Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

2.

010 Fall 2000 Solution of Homework Assignment 6

1. Hydraulic Servomechanism. Creating a model of the hardware, or plant, for


this system is a fairly challenging task. First we identify the major parts: the rotating
table (its moment of inertia It will be an important dynamic parameter), the gear
transmission (its ratio N will undoubtedly be required), the hydraulic motor (the
relations between the input flow and pressure and the output torque and speed will
be required), and finally the servovalve (which permits the flow-pressure relation to
be controlled by an input current). The problem statement suggests that we start
with lossless transduction elements and then add one or more separate dissipative
elements to account for the unavoidable losses in real equipment.

Study the motion. Let the angular position of the hydraulic motor output shaft be
denoted by the angle θ. The corresponding angular velocity is ω = θ̇. The angular
velocity of the table has magnitude ωt = ω/N (in the opposite sense to ω).

Constitutive equations. Consider, first, an ideal positive-displacement hydraulic mo-


tor. We are told that the volume V of fluid displaced is proportional to the angular
displacement of the motor shaft V = Dθ. The time derivative of this relation
dV dθ
Q= = D = Dω (1)
dt dt
states that the input volume rate of flow Q is proportional to the output angular
velocity ω. Now in an ideal transducer, the power in is equal to the power out. For
the hydraulic motor the power in is the product of the input pressure P and the input
volume rate of flow Q while the power out is the product of the output torque τm and
the output angular velocity ω, so
P Q = τm ω (2)
when (1) is inserted in (2) we learn that τm = DP . These relations are summarized
in the following iconic representation of the motor.

P τm = DP τm

Q Q = Dω ω

Figure 1: Positive Displacement Hydraulic Motor.

A similar icon can be drawn for the reduction gear. At the pinion input the torque is
τm and the angular velocity is ω, while at the table output, the torque is τt and the

1
angular velocity is ωt . The gear ratio is N . The icon summarizing this information
is shown below.

τm τt = Nτm τt

ω ω = Nωt ωt

Figure 2: Reduction Gear.

The servovalve supplying the flow Q and the pressure P to the hydraulic motor is
described by the given valve characteristic
Q i P
= −
Qmax imax Pmax
In the present application the current i is the control variable and the flow Q is fixed
by the angular velocity of the motor shaft, so that the valve characteristic can be
considered to be the constitutive equation for determining the pressure P . When it
is is solved for P , we have
Pmax Pmax
P = i− Q (3)
imax Qmax
which can be represented diagramatically as follows

i
Pmax
imax
+
Σ P
- Pmax
Q
Qmax

Figure 3: Hydraulic Servovalve.

The rotating table rotates with the angular velocity ωt and acts as an energy storage
element. None of the elements discussed so far involve any energy dissipation. To
illustrate how some dissipation can be included in our model, let us consider some
friction acting to retard the table rotation, modeled by a linear damper with damping
coefficient Bt . Then the torque τt delivered by the large gear would have to overcome
the friction torque Bt ωt as well as increase the angular momentum It ωt of the table
τt = Bt ωt + It ω̇t
This time-domain differential equation can be represented by the transfer function
ωt (s) 1
=
τt (s) Bt + sIt

2
1
τt ωt
Bt + sIt

Figure 4: Transfer Function from Torque to Angular Velocity.

and displayed in iconic form as shown in Fig. 4.

(a) To develop a model to describe the the dependence of the table angular velocity
ωt upon the input current i, we assemble the above elements into the following
circuit diagram

i Pmax
imax
+
Σ P τm τt 1
Bt + sIt
ωt

- D N
Pmax
Qmax Q ω ωt

Figure 5: Diagram of System Relating Input Current to Output Angular Velocity.

