0% found this document useful (0 votes)
48 views34 pagesSurekha B3
book for process control
Uploaded by
smitthakor1415Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
0% found this document useful (0 votes)
48 views34 pagesSurekha B3
book for process control
Uploaded by
smitthakor1415Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
/ 34
ee
CHAPTER
A
Closed-loop Response
4.1 INTRODUCTION
Closed-loop response of a process pertains to th
whole system (including controller, measuring clement, final control element),
after an upset which can be caused either by a change in the set point or by a
change in the load/disturbances acting on the process. The characteristics of
every element present in the loop affects the performance of the process. The
combined effect of different types of controller and different types of process is
evaluated on the basis of amount of overshoot, final steady state error, and the
time taken for the controlled variable to reach the set point. Depending upon the
type of process, different controller modes may be required to achieve optimum
performance.
he dynamic behaviour of the
4.2 TRANSFER FUNCTIONS OF CLOSED LOOP
Figures 4.1 and 4.2 depict the block diagram and the signal flow diagram of a
closed loop respectively. To simplify calculations, unity transfer function is as-
sumed for sensing and final control element, Further, following assumptions are
made for simplifying analysis.
1. Measuring lag of sensor is zero,
variable, i.e. b = ¢
2. The controller lag is zero.
so that controlled variable is the feedback
Fig. 4.2 Signal flow diagram of closed loop
system with unity feedback
ed loop,
fi f clos
Fig. 4.1 Block diagram of
system with unity feedback ‘The process equation is given by:
c= Gyn + XN; (4.1)
= Gy (Gye) + LN;
G, = Controller transfer function
Gy = Process transfer function
N, = Process load function for disturbance 1,
m= Gye (Controller equation)
If u=0, then £ = (4.2)
,
and 2 =—1_ (4.3)
v 13,
cl N
If v=0, then £ =“ _ (44)
. no Te Gy
and! x £,5,-—% (4.5)
u 14+GG,
: Combining the effect of both ‘set point’ and ‘disturbance’:
" GG N
' = F1_ wy, 4.6
+aG,* aa,” G9)
I N
= —— _(v) - —__ (4.7
“T6G,- haa, a
If there are more than one load variable, say n, the process and error equations
ty would be:
it 6
" = 1G yey —M 48
OO Teean) Linea * Ce
1 n oN,
= y- yy, 49
nae,” zie" a)
From above equations, it is clear that there can be two sources of deviation.
(i) Variation of load variables (w)
i) Variation of set point (v)
In servo control, the controlled variable ‘c’ is expected to follow the set point
with zero or minimal deviation; while in regulatory control, the controlled variable is
expected to remain constant despite the changes in load variables. The error in
_ the controlled variable resulting from the change in set point and upon distur-
sbance is the basis on which different controllers are comp:
dieronterms a
sulting from the step change in the load — Offset error (E,)
Iting from the step change in set point — Static error (E,)
from the linear change in set point — Velocity error (E)
ared, Three commonly (43°) PROPORTIONA
Closed-loop Response 89
iow, let us take a closer look at different processes and contol modes to study
their closed-loop response.
L CONTROLLER IN CLOSED LOOP
portional controller amplifies the error signal to gener-
rete contol signal. If the error signal isin voltage, and the contol signal is also
se eltage, then the proportional controller used in the loop is just an amplifier.
Figure 43 depicts a level control system with mr and, s the inflow and outflow
respenively. The downstream head w acs as the disturbance and ¢ is the level in
the tank, 4 is the cross-sectional area of the tank.
4p analog system, 8 Pro
de
ae ES 10)
m= 47 AT 4.10)
enw
= 1
0 4.11)
Where 2 is the resistance of the outflow pipe
‘The mass balance equation (Eq. (4.10)] can
circuit shown in Fig, 4.4, with the governing equation as:
be represented by the electrical
cl =1-4
dt
Capacitance is equivalent to area of the tank. Inflow and outflow are equiva-
potential (V). If, flows through
ent to current, and level is equivalent to electric
resistive element as shown in Fig, 4.5, then J) =
tance element having its other end at different poten
If flows through resist
y
tial V;, as shown in Fig. 4.6, then J) = —p
Nowe let us find the process equation for Fig. 43 in Laplace domain.
(3) ~ qo(s) = 48 (8) (4.12)
als) = (els) =u] (4.13)
R
(4.14)
Therefore, es) = [4a m(s) + a us)
(downstream head)
eat 0
R
Fig.4.3 Level controlsystem 90. Process Control: Principles and Applications
jivalent when down-
i i ivalent of Eq. (4.10) Fig.48 Electrical equiva e
Tae eps asa ° stream flow through resistance
Ris open to atmosphere
R
Ae
Fig. 4.6 Electrical equivalent of Eq. (4.11)
= e(s) = (al m(s) + ea} (4.15)
where T= time constant of the vessel = AR; it depends on the capacity (C = A)
and resistance, R. Figure 4,7 depicts the use of proportional controller in a closed-
loop system.
