ENG450 - ENGINEERING INTERNSHIP FINAL REPORT by Isaac Correia Supervised by: Dr Parisa Bahri – Academic Supervisor Kristian Sikora – Industry Supervisory A thesis submitted to the School of Engineering and Energy, Murdoch University in partial fulfilment of the requirements for the degree of Bachelor of Engineering 2008
73
Embed
ENG450 - ENGINEERING INTERNSHIP FINAL REPORT · 2 C hapter 1 1. Introduction This is a report covering the internship carried out at Alcoa’s Wagerup refinery commencing on the 11
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
ENG450 - ENGINEERING INTERNSHIP FINAL REPORT
by Isaac Correia
Supervised by:
Dr Parisa Bahri – Academic Supervisor Kristian Sikora – Industry Supervisory
A thesis submitted to the School of Engineering and Energy, Murdoch University in partial fulfilment of the
requirements for the degree of Bachelor of Engineering
2008
We are satisfied with the progress of this internship project and that the attached report is an accurate reflection of the work undertaken. Signed: Industry Supervisor Signed: Academic Supervisor
MURDOCH UNIVERSITY
ALCOA WORLD ALUMINA, AUSTRALIA
ABSTRACT
This report covers the work undertaken during the internship at Alcoa’s Wagerup
refinery performed throughout the period of 11th August to 5th November 2008.
The report will cover the work conducted during the internship, primarily
focusing on the main project. The aim of the project was to conduct a
performance of all control loops, instrumentation and code in precipitation and
fix issues where possible in preparation for the implementation of an advanced
controller.
The project was split into two parts, the software review which involved the
analysis of all necessary programs and a loop analysis to analyse the performance
of the controllers using performance indices such as rise time, settling time, decay
ratio and Integral Square Error (ISE). Saturation time and mode analysis was also
required and the incorporation of the valve characteristic plot allowed for an
effective way to detect instrument problems such as valve stiction.
The review has so far covered 90% of all the necessary loops required to be
analysed. Of those analysed so far over 65% exhibited less than satisfactory
performance
Acknowledgments
I wish to thank academic supervisor Prof. Parisa Bahri for making this internship
possible, and also for her continual help and guidance along the way.
I would also like to thank Kristian Sikora, for his guidance during this time and
for providing the opportunity to work in an industrial environment.
Special mention should go to the operations and maintenance staff at Operations
Centre 3 (Precipitation) at the Wagerup alumina refinery.
i
TABLE OF CONTENTS
Acknowledgments ............................................................................................................... ii Table of Contents ................................................................................................................. i 1. Introduction ........................................................................................................... 2 2. Background ............................................................................................................ 3
2.1. Alcoa World Alumina, Wagerup Refinery ................................................ 3 2.2. The Bayer Process ......................................................................................... 3 2.3. Overview of Precipitation ............................................................................ 5
3. Overview of Major Project ................................................................................. 7 3.1. Tools/ Inventory............................................................................................ 8
4. Software Review .................................................................................................. 10 4.1. Total Bank Flow Calculation ..................................................................... 10
4.1.1. Total Green Liquor flow down the bank ................................ 12 4.1.2. Total Coarse Seed flow down the bank .................................. 15 4.1.3. Total Tray Seed flow down the bank ....................................... 20 4.1.4. 50E Return Filtrate to Precipitation ......................................... 24 4.1.5. Summary........................................................................................ 26
5.2.1. Rise Time ...................................................................................... 29 5.2.2. Settling Time ................................................................................. 29 5.2.3. Integral Square Error .................................................................. 30 5.2.4. Control Valve Issues ................................................................... 31
5.3. Presentation of Data.................................................................................... 37 5.4. Methodology ................................................................................................. 39 5.5. Analysis of Inter Stage Cooler Controllers ............................................. 43 5.6. Analysis of Green Liquor Controllers ...................................................... 46 5.7. Analysis of Coarse Seed Controllers ........................................................ 53
6. Conclusion ............................................................................................................ 62 6.1. Progress and Future Work ......................................................................... 62 6.2. Problems Faced ............................................................................................ 63
2
C h a p t e r 1
1. Introduction
This is a report covering the internship carried out at Alcoa’s Wagerup refinery
commencing on the 11th August 2008 and ending on the 28th November 2008.
