Design and Co-simulate control of an Extractive Distillation Column Using Aspen plus Dynamics with MATLAB and Simulink Toolbox Thesis Report ENG470 – Engineering Honours Thesis A report submitted to the School of Engineering and Energy, Murdoch University in partial fulfilment of the requirements for the degree of Bachelor of Engineering. Muhammad Syahmi Khairul Sham 1/28/2016 Unit Co-ordinator: Dr. Gareth Lee Thesis Supervisor: Dr. Linh Vu
85
Embed
Design and Co-simulate control of an Extractive ...researchrepository.murdoch.edu.au/id/eprint/30806/1/whole.pdf · Design and Co-simulate control of an Extractive Distillation Column
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Design and Co-simulate control of an Extractive
Distillation Column Using Aspen plus Dynamics
with MATLAB and Simulink Toolbox Thesis Report
ENG470 – Engineering Honours Thesis
A report submitted to the School of Engineering and Energy, Murdoch University in partial fulfilment of
the requirements for the degree of Bachelor of Engineering.
Muhammad Syahmi Khairul Sham
1/28/2016
Unit Co-ordinator: Dr. Gareth Lee
Thesis Supervisor: Dr. Linh Vu
I, Muhammad Syahmi Khairul Sham, submit this document to School of Engineering and Energy
completing the requirements of the undergraduate course at Murdoch University. I with this declare
that this thesis document is my own work except the idea of the project which is referenced.
Furthermore, this document has not been submitted to any other school or academic institution.
___________________________ Date: _____________
Muhammad Syahmi Khairul Sham
i
Abstract
This case study investigates the co-simulation of an extractive distillation column using Aspen Dynamics
together with MATLAB Simulink toolbox. This extractive distillation column separates Methyl Cyclo
Hexane (MCH) from Toluene by using input Phenol as a third component (entractant) to move the
ternary system beyond the azeotropic point. The study started with testing the steady state model of
the process in Aspen Plus; then continued with importing and testing the process dynamic model in both
manual and automatic modes using Aspen Dynamics. Finally, the process model in Aspen Dynamics was
connected to the built-in controllers in Simulink then the co-simulation of the controlled process was
performed using Aspen Dynamics together with the MATLAB Simulink toolbox.
The case study was an example taken from Aspen Dynamics version 8.4v. With the newest version of
Aspen Dynamics and Simulink version 8.4 operating platform Windows 7, it is required to install the 32
bit MATLAB to address compatibility issues between Aspen Dynamics and MATLAB.
The same control system design including four conventional controllers was implemented by Aspen
Tech in two different software package structures: in Aspen Dynamics stand-alone simulations and in
Aspen Dynamics – Simulink co-simulations to control the feed tank level, reboiler level, reflux drum level
and top stream pressure of column by adjusting feed 2 flowrate, coolant flowrate to condenser, bottom
(Toluene and Phenol) flowrate and product (MCH) flowrate. Then a new controller was developed in
Aspen Dynamics and co-simulation to control the product (MCH) purity by adjusting entrainer (Phenol)
flowrate. Advanced controller (DMC) has tried to be developed in co-simulation to replace PI controller.
However, attempts to develop DMC had failed after few trials.
All conventional controllers were tuned using auto tuning method in Aspen Dynamics using a special
tool, which is 'tuning' tool. It gives the best control parameters to achieve the best possible control
response. Set point changes and disturbance changes have been made to PI controllers and variables
respectively, and it is intended to investigate the effect on product purity. The new controller is very
helpful in improving the level of product purity. All run shows that Aspen Dynamics stand alone, or co-
simulation gives the same results in every test.
ii
Before developing Dynamics Model control (DMC) in co-simulation, DMC examples exercises from ‘ENG
420’ was implemented on a simple first order system in MATLAB to understand the basics of the
predictive control strategy along with the effect of design parameters.