1
1+ Ts 1+Ts
|: ' (4.16)
Gy= Key =
| 1 N
| ee —
[+a 1+ G,G,
Substituting G, and G, in Eq. (4.16), we get,
eo) = CAD =uls)
[¥ Ts KR 1+K.R
| T, < T, ic., time constant of closed loop system is less than process time con-
stant, making the response faster.
For a fixed set point ‘v’ and for a step change in load of magnitude “U’,
where 7, = (4.17)
“= (4.18)
—__ 3s"
(0+ 790+ KR)
Block diagram representation of system shown in Fig. 4.3. Closed-loop Responee 94
(eh
10 \
= Kft 0 (0 conc
+ Ke
~ Khe
ee
—— Tine
Fig.4.8. Eflectofk, Ron offset
Step response
Slate ror J
09
08
07
06
0s
04
03
02
ou
5 Ts as 3
Time (see.)
Fig. 4.9 Effect of K, on static error
K-10
-u
Fy = offset = Lt s els) = ,
fo offset = Ls) TKR a)
Figure 4.8 depicts the effect of K,. on offset.
Tt can be observed from Fig. 4.8 that with the introduction of proportional
control, the offset is reduced; and a gradual increase in K, reduces the offset
further.
AR j
Stabilization time (99% response) = 5T;= 7 gs italso decreases as Kis
increased.
Figure 4.9 illustrates the
It can be observed that as K, increases,
fect of proportional control mode on servo operation.
the static error decreases.
Example 4.1. Calculate offset and stabilization time for the level control system
shown in Fig, E4.1 (a) in which outflow ‘x’ is through a pump,
Lu
Fig, E4.1 (2) Liquid level control system {
PH
92 Process Control: Principles and Applications
Given: v= 2 m/s
1 m/s step increase in outflow
m’/s/m [Proportional Controller]
A=10m?
‘Solution
moun sk
di
i
Taking Laplace transform,
m(s) ~ u(s) = As e(s)
Therefore, o(s) = mo @
is As
Block diagram of the closed- I-loop system is shown in Fig. E4.1 (b)
Fig. E4.1 (6) Block diagram representation ofthe system in Fig, E4.1 (a)
with proportional controller
N
v(s) — GG, us)
1
On Ge, 7
1
hs) + (Fk, u(s)
ae +05
where T= A/K,
uit
For a step change in outflow, E(s) = + 0K
U
Offset Ey = a (s E(s)) = Kk o = 0.05 m which causes decrease in level
: a
(since outflow has increased),
A
Stabilizing time = $-r= 5.2 = 5 x 2 = 25 sec,
fi 2 a
Note that as K,, increases, both offset and stabilization time decrease,
Example 4.2
. R N R
Given: G,=K, Gr Fale Bo) Tals Es
(a) Calculate kK, for critical damping.
(b) Calculate the error due to step change in load and in set Point, Closed-loop Response 93
KAR,
1+
Ts (1+ Ts)
Ts(1+Ts)
Denominator polynomial when set equal to zero is called the characteristic
equation because roots of this equation determine the character of time response.
Comparing the denominator with the second order characteristic equation,
2¢
—-s+1 =0, we have:
een
md ER
7,
Hence = le eg
A
If K, increases, damping factor decreases; which implies that employing a
proportional controller will turn the overdamped system into an underdamped
system
Tq,
For critical damping, ¢= 5K, 8
T,
> GRE)
(b) Error due to step change in set point:
Ts+hs) Vv
Static error = e, = Tas +hst+K,R, :
E,= lt se,
530
Error due to step change in load variable: e
at |
+H stk, RS
Offset = Eo = M, Seu= Hy Th 94 Process Control: Principles and Applications
Example 4.3 The transfer function of a 3-tank system is
6
G(s) = ———__° __
n= ee hde rhea
The system is under proportional control. Find the values of K,, for which the
closed-loop ins stable using method of (i) Routh’s eriterion ang
(ii) dire
Solution
6K,
() Characteristic equation= 1+ G.(s) G(s) = 14 0
Qs+1)(4s+ 1) (65+ 1)
= ABs" 4 dds? + 1284 (146 K,) = 0
Checking stability using Routh’s stability criterion:
48 12
44 146K,
120- 72K,
——* 0
MW
1+6K, 0
The first column should contain Positive elements, therefore,
120 _ 5
120>72 Kk, K,< = =2
- 22 3
1+6K,>0 = K,>o1
6
1 5
Hence, ~i ee 2
Gh an a
Integral controller function = G, =
Ts
R
Process transfer function = G) = —~——
“ 1+Ts
1
1+Ts
Load transfer function = N =
where c= controlled variable
T = vessel time constan'
R = resistance of outlet valve
C= vessel capacitance = A = area
m= manipulated variable = inflow
= load variable = downstream head
= —L(y—c) where 7; = integral time of
$s
8. 1
The control equation is: m= 7—e
Ts
integral controller.
Block diagram of the ¢
integral controller is shown in Fig. 4.10.