The objective of the internship is to “provide the intern exposure to the applied world of
engineering design or research through a period of workplace employment in a relevant area of
engineering” (Murdoch University, 2008). While at Wagerup, the intern was part of
the Process Control Team which is a division of the refinery’s Technical
Department. This team is responsible for maintaining the DCS and all control
applications at the refinery. During the course of the internship, the intern was
granted many responsibilities typical to those of a qualified instrumentation and
control engineer. These included the building and modification of operator
displays, control scheme modifications, documentation development and
controller tuning.
This report will concentrate on the major project undertaken during the
internship.
The aim of this project was to ensure all control loops and instrumentation in
precipitation are performing satisfactorily before the installation of an advanced
controller. This required conducting a review of all control loops, instrumentation
and code that will become apart of the new controller and fix problem where
possible. Once the review is done, a report documenting the problems found will
be provided to the refinery in order to repair such issues.
The intern was also part of the team commissioning secondary feed system.
Details on this project are available in Appendix 1.
3
C h a p t e r 2
2. Background
2.1. Alcoa World Alumina, Wagerup Refinery
The Wagerup refinery is one of three Alcoa refineries in Western Australia. It was
first opened by the Premier of WA, the Honourable Brian Burke in 1984 and
today employs 900 personnel at both the refinery and the minesite.
The refinery is located approximately 130km south of Perth, and is in close
proximity to the Willowdale minesite which supplies the bauxite ore to the
refinery. Its current production capacity is 2.6 metric tons of alumina per year.
The alumina produced is exported to aluminium smelters around the world via
the Bunbury port, which is one of two Alcoa ports in WA.
The refinery is heralded as “the most environmentally advanced alumina refinery in the
world, with the application of advanced technology contributing to recent improvements in
environmental management.” (Alcoa Inc. 2008)
2.2. The Bayer Process
The Wagerup refinery uses the Bayer process to refine the alumina out of bauxite
ore. The Bayer Process was first developed in 1888 by a German chemist Karl
Bayer. It works on the principal that “a hot solution of liquid caustic soda can dissolve the
alumina from bauxite and the alumina will remain in the solution as long as the solution stays
above 100°C.” (Goodman, A. 2003). The Wagerup refinery uses a slightly altered
version of the original Bayer process but the four key stages still remain;
digestion, clarification precipitation and calcination.
The following is an overview of the process used at the Wagerup refinery. Note
that the Precipitation process will be described in greater detail in the following
chapter. (Goodman, A. 2003):
4
1 The process is basically a continuous recycling caustic soda stream of liquor,
which starts and ends at the Spent Liquor Storage tanks (or 30A tanks). At
this point, the liquor is “spent” liquor (i.e.: the liquor is not saturated with
alumina). From here, two streams of spent liquor leave the tanks; one to
milling (known as liquor to milling or LTM) and the other to digestion
(known as liquor to digestion or LTD)
2 Bauxite ore is then mixed with the LTM and fed into a ball mill where the ore
crushed and ground in a ball mill. The product from the mills is slurry of
caustic and bauxite which it then pumped into the digestion building.
3 Digestion: The slurry is mixed with liquor heated to 150°C and put through a
series of five tanks called digesters. The digesters maintain the liquor at a high
temperature and pressure in order to dissolve the alumina from the bauxite in
the liquor. Once it has left the digesters, the pressure in the lines is reduced to
atmospheric and the slurry is cooled to 104°C. The liquor solution leaving the
digesters are now saturated with alumina is now known as “green” liquor.
4 Clarification: This stage of the process is where solid residues (sand and mud)
are removed from the green liquor. The green liquor first enters through the
bottom of a balloon shaped pressurized vessel know as a sand trap. The
coarse sand, being heavier, falls to the bottom of the tank and is pumped to
residue disposal whilst the remainder leaves through outlets at the top and
enters into two large flat tanks known as thickeners. Here the liquor is mixed
with flocculent which combines the small mud particles together to form
larger ones and settles to the bottom of the tank. The mud is then raked to
the centre of the tank where it is sent the mud washers for further processing.
The rest of the green liquor leaves the thickeners and enters the filtration
building where the last portion of the mud is filtered out before entering
precipitation.
