Anda di halaman 1dari 18

Introduction

Real world processes often can be modelled using computer, enabling simulation of the
process without using actual equipment. This greatly reduces the cost as the simulation can
be done multiple times. Computer generated simulation help process control engineers to
derive an effective controller without having to take risk of changing real process conditions
that could cause upset. Simulink is an interactive programming within Matlab for carrying
out dynamic simulation. In this experiment, Simulink is used to run the steady state analysis
and transfer function model is derived from frequency response data for a distillation
column. Besides that, closed-loop simulation using PI controller is also performed.

Section 3.1 and 3.2

The activities were completed according to the instructions given in the manual. The results
were not recorded.

Section 3.3

This section was run by using the Simulink model provided and followed the instructions
given in the lab manual. The system has two controlled variables which are Overhead MeOH
Composition (y d) and Bottom MeOH Composition (xd) and the input variables are Reflux
Ratio (R), Feed MeOH Composition (d1), Feed Flowrate (d2) and Vapour Flow rate (V). This
can be expressed as below:

The Reflux Ratio was adjusted while keeping the Vapour Flow rate constant at nominal value
and this gives the following results.

R yd xb R yd xb
1.75 (nominal) 0.84952 0.15037 0 0 0
1.85 0.86308 0.16159 0.1 0.01356 0.01122
1.95 0.87711 0.1728 0.2 0.02759 0.02243
1.55 0.82487 0.12294 -0.2 -0.02465 -0.02743

Now, the Reflux Ratio was kept constant at nominal value while the Vapour Flow rate was
adjusted and the results obtained as below.

V yd xb V yd xb
0.033 (nominal) 0.84952 0.15037 0 0 0
0.040 0.80637 0.01944 0.007 -0.04315 -0.13094
0.036 0.84336 0.08276 0.003 -0.00616 -0.06761
0.032 0.85199 0.17448 -0.001 0.00247 0.02411
In this case, d1 (Feed MeOH Composition) and d2 (Feed Flow rate) are ignored because
they are fixed for all the permutations of the system, which leaves only R and V as
variable of concern. Therefore, the equations can be rearranged to:

and .

When V is constant (ie. V=0), the steady-state gains of G11(0) and G21(0) can be obtained
from the slopes of the graphs of the both compositions of overhead and bottom vs R.
Likewise the steady-state gains of G12(0) and G22(0) can be obtained by constant the R (ie.
R=0). (See appendix for the graphs) Overall, the values of the steady-state gains for the
systems as follow:

Transfer Function Steady-state Gain


G11 0.1302
G12 -5.6157
G21 0.125
G22 -19.381

Section 3.4

Frequency Response Analysis menu was selected from mainmenu interface in which the
flowsheet and column strip will popped up on the screen. Firstly, the simulation of the
column was executed using default arrangement. Then, the nominal values of the overhead
and bottom composition were saved as other variables from MATLAB prompt. After that,
the rectangular pulse block was connected to Reflux Ratio channel and the workspace
module. The values used for rectangular pulse was applied as in table

Channel Reflux Ratio Vapor Flow Rate


Magnitude 4.5 0.15
Width 35 33

The simulation was started. After the simulation was completed, output vectors of oMeOH
and bMeOH were generated then normalised to a deviation vectors.

dxMeOH = xMeOH xnominal

where xMeOH is either oMeOH or bMeOH and xnominal is the nominal value of the
overhead and bottom composition of MeOH without rectangular pulse. Then bode plots
were generated by applying pre-made pcmfft function.

[AR, phi, ] = pcmfft(input, output, dt)


where input was rrpulse or vfrpulse, output was dxMeOH and dt was equal to 1. The
transfer functions were determine using previous values of steady state gain and bode plot
analysis for time constant, and time delay, .

The transfer functions for the system are shown below.

Section 3.5

From steady-state analysis, it showed that reflux ratio affects more on y d while the vapour
flowrate affects xb most. Therefore, the process can be approximated with

The control loop C1 (controlling y d by manipulating R) and C2 (controlling xb by


manipulating V) were designed and the proportional KC and integral KI value were derived
from the transfer function G11 and G22 by using Ziegler-Nichols and IMC based on PID control
setting table (week 6 lecture note). Hence, the value of KC and KI obtained as below.

KC KI
C1 14.54 0.018
C2 -0.1032 -0.00008256

The control loops were constructed to have negative feedback loop as instructed in the
guidelines.(See appendix for the block diagram) The value of K C and KI were applied and the
system found to be stable. Step changes in the feed composition were adjusted from 0.5 to
0.4 while keeping the set point of overhead composition constant as 0.85. After that, the
feed composition was set to be constant to 0.5 while the set point of overhead composition
was varied from 0.85 to 0.83. The simulation was run separately and the control error was
recorded and plotted. (See appendix for the control error plot).