All results obtained are discussed in Section Results and Discussion. Guideline for the next thesis
student has been outlined at the end of this report. Overall, most of the main objectives of this thesis
was achieved with very satisfying results. However due to unforeseen circumstances and time
constraints, DMC controller is not fully functional.
iii
Acknowledgements I welcome this opportunity to acknowledge, and to express my gratitude and appreciation towards my
supervisor, Dr. Linh Vu for her passion for helping, guiding and spend much time throughout this
project. I consider myself fortunate to be able working with her as she is entirely dedicated to giving
opinions and advice in preparing this report, and pushing me further to give my best effort. Also, I would
like to thank Will Stirling for his support and technical assistance during this project was carried out. He
had provided necessary facilities during the running of this project. Thank you also to all my colleagues,
especially Mohd Umair for his help and support during my project work.
Lastly and never forgotten, my deepest gratitude goes to my family for being my backbone throughout
my journey in completing this project.
iv
Table of Contents Abstract .......................................................................................................................................................... i
Acknowledgements ...................................................................................................................................... iii
1 Introduction and Layout of the Project ................................................................................................ 1
2 Background, Scope, and Aim of the Project ......................................................................................... 3
8 Future Work ........................................................................................................................................ 55
8.1 Review the Composition Controller and the Dynamic Matrix Control (DMC) ............................ 55
8.2 Implementing the Solvent Recovery Column ............................................................................. 55
8.3 Relative Gain Analysis (RGA) ....................................................................................................... 56
10 Works Cited ..................................................................................................................................... 60
6.2 Open Loop System Testing Aspen Dynamics Stand-Alone and Co-Simulation For the purpose of analysis, the system in the Aspen Dynamics and co-simulation was tested with the
open loop system. The testing of open loop system was done by implementing step on each
manipulated variable individually, and the deviation of 10% from steady state was used as the step
magnitudes. Table 8 shows the results of the purity product was obtained when the step was introduced
in manipulated variable.
Table 8: Purity Response (Open Loop System)
Step size Purity (Molar)
Phenol Flowrate (1000lbmol/lb) 60% 0.975
40% 0.83
Coolant Flowrate 60% 0.92
40% 0.9735
MCH Flowrate 60% 0.865
40% 0.9725
Toluene and Phenol Flowrate 60% 0.973
40% 0.973
The purity of the product was at the highest level of 0.975 molar when the step up was introduced in
phenol flowrate. The purity of the product was at the lowest level of 0.83 molar when the step down
was introduced in phenol flowrate. Most of the product purity above than the minimum purity (0.97257
molar) when the steps are introduced except when coolant Flowrate and MCH Flowrate were stepped
up, and phenol Flowrate was stepped down. Therefore, the closed loop system is developed to maintain
product quality.
6.3 Set Point Tracking Aspen Dynamics Stand-Alone and Co-Simulation The ability of feed tank level, reboiler level, reflux drum level, and top stream pressure for an extractive
distillation to track the set point changes were tested in Aspen dynamics and co-simulation using PI
control scheme, where the purity of distillate was run in open loop system. The testing of the controllers
was done by performing step on set points individually. The deviations of 5%, 10% and 25% from
the steady state were used as the step magnitudes. All tests were run with the same parameters as
shown in Table 6 in Section 4.4.1.3 Controller Tuning. The period of the simulation was kept constant
for all controller performance tests; 300 units.
In the other hand, the purposes of set point changes were implemented are to investigate the effect of
set point changes in each variable towards distillate and purity of the product. Figure 12(a)(b) and
Figure 13(a)(b) show the responses of the set point changes +10% in the Feed level and -10% in the top
34
stream pressure respectively, as they are most noticeable affect the distillate and purity of the product.
For the full results of the set point changes, please refer to Appendix C.
The minimum requirements to achieve excellent purity must be at least 0.97 molars, and purity will
consider as unsatisfactory if it is lower than the minimum requirement. From the tests that have been
carried out, the results obtained in Aspen Dynamics stand-alone and co-simulation is the same. The
results have been illustrated in Figure 12(a)(b), and it is clearly shows the purity of distillate is at the
minimum requirement with 0.97 molars when the step changes of +10% is introduced to feed level.