5
5 Precipitation: This stage of the process basically the extraction of dissolved
alumina from the green liquor in the form of alumina trihydroxide, better
known as hydrate
6 Calcination: The hydrate from precipitation enters the calciners. Here the
hydrate is basically “cooked” at temperatures exceeding 1000°C to break the
tight bonds between the water molecules and the alumina of the hydrate and
releasing the water as steam, leaving behind the alumina crystals. This alumina
product is cooled and railed to the Bunbury port as product.
2.3. Overview of Precipitation
The project that is described in this report is based in the Precipitation area of the
plant. A graphical overview of the precipitation process is depicted in Appendix
2 of this report.
Before the green liquor is sent to precipitation, it passes through the heat
interchange building where the green liquor leaving 35X (filtration building) is
cooled down from 100°C to 80°C, the desired temperature required to begin the
alumina re-crystallisation process. The green liquor is then sent to Building 45
where precipitation begins.
From the 45A tank, the green liquor is usually sent to the first tank of each bank.
There a total of four banks, each with twelve Precipitator tanks in series. These
banks are then split into 2 units where banks 1 and 2 are unit 1 and banks 3 and 4
are unit 2. The figuration of the banks is shown in the Figure 2-1.
6
10 20
11 21
12 22
13
14
15 25
23
24
40 30
41 31
42 32
43
44
45 35
33
34
50 60
51 61
52 62
53
54
55 65
63
64
80 70
81 71
82 72
83
84
85 75
73
74
Bank 2 Bank 3 Bank 4Bank 1
10 20
11 21
12 22
13
14
15 25
23
24
10 20
11 21
12 22
13
14
15 25
23
24
40 30
41 31
42 32
43
44
45 35
33
34
40 30
41 31
42 32
43
44
45 35
33
34
50 60
51 61
52 62
53
54
55 65
63
64
50 60
51 61
52 62
53
54
55 65
63
64
80 70
81 71
82 72
83
84
85 75
73
74
70
81
82
Unit 1 Unit 2
10 20
11 21
12 22
13
14
15 25
23
24
10 20
11 21
12 22
13
14
15 25
23
24
40 30
41 31
42 32
43
44
45 35
33
34
40 30
41 31
42 32
43
44
45 35
33
34
50 60
51 61
52 62
53
54
55 65
63
64
50 60
51 61
52 62
53
54
55 65
63
64
80 70
81 71
82 72
83
84
85 75
73
74
80 70
81 71
82 72
83
84
85 75
73
74
Bank 2 Bank 3 Bank 4Bank 1
10 20
11 21
12 22
13
14
15 25
23
24
10 20
11 21
12 22
13
14
15 25
23
24
40 30
41 31
42 32
43
44
45 35
33
34
40 30
41 31
42 32
43
44
45 35
33
34
50 60
51 61
52 62
53
54
55 65
63
64
50 60
51 61
52 62
53
54
55 65
63
64
80 70
81 71
82 72
83
84
85 75
73
74
70
81
82
Unit 1 Unit 2
Figure 2-1: Precipitator Bank Configuration
The precipitators in each bank decrease in height as you go down. This allows the
green liquor to be gravity drawn, rather than pumped to the next precipitator
effectively, thus reducing energy usage. The green liquor travels from one
precipitator to the next through open troughs known as launders. As it goes
through the precipitator bank, it is cooled and agitated which allows some of the
alumina crystals to precipitate.
In order to and speed up this process and effectively control the size of the
crystals, seed crystals are added to the first and second tank of each bank. These
seed crystals are fine crystallised material recycled after it’s gone through the
precipitators because it’s too small to be processed into alumina. The seed acts as
nuclei for the dissolved alumina to precipitate onto.
The hydrate produced after leaving the last precipitator is then classified through
cyclones. The larger particles move on to calcination whilst the smaller particles
are sent through to the filtration building (building 44-1 and 44-2) which filters
the seed from the spent liquor. The small hydrate particles are recycled back as
seed whilst the spent liquor is sent to evaporation and recycled to digestion.
Author
Move to Appendix 1
7
C h a p t e r 3
3. Overview of Major Project
Alcoa embarked on a process to install multivariable controllers in its refineries
around the world approximately 8 years ago. Since then, those projects have
demonstrated significant benefits to the business. Alcoa has now started the next
phase of this work, and as part of that, Wagerup will be getting a new controller
for Precipitation.
