Sei sulla pagina 1di 10

Methods of Controller Tuning

i)

Ziegler-Nichols Method

This method is used as a rule where the rule of heuristic PID produce good values for three PID
gain parameters. Those three PID gain parameters are the controller path gain, Kp, the
controllers integrator time constant, Ti and the controllers derivative time constant, Td.
From the measurements it is derived the two measure feedback loop parameters which are the
period of the oscillation frequency, Tu at the stability limit and the ultimate gain margin Ku for
loop stability. Using the characteristic equation of 1+ GOL = 1 +GvGmGpGc = 0
From the value of Ku and Tu that obtained from the derivation, the value of Kc,

I and

can be calculated by using the Ziegler-Nichols Tuning Chart:


Kcu

P control

K u /2

PI control

K u /2.2

T u /1.2

PID control

K u /1.7

T u /2

D
-

T u /8

The other way of using this method is by using the formula;


=

( Pu2 )
Pu
tan (
)
2
Pu

Pu
K K cu )21
(
2

ii)

Direct Synthesis

Direct synthesis method are based upon predescribing a desired form for the systems response
which then a controller strategy and parameters is found to give the response.
For first order process of direct synthesis controller strategy, the formula is given by

Kp
~
G=
p s +1
And setting as PI control it gives
K c=

p
K p c

Where i= p

As for first order plus time delay (FOPTD) Model, the standard model is given by
s

K p Ke
~
G (s)=
s+1

And the controller settings is


K c=

1
K + c

Where i=

iii)

Internal Model Control (IMC) Method

IMC method is similar like DS where the controller is based on the assumed process model and a
desired closed-loop transfer function which leads to analytical expressions for the controller
settings.
In order to get the values for the controller settings, 1/1 Pade approximation and the First Order
Taylor Series approximation can be used.
1) 1/1 Pade Approximation


1 s
2
s
e =

1+ s
2
D=

K c=

1
K

( )+1

( )+1
c

I= +
2

2
+1

()

2) First Order Taylor series approximation


s

=1s

K c=

1
K c +

I =

Selection of design parameter, c


In choosing the design parameter,
method. Generally, when
because

Kc

, it is a key decision for both DS and IMC design

increases, it produces more conservative controller. That is

decreased when I increases.

The guidelines for choosing the c

is as follows

1. c/ > 0.8 and c > 0.1(Rivera et al., 1986)


2. > c >
(Chien and Fruehauf, 1990)
3. c =
(Skogestad, 2003)
RESULT
i)

Ziegler-Nichols Method

By using the formula,


1 + GOL = 1 + GvGmGpGc = 0

--------------- (1)

Assuming Gm=Gv= 1,
Given Gp =
.

1
e0.5 s , which then divided by 3 to make in general form of , Gp =
2 s+3

K
s+1

This gives, Gp =

0.333
0.6667 s+ 1

0.5s
By using Pade approximation, e
=

Gp =

10.25 s
1+ 0.25 s

0.333
10.25 s
(
)
0.6667 s+ 1 1+0.25 s

Z-N PID Controller

0.3330.0833 s
2
0.6667 s + 0.917 s+ 1

(0.3330.0833 s) Kc
1+ 0.6667 s2 +0.917 s+1

From (1),

=0

Substituting s = i., and replace i2 = -1,


-0.167 2 + 0.917 i. +1 + 0.333Kc 0.0833(i.)Kc=0
Separating the imaginary and real terms,
(1+0.333Kc-0.1672) + (0.917-0.0833Kc)i=0
0.917-0.0833Kc=0
Kc=11.01
1.188

1+0.333Kc-0.1672

=5.29

=
=

2
Tu

=> Tu

Tu =

PV vs Time Graph (Ziegler-Nichols)

Choosing PID mode, the ZN table;

Kcu

Mode

K c /1.7

PID
17

6.476
3.238

T u /2

T u /8

0.594
0.297

0.1485
0.07425

Error (ISE)

Error (IAE)

Display

7.466E34
3.163

8.771E17
10.1

-1.644E17
1

The first value calculated results in oscillated graph (as shown in Figure 1a) with large value of
error. The graph oscillated with large amplitude therefore it is not stable. The value is then tuned
by dividing each value with 2 which gives ISE, IAE error responses of 3.163 and 10.1
meanwhile the PID controller response is 1. This value is stable as shown in Figure 1b as it
reached the set point value.
10

2
1.5

Z-N PID Controller

0.5
0
-0.5
-1
-1.5
-2

Figure 1a: The response of ZN PID controller before tuning


0 PV vs Time
50
100
150
(Ziegler-Nichols)

200

250

Time (seconds)

1.4
1.2
1

0.8

ii)

0.6

Figure 1b: The response of ZN PID controller after tuning

Direct
0.4 Synthesis Method

In order to get 0.2


the values of Kc and the PI controller, the First Order Plus Time Delay (FOPTD)
formula is used.
0

50

100

150

Time (seconds)

200

250

K c=

1
K + c

and

I =

From the transfer function, K = 0.333, = 0.667, = 0.5;


Value of c

based on Rivera et als theory, c/ > 0.8 and c > 0.1. It is calculated that c needs

to be bigger than 0.4.

c
0.5
0.6

Kc
2
1.818

Error (ISE)
2.045
2.098

Error (IAE)
4.5
4.5

Display
1
1

With c of 0.5, it gives the value Kc of 2 and it gives a smaller ISE error compared to when c is
0.6.

