CHEM ENG 3P04 ASSIGNMENT 6 DUE 4:00pm Tue 03/31/15

QUESTION 1

Multivariable control is to be applied to a wet-grinding circuit shown in the schematic

below. For this particular circuit, the following variables were decided to be con-
trolled: the torque required to turn the mill (TOR) and the flow rate from the mill
(FML). The manipulated variables for this system are the feed rate of water to the
mill (MW) and the feed rate of solids (SF).

By means of open-loop testing, the following transfer functions relating the con-
trolled and manipulated variables were obtained:

MW (kg/s) SF (kg/s)

TOR (Nm) 153 119

337s + 1 217s + 1
FML (m3 /s) 0.000767 0.000370
33s + 1 500s + 1

All time constants and dead times are in seconds.

(a) Use the relative gain array to determine an appropriate loop pairing for this
system

1
 (b) Draw a clearly labelled block diagram of the complete control system. Vari-
ables and transfer functions should be clearly indicated.
(c) Design a decoupler for this system, and show its implementation on your block
diagram.
(d) Looking at the schematic of the process, do you think that your control system
would be sufficient for the process to be practically operable? Explain, and
suggest a remedy if warranted.

QUESTION 2

The process represented by the transfer function model

    
1 2
y1 (s)  u1 (s) 
    
  



 =

 2s + 1 s+1 
 

    
 
2 
5  
y2 (s) u2 (s)
    
2s + 1 s + 1
is to be controlled using a multi-loop proportional control system with pairing y1 −
u1 /y2 − u2 . If the gain of the y1 − u1 pair is Kc1 = 5, determine the range of
controller gains for the y2 − u2 loop (Kc2 ) for which the system will be stable with
both control loops closed. Will the system remain stable for Kc2 within this range
under conditions where only one of the control loops is operating under feedback
control, with the other loop on manual (i.e. with the corresponding input to the
plant being constant)? Justify your answer.

QUESTION 3

The temperature T , on a critical tray in a pyrolysis fractionator, is to be controlled

as indicated in the accompanying diagram, employing a feedforward controller that
manipulates the underflow reflux flowrate R on the basis of feed flowrate measure-
ments F . By experimental testing, the following individual transfer function models
were obtained for the response of the critical tray temperature to (1) feed changes
and (2) reflux flowrate changes:
0 20e−15s 0
T (s) = F (s) (1)
(30s + 1)
−5e−2s 0
T 0 (s) = R (s) (2)
(6s + 1)

2
with variables expressed as deviations from steady-state values. The time scale is
in minutes, the flowrates and temperatures are, respectively, in Mlb/hr (thousands
of pounds/hour) and ◦ F.

(a) Assuming that the process model is accurate, design a feedforward controller
for this system to achieve “perfect ” feed disturbance rejection, and show in
a block diagram how it would be implemented. Is the feedforward controller
realizable? Explain.

(b) Consider now the situation in which the model is inaccurate, with the “actual”
process given by

−4e−2s 0 25e−15s 0
T 0 (s) = R (s) + F (s)
(6s + 1) (30s + 1)

Using the feedforward controller designed in (a), simulate (using Simulink)

and plot the response of the process, T 0 , under feedforward control to a step
change of 2 Mlb/hr in the feed rate (disturbance).

(c) Design a feedback controller for this process using Direct Synthesis, with τc
chosen as 25% of the open-loop time constant. In the design, use the param-
eters of the model assumed to be correct and given in Eqs (1) and (2). Apply
this controller to the “actual” process given in (b) without the feedforward
controller, and simulate the closed-loop response to a step change of 2 Mlb/hr
in the feed rate (disturbance).

(d) Now apply (to the “actual” process) a feedforward-feedback control scheme
using the feedforward controller designed in (a) (and used in part (b)), and
the feedback controller designed in part (c). Simulate the closed-loop response
to a step change of 2 Mlb/hr in the feed rate (disturbance).

3
(e) Briefly discuss your findings with regard to the relative performance of the
three control schemes above.