Besides, the change in feed level has affected the pressure, reflux drum and reboiler level but the
controller managed to control them back to the desired set point. The purity in Figure 13(a)(b) shows
unsatisfactory when the pressure set point was reduced by -10% because it is less than the minimum
requirement with 0.92 molars.
35
Figure 12(a): Step Change in Feed Level +10%
36
Figure 12(b): Step Change in Feed Level +10%
37
Figure 13(a): Step Change in Pressure -10%
38
Figure 13(b): Step Change in Pressure -10%
39
In conclusion, the results obtained in Aspen Dynamics stand-alone and co-simulation is the same. This
shows that the link is successfully made to connect the co-simulation. Furthermore, the performances of
the controllers in both strategies are at satisfactory levels.
6.4 Disturbance Rejection Aspen Dynamics Stand-Alone and Co-Simulation The ability of the controller in each variable to reject disturbance in the stream was tested using PI
controller. The method for performing disturbance changes was slightly different than the set point
tracking, where Phenol flowrate and coolant flowrate to the condenser were set as a disturbance in
order to investigate the effect of disturbances toward purity of the product. The feed level, pressure,
reflux level and reboiler level have been set to automatic mode when the disturbance in phenol flow
rate was introduced. The feed level, reflux level, and reboiler level have been set to automatic mode
when the disturbance in coolant flowrate to the condenser was introduced. The ±25% deviations from
the steady state were used for disturbance rejection. Also, the effects of disturbance toward the
distillate and the purity product were considered and analyzed. The minimum level of purity is 0.95
molar, which is the same as in set point tracking.
6.4.1 Phenol Flowrate The first disturbance rejection involving the Phenol flowrate was a step up of25%. This has affected the
reboiler level because the flow has been affected by the feed stream which remains at the initial steady
state and disturbance changes in phenol flowrate. Due to the change in the reboiler level, then it causes
changes in pressure, reflux level, distillate and purity of product.
Observe Figure 14(a)(b) where the controllers managed to eliminate the disturbance to achieve the
desired set point. Besides, although the purity of phenol was affected by the changes in the flowrates at
0 (s) to 20 (s), but the purity of product was below the minimum condition with result of 0.96 molar.
Overall disturbance rejection is as expected; the purity and amount of distillate are at the effective level.
40
Figure 14(a): Step up Phenol (entractant) Flowrate (Disturbance Rejection)
41
Figure 14(b): Step up Phenol (entractant) Flowrate (Disturbance Rejection)
42
Figure 15(a): Step down Phenol (entractant) Flowrate (Disturbance Rejection)
43
Figure 15(b): Step down Phenol (entractant) Flowrate (Disturbance Rejection)
44
The second disturbance rejection involving the Phenol flowrate was based on a step down of 25%. The
controllers on all variables managed to control and eliminate the disturbance when it was introduced in
the phenol flowrates. As can be seen in Figure 15(a)(b), the purity of the product (0.91 molar) is affected
when the step down was introduced in the phenol flowrate. The result was expected because as
discussed in Section 6.1 Sensitivity of Purity, the purity depends on the phenol flowrate. Therefore, the
purity of distillate is increase if the Phenol flow rate increased as shown in Figure 14(a)(b), and the
purity will decrease if the phenol flow rate is decreased as illustrated in Figure 15(a)(b). Therefore, the
changes in phenol flowrate must be considered in order not to affect the purity. Lastly, the feed flowrate
was not affected by the disturbance change in the phenol flowrate nor water input flowrate at the
pressure.
6.4.2 Coolant Flowrate The coolant flowrate to the condenser was used as the disturbance for this case, by using twenty five
percent steps up. This would inadvertently affect the reflux level and reboiler level. The response in
reflux and reboiler did overshoot at the beginning then they started to reach the desired set point after
10 (s). From the results obtained in Figure 16(a)(b), the purity of the product decreased to 0.86 molar
when the top stream pressure for an extractive distillation drops as it had affected the reflux ratio. The
reflux was changed to get the appropriate temperature at the top tray to get the absolute purity of
product. Some of the condensate in reflux drum is recycled back into the column, while the remaining is
discharged as top stream product. This becomes even worse when the flowrate of distillate products are
at a high level (3 kg/s) when the purity drops.
