CHAPTER 10 NONLINEAR CONTROL SYSTEMS
10.1 Describing functions of some entries of Table 10.2 have been derived in Section 10.4 and Section 1
10.2 Revisiting Example 10.1 (Fig. 10.19) will be helpful.
G(jw) = 4
1
1 jw( + jw)2 -N E
; ( ) = - p E M 4 – 90∞ – 2 tan–1w1 = – 180∞
This givesw1 = 1 rad/sec
- 1
N E
a f
1 = |G(jw1)| = 2This gives E1 = 8M/p
y(t) = – e(t) = – 8M p sin t
10.3 Revisiting Example 10.1 (Fig. 10.22) will be helpful.
G(jw) = 5
1 0 1 2
jw( + j . w) ; N(E) = 4
1 0 1 2 p E -
FH IK
E. – 90∞ – 2 tan–1 0.1w1 = – 180∞This gives w1 = 10 rad/sec.
For a stable limit cycle, 2D < E < •;D = 0.1
– 1
N E
a f
1 = |G(jw1)| = 0.25This gives E1 = 0.3
Amplitude of limit cycle is 0.3 and frequency is 10 rad/sec.
10.4 Revisiting Example 10.1 (Fig. 10.22) will be helpful.
For a limit cycle to exist, 2D < E < •;D = 0.2 –G(jw1) = – 180∞ for w1 = 10.95. |G(jw1)| = 0.206
|G(jw1)|π - 1
N E( ) for any value of E
The intersection of the two curves does not occur; therefore no limit cycle exists.
Intersection of the two curves occurs for
pD
2 < 0.206 or D < 0.131
10.5 G(jw) = 10
1+ 0 4 1 2 ( j . w) +( j w)
N (E) = 4 1
2
p E
H
E j H
-
FH IK
- EL NMM O
QPP
; H = 0.2The following figure shows G(jw)-plot and -( )
1
N E locus.
E H= Im
Re
G j( )w
E increasing -1
pH
N E( )
4
At the point of intersection of G(jw)-plot with -( )
1
N E locus, measure the angle of G(jw). This comes out to be –115.7∞.
–G(jw1) = – tan–1 0.1w1 – tan–1 2w1 = – 115.7∞ This gives w1 = 5.9 rad/sec
|G(jw1)| = 0.33
- 1
N E
a f
1 = p E41 = 0.33This gives E1 = 0.42
10.6 G( jw) = K jw(1+jw)2
N(E) = 2 2
1 1
1 1
1 2
p
L
p - - -FH IK
NMM O
-
QPP
sin E E E
=1 2 1 1
1 1
1
-
L
+ - 2NM O
-
QP
p sin
E E E = 1 – Nc(E)
The function Nc(E) is listed in Table 10.3.
For any K > 0, one of the following can happen.
(i) G(jw)-plot does not intersect -( )
1 N E locus (ii) G(jw)-plot intersects the
-( ) 1
N E locus; but the point of intersection corresponds to an unstable limit cycle.
10.7 Revisiting Review Example 10.3 (Fig. 10.40) will be helpful.
G(jw) = 10
1 1 0 5
jw
a
+ jwfa
+ j . wf
N(E) = 2 1 1
1 1
1
p
L
sin- + - 2NM O
QP
E E E = Nc(E) The function Nc(E) is listed in Table 10.3.
–G(jw1) = – 180° gives w1 = 2
|G(jw1)| = - 1
N E
a f
1 gives E1 = 4.25The limit cycle with amplitude 4.25 and frequency 2 rad/sec is a stable limit cycle.
10.8 Revisiting Example 10.2 (Fig. 10.24) will be helpful.
G(jw) = K
jw(1+ jw) +(1 j0 5. w) For K = 1, G(jw)-plot does not intersect
-( ) 1
N E locus. For K = 2, two intersection points are found. One corresponds to unstable limit cycle and the other corresponds to a stable limit cycle of amplitude 3.75 and frequency 1 rad/sec.
10.9 (a) Characteristic equation has two real, distinct roots in the left half of s-plane.
The origin in the (y, y) plane is a stable node.
(b) d y dt
d y dt
2 2
1 5 1
-( ) + ( - )
+ 6 (y – 1) = 0
The singular point is located at (1, 0) in the (y, y) plane. The charac-teristic equation is
s2 + 5s + 6 = 0 = (s + 3) (s + 2) The singular point is a stable node.
(c) d y dt
d y dt
2 2
2 8 2
- -
-a f a f
+ 17(y – 2) = 0The singular point is located at (2, 0) in the (y, y) plane. The charac-teristic equation is
s2 – 8s + 17 = 0 with complex roots in right half s-plane. The singular point is an unstable focus.
10.10 e = f(qR – q) = sin (qR – q) = sin (– q)
q + aq + K sin q = 0
With x1 = q and x2 = q, we have
x1 = x2
x2 = – K sin x1 – ax2
Singular points are given by the solution of the equations x2 = 0
– K sin x1 – a x2 = 0
This gives x1 = kp; k in an integer. We therefore have multiple singular points.
Linearized equation around singular point (0, 0):
q + aq + Kq = 0 The characteristic equation is
s2 + as + K = 0
The singular point is stable and is either a node or a focus depending upon the magnitudes of a and K.
Linearization about the singular point (p, 0) gives
q + aq – Kq = 0 The characteristic equation is
s2 + as – K = 0
The singular point is a saddle point.
10.11 d y dt
d y dt
2 2
1 2 1
-( ) + z ( - ) + y – 1 = 0
(i) z = 0; singularity (1, 0) on (y, y) plane is a centre
(ii) z = 0.15; singularity (1, 0) on (y, y) plane is a stable focus.