The circuit diagram is a shorthand device for representing a large number of


relations between many variables. The desired transfer function is obtained by
eliminating all the intermediate variables to get a single relation between i(s)
and ωt (s). The loop structure of the circuit requires an organized attack. One
approach is to write the servovalve equation in the form (3) and then introduce
auxiliary relations connecting P (s) and Q(s) to the desired response ωt (s). From
Fig. 5 we see that
DN
ωt (s) = P (s) (Aux.1)
Bt + sIt
and
Q(s) = DN ωt (s) (Aux.2)
Now , if P (s) from (3) is inserted for the P (s) in (Aux. 1) and Q(s) from (Aux.
2) is inserted for the Q(s) in (3) the result
à !
DN Pmax Pmax
ωt (s) = i(s) − DN ωt (s) (4)
Bt + sIt imax Qmax

is a single equation relating ωt (s) to i(s).

In a derivation this complicated it is easy to make minor mistakes. One simple,


but very valuable check, is to see if the result is dimensionally consistent. In
equation (4) the first term in the parenthesis has the dimension of pressure.
For consistency the second term must also have the dimension of pressure. It

3
is worthwhile to verify that this is indeed the case. The parameter D has the
dimension of volume and the parameter N is dimensionless , so the term DN ωt
has the dimension of volume per unit time; i.e., the dimension of volume rate of
flow, which cancels the dimension of Qmax , so that the dimension of the second
term is that of Pmax which agrees with the dimension of the first term.

Continuing the dimensional consistency check, consider the fraction in front of


the parenthesis on the right-hand side of the equation. The denominator has two
terms . For consistency, they both must have the same dimension. The friction
coefficient Bt has the dimension of torque per unit angular velocity or [F · L · T ]
while the second term sIt has the dimension of angular velocity times moment of
inertia [1/T · M · L2 ], which on substituting the dimensional relation underlying
Newton’s second law [M ] = [F · T 2 /L], reduces to [F · L · T ] in agreement with
the first term. The numerator DN of the fraction has the dimensions of volume
[L3 ], so that the dimension of the fraction is [L3 ]/(F · L · T )] or the dimension
of angular velocity [1/T ] per unit pressure [F/L2 ]. The dimension of the right-
hand side of Eq.(4) is thus simply [1/T ] in agreement with the dimension of the
angular velocity ωt on the left-hand side of the equation.
When rearranged, Eq.(4) provides the desired transfer function
Pmax DN
ωt (s) i I
= max t (5)
i(s) s + λi
where à !
Pmax
2 2 1
λi = Bt + D N
Qmax It
The break frequency of the first-order dynamic response of Eq.(5) is λi which
was observed to be 1.0 Hz or 6.28 rad/sec. The decay time constant τd is the
reciprocal of the break frequency so
It
τd = Ã ! = 0.1592 sec
2 2
Pmax
Bt + D N
Qmax

The motion of the table is controlled by an angular position sensor which pro-
duces a voltage esensor = M θt . The controller generates a current proportional
to the error voltage eerror which is the discrepancy between the reference voltage
eref and the sensor voltage esensor

i = Geerror = G(eref − esensor )

The circuit diagram for the system plus controller is shown in Fig. 6.
To answer Parts (b) and (c) we need the response of the error voltage

eerror (s) = eref (s) − esensor (s) (6)

4
eref
+
Σ eerror
G
i Pmax
imax
+
Σ P τm τt 1
Bt+Its ωt
- D N
-e sensor Pmax
Qmax Q ω ωt
1
s
esensor θt θt
M

Figure 6: Diagram of System and Controller.

to the input reference voltage eref (s). The sensor voltage can be eliminated
from (6) by introducing the open-loop transfer function

esensor (s)
OLT F (s) = (7)
eerror (s)

Substitution of (7) in (6) yields

eerror (s) 1
= ERT F (s) = (8)
eref (s) 1 + OLT F (s)

Here the open-loop transfer function is readily obtained from Fig. 6 by incor-
porating the transfer function (5) from the current i(s) to the angular velocity
ωt (s) with the auxiliary relations

i(s) esensor (s) 1


=G and = M
eerror (s) ωt (s) s
to get
Pmax DN
esensor (s) ωt (s) i(s) i I 1
OLT F (S) = = GM max t (9)
ωt (s) i(s) eerror (s) s + λi s
Insertion of (9) in (8) yields the desired error transfer function

eerror (s) s(s + λi )


= (10)
eref (s) Pmax DN
s(s + λi ) + M G
imax It

(b) The system of Fig.6 has one free integrator, and therefore is of Type One. As
a consequence, the steady-state error in response to a step input is zero. The
same result is obtained by applying the Final-Value theorem to the transfer
function (10).