Josed-loop system for the process in Fig. 4.3 with 96 Process Control: Principles and Applications
IT a)
han F
Tse Ty T.s(14 Ts)
Ts(l+ Te)
The shsek
(4.21)
S14 Ts)
u(s) (4.22)
where, +
(4.23)
Static error = 0 and offset = 0
It may be observed that introduction of, integral action has increased the order
of the system. Because the integral action forces the controlled variable to retum
{o the set point whenever there is a change in load, it climinates the offset com-
Pletely: On the other hand, integral action makes the system to oscillate around
the set point and increases the maximum deviation, Ultimately, the stabilization
time is increased due to increased order of overall system, Closed-loop Response 97
‘The integral time 7; of the controller must be sel
The ime T, selected to provide proper damp-
ing, Since theres no ast, a moderate damping atumayrechae ne
- :
of about = is usually adequate. The integral time for damping ratio of + can be
computed from the following equation: :
T,= 48 RT (4.24)
Figure 4.11 depicts the effect of 7; on the system response, It can be observed
that small value of 7; increases the range of oscillations.
Step response
‘Amplitude
0 3 10 15
Time (sec.)
Fig. 4.11 Integral response of first-order system to unit step change in set point
Example 4.4 Show that for the process shown in Fig. E4,1(a), instability is
possible with integral control.
Solution
a de
The process equation is: m—u= AT
Taking Laplace transform,
_1 1
es) = ols) — ems)
where c= controlled variable = Level in vessel
A= C= capacitance of vessel = Area
m= manipulated variable = Inflow
= load variable = Outflow
1 N
—1_y-—
14#G,G, 1+G,G;
(E4.4.1)
AT, 5°
2 AUS ss) 4
6” Artal 98 Process Control: Principles and Applications
Fig. £44 Block diagram of Fig. E4-1(a) with integral controller
The transient solution to the characteristic equation (denominator of Ey.
(E4.4.1)) is a sinusoidal function without damping (there is no ‘s* term), hence
Producing a continuously oscillating controlled variable, Outtov in this process
does not depend upon the head: liquid is pumped out at a constant rate, So, it
inflow increases or decreases, the tank will overflow or go emply. The process
itself is not self-regulating and the addition of an integral controller to such a
system will make it unstable. Recommendation: Do Not use integral controller
for processes that are not self-regulating,
sr CONTROLLER IN CLOSED LOOP.
‘roportional and integral controls are often combined in order to obtain the ad-
vantages of both the modes ~ inherent stability of proportional control and offset
temoval of integral control. P/ is the most commonly used mode of controller itt
industries. Consider the same process of Fig. 4.3. The block diagram of the
closed-loop system with PY controller is shown in F 4.12,
R
The process equation is: o(s) = mts) + ys)
Tt
Ts+1
where
controlled variable
Fig.4.12 Block diagram of the process of Fig. 4.2 with PYcontroller Closed-loop Response 99
R= resistance of outlet valve
T= vessel time constant = RC
Ge
‘m= manipulated variable = inflow
A= vessel area
u= load variable = downstream head
Substituting values of Gy, Gy, and Nin Eq, (4.16),
Ts(Ts+l)
y= __ypg
TaTi+ + RK, CoD)” Tater + RK, hse) 8)
Reducing the denominator to the standard form:
2g 22
Ss41=0 of Ps +20 s+; peb
“a,
where
oe (14 RK, 2 (1+ RK),
4RK,?
agrt=
S (RK,
(RK, +I _
sy TRK,T = Damping ratio
itis noticeable tat the order ofthe sytem has increased; K, and 7, should be
chosen carefully so thatthe system doesnot become highly underdamped resull-
ing in oscillations.
integral control of the process shown in the
jonal-
sensitivity and integral
Example 4.5 For proport
Fig, E4.1(a), calculate the relation between proportional
1
time for a damping ratio of = -
Solution
1 1
The process equations: (9)= 5-19) —3 10) Characteristic equation = 7,
where
¢= damping factor = (
For a damping ratio of i. c=
Example 4.6 For the process shown in Fig, E4.1 (a),
manipulated variable ‘m’,
disturbance.
2m’, mea
inflow to tank is the
Outflow ‘u' through a pump on outflow line acts as
Controller is P/ having a proportional gain of 4, area of the tank is
ring element has unity transfer function
(a) Draw the block diagram of the process control loop.
(b) Find @, and the value of K, which will result in a closed-loop response
having a damping factor of 0.5.
Solution
Block diagram of the controlled system is shown in Fig. E46
h
Fig E4.6 Block diagram of the process of Fig. E4.1(a) with Pl controller
Characteristic equation = 1 + G, G, K,=4, A=2m?
we 2s? +45 + 4K, = 0 as t2s+2Ke0
Comparing with standard form
84 Was + a= 0, Se, = 2: tor C= 0,8
@, = 2 rad/see
> 2
2K,= @ or K,~ &e
Example 4.7 A process control loop has G, as P/ controller, Process i
grating element with a prov in of K. Control valve and measuring element
have unity transfer funetion, Find the relationship among, K, K,q and K, for a
damping factor of 0.5, cg ekiaie o
Solution
> 8 +sk,K+K,/K,K=0
Comparing with the standard equation, s? + 2¢¢,8 + 2 0, we have:
a, = (KK, K
260, = Ky K
Given that ¢= 0.5
K, K
= ¢el 2 =05
= K,K=K,
: m : :
Example 4.8 The servo motor has a gain of ar and a time constant of
vol
20 seconds. The speed sensor changes its output linearly from 0-8 V for a speed
change of 0-80 rpm. The controller is of P/ mode with 0-10 V output and 50%
7 _ 6 (% /min)
PB. The constant of integration is ——~—— - Output of the controller goes
through an amplifier of gain 2 to drive the’ motor. Set point is 55 rpm.