This controller will be implemented to help maximise production through yield
and flow improvements. The final detailed design has not yet been completed,
but the conceptual design has identified the general inputs and outputs to be
used.
Advanced Process Control (APC) projects are a significant investment in time
and money. Traditionally, the controller design began before any detailed
instrumentation review was carried out. This resulted in some projects being
delayed with large cost blow-outs. To reduce the time required to implement the
Precipitation controller and to maximise the probability of success, it was decided
to review the plant before the project was formally started.
The main internship project was to perform this review, and to provide
recommendations to the plant for maintenance of the physical equipment as well
as any software changes required.
To do this, the project was split into to parts:
1) Software Review
2) Loop Analysis
The software review involved going through all necessary programs,
understanding the code then documenting what it does, how it works; any errors
found and suggested changes to be made.
8
The Wagerup refinery, along with all Alcoa refineries, has incorporated programs
which use Control Language (CL) code. CL is a Honeywell proprietary
programming language developed specifically for control applications on
Honeywell DCS. Programs in CL can be used to perform automatic calculations
based on the point tags contained in the DCS.
The Loop analysis involved conducting performance analysis on all control loops
and their related instrumentation.
3.1. Tools/ Inventory
DOC4000
DOC4000 is the refinery’s database for management and documentation of the
Honeywell DCS system. It was used extensively during the internship for:
• Identification of tags for controllers and instruments
• Identifying the manipulated and process variables of the controllers using
Map Builder
• review of CL code using its CL viewer
Opertune
Opertune is an automated, closed loop, PID tuning technology for Experion
CEE- based controllers (embedded) and any PID controller. It was used to assist
with the tuning of any control loop deemed necessary during the review.
Uniformance Process Trend
Uniformance Process Trend is a desktop trending and analysis tool used to
retrieve and view data from the refinery. It is tag-based, meaning all information
retrieved and viewed is based on the tags selected and how they are customised
by the user.
Process Trend was used to assist with the review and tuning of base level control
loops, in order to identify if the controller needed tuning, the performance of the
9
controller as well as disturbance rejection. It was also used for the
instrumentation review. By analysing the trend of the process, manipulated and
disturbance variable(s), possible instrument faults can be identified for further
investigation.
Loop Scout
Loop Scout sits on the GUS and allows for fast data recording over a specified
time period which is then sent to the server then sends back a report to the
refinery. This report provides:
• An analysis of the controller performance or behaviour (ie: aggressive
control action, etc).
• Advice on how to improve the performance of the controller.
• PV vs OP plot which can be used to identify problems such as valve
stiction.
10
C h a p t e r 4
4. Software Review
The first task was to conduct a review of all CL programs that would become a
part of precipitation APC. This involved providing a thorough description of the
code’s operation and its purpose. The strategy used was as follows:
1) Split the code into relevant sections.
2) Research the process that each section of code covers. This required
investigating the inputs and outputs of the code, the importance and
sources of the read in reading in and/ or writing to. This can be done
using DOC4000.
3) The functionality of each section was then defined. Code was also
checked for errors and irrelevant lines of code.
There were a number of programs that the software review needed to cover. The
main three programs were:
• Total Bank flow calculation
• Current Seed Charge Calculation
• Total Green Liquor to Bank
Each of these programs in which the code was reviewed in a similar manner, and
therefore this report will only cover the review of the total bank flow calculation.
4.1. Total Bank Flow Calculation
The purpose of the total bank flow calculation is to determine the total input flow
into each bank. As stated before in section 1.5, the green liquor flows from tank
to tank through launders which are situated at the top level of the precipitator
banks. If the total flow into these banks increases above a certain limit, the
11
launders will overflow, resulting in flow losses as well as a potential hazard for
personnel working in the lower levels.
There are four CL programs which calculate the total bank flow, one for each
bank. These are:
• F1BANK1
• F1BANK2
• F1BANK3
• F1BANK4
The code for each of these programs is exactly the same with only the tag names
changed to coincide with the specified bank. For this reason, the analysis was
only conducted on F1BANK1.