45
Figure 16(a): Step up Coolant Flowrate (Disturbance Rejection)
46
Figure 16(b): Step up Coolant Flowrate (Disturbance Rejection)
47
Figure 17(a): Step down Coolant Flowrate (Disturbance Rejection)
48
Figure 17(b): Step down Coolant Flowrate (Disturbance Rejection)
49
The final disturbance rejection involved the use of the coolant flowrate with the step down of 25%. The
purity of the product was reduced from 0.974 molar to 0.969 molar when the disturbance was
introduced. However, the purity is still considered good because it is above the minimum level of purity.
From the results obtained in Figure 16(a)(b) and Figure 17(a)(b), it can be seen that the purity of the
product produced a better response when the step down of disturbance was introduced compared to
the step up. However, the flowrate of distillate product has decreased from 2.6 kg/s to 0.7 kg/s when
the step down of coolant flowrate was conducted. The result was as expected because as discussed
earlier, the output stream of reflux drum is divided into two streams. The first stream is withdrawn as a
final product, and another stream is recycled back to the distillation column. Most of the condensate in
the reflux drum was frequently recycled for multiples times to get a better purity before the condensate
is withdrawn as a final product.
To sum up, the results of disturbance rejection is observed the same as in Aspen Dynamics stand-stand
alone and co-simulation. This proves that when the link between Aspen Dynamics and MATLAB Simulink
is done properly, then the co-simulation works successfully. The advantages of co-simulation are able to
develop advanced controller strategy such as Dynamic Model Control (DMC), and perform real time
simulation.
50
6.5 Effects of Phenol Flowrate on Product Purity As discussed in Section 6.1 Sensitivity of Purity, phenol flow rate plays a significant role in controlling
product purity. Besides, the results of the disturbance change of phenol flowrate in Section 6.4.1 Phenol
Flowrate has shown that the flow rate of phenol has affected the purity of the product. Therefore, a lot
of time was spent in constructing phenol flow rate controllers in Aspen Dynamics and co-simulation.
This caused some consequential delays on the overall progress, and generally too much time was
consumed comparing to what was planned previously in the ‘Gantt Chart’. The process of identifying
problem and solution resulted in the redesign of the controller code and configuration. After several
attempts, the closed loop system has been successfully developed.
PI controller with tuning parameters K = 32.38 and τ =13.2 has been used as a result of tuning, which
was obtained in Section 4.4.1.3 Controller Tuning. From Figure 18 it can be seen the product purity
achieved the desired set point and yielded good results. However, the results of phenol flowrates to
achieve the desired purity in Table 7 in Section 6.1 Sensitivity of Purity are not the same as the results
obtained in Aspen Dynamics and co-simulation. The comparisons between these results are as shown in
Table 9.
Table 9: Comparison between Phenol Flowrate in Aspen Plus, Aspen Dynamics, and Co-simulation
Aspen Plus Aspen Dynamics and Co-simulation
Status Case MCH Purity (molar) Phenol Flowrate (lbmol/hr)
Phenol Flowrate (lbmol/hr)
1 0.97257 1200 1200
2 0.97535 1300 1214.8631
3 0.97771 1400 1233.431
4 0.97971 1500 1255.5827
5 0.98143 1600 1280.9711
6 0.98292 1700 1308.7742
7 0.98422 1800 1338.1374
8 0.98534 1900 1367.7391
9 0.98634 2000 1397.9337
51
Figure 18: Step up Purity from 0.97257 to 0.98634 in Aspen Dynamics and Co-simulation
52
From Table 9 it can be seen that phenol flowrates in Aspen Plus require 2000 lbmol/hr to achieve
0.98634 molar purity of the product. However, Aspen Dynamics and co-simulation only requires
1397.9337 lbmol/hr of phenol flowrates to achieve 0.98634 molar purity of the product. Out of nine
status cases, only the first status cases have the same phenol flowrates to achieve 0.97257 molar purity
of product. At first, it was expected to get the same flowrate for all product purity.