(Draw the process control loop with transfer function inside each block and
range of values with appropriate units outside each block.
(ii) Find natural frequency of the closed-loop system.
Solution
Figure E4.8 shows the block diagram of controlled system with proper values
and units and transfer functions. 02 Process Control: Principles and Applications
é S410" Controller Amplifier
s sssg yf oft story 7 pe
‘ 7 . ai} - ed
‘ “1! G,
6
Measuring element
- i
ORV 10
Fig. £48
y= Noten _ 19 as
SO, 5
50% impor ($0 xslv
(100
inc 10)
00 =
nt = Lsc0
60sec l xs) a
(100
‘Characteristic equation of the closed-loop transfer function 1s
1*GG,
)
= tr2s{1+ L Ja 2 x(t} -0
&y 20s+1) (10)
" o» 3498), x2
8” 08+ 1)
= BS (20s+ 1) + Las + 2=0
= 100s? + ks + 16s +2 =0
= 160
=> Bas?
= = 0.11 rad see
Example 4.9. The liquid level in a tank (cross-sections! ares of tank & 8 me,
Steady state height of tank is 4m ) is controlled by cestroliing the inflow theough
s valve on inlet pipe which ih tum is acted upon by a P/ coatruller The cutflier
4s equal to 24°°. Transfer function of the contiol valve is units acd treme
function of the measuring element is a first-order systems with steady state gi
and time constant ay unity. Draw the process coat! foop. Fad the value of
that will make the closed loop oscillaty at a coastant freaqucacy, when G) 4, >
(ii) K,, ~ 10. Find the frequency of osciltation in boch the casce a Closed-loop Response 1903
di Vo = ath
dt
q:- 28? = 4th
7 dt
Using Taylor’s series,
= 3/20
Fo = Gos * (2RY Wy Ch hs h = din
90 Gos * 3h, (h~h,)
In terms of deviation variables,
Q,=6H
2: (s) ~ Q, (8) = AsH (s)
Q;(s) - 6H (s) = 9sH (s)
AO lt
Qs) As+3fh, 6495
Here,; » G,= controller transfer function = K, (: +h
. 5
G, = Process transfer function =
+95
; 1
G,, = Measuring element transfer function = Tes
+s
Characteristic equation = 1 + G, G, G,, = 0
> 953 + 155° + s(6 + K,) +K, K/=0
Case I: K, = 5
The characteristic equation becomes 98° + 15s” + IIs + 5K;=0
Applying Routh’s stability criterion,
9 ll
15 5K;
165-45K, g
15
5K, 0
165-45K, _g
To oscillate at constant frequency, —— 7 ~~
au
"3
tion method characteristic equation (CE)
Using direct substitut ‘
15a + Ijat 5K,
(cE) = 97 - 104 Process Control: Principles and Applications
vil rad
K,= 30 and @=
g 3 sec
So k=
Case I: k,
The characteristic equation becomes
9s? + 155? + 16s + 10K, =0
4 rad
J@and equating real and imaginary parts, o= 4
3 sec
Substituting s =
K-38
3
Observation: High value of K,, results in reduced K.
we CONTROLLER IN CLOSED Loop
controllers offer the f it
ative action
a very large number
Pacitance process as
u
Fig. 4.13 Two capacitance! non-interacting process with PD controller
The system equations are: Closed-loop Response 105
m(s) — 4(s) = Cz 5 Mts)
gs) =
gs) — uls) = C, s els)
From the above equations, we have:
his)
Ap Mls) = C1 8 68)
R
=> hs) = Tears"
> m(s)— ae sh(s)
1
m(s)—u(s) = C, s ets)
1+ G Rs
oa Rtn
= 9" RasaeGRn™
=> ols) = n(s) ———u(s)
Rsd+hs). Gs
where T, = RC, and T; = RC,
‘The block diagram of the controlled system is shown in Fig, 4.14.
i
Here, G, = K, (1 + Ts); @ = ————s N= - a
i= Ky 1+ Ta = Fae Ray Gs
1
Now, e = ——>’-— 7
1+G,G, | 1+G,G;
v(s) + ce u(s)
=> &s)= —————_+—_ 3
Thee
14K, +L FG Re
N
aes
w Gs
vitae r R 3 7
CO] Ko + T4 | Tys (1 + Ta) +
a a
aller
4.14. Block diagram of process of Fig. 4.13 with PD contr
Fig. 406 Process Gontrol: Principles and Applications
=
ae (4 Ts)v(s) + : (1+ F s)u(s)
Kk Ky :
=> y= RK, RE (4%
reas stl
KR
T+ K, RT;
gt Ky Rl = 207,
5 2
(RK, T+) _
4K, RTT,
s(l+ Ts) 1 (+%s)
ee el
x +2¢T, se“) K, Ts? +207.s+1
The addition of derivative control to proportional control doesn’t increase th
order of the system. Instead, it increases damping of the system and makes th
response sluggish. Derivative mode does not have any direct effect on the offse
but it can reduce the offset indirectly by allowing K, to increase without causin
oscillations.