The analysis began at the very end of the code. It was here that the flows that
form the total bank flow and the coefficients that make up their calculation were
identified. Working backwards from here it is possible to find how and where
each of the coefficients was defined. The final calculation for the total bank flow
is given below:
-- calculate total bank flow SET total = lb1 + sb1 + tsb1 + 50e
The variable, total, is the total bank flow. The other variables are defined as
follows:
-- calculate liq flow down bank SET lb1 = lp10 + lp20 -- calculate total seed flow from 44-2 down bank SET sb1 = sp10 + sp20 -- calculate total tray seed flow down bank SET tsb1 = tsp10 + tsp20
12
From this, it was determined that there were four sections
• Total green liquor flow down the bank
• Total seed flow from 44-2
• Total tray seed flow
• 50E return flow.
The results are then stored in the following locations.
-- store results
SET ftotb1.PV = total
SET fyglp10.PV = lp10
SET fyglp20.PV = lp20
SET fyglb1.PV = lb1
SET fysp10.PV = sp10
SET fysp20.PV = sp20
SET fxtsp10.PV = tsp10
SET fxtsp20.PV = tsp20
SET fysdb1.PV = sb1
SET fytsdb1.PV = tsb1
Before the code could be analysed it became apparent that the process needed to
be fully understood. Therefore, discussions with other process control engineers
were conducted and schematics were developed for each of the 4 sections. The
following sections of this chapter will cover the analysis of each of these sections
of code in detail.
4.1.1. Total Green Liquor flow down the bank
This section deals with the part of the program which calculates the total green
liquor flow to the precipitator banks. An overview of the green liquor circuit is
provided in Figure 4-1.
13
45 A
60 50 80 70
FCLP60
10 4020 30
FCLP10
50 804010
FCLP10S FCLP40S FCLP80SFCLP50S
FCLP20 FCLP40 FCLP30FCLP70FCLP50 FCLP80
Line 2 – Main GL Line, Unit 1 Line 1 – Main GL Line, Unit 2
Spare GL Line, Both Units
Bank 1 Bank 2
Bank 3
Bank 4
Bank 4
Bank 3
Bank 2Bank 1
45 A
60 50 80 70
FCLP60
10 4020 30
FCLP10
50 804010
FCLP10S FCLP40S FCLP80SFCLP50S
FCLP20 FCLP40 FCLP30FCLP70FCLP50 FCLP80
Line 2 – Main GL Line, Unit 1 Line 1 – Main GL Line, Unit 2
Spare GL Line, Both Units
Bank 1 Bank 2
Bank 3
Bank 4
Bank 4
Bank 3
Bank 2Bank 1Figure 4-1: Green Liquor Circuit to Precipitation
The green liquor (GL) circuit begins at the 45A tanks. From here, the green
liquor is pumped into the first precipitator of each bank. There are 3 GL lines:
two main lines (lines 1 and 2) and one spare. Line 1 supplies GL to unit 2 whilst
line 2 tends to unit 1. The spare line can supply to both units. This line is usually
used when one or both of the main lines are taken off circuit for maintenance.
When the whole precipitator is taken offline, the GL is rerouted to the second
precipitator on the main line via the isolation valve to that precipitator.
As shown in the previous section, the equation for the total GL to bank 1 was
given as:
SET lb1 = lp10 + lp20
Where lp10 is the GL flow to liquor flow to precipitator 10 and lp20 is the GL
flow to precipitator 20. The values of both of these variables are defined in the
code shown below:
SET lp10 = ( WHEN fsglp10.PV = on_line AND
& fsglp10s.PV = on_line : fclp10s.PV + fclp10.PV;
14
& WHEN fsglp10.PV = on_line : fclp10.PV ;
& WHEN fsglp10s.PV = on_line : fclp10s.PV ;
& WHEN OTHERS : 0 )
SET lp20 = ( WHEN fsglp20.PV = on_line : fclp20.PV ;
& WHEN OTHERS : 0 )
Where:
• Fsglp10 = Green Liquor to Precipitator 10 Line status
• Fsglp10S = Green Liquor to Precipitator 10 Spare Line Status
• Fsglp20 = Green Liquor to precipitator 20 Line status
• Fclp10 = Green Liquor flow through main line to Precipitator 10
• Fclp10s = Green Liquor flow through spare line to Precipitator 10
• Fclp20 = Green Liquor flow through line to Precipitator 20
The code filters the GL flow readings based on the status of the line. The
Figure 5-8: The trend and Valve Characteristic of FCWISC11
Other performance indicators also suggested excellent control. Over a period of a
month, the controller was in its normal mode 93.62% of the time and was only
saturated 33.88% of the time, which is quite significant. Overall performance
however was excellent and thus no action was recommended to be undertaken.