(c) The Laplace transform of a unit ramp is 1/s2 . The Final-Value theorem applied
to the response of the error voltage to a unit ramp indicates that there is a

5
steady-state tracking error
" #
1 eerror (s) λi
lims→0 s( 2 ) =
s eref (s) Pmax DN
MG
imax It

(d) The system is now assumed to have additional dynamics. Where the servovalve
used to operate directly on i(s) in the s-plane, it now operates on
à !2
λv
i(s)
s + λv

and where the sensor voltage was simply proportional to the angle θt (s), it now
is proportional to the following operation on the angle
λs
esensor = M θt (s)
s + λs
The open-loop transfer function will now contain these additional operations

esensor (s) Pmax DN λ2v λs


OLT F (s) = = GM (11)
eerror (s) imax It s(s + λi )(s + λv )2 (s + λs )
The altered system is still a Type One system. It has one free integrator. The
steady-state error in step response is still zero.

(e) The gain margin is obtained from the Bode plot of the open-loop transfer func-
tion. It is easy to see that the Bode plots for (9) and (11) will be quite different
and that therefore introducing the servovalve and sensor dynamics will affect
the gain margin. The answer is “Yes”.

2. Rate Feedback vs. PD Control. The hardware, or plant, in this problem is


the ultimate in simplicity. A mass m moves along the x-axis under the influence of a
controlled force f . There is no friction and there are are no springs or dampers. The
emphasis is on the nature of the control.

(a) Proportional Control. The system is diagrammed in Fig.7 below.


The open-loop transfer function is
ex (s) GDM
OLT F (s) = = 2 (1)
eerror (s) sm
Let us check that this is dimensionally consistent. The OLTF is the transfer
function from a voltage to a voltage, and thus must be dimensionless. The
dimensions of the terms in the right-hand expression in (1) are

G : [0] D : [N/volt] M : [volt/m] s : [rad/sec] m : [kg]

6
er +
Σ eerror
G
ec
D
f 1
ms
v 1
s
x

-
ex
M
ex

Figure 7: Proportional Control.

When these are combined as in (1) the dimension of the result is


GDM
: [N-sec2 /(m-kg)] = [0]
s2 m
The closed-loop transfer function can be obtained from the open-loop transfer
function by writing
ex (s) = 0LT F (s) · eerror = 0LT F (s) · (er − ex )
and solving for
GDM
ex (s) OLT F (s) s2 m GDM/m
= = = 2
er (s) 1 + OLT F (s) GDM s + GDM/m
1+ 2
sm
Note that the poles of this transfer function lie on the imaginary axis of the
s-plane. This implies that the natural motions of the system are undamped sinu-
soids. The response to a step input will excite tne natural motions which cause
undamped oscillations about the steady-state position. Changing the gain G
changes the frequency, but not the amplitude of the oscillations. The magnitude
of the overshoot is uncontrollable with this simple proportional control.
(b) Proportional Control with Rate Feedback. The system is diagrammed in Fig.8
below.
The closed-loop response transfer function can be obtained by writing the con-
troller equation
ec = Gx (er − ex ) − Gv ev
plus the transfer functions from ec to ex and from ev to ex in the form
DM
ex =
ec and ev = sGD ex
s2 m
and then eliminating ec and ev to get the transfer function from er to ex
DM Gx
ex (s) s2 m DM Gx /m
= = 2
er (s) DM Gx DM Gv Gd s + sDM Gv Gd /m + DM Gx /m
1+ 2 +
sm sm
7
Σ Σ
er + eerror + ec f 1 v 1 x
Gx D ms s
- - M
ex
Gv Gd s ex
ev dex
dt
ex

Figure 8: Proportional Control with Rate Feedback.