Example 4.10 For the control system shown in Fig. E4.10, given
1 <,
2s 43541)
Gs) G(s) = 1; 6.) =K,
Gs) = $= 1.06
Ue)
+O Guy Guy MS Gs)
= 2 + eh (8 * DOs + 1]
¥)
Fig. E4.10
Design a control system which makes the offset resulting from a unit step!
0,1 and clased loop damping > 0.6 Closed-toop Response 107
Solution
Ys) _ G,(s)
U(s) T¥ GG, (5)
a = 1
5) Pee
. (IG, (8) FE 3y +bR)
Putting in standard form, a
Offset = +K 3K, 29 will make offset $0.1
For K, = 9, @, = 2.237 rad/sec, and C= 0.335
If ¢ = 0.6, K, = 2.125, which wi
ill make offset = 0.32. This implies that
proportional control alone wi
ill not satisfy both the requirements of offset and
damping. To increase damping, PD mode can be tried,
_l
1+K,
73+ KI
14K,) * 1#K,
2 ~ tT og
, «VIFK, Sat KD ,
for offset < 0.1
U(s)
With PD controller, 2). =
s+
wos
K,
K,
p
-t_|2 = [2 oa
a neds 10
$= 0.335(1 +3 Ty)
= -T, = 0.264 sec for = 0.6
0
«, (= 9), derivative action has increased damping,
Note: With the same K,, (= 9), derivative
—\ iD LOOP
\ ILLER IN’ CLOSEI
(FID CONTRO!
Tet der the process of Fig. 4.3 and analyze how the combination of
-€t us again consi : :
Process and P/D controller will behave.
G)=G.=K, (+ Kas + Kil) oY
108 Process Control: Principles and Applications
u(s)
1
+
741) T+
e(s) =
= —_s)- "ws
e(s) aa, v(s) GG, H(
- 3(Ts-+1) co
“O" Sa K,KaR)+(+K,Rs+K,KR
I SS + u(s)
S(T+K,K, R)+(+ K,R)s+KK;R
Both static error and offs
et would be zero. Now, we have three tunin 1g Parameters
10 get optimal response.
Example4.11 For the process Sf Fig. 4.3, following data is given:
4; Ky= 1 see; K,=8 see €=0.5; and ©, = 2 rad/sec. Find the area of
the tank and resistance R,
Solution
KyKR 2¢ RK
4
“\THK RR ©, TEKKR
Substituting given data, 4 = 4 mand R= 1 see
‘sical limits of fh nl
i ly oper
And if integral terms keep on
0%, 150%, 50% . This
Capabilities of Curent Mere WA
Integrator windup is Mustrated in Fig. 4.55, which shew si
system where the control signal saturates at a level of 20.41.
‘The control signal saturates iramediately when step error is zp
trol signal then semains at saturation level, and the integgal part on
inereave because the error (SP ~ PV) is genitive, Thee intepral part
crease only when the process output (PV) has beourne
(SP), but the process output remains sturated becauxe of th
Slowly the process output decreases towards the set pwiint. The net
there is a large overshoot.
4,16 (a) shows the integral windup phenomens duc to actuzter vz
in block diagram fashion,
“Anti-windup” in an absolute must for preventing integral windup. So
ods limit controller output at its marimurn/eniniraum value, of stop the
up of error when controller saturates, Fig. 4.16 (by shew the
which manipulates the integral part and keeps it at a proper
actuator saturates; so that the controller is ready to resume action, 2 soon
error changes sign.
Contcoller winds
Process output
ez
ONS,
jgnal
b 0.05
Control si
-
0.15
03
4°
a ol
zg 0
0.1 “CO 40 50.
time
control signal (¢) integral responce
Fig. 415. (0) process output (b) 4140 Process Control: Principles and Applications
Fig. 4.16(a) Integral windup duo to actuator saturation
wl)
eT be Sear
pagel
Fig. 4.16(b) Antiwindup scheme
4.9 COMPARISON OF VARIOUS CONTROLLER
CONFIGURATIONS
Alier discussing so many combinations of controllers, we are in a position to
consolidate the observations, Integral controller alone or in combination with
other controllers will always result in zero offset, Offset of P-controller would
be more than that of PD-controller, Deviation from the set point will be minimum.
for PD-controller, and maximum for /-controller. P/ will reduce the maximum
ion, Table 4.1 depicts the comparison of various control modes.
dev
Figure 4.17 depicts the response of some contro! modes.