In contrast, FCGLPC1 which controls one of the GL flow from the third
precipitator, is a classic example of poor performance. Although it is never
saturated, the controller is hardly ever in its normal mode and has a stiction
probability of 100% and this is easy to see when examining the plots in Figure 5-
9.
45
Figure 5-9: The trend and valve characteristic of FCGLPC1
There are two problems here. Firstly the flow reading is very noisy and since the
controller action (OP) is fairly smooth, it does indicate issues with the flowmeter.
Furthermore, it is very clear that there is no relationship between the flow and the
OP. Not only by the shape of the valve characteristic, but also by the fact that the
OP resembles a saw tooth function and the flow is a square wave. This is the
typical signature of stiction. To confirm this, the valve stroking test was
conducted. The test was conducted with someone out in the field to observe the
valve movement with a walkie talkie in hand to communicate with the operator in
OC3 and notify them to conduct the step change. Although the valve did move
for the initial step from 55% to 25%, it would not move any further. The OP was
stepped from 25 to 30%, then to 35% then to 40%, each time the valve did not
move. Thus the valve is severely jammed and a recommendation for it repair was
made.
Although this was the most severe case of stiction observed, similar results on the
other GL flow to interstage cooler controllers all tested positive for stiction. The
reason for this is the fact that the setpoints of theses controllers are hardly ever
changed, thus once the PV is at setpoint, the controller doesn’t need to move the
valve that much to keep it there. This allows scale to build up within the valve
quicker.
46
5.6. Analysis of Green Liquor Controllers
There are a total of 16 green liquor flow controllers in precipitation. Of those:
• 2 exhibited Excellent performance,
• 3 exhibited Fair performance
• 3 exhibited Ok performance
• 2 exhibited Poor performance, whilst
• 6 had not been in use in the last year
Referring to Figure 5-1 of this report, there are 12 main control loops that feed
GL to the precipitators, split evenly between 3 lines. This report will cover the
analysis of the controllers on line 2, the main line to Unit 1.
There are three green liquor controllers for bank 1 whilst banks 2 to 3 had four.
The extra controllers in the other banks dated back to when the green liquor
could also go into the third precipitator of each bank. This was removed from the
physical system quite a while now, thus making these particular controllers
redundant. The names of these controllers are:
• FCLP41
• FCLP51
• FCLP81
Since these controllers do not exist in the physical setup, it was recommended
that they be eliminated from the DCS.
The GL to the precipitator banks is controlled via a total flow controller,
FCLTPTOT. The controller obtains the result of the total GL to each bank from
the total bank flow calculation programs discussed earlier in this report. The
setpoint of this controller is set by the operator. The output of the controller is
47
then the setpoint for the GL controllers. Since it is a total flow controller, its
performance can not be analysed until the performance of its underlying
controllers improve.
The main issue faced with all of these controllers was valve stiction caused by
green liquor scale. In order to clear the scale, the line must be washed using either
high pressure water or hot caustic soda, depending on the hardness of the scale.
To do this however, the line has to be taken offline which could only happen at
the end of its cycle (28 days). Thus after confirming valve stiction, there is a
waiting period of up to 2 months before the line is back online again, and the
performance can be retested to determine if any require retuning.
One of the controllers tested was FCLP40, the green liquor to precipitator 40.
During the initial analysis, the controller performed fairly well with an ISE score
of 5.17E+03 and a low stiction probability of 7.93%. However, when examining
the valve characteristic shown in Figure 5-10, it became evident that the control
valve has a significant deadband of between 8% and 10% of the OP which is far
too large to be considered as the natural deadband.
FCLP40 Valve Characteristic
295
325
355
385
415
445
475
505
535
565
22 27 32 37 42 47 52 57 62 67 72OP
PV
Figure 5-10: Valve Characteristic of FCLP40
48
Furthermore, for a setpoint change from 150 to 200 (refer to Figure 5-11), the
controller has an initial delay of one minute, but the PV delayed by about 3 and a
half minutes. The delayed control response may be due to the 0.2 minute lowpass
filter. As for the extra 2 and a half minutes on the PV, this has to be due to
stiction as the OP increases from 14% to 20% before any change in the PV is
achieved, thus the valve could not have been moving.