As the results arisen with some problem, some assumption had been made. The first assumption was:
probably a mistake occurred while performing some changes during the redesign and configuration of
the controller code. The second assumption: maybe the dead time need to be installed on composition
controller, this is to avoid the limitation of the doable response in the system. However, a full analysis
was unable to be performed due to time constraints.
53
6.6 Developing Dynamic Matrix Control (DMC) Dynamic Matrix Control (DMC) has been trying to implement in co-simulation because it can perform
advanced controller. Initially, DMC examples exercises from 'ENG420' was implemented on a simple first
order system in MATLAB to understand the basics of the predictive control strategy along with the effect
of design parameters [25].
However, after repeatedly trying to develop DMC ended unsuccessfully because of the system was
showing a "linearization" error as illustrated in Figure 19. Further analysis was not done because of time
constraints and will be recommended for future work.
Figure 19: Error When Implementing DMC
54
7 Conclusion This section discusses the achievements and efforts made throughout this project. The aims will be
reviewed again, and achievements will be presented.
Studying and understanding the Aspen Plus, Aspen Dynamics and Co-Simulation has been done since the
beginning of this project, as they are sophisticated software. The results indicate the software used are
very helpful especially in developing and testing process simulation by providing a comprehensive
system.
Aspen Plus was used to obtain the results of the process model as it can predict the behaviour and
calculate the steady state value. The first task is to conduct sensitivity analysis in Aspen Plus. Steady
state simulation has given good results. The results show the flowrate of phenol was substantially
affecting the composition of the product. Increasing the phenol flowrate will increase composition of
the product. Therefore, the new controller has been developed to control the product composition.
It is possible for a complex system such as extractive distillation to perform dynamics simulation in
Aspen Dynamics. The new controller has been designed, and all controllers have been tuned to ensure
the best response is achieved. Phenol flowrate controller has been developed to control the quality of
the product. Tyreus Luyben tuning method was used because it is less aggressive than other tuning
methods. The special tool which available in Aspen Dynamics has been used to perform auto tuning, and
it has given the optimal parameter values. The set point changes and disturbance changes have been
introduced to the controllers and variables respectively.
After that, the Co-Simulation has been developed by linking Aspen Dynamics and MATLAB Simulink.
Steps to perform the link have been described in detail. Co-simulation are using the same controller
design and tuning parameters as in Aspen Dynamics. The set point changes and disturbance changes
have been implemented in co-simulation using the same method as in Aspen Dynamics. Then DMC has
been tried to be developed to replace PI controller as it is able to develop the advanced controller.
However, attempts to develop DMC had failed after a few trials. Further analysis was not done due to
time constraints.
Although failing to develop DMC in co-simulation, the co-simulation has its advantages, where it can
perform real time simulation, provide rigorous dynamic simulation and is able to optimize the process
performance. This project is a privilege for me to gain valuable knowledge in exploring the software. The
software is one of the most popular software in the industry and is often used by control engineers.
55
8 Future Work This section emphasizes on potential directions that should be of the concerns by the next students
moving on this project.
8.1 Review the Composition Controller and the Dynamic Matrix Control (DMC) From the results obtained in Section Sensitivity Analysis and Section Effects of Phenol Flowrate on
Purity shows that they have different flowrate to achieve desired MCH purity. In contrast, flowrate from
both sections should be the same to achieve the desired MCH purity. This problem is assumed that the
controller code configuration caused it. Besides, death time might be installed on composition controller
as it helps to avoid response limitation inside the system.
DMC was failed to be developed in this case study as the system was showing a "linearization" error.
Hence, this issue must be solved to ensure the process to be controlled by DMC. This error was probably
caused by the matter of setting and configuring the variables in MATLAB Simulink. If this problem could
be solved, comparison control strategies in Aspen Dynamics and co-simulation can be done through
performance criteria method.