Pecontrollers have a greater maximum deviation than controllers with derivative
action because P-controllers do not have the benefit of the stabilizing effect of
derivative action, Offset is also larger in P-controllers. PY has no offset because
of integral action, The destabilizing effect of integral action is reflected in its large
maximum deviation and persisting deviation, Integral control is best suited for
control of processes having little or no energy storage. The simplest mode
providing adequate control is usually the most desirable, Addition of modes adds
Table 4.1 Comparison of various control modes
Controller Mode | Seung Time J Ofiset_[ Max. deviation Rom set pont
P ut 1 ul
Pr HW Zz "
PD L un L
PID ML Zz ML
LH = lower than highest ML = more than lowest Z = zero
H= highest L = lowest Closed-loop Responso 114
Controlled
variable
ZL Uncontrolled
= Integral control
Proportional control
Proportional plus
Sar Yor
Set point Time
Fig. 4.17 Response of various control modes
to the complexity of the process and makes tuning of parameters difficult, Given
below are the positive and negative features of some commonly used control
modes and the general guidelines.
PControl PIControl PID Control
‘© Fast closed-loop * Zero steady state © Zero offset
response error (offset)
* Non-zero steady state] * Most common controller | © Derivative action
error (offset) form reduces oscillation
* Integral action may * Measurement noise
induce close-loop amplified
instability
Guidelines
* Use P-controller if offset is acceptable
* If offset is intolerable, use P/
© PiD-controller is good for sluggish processes and those with “small”
measurement noise
4.410 CONTROLLER TUNING
The tuning of a controller can be described as adjusting the parameters of the
controller so that the system response is good or what is wanted by the designer,
Examples include tuning of a TV’s volume, color/contrast, brightness, intensity
of RGB beams, ete. TV response is immediate while process response takes
time, making the tuning more tedious and time consuming. With more knobs,
tuning will become more complex, but better response is achieved
There are a variety of performance criteria for controller tuning that may be
adapted, some of which are: :
© Keep the maximum deviation (error) as small as possible.
© Achieve short settling times.
© Minimize the integral of errors until the process has settled at its desired set
point. 412 Process Control: Principiss and Applications
4.10.4 Performance Criteria
ssocess control application, nvo types of performance criteria ares,
0
performance criteria, and
erform
dy-state performance criterion usually is zero error at steady
‘ns. the proportional controller cannot achieve zero steady.
Je a Pf controller can. Also, for proportional control. the steady.
to zero as K. > =.
ion of the dynamic performance of a closed-loop system is based
used criteria:
of common;
1. Criteria that use only’ a few points of the response. Such criteria are sim-
bur only approximate.
Criteria that use the entire closed-loop response from time ¢ = 0 until re-
sponse seitles down.. These are more precise but more cumbersome to use.
¢ simple performance criteria based on the closed-loop response of
Some of:
5
mare:
© Overshoot
e Rise time (time needed for the response to reach the desired value for the
first time)
© Settling time (time needed for the response to settle down within 25% of
the desired value)
# Decay ratio
«Frequency of oscillation of the transient response
“All of the above characteristics could be used by the designer as the basis for
selecting the controller and the values of its adjusted parameters. Thus, we could
design the controller in order to have minimum overshoot, or minimum settling
time, and so on. The most common criterion applied is to adjust the controller
parameters so that the system response curve has a decay ratio of 1/4, ie. the
ratio of overshoot of the first peak to the overshoot of second peak is 4:1. A
decay ratio of 1/4 is a compromise between a rapid initial response and a quick
settling time, This criterion is usually known as the one-quarter decay ratio extern,
The classical methods of tuning the loop by Ziegler-Nichols and Cohen-Coon
are still in wide use in industry, They provide values of different controller pa-
rameters for different controller configurations.
Various criteria are used to get well-tuned controllers, depending on the
application, the past experience of the engineer and the operator’s personal pref
erences. These include:
(i) Fast response to set point changes, with minimal overshoot and fast set-
tling times. This is particularly important for servo control where the set
point is constantly changing. Cosed-loop Response 142,
(ii) Fast recovery from disturbances. Most disturbances are rendom in nature,
This introduces another level of uncertainty in the design because the wn
for one set of disturbances may be quite differeet
Fast recovery may be characterized in the
decay ratio, settling time, rise time, ov
transient, oF 2 combination of these
If the process is nonlinear (which usual
from one operating point to another, and thereby 2 pantieu
eters can produce the desired response at only one operating point I
ing point changes, controller needs to be retuned. Adee ers or self
tuning controllers are designed and fine-tuned to take care of such requirements.
Adaptive controllers, discussed in Chapter 11, provide “best” adjustment of con-
weoller parameters in face of changing process dyn
form of an integral squ
fershoot, frequency of oscill
operat-
jamics.
Tuning Based on Integral Criteria
he shape of the complete closed-loop response from 1 = 0 until steady tate has
been reached could be used for the formulation of a dynamic performance crite-
tion. Unlike the simple criteria that use only isolated characteristics of the dy-
namic response (c.g., decay ratio, settling time), the criteria for this categ
based on the entire response of the process.