#1 (A) PRP.FCLP40.OP 18/09/2008 16:33:38 23.74 % (Raw) *G.L. TO PRECIPITATOR 04 0 / GLIQ TO PRECIP FLOW**#2 (A) PRP.FCLP40.PV 18/09/2008 16:33:38 200.16 kl/h (Raw) *G.L. TO PRECIPITATOR 04 0 / GLIQ TO PRECIP FLOW**#3 (A) PRP.FCLP40.SP 18/09/2008 16:33:38 200.00 kl/h (Raw) *G.L. TO PRECIPITATOR 04 0 / GLIQ TO PRECIP FLOW**#4 (A) PRP.FCLP40.MODE 18/09/2008 16:33:38 1.00 kl/h (Raw) *G.L. TO PRECIPITATOR 04 0 / GLIQ TO PRECIP FLOW**
30.00
10.00
250.00
100.00
250.00
100.00
0.79
-0.21
18/09/2008 15:33:39 18/09/2008 16:33:39
New Plot Title @ 1h0m 0s
Figure 5-11: Response to setpoint change
The reason for the low stiction probability may be due to the sluggish nature of
the controller, which is evident in that the rise time for the same step was just
over 22 minutes, extremely slow for a flow controller. If one was to wind up the
controller gain from 0.1 to say 0.3, the results that would come from the ISE and
the stiction probability would confirm that it was stiction. In saying this, there
was evidence of stiction in Figure 5-12. There are numerous spots on the trend
where the OP was changing but the PV remains unchanged. So the valve will
have to be tested for stiction.
49
#1 (R) PRP.FCLP40.OP 13/09/2008 03:48:54 53.10 % (Raw) *G.L. TO PRECIPITATOR 04 0 / GLIQ TO PRECIP FLOW**#2 (R) PRP.FCLP40.PV 13/09/2008 03:48:54 514.46 kl/h (Raw) *G.L. TO PRECIPITATOR 04 0 / GLIQ TO PRECIP FLOW**#3 (R) PRP.FCLP40.SP 13/09/2008 03:48:54 463.02 kl/h (Raw) *G.L. TO PRECIPITATOR 04 0 / GLIQ TO PRECIP FLOW**#4 (R) PRP.FCLP40.MODE 13/09/2008 03:48:54 2.00 kl/h (Raw) *G.L. TO PRECIPITATOR 04 0 / GLIQ TO PRECIP FLOW**
80.00
30.00
600.00
350.00
600.00
350.00
2.00
0.00
12/09/2008 23:48:54 13/09/2008 03:48:54
New Plot Title @ 4h0m 0s
Figure 5-12: Evidence of Stiction - FCLP40
When testing, the controller was put in manual along with the FCLP20, the only
other controller that was online on the same header. Now FCLP20 was a very
special case indeed. If one refers to Figure 5-1, FCLP20 branches off the line to
FCLP10, the GL to the first precipitator of bank one through the main line.
Usually, the GL will go into precipitator 10 only through either the mainline or
the spare line. However, precipitator 10 had just been taken offline for
maintenance, so precipitator 20 now became the first tank and FCLP20 was now
in use for the first time in over 2 years.
Since the controller had been offline for such a long time, no analysis had been
conducted on it, thus it was uncertain what was to be expected. When the
controller first came online, it displayed very strange behaviour as depicted Figure
5-13. Output resembled something like sound frequency.
50
#1 (R) PRP.FCLP20.OP 23/10/2008 06:51:02 40.98 % (Raw) *G.L. T O PRECIPIT AT OR 020 / GLIQ T O PRECIP FLOW**#2 (R) PRP.FCLP20.PV 23/10/2008 06:51:02 442.98 kl/h (Raw) *G.L. T O PRECIPIT AT OR 020 / GLIQ T O PRECIP FLOW**#3 (R) PRP.FCLP20.SP 23/10/2008 06:51:02 522.05 kl/h (Raw) *G.L. T O PRECIPIT AT OR 020 / GLIQ T O PRECIP FLOW**#4 (R) PRP.FCLP20.MODE 23/10/2008 06:51:02 2.00 kl /h (Raw) *G.L. T O PRECIPIT AT OR 020 / GLIQ T O PRECIP FLOW**
100.00
0.00
800.00
0.00
800.00
0.00
2.00
0.00
23/10/2008 02:51:02 23/10/2008 06:51:02
New Plot Title @ 4h0m 0s
Figure 5-13: The bizarre behaviour of FCLP20
On closer examination, it became clear that this behaviour was due to aliasing.