From the error plot shown in the appendix, the system remained stable under the
disturbances. The control displayed control error within 0.08 of the desired set point. In
both responses to the inputs, a small degree of overshoot was detected before the system
stabilised.

Discussion

1. What are the ideal width and height for the rectangular pulse signal that you used in
section 3.4? Why?

Ideal width and height of the rectangular pulse signal can be determined by trial and
error. However, the pulse must have sufficient frequency components in order to excite
the process. For rectangular pulse, the dimensionless frequency t should be less than
5. Therefore, the width and height for G 11 and G21 were determined to be 35 and 4.9
respectively. On the other hand, the width and height for G 12 and G22 were determined to
be 33 and 0.15 respectively. The reason for the difference value in magnitude is because
the simulation would crash during the second round of experimentation.

2. Are the steady-state gains you obtained in section 3.3 at different steady-state
consistent? What does this imply?

The results obtained were consistent and R2 values for the linear correlation between
composition and input variables are more than 0.9. By comparison, the reflux ratio
seemed to be fairly proportional with overhead and bottom compositions while the
vapour flowrate showed weak linear relationship with bottom composition. The effect of
reflux ratio is more significant than vapour flowrate on bottom composition.

3. Why the Bode plots appear more noisy at the high frequency?

The noise at high frequency is the effect of the pulse. The rectangular pulse contains
more useful data at lower frequency than at higher frequency. Therefore, noise will
produce because it is impossible to separate the high frequency component from the
pulse.

4. Do you need to re-tune the PID controllers designed for individual loops such that they
can be used simultaneously? Why?

There is no need to re-tune the PID controllers as the value of KC and KI have been
calculated using either IMC or Ziegler-Nichols method. If direct synthesis method is
applied, then the PID controllers have to be re-tuning individually.

5. Should you use the derivative action? Why?


The derivative action is not suitable for this simulation because of the presence of high
frequency noise. This will prevent the derivative action from predicting disturbances that
occured in the process. Instead, noise would be amplified and will affect stability of the
system. To allow the use of derivative action, filter must be added into the system. Since
the system reached stability under proportional and integral action thus, the
implementation of derivative action and filter are pointless.

Conclusion

The experiment started by carrying out steady-state analysis of the distillation column
system to calculate steady state gains for each component in the system matrix. After that,
by running the frequency response program in the Simulink, Bode plots were constructed to
derive the approximated transfer functions for the distillation column system. Then, using
both results from steady state and frequency response analysis, Ziegler-Nichols and Internal
Model Control were applied to develop suitable PI controllers for the system. Finally, the
controller is added to the closed-loop simulation and was proven to be effective in
maintaining the stability of the process.
Appendix

The graphs obtained from Section 3.3.

For G11(0):

Overhead composition vs Reflux ratio


0.04 y = 0.1302x + 0.0009
0.03 R = 0.9989
Overhead composition

0.02

0.01

0
-0.3 -0.2 -0.1 0 0.1 0.2 0.3
-0.01

-0.02

-0.03
Reflux Ratio

For G12(0):

Overhead composition vs Vapor Flowrate


0.01

0
Overhead composition

-0.002 0 0.002 0.004 0.006 0.008


-0.01

-0.02

-0.03
y = -5.6157x + 0.0009
R = 0.9002
-0.04

-0.05
Vapor Flowrate
For G21(0):

Bottom composition vs Reflux ratio


0.03
y = 0.125x - 0.0016
0.02 R = 0.9969
Bottom composition

0.01

0
-0.3 -0.2 -0.1 0 0.1 0.2 0.3
-0.01

-0.02

-0.03
Reflux ratio

For G22(0):

Bottom composition vs Vapor Flowrate


0.04
0.02
0
Bottom composition

-0.002 -0.02 0 0.002 0.004 0.006 0.008


-0.04
-0.06
-0.08
-0.1
-0.12
y = -19.381x - 2E-06
-0.14 R = 0.9909
-0.16
Vapor Flowrate
Bode plot for section 3.4

G11
G21
G12
G22
Section 3.5

PI Controller Calculations

C1 Proportional and Integral Settings

By Ziegler-Nichols method

Calculation of Kc:

Kc = 14.54

Calculation of KI:

C2 Proportional and Integral Settings

G22 calculation using IMC Method:

K = -19.381, = 1250 s
Assume c = 0.5
KKc = / c
KKc = 1250 / 625
KKc = 2
Kc = 2 / -19.381
Kc = -0.1032

I = = 1250 s
KI = K c / I
KI = -0.1032 / 1250
KI = -0.00008256

1) Change in Disturbances:
Control Error Plot
Feed Composition Step Change
Bottom Control Error
0.02

0.015

0.01
e(t)

0.005

0
0 5000 10000 15000 20000

-0.005
time (s)

2) Change in set point:


Control Error Plot

Anda mungkin juga menyukai