Based on the closed-loop response. these methods minimize the area under
error vs. time curve. Tuning relationships are intended to minimize the integral of
errors, so they are referred as minimum error integral tuning. Integral of error
cannot be minimized directly because a very large negative error would be the
minimum value, so cither absolute value or ‘Square of the error is taken and
minimized.
Y are
Integral of squared area : ISE = * (1) dt (4.31)
Integral of absolute error : IAE = J |e(1)| dt (4.32)
Integral of time multiplied by absolute value of error:
ITAE =J 1 Je(1)| dr (4.33)
e(t) = ysp0) —3(1), is the deviation (error) of the response from the desired set
point. For computational purposes, the upper limit of the integral may be re-
placed by settling time (1,). , :
The following are some general guidelines for selecting one of the three crite-
tia above for the best controller tunin;
* If rejection of large errors is critical, ISE is better than IAE because the
errors are squared and thus contribute more to the value of the integral.
If rejection of small errors is critical, IAE is better than ISE because when
wwe square small numbers (smaller than one) they become even smaller.
IAE gives equal weight to both positive and negative errors
If rejection of errors that persist for long times is critical, the ITAE ce
rion will tune the controllers better because the presence of a large “7
amplifies the effect of even small errors on the value of the integral
© Different criteria lead to different controller designs. 114 Process Control: Principles and Applications
ion, different input changes lead t0 differen,
© For the same time integral crit
desi
40.3 Process Reaction Curve (PRC) Method
This method of controller tuning was developed by Cohen and Coon; it is also
called open-loop method. Consider the control system shown in Fig. 4.18, which
been “opened” by disconnecting the controller from the final control cle.
iment. Introduce a step change of magnitude 4 to the variable m which actuates
the final control element. In the case of a valve, m is the stem position, Record
the value of the output with respect to time. The curve y,,(0) is called the process
reaction curve,
m(s)
Equation (4.34) shows that the process reaction curve is affected not only by
the dynamics of the main process but also by the dynamics of the measuring
sensor and final control clement.
Cohen and Coon observed that the response of most processing units to an
input change, such as the above, has a sigmoidal shape (Fig. 4.19), which can be
adequately approximated by the response of a first-order system with dead time
(FOPDT)
Gras)
Gpncls) GS) Gy (5) Gy (8) (4.34)
(4.35)
Ks)
O- a fe L G [4 O20
Fig,
4.19 Process reaction curve Closed-loop Rosponso 118
Cohen and Coon used the approximate model of Eq. (4.35) and estimated the
values of the parameters: static gain A, dead time 4 and time constant,
B
4.30)
1 (4.36)
is the slope of the sigmoidal response at the
point of inflection (the point at which slope of the curve
changes sign)
ty = time elapsed until the
tent responded
1. For P-contnoliiers, use
Kn tlie 4.37
7 yl (4.37)
For Pl-contraliers,
(4.38)
(4.39)
(4.40)
(4.41)
(4.42)
Ir may be observed that the gain of the P/-controller is less
P-controller. This is due to the fact that the integral action makes the system
ve (may even lead to instability) and thus the gain value needs to be
The stabilizing effect of the derivative control mode allows the use of higher
ns in the PzD-controller (higher than the gain of P-and P/-controllers)
Ite, <= then FOPDT may be treated as FOP.
than that of the 116 Process Control: Principles and Applications
Ziegler-Nichols Tuning Method
Sposed by John G. Ziegler and Nathaniel B. Nichols, in 1942, this popula
method is based on frequency response analysis of the process. It is also known
as online tuning method or ultimate gain method or ultimate cycle method. This
method is primarily experimental and uses real process data from the sysiem’s
response. Unlike the process reaction curve method which uses data from the
open-loop response of a system, the Ziegler-Nichols tuning technique is a closed.
loop procedure. It employs the following steps:
1. Place the controller in the closed-loop with low gain; no reset and no
derivative contribution.
2. Using only the proportional control and with the feedback loop closed,
introduce a set point change and vary the Proportional gain until the system
oscillates continuously (Fig. 4.20). The frequency of continuous oscilla-
Hon is the crossover frequency, ,,. K,,is the gain that causes a system with
a P-only controller to be marginally stable (“u” stands for “ultimate"). The
ultimate gain is found by a trial and error process. Set K, to some low value
(K; and K, are 0 at this stage). Test the system with this K, to see if it
oscillates continuously (marginally stable). If the oscillationg decay, keep
on increasing X,, If the oscillations increase in amplitude (unstable
tem), reduce K,, Do this until the system is marginally stable,
3. Note the gain (ultimate gain, K,
shown in Fig, 4.20,
4. Using the values of K, and P,,
following tuning parameters for vi
sys-
») and period (ultimate period, P, = 27/0)
Ziegler and Nichols recommended the
‘arious modes of controllers:
Table 4.2 Tuning parameters according to Ziegler-Nichols method
Ky Ty Ty
P 05K, = =
PL 045K, | P,/12 =
PID 06K, [ P/12 PIS
The settings above reveal th,
1. For Prcontroller, use a loy
integral control mode
rationale of the Ziegler-Nichols met!