QUESTION 4

4
(a) Cascade control is a popular strategy for dealing with the control problems
created by such steam pressure fluctuations. Draw a block diagram for this
process under such a cascade control strategy; include all the given transfer
functions (and controllers), and label all the signals.

(b) Consider the cascade control system in (a). Determine the offset to a unit step
change in the steam supply pressure, if proportional controllers were used in
both the primary and secondary loops, with the controller gains (Kc ) in both
controllers set to 2.

(c) Because the steam supply pressure is measured, it is also possible to config-
ure this process for feedforward/feedback control instead of cascade control.
Reconfigure the process for this new control structure, draw a block diagram
showing this new configuration, and obtain an expression for the feedforward
controller to be implemented.

(d) Show the (i) cascade control scheme of part (a) and (ii) feedforward/feedback
control scheme of part (b) on replicas of the process flow diagram of Fig P16.3.

5
CHE 3P04 Process Control: Assignment 6
Due on 31/03/2015

Dr. C.L.E. Swartz

Assignment 6 Solution Guideline

Assignment 6 Solution Guideline CHE 3P04 (Dr. C.L.E. Swartz): Assignment 6

Problem 1:

(a)

1
λ11 = k12 k21
1− k11 k22
= −1.63
 
−1.63 2.63
Λ=
2.63 −1.63

∴ pair TOR-SF and FML-MW.

(b-c)

T ORset + − SF ∗ SF 119 T OR
gc1 (s) + (217s+1) +

153
D1 (337s+1)

0.00037
D2 500s+1

F M Lset + MW∗ MW 0.000767 FML

gc2 (s) + +
− 33s+1

For D2 set SF ∗ = 0
119 153
T OR = 0 = M W ∗ D2 + MW∗
217s + 1 337s + 1
−1.28(217s + 1)
∴ D2 =
337s + 1
For D1 set M W ∗ = 0
0.000767 0.00037
T OR = 0 = SF ∗ D1 + SF ∗
33s + 1 500s + 1
−0.482(33s + 1)
∴ D1 =
500s + 1

Problem 1 continued on next page. . . Page 1 of 8

Assignment 6 Solution Guideline CHE 3P04 (Dr. C.L.E. Swartz): Assignment 6

(d)

No. Since we would want to control the level in the tank to ensure that it does not overflow.

Problem 2:

y1,set + − u1 y1
gc1 g11 +

g12

g21

y2,set + u2 y2
gc2 g22 +
−

For stability with both loops closed, roots of g(s) need to be in left-half complex plane where

g(s) = (1 + gc1 g11 )(1 + gc2 g22 ) − gc1 gc2 g12 g21
(Seborg p. 346)

5 
 
5 2

k >

c1 k c2 5 2
=⇒ 1 +   1+ −
kc1
>k

c2 =0
2s + 1 s+1 s + 1 2s + 1
(2s + 1 + 5)(s + 1 + 5kc2 ) − 20kc2
=⇒ =0
(s + 1)(2s + 1)
=⇒ 2s2 + 6s + (1 + 5kc2 )2s + 6(1 + 5kc2 ) − 20kc2 = 0
=⇒ 2s2 + (8 + 10kc2 )s + 6 + 10kc2 = 0

For stability

8 + 10kc2 > 0
=⇒ kc2 > −0.8

Problem 2 continued on next page. . . Page 2 of 8

Assignment 6 Solution Guideline CHE 3P04 (Dr. C.L.E. Swartz): Assignment 6

6 + 10kc2 > 0
=⇒ kc2 > −0.6

So, kc2 > −0.6.

For stability of individual loops (with other loop on manual), require

1 + gc1 g11 = 0
5
1+ =0
2s + 1
=⇒ 2s + 6 = 0

Both coeffs +ve =⇒ stable

1 + gc2 g22 = 0
kc2 5
1+ =0
2s + 1
=⇒ s + 1 + 5kc2 = 0
=⇒ 1 + 5kc2 > 0 or kc2 > −1/5

Problem 3:

(a)

0 5e−2s 0 20e−15s 0
T (s) = R (s) − F (s)
6s + 1 30s + 1

0 20e−15s
R (s) Gd 30s+1 4 (6s + 1) e−13s
Gf f = 0 =− = 5e−2s
=
F (s) Gp (30s + 1)
6s+1

The above controller is realizable. It does not have any right half plane poles or require any
future output values to compute a current control move.

