rad]

0.6

1 Response FB (PID)
0.4 Reference
0.9
Response FF+FB (PID)
0.8
0.2
0.7

0.6
0
[rad]

0 0.5 1 1.5 2 0.5 2.5 3

[s]
0.4

0.3

0.2

0.1

0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
[s] 20

10