Wer proportional gain because
introduces additional phase
hodology,
use the presence of the
‘ag in all frequencies with
bp
Fig. 4, i
ig. 4.20 Conn oetton Zeger Meno ning Closed-loop Response 117
destabilizing effects on the system. Therefore, a lower K,, maintains ap-
proximately the same gain margin.
The presence of derivative control mode introduces strong stabilizing
fects on the closed-loop response. Consequently, the proportional gain K,,
of a PlD-controller can be increased without threatening, stability of the
system.
function is given by
s trans|
Examplg4.12 Fora unity feedback system, proces
eam :
—————. The controller is of PID mode. Calcu
(s+ 1)(s +5)
values of controller parameter based on ultimate cycle method of tuning.
te the optimal
Solution
Characteristic equation is: 1+ G, G, = 0
Using P-controller alone, the characteristic equation becomes:
9 +62 +5s+K,=0
Putting s = jo, the characteristic equation becomes
(oy + 6 (jay? + 5 (jw) + K,=0
(K, - 6 a) +j (S@- a) =0
Equating real and imaginary parts to zero, we have
@ = 5 or @, = (5)'” rad/sec = W,,
K, = 6 @ = 30 and P, = (277/a,) = 2.8099 sec
PlD-controller parameters are:
0.6K, = 18, 7,=P,/1.2 = 2.34 sec, and Ty = P,/8 = 0.35 sec.
6
i -tank system is ————___-.
Example 4.13. The transfer function ofa 3-tank system is es OGs + Hes)
Stroller is proportional and all other elements have unity transfer function.
Calculate the optimal values of P/D-controller parameters based on ultimate cycle
method of tuning.
Solution
Characteristic equation:
1+GH=0
; 6K, _
~ (s+ D(4s+D(6s+1)
= 48s? + 44s? + 128+ (1+6K,)=0 118
Putting s
Hence, @,
Optimal values of P/D-controller
k,
comommcnmmmmat RECAPITULATION | sescscmmmmnmm
Process Control: Principles and Applications
jewand equating real and imaginary parts to zero, we get
48a = 120, = @, = 0.5 rad/see
-Me +(1+6K,)=0 = K,=5B
3 and P, ~ (22/e,) = 12.56 sec
0.5 rad/see K,= 5
‘rameters are given by:
0.6K,= 1: 7,=P,/1.2= 10.46 sec; Ty = P,/ 8 = 1.57 sec
Closed-loop response of a process is the dynamic behaviour of the process
. final control element) fo:
control loop (including controller, measuring cleme
change in the set point or in the load/disturbances.
Error resulting from the step change in the load is termed as offset error, while
error resulting from the step change in set point is termed static error.
Integral controller alone or in combination with other controllers will alway:
result in zero offset. The destabilizing effect of integral action is reflected in it:
large maximum deviation and persisting deviation. Offset of P-controller woule
be more than that of PD-controller. Deviation from the set point will be minimurr
for PD-controller, and maximum for /-controller.
Controller tuning refers to “best” adjustment of controller parameters.
The most common criterion known as the one-quarter decay ratio criterion i:
used for tuning the controller parameters.
Based on the closed-loop response, minimum error integral tuning method:
minimize the area under the error curve using IAE, ISE, LATE criteria.
Process Reaction Curve (PRC) Method developed by Cohen and Coon is alse
called as open-loop method.
Ziegler-Nichols tuning technique is a closed-loop procedure.
memnemeesesess EXERCISES smscsomememsee
Review Questions
1.
2s
3.
4.
Define static error, offset and velocity error.
What is the advantage of using substitution method over Routh's criterion for
finding stability?
Name and write the expressions of the three methods used for tuning controllers
based on entire response of the process, and compare them.
Idemtify the controller mode with maximum and minimum (i) deviation from set
Point (ii) offset. Justify your answer. Closed-loop Response 119
Numerical Problems
4d
4.2.
4.3.
4.4,
4.5.
4.6.
47.
Liquid level system in which in flow is m, outflow w is through a pump. Draw the
process and block diagram of the controlled system. Prove that a level controller
configured as integral controller alone would lead to sustained oscillations,
In a liquid level control system, the tank is of area A (m2), inflow is m ( m/sec),
outflow is q, (m'Vsee), disturbance is downstream head u (m), linear resistance at
outflow pipe is R (sec/m*). Controller is P/. Measuring and final control elements
have unity transfer function
a. Draw diagram of the level control system.
b. Write expressions for error and controlled variable in terms of disturbance
and set point.
Draw block diagram of the process control loop.
Find static error and offset.
€. Derive expression for reset rate of Pi-controller for damy
closed-loop system to be 1/3.
Using direct substitution method, find stability range of K,, for the system hav-
4
10s +185? +8541
ng coefficient of
ing forward path transfer function as
and feedback path
\ 134
element transfer function as unity. Ans: (;