Problem 3 [(a)] continued on next page. . . Page 3 of 8

Assignment 6 Solution Guideline CHE 3P04 (Dr. C.L.E. Swartz): Assignment 6

−Gd
Gf f = Gp F0

20e−15s
Gd = 30s+1

+
y set + E1 + R0 (s) −5e−2s + T0
G c1 Gp = 6s+1
− +

(b-d)

Note: Feedback controller (part-c)

1 6
Kc = − = −0.343, where τc = 0.25 × τp = 1.5
5 2 + τc
τI = 6

Closed loop response for each controller to a step change in F of 2 Mlb/hr:

Problem 3 [(b-d)] continued on next page. . . Page 4 of 8

Assignment 6 Solution Guideline CHE 3P04 (Dr. C.L.E. Swartz): Assignment 6

Figure 1: Closed loop response for each controller

(e)

As observed in the figure for the previous part of this question, there is offset when imple-
menting the feedforward controller. This is expected since there is plant-model mismatch
(i.e. models used to design controller is different than “actual” process and disturbance
model). Even in the presence of plant-model mismatch, the feedback controller steers the
process back to the setpoint; however, when implementing both the feedback and feedback-
ward controller, the closed loop performance is improved (i.e. smaller deviation from the
setpoint).

Problem 4:

Problem 4 continued on next page. . . Page 5 of 8

Assignment 6 Solution Guideline CHE 3P04 (Dr. C.L.E. Swartz): Assignment 6

(a)

d1
1.5
Gd = 0.5s+1

+
y set + E1 uset
1 + E2 u2 2.2 + u1 0.25
y
G c1 Gc2 Gu = 2s+1 Gp = 10s+1
− −

(b)

Condensed inner loop

1.5
u1
= 0.5s+1
d1 1 + k2s+1
c2 2.2

1.5(2s + 1)
=
(0.5s + 1)(2s + 1 + (2)(2.2))
3s + 1.5
=
(0.5s + 1)(2s + 5.4)

c2 k 2.2
u1 2s+1
=
u1set 1 + k2s+1
c2 2.2

(2)(2.2)
=
2s + 1 + (2)(2.2)
4.4
=
2s + 5.4

Figure 2: Condensed Block Diagram

Problem 4 [(b)] continued on next page. . . Page 6 of 8

Assignment 6 Solution Guideline CHE 3P04 (Dr. C.L.E. Swartz): Assignment 6

For Gc1 = Gc2 = 2, and d1 = 1s :

0.25 3s+1.5
y 10s+1 (0.5s+1)(2s+5.4)
= 4.4 0.25
d1 1 + 2 2s+5.4 10s+1
, Gd

  0.25 3s+1.5
1 10s+1 (0.5s+1)(2s+5.4)
y(t=∞) = lim sy(s) = (s) 4.4 0.25 = 0.049◦ C
s→0 s 1 + 2 2s+5.4 10s+1

Offset = y set − y(t=∞) = 0 − 0.049 = −0.049◦ C

(c)

−Gd
Gf f = Gu d1

1.5
Gd = 0.5s+1

+
y set + E1 + u2 2.2 + u1 0.25
y
G c1 Gu = 2s+1 Gp = 10s+1
− +

In this process, the disturbance affects our control action u1 . Therefore, the feedforward
controller should adjust u2 in order to negate the impact the disturbance has on u1 . The
expression for Gf f is given as:
1.5
−Gd −0.681(2s + 1)
Gf f = = − 0.5s+1
2.2 =
Gu 2s+1
0.5s + 1

Problem 4 continued on next page. . . Page 7 of 8

Assignment 6 Solution Guideline CHE 3P04 (Dr. C.L.E. Swartz): Assignment 6

(d)

Page 8 of 8