You are on page 1of 5

Solutions to Selected Problems in Chapter 7

February 9, 2006

Refer to the MATLAB le ch7a.m for the code used to synthesize the controllers.

Problem 7.5

The specied goals lead to the following structural constraints: The controller must have integration to achieve zero steady state error for constant references. The controller must have poles at s = j0.25 to achieve zero steady state error for sine wave disturbances of frequency 0.25 [rad/s]. The closed loop polynomial should be chosen of minimum possible degree to have a biproper controller, i.e Acl (s) must be of degree 6. To make the problem as simple as possible, Acl (s) is chosen to cancel the plant poles. This will force the factor (s + 1)(s + 4) into P (s). Say we choose Acl (s) = (s + 1)(s + 4)(s2 + 4s + 9)(s + 10)2 . Then the Diophantine equation is

(s + 1)(s + 4) s(s2 + 0.0625)(s + 0 ) +


L(s)

(s + 4) (s + 1)(s + 4)(2 s2 + 1 s + 0 ) =
P (s)

(s + 1)(s + 4)(s2 + 4s + 9)(s + 10)2 (1) 1

Using the paq.m MATLAB subroutine we obtain C(s) = (s + 1)(s + 4)(97.074s2 + 199.358s + 225) s(s2 + 0.0625)(s + 121.074 (2)

problem 7.6

7.6.1 We need to solve the pole assignment equation (s + 2)(s + 4) s(s + 0 ) +8 (2 s2 + 1 s + 0 ) = (s + a)2 (s + 5)2
L(s) P (s)

(3)

We then solve for a = 0.1, using paq.m, and obtain C(s) = 0.7737 and, for a = 10 (s + 3.7803)2 + 16 s(s + 24) (s + 4.613)(s 0.0088) s(s + 4.2) (4)

C(s) = 21.625

(5)

7.6.2 We observe that for a = 0.1, the controller exhibits a NMP zero, i.e. a zero located in the RHP. Furthermore, the RHP zero is located at 0.0088, which is much closer to the imaginary axis than the dominant closed loop pole (s = 0.1). This will lead to large undershoot in the step response, as explained in Chapter 4. This is not the case for a = 10. The source of this dierence is that all poles and zeros in the plant nominal model are located to the left of the specied closed loop double pole at 0.1. Using root locus arguments, we see that if we force an open loop pole at the origin, we will need to have one branch closer to the origin than the plant dominant pole. However we still need a branch attractor to position the second pole, this can be achieved with a controller zero in the RHP.

The eect of the double pole at 0.1, compared to a simple pole at the same location, can be better appreciated if we choose Acl (s) = (s+0.1)(s+5)3 . Then the pole assignment synthesis yield 1.7375s2 + 7.4625s + 1.5625 s(s + 9.1)

C(s) =

(6)

The eects of one and two closed loop poles can be appreciated in the step responses shown in Figure 1.
1 0 Step responses 1 2 3 4 Acl(s)=(s+0.1) (s+5)
2 2

Acl(s)=(s+0.1)(s+5)3

10

20

30 Time [s]

40

50

60

Figure 1: Eect of a single and and a double closed loop pole at s = 0.1.

The dierences in Figure 1 are clear. We emphasize that the problem, in this case, lies in the fact that we have specied two closed loop poles which are much slower than the plant poles.

Problem 7.7

To achieve a strictly proper controller the degree of Acl (s) has to be chosen larger than the minimum degree. In this case that minimum degree is 3 (= 2n 1). Say we choose k = 2, then the degree of Acl (s) turns out to be 4. The Diophantine equation then becomes

(s + 1)2 s(s2 + 1 s + 0 ) + 1 s + 0 = (s2 + 4s + 9)(s2 + 4s + 4)


L(s) P (s)

(7)

The subroutine paq.m gives L(s) = s2 + 6s + 16 and P (s) = 14s + 20. The same idea can be applied if we force integration in the controller, then k should be chosen equal to 3.

Problem 7.8

We rst notice that a minimum degree biproper controller (with integration) requires Acl (s) of degree 4 (= 2n). We thus choose Acl (s) = (s2 + 7s + 25)(s + 10)2 (8)

The choice of the double pole at s = 10 is arbitrary but for the requirement that they should generate modes1 faster than those2 produced by the factor s2 + 7s + 25. The associated Diophantine equation is (s2 s 2) s(s + 0 ) +(1) (2 s2 + 1 s + 0 ) = (s2 + 7s + 25)(s + 10)2
L(s) P (s)

(9)

We thus obtain L(s) = s(s + 78) and P (s) = (295s2 + 1256s + 2500).

Problem 7.10

A Smith predictor is shown in Figure 7.1. We then need to synthesize a controller considering only the rational part, Go (s), of the nominal model, Go (s), where Go (s) = s+5 (s + 1)(s + 3) (10)

The nominal complementary sensitivity is then


1 2

Those modes are e10t and te10t Those modes are K1 e3.5t cos( 12.75 t + K2 )

To (s) =

e0.5s C(s)Go (s) 1 + C(s)Go (s)

(11)

If the dominant closed loop poles are 2 j0.5, and we require integration, we can build a closed loop polynomial of the form Acl (s) = (s2 + 4s + 4.25)(s2 + 8s + 16). Thus

(s+1)(s+3) s(s + 0 ) +(s+5) (2 s2 + 1 s + 0 ) = (s2 +4s+4.25)(s2 +8s+16) (12)


L(s) P (s)

The solution of the above equation yields L(s) = s(s + 4.7687) and P (s) = 3.2313s2 + 14.0187s + 13.6.

You might also like