Therefore, it is suggested that future student to try to solve this first before moving forward on any
other recommendations.
8.2 Implementing the Solvent Recovery Column As stated earlier, this project consists of two parts as a whole; the extractive column is part one, and
part two is the solvent recovery column. However, this project is focused on the first part, and it has
been tested with the steady state and dynamic process. In the second part, they have to separate the
product from part one, namely Toluene and Phenol (entractant). Therefore, Phenol (entractant) will be
circulated back to the extractive distillation column for reuse, while Toluene would be a product. The
steps to be taken on the second part is the same as in the first part where it needs to be tested by the
open-loop system, closed loop system, a step change, and performance criteria.
56
8.3 Relative Gain Analysis (RGA)
It should be noted that the control loop pairings in this project were chosen from a logical basis rather
than following the RGA approach. This was the case to ensure time was spent on design,
implementation, and analysis of the conventional controller achieved. Therefore, relative gain analysis
(RGA) is used to define the optimal input-output variable pairings. Through the use of a gain matrix, it is
possible to derive the use of the pumps and valves for the project about the impact each pump and the
valve has on each other.
For this section, the relative gain analysis will be based on the distillation column, reboiler, and
condenser. Before getting straight into the matrix calculations, the gains of the pump and valves must
be determined. This is done by stepping up each valve or pump individually. Stepping the valves or
pumps up separately allows the gains to be calculated at the temperatures, flows or levels that interact
with it.
To calculate the gains of each response, regression analysis is to be used. Regression analysis has been
used before in determining the mathematical models for the system. When each gain value is obtained,
they will be placed in a gain matrix. This gain matrix will then be inversed and transposed based on the
relative gain analysis equation, ( ) , where G is the gain matrix and T is the transpose of the
inverse matrix.
Once the matrix has been inverse and transposed, it will then be multiplied by the gain matrix. However,
this is not a normal matrix calculation as its element number will multiply each element. An example is
based on the matrix below, gain matrix. When conducting the multiplication, , will be multiplied by
the corresponding value in the transposed matrix, i.e. . This will occur for every element in the matrix
[26].
Figure 20: Gain Matrix (G)
57
8.4 Aspen Custom Modeler (ACM)
Aspen has created different types of software such as Aspen Dynamics, Aspen Plus, and Aspen Custom
Modeler (ACM). For future works, it is encouraged if future students can implement this project in
Aspen Custom Modeler because students can identify the software and compare this software other
types of software. Aspen Custom Modeler (ACM) is very different from the software used in this project,
where ACM is a standing alone software; in which it does not require other software to do steady state
process, dynamic process, and optimization because it can do it by itself. Therefore, this software is very
beneficial to students because it does not require different types of software to perform and test a
process. Apart from that, students can analyze and compare controller in Aspen Custom Modular with
software used in this project, whether it unnecessarily aggressive or too lagging.
58
9 Bibliography B. A. Ogunnaike and W. H. Ray, Process Dynamics, Modeling, And Control, Oxford, New York: Oxford
University Press, 1994
This book is very useful and often serves as a reference throughout the study at Murdoch University and
especially in this thesis. Among chapter that has been referred were chapter 14 (Feedback control
system) and chapter 15 (Conventional feedback control design) on page 461 and 513 respectively.
Chapter 14 had become the reference for a better understanding of the concept of feedback control and
conventional feedback controllers. Meanwhile, chapter 15 had been referred as to study the analytical
method of performance criteria of the controller. Through this method, the performance of the
controller can be identified, whether it is too aggressive or lagging when the set point change is made.
aspentech, "Jump Start: Getting Started with Aspen Plus V8," AspenTech, Bedford, 2015.
Aspen Plus is smart and sophisticated software. Therefore, this document has been used to understand
how it operates and the basic operation of the software. Apart from that, this document also describes
in detail ways to use this software, and this helps the reader to understand more about this software.