The denominator has the standard second-order form with controllable behav-
ioral parameters
DM Gx m 1 kg
ωn2 = = (10 rad/sec)2 if Gx = ωn2 = (10 rad/sec)2 = 10
m DM 10 N/volt × 1 volt/m
and
2ζωn = DM Gv Gd /m = 2(0.707)(10 rad/sec) = 14.14 rad/sec
if
m 1 kg
Gv = 2ζωn = 14.14 rad/sec = 1.414
DM Gd 10 N/volt(1 volt/m)(1 volt per volt/sec)
With these values, the closed-loop transfer function is
ex (s) 100
= 2 (2)
er (s) s + 14.14s + 100

(c) Proportional plus Derivative Controller. The system is diagrammed in Fig.9


below.
Here the controller equation is

ec (s) = G(1 + sKd )[er (s) − ex (s)] (3)

A second equation relating ec and ex is


DM
ex (s) = ec (s) (4)
s2 m
When ec is eliminated by inserting (3) into (4), the result is the closed-loop
transfer function
ex (s) DM GKd /m(s + 1/Kd )
= 2
er (s) s + sDM GKd /m + DM G/m

8
Σ Σ
er + eerror + G ec f 1 v 1 x
Kds D ms s
- + M
ex
ex

ex

Figure 9: Proportional plus Derivative Control.

Note that this has a zero at s = −1/Kd as well as a denominator with the
standard second-order form with controllable behavioral parameters
DM G m 1 kg
ωn2 = = (10 rad/sec)2 if G = ωn2 = (10 rad/sec)2 = 10
m DM 10 N/volt × 1 volt/m

and
2ζωn = DM GKd /m = 2(0.707)(10 rad/sec) = 14.14 rad/sec
if
m 1 kg
Kd = 2ζωn = 14.14 rad/sec = 0.1414 sec
DM G 10 N/volt(1 volt/m)(10)

With these values, the closed-loop transfer function is

ex (s) 14.14(s + 7.07)


= 2 (5)
er (s) s + 14.14s + 100

Note that (2) and (5) have the same denominator, and therefore , the same
poles. The difference between proportional control with rate feedback (2), and
P-D control (5), is that the former has no zeros, while the latter has a zero.

9
Pole-zero map Bode Diagrams
10

5
0
Imag Axis

0
-5

-5

Phase (deg); Magnitude (dB)


-10

-10
-8 -6 -4 -2 0 -15
Real Axis

Step Response
0

1.2
-20
1
Amplitude

0.8 -40
0.6
-60
0.4

0.2 -80
0 0 1 2
10 10 10
0 0.16 0.32 0.48 0.64 0.8
Time (sec.) Frequency (rad/sec)

Figure 10: MATLAB Plots for ζ = 0.707.

(d) The following MATLAB session produces the plots displayed in Fig. 10.

sysd = tf([14.14 100], [1 14.14 100]);


subplot(2,2,1),pzmap(sysd)
subplot(2,2,3),step(sysd)
subplot(1,2,2),bode(sysd)

(e) The following MATLAB session produces the plots displayed in Fig. 11.

syse = tf([40 100], [1 40 100]);


subplot(2,2,1),pzmap(syse)
subplot(2,2,3),step(syse)
subplot(1,2,2),bode(syse)

10
Pole-zero map Bode Diagrams
1

0.5 0
Imag Axis

-5
0
-10

-0.5 -15

Phase (deg); Magnitude (dB)


-20
-1
-40 -30 -20 -10 0 -25
Real Axis

Step Response
0

1 -20
0.8
Amplitude

-40
0.6

0.4 -60

0.2 -80

0 -1 0 1 2 3
10 10 10 10 10
0 0.16 0.32 0.48 0.64 0.8
Time (sec.) Frequency (rad/sec)

Figure 11: MATLAB Plots for ζ = 2.0.

The system in (d) is underdamped (ζ = 0.707) and the step response has
considerable overshoot. The Bode plot has a mild resonant peak (≈ 3dB). The
system in (e) is overdamped (ζ = 2.0), but the slow-response pole is almost
canceled by the zero with the result that there is only a small overshoot in the
step response, and the Bode plot shows almost no resonant peak (< 1dB).

11

You might also like