10.12 Revisiting Example 10.4 (Fig. 10.34) will be helpful.
Jq + Tc sgn
c h
q = KA K1e; e = qR – q Thereforesgn
e K
J e T Jc e
+ + ( ) = 0; K = KAK1
sgn
e e T
K e
n
+w2
LNM
+ c ( )OQP
= 0; wn = K J/With x1 = e and x2 = e/wn, we have
x1 =wn x2
x2 = – wn x T
Kc nx
1+ 2
LNM
sgna f
wOQP
For Tc = 0, we have
x1 =wnx2; x2 = – wnx1 d x
d x
2 1
= – x x
1 2
; this gives x21 + x22 = x21(0)
The effect of the Coulomb friction is to shift the centre of the circles in phase plane to + Tc/K for x2 < 0 and to – Tc/K for x2 > 0. The steady-state error found from phase trajectory is – 0.2 rad. From the phase trajectory, we see that
(a) the system is obviously stable for all initial conditions and inputs; and (b) maximum steady-state error = ± 0.3 rad.
10.13 With x1 = y and x2 = y, we get
x1 = x2
x2 =
- - <
- - - >
- - + <
-R S|
T|
x x
x x x
x x x
2
2 1 1
2 1
1
2 1 1
2 1 1
;
;
;
Region I 1 <
Region II Region III
1
1
a f
a f a f
a f a f
From the phase portrait shown in the figure below, it is seen that the system is stable in the equilibrium zone.
10.14 Revisiting Review Example 10.6 (Fig. 10.42) will be helpful. The system equation in the linear range is
e + e + e = 0 The characteristic equation
s2 + s + 1 = 0
has complex roots in left half of s-plane. Therefore, on the (e, e) plane, the origin is a stable focus.
The system equation in saturation region is
e + e = sgn (e)
The trajectories are asymptotic to the ordinates – 1 and + 1 depending on whether the error is +ve or – ve.
From the phase trajectories plotted from the given initial conditions (with and without saturation) we find that velocity is always greater for trajecto-ries without saturation; thus the error saturation has a slowing down effect on the transient.
10.15 With x1 = e, x2 = e and u = f (e), we have
x1 = x2
x2 = – x2 – f (x1)
Without hysteresis, the system behaviour is oscillatory with decreasing (tending towards zero) period and amplitude of oscillations (refer Fig.
10.32). With hysteresis, the system enters into a limit cycle, as is shown in the figure below.
10.16 (a) e = – u = – f(e); x1 = e, x2 = e
The trajectories corresponding to x1 > 0 are given by (refer Eqns (10.24))
x1(t) = – 1 2 2
x2(t) + x1(0) + 1 2 2
x2(0) and trajectories for x1 < 0 are given by x1(t) = 1
2 2
x2(t) + x1(0) – 1 2 2
x2(0)
The system response is a limit cycle as shown in the figure.
(b) e = – u = – f (e + KDe); x1 = e, x2 = e
The trajectories corresponding to (x1 + KDx2) > 0 are given by x1(t) = – 1
2 2
x2(t) + x1(0) + 1 2 2
x2(0)
and the trajectories corresponding to (x1 + KDx2) < 0 are given by x1(t) = 1
2 2
x2(t) + x1(0) – 1 2 2
x2(0)
With derivative feedback, the limit cycle gets eliminated as shown in the figure.
(c) Constructing a trajectory for the case of large KD proves the point. An illustrative trajectory is shown in the figure.
Slope = - 1 KD
Trajectory
Trajectory Trajectory
Switching line
Switching line
x2
x2
x2
x1
x1
x1
10.17 q + 0.5q = 2 sgn(e + 0.5q) With x1 = e and x2 = e, we have
x1 = x2
x2 = – 0.5x2 – 2 sgn(x1 + 0.5 x2)
As seen from the phase trajectory in the figure, the system has good damp-ing, no oscillations but exhibits chattering behaviour. Steady-state error is zero.
+ e +
-
-qR= const
x1
x2
q q
1 1
s + 0.5 0.5
s 2
-2
1
2
x1+ 0.5x2= 0
10.18 (a) With x1 = e, x2 = e and u = f (e), we have
x1 = x2
x2 = – x2 – f(x1)
The system oscillates with everdecreasing amplitude and everin-creasing frequency (refer Fig. 10.32).
(b) A rough sketch of the phase trajectory is shown in the figure.
(i) Deadzone provides damping; oscillations get reduced.
(ii) Deadzone introduces steady-state error; maximum error = ± 0.2.
(c) x1 = x2
x2 = – x2 – f x1 1x2 +3
FH IK
A rough sketch of the phase trajectory is shown in the figure. By derivative control action (i) settling time is reduced, but (ii) chattering effect appears.
10.19 Section 10.10 provides solution to this problem. Optimum switching curve is given by Eqns (10.48). Figure 10.37 shows a few trajectories.
10.20 e + e = – u; x1 = e, x2 = e For u = +1 (refer Eqns (10.26)) x1 – x1(0) = – (x2 – x2(0)) + ln 1
1 0
2 2
+
F
+HG I
x
KJ
x ( )
The trajectories are asymptotic to the ordinate – 1.
For u = – 1
x1 – x1(0) = – (x2 – x2(0)) – ln 1
1 0
2 2
-F
+HG I
x
KJ
x ( )
The trajectories are asymptotic to the ordinate + 1.
For x1(0) = x2(0) = 0,
x1 = – x2 + ln (1 + x2); u = + 1 x1 = – x2 – ln (1 – x2); u = – 1 Switch curve is given by
x1 = – x2 + x x
2
| 2| ln 1 2
2
2
F
+HG I
x
KJ
x
| |
The figure given below shows the switching curve and a few typical mini-mum-time trajectories.