Figure 2:
response

The
of DS
PI controller with Kc equals to 2.

iii)

Internal Model Control (IMC) Method

In order to get the value Kc, I and D , Pade approximation is used. The formula is;

With K = 0.333, = 0.667, = 0.5,

Kc=

1
K

2
2

( )+1 = + =
c
+1

( )

2
+1

()

1=0.917 and D=0.182

c
0.5
1
1.5

Kc
3.667
2.2
1.571

Error (ISE)
1.627
1.636
1.779

At c
the
gives
value
2.2
the
error

shown when response has approached the set point.

Error (IAE)
3.394
3.277
3.278

Display
1
1
1
of 1,
Kc
the
of
has
least

Figure 3: The graph of IMC PD controller at c of 1 and Kc of 2.2

DISCUSSION
The methods of controller tuning are Ziegler-Nichols (ZN), Direct Synthesis (DC) and
Internal Model Control (IMC). With the given transfer function and simulation stop time of 250
second, the study is to observe and determine which method gives the best optimum value
according to the response and error value obtained.
The first method of controller tuning is Ziegler-Nichols. For Zigler Nichols method we
used PID controller to obtain the response. We used 1+G OL = 0 to find the value of K c and Tu and
using both value to find K , I , and D respectively. From calculation, value of K = 11.01
cu

and Tu = 1.188. Based on the Figure 1a, the graph that we obtained was constant at zero before
oscillated start at 200s. The oscillation increases with time. Thus, the process did not achieve
stable response as it was not at its optimum value, 1 and resulting the error to be huge in number
with 7.466E34 and 8.771E17 using ISE and IAE respectively. Therefore, tuning is needed to be
done to achieve a stable response. Refer to Figure 1b, tuning has been done and it showed that
the graph reached it setlling time around 50s with a slightly overdamped and achieved stable
I
response at 1. The errors calculated were 3.163 for ISE and 10.1 for IAE. The value of K
cu,

and D for figure 1b was half the value of figure 1a.


The second method of controller tuning that we used was Direct Synthesis (DS). A PI
controller is used in this method because the process was first order. The values of K = 0.333, =
0.667 and = 0.5 were taken from the transfer function. Using a given equation with different
values of c,, the value of Kc can be calculated. Values for c were assumed above 0.4 as we used
Rivera et als theory. The errors calculated for K c = 2 were 2.045 and 4.5 for ISE and IAE
respectively. Meanwhile, when Kc = 1.818 the errors were 2.098 and 4.5 for ISE and IAE
respectively and both values of Kc achieved the optimum value, 1 with slightly overdamped
before achieving the stable response but Kc = 2 was more desirable as the error was small
compared to Kc = 1.818.
The last method of controller tuning that has been used was Internal Model Control
(IMC). A PID controller was used in this method. Before calculating the values of K , I and
c

D , Pade approximation was used and values of were assumed with values of 0.5, 1 and
c
1.5. Thus, values of Kc calculated were 3.667, 2.2 and 1.571 respectively and the lowest error
obtained for ISE was 1.627 for Kc = 3.667 and error for ISA was 3.277 for Kc = 1. Regarding of
that, the graph obtained stable response of optimum value, 1 with slightly overdamped for all
three values of Kc but the desirable one was Kc = 2.2 as the average error value for it was the
smallest compared to other two Kc values.
In a nutshell, the best method that gives least error is IMC compared to Z-N and DS. The
larger the process Dead-Time ( D relative to the process Time Constant, the more difficult
the associated process will be to control. As the

exceeds the Time Constant, the speed by

which the controller can react to any given change in that same process is significantly
decreased. That undermines the PID controller's ability to maintain stability. Even though DS did
not have delay time but the error was still big compared to IMC since the controller used was PI
since it was a first order reaction and PID controller is to maintain stability. Thus, IMC method
reached stability much faster than DS. Meanwhile, even though Z-N and IMC both used PID
controller, Z-N error was bigger than IMC because Z-N method leaves the loop with very little
robustness, which can lead to loop instability. The rules are very sensitive to an accurate
measurement of dead time, which is difficult on lag-dominant processes with short dead times.
Thus, Z-N is a poor choice for process stability.
RECOMMENDATION
As to get the optimum value for lessen the error for any PID tuning, firstly select the tuning rule
according to the process and operating objective. The transfer function given might be one of the
factors for the different kind of response. Therefore, the method of getting the optimum value
might be different. To get more stable response, PID controller is recommended as it has the
ability to maintain the stability. By testing which method is suitable, there will be needing some
trial and error method in order to observe which method has the best response. When tuning is
properly done, loop balances the demands of stability, responsiveness and produces low
overshoot. By adjusting the three tuning parameters, the loop can responds well and produces a
steady-state situation.

CONCLUSION
The methods of controller tuning are Ziegler-Nichols (ZN), Direct Synthesis (DC) and Internal
Model Control (IMC). From the PID controller diagrams, the integral of squared error (ISE) and
the integral of absolute value of error (IAE) is obtained from running the simulink. Based on the
calculated values, and with some tuning when using ZN method, it shows that all of the response
have reached the set point. From the comparison of all methods used with the selection of c, it
shows that IMC has the least ISE and IAE compared to ZN and DS response. This is due to the

time delay which adds the stability to the loop. As the conclusion, all of the methods can be used
with the given transfer function and IMC has the best response.

Potrebbero piacerti anche