The sampling time for the controller trend data was 1 minute. The period of
oscillations was less than this resulting in the observed pattern. This is shown in
Figure 6-14, which shows the time when the sampling time was increased to 5
seconds. At first it was believed that the controller was too aggressive, so an
attempt was done to retune it using Opertune. However, the oscillations
continued and a pattern emerged that suggested stiction. Furthermore there was a
delay of about 15 to 30 seconds between the changes in OP to a change in flow.
Thus the valve had to be tested for stiction .
51
#1 (A) PRP.FCLP20.OP 23/10/2008 08:13:17 26.77 % (Raw) *G.L. T O PRECIPIT AT OR 020 / GLIQ T O PRECIP FLOW**#2 (A) PRP.FCLP20.PV 23/10/2008 08:13:17 519.62 kl /h (Raw) *G.L. T O PRECIPIT AT OR 020 / GLIQ T O PRECIP FLOW**#3 (A) PRP.FCLP20.SP 23/10/2008 08:13:17 506.40 kl /h (Raw) *G.L. T O PRECIPIT AT OR 020 / GLIQ T O PRECIP FLOW**#4 (A) PRP.FCLP20.MODE 23/10/2008 08:13:17 2.00 kl/h (Raw) *G.L. T O PRECIPIT AT OR 020 / GLIQ T O PRECIP FLOW**
110.00
-10.00
1000.00
0.00
1000.00
0.00
2.00
0.00
23/10/2008 07:13:18 23/10/2008 08:13:18
New Plot Title @ 1h0m 0s
#1 (A) PRP.FCLP20.OP 24/10/2008 13:40:11 31.91 % (Raw) *G.L. T O PRECIPIT AT OR 020 / GLIQ TO PRECIP FLOW**#2 (A) PRP.FCLP20.PV 24/10/2008 13:40:11 684.74 kl /h (Raw) *G.L. TO PRECIPITATOR 020 / GLIQ TO PRECIP FLOW**#3 (A) PRP.FCLP20.SP 24/10/2008 13:40:11 596.32 kl /h (Raw) *G.L. TO PRECIPITATOR 020 / GLIQ TO PRECIP FLOW**#4 (A) PRP.FCLP20.MODE 24/10/2008 13:40:11 2.00 kl/h (Raw) *G.L. TO PRECIPITATOR 020 / GLIQ TO PRECIP FLOW**
110.00
-10.00
1000.00
0.00
1000.00
0.00
2.00
0.00
24/10/2008 13:28:18 24/10/2008 14:28:18
Figure 5-14: Proof Of Ailiasing: (left) FCLP20 with 1 minute sampling time. (Right) the same controller with 5 second sampling time
Before testing was conducted, both controllers were set to a 5 second sampling
time and filter time constant was reduced from 0.2 minutes to 0. This was done
on the advice given by the internship industry supervisor that any controller with
a filter constant greater than 0.1 minutes (6 seconds) should be tested for noise.
During testing, a restriction was put in place such that the total bank flow limit
was not breached as this will result in the launders overflowing. Furthermore, the
level of the 45A tank had to be monitored to ensure that it did not overflow. For
this reason the operators were notified prior to testing so the level of the tank
could be brought down to between 40 to 60%. This provided a 50% buffer zone
for testing.
There were two tests conducted, the first of which was called the noise test to
evaluate how noisy the PV was without the filter. Both control loops had a
magnetic flow indicator (mag-flow) to measure the flow.
The test involving placing both controllers in manual, holding the output
constant and evaluating the level of noise. Both controllers passed the test, with
hardly any noise present.
The second test was valve stroking. The first controller tested was FCLP20. The
OP was initially set 33%. From there the OP was stepped down by 1%, waited
for the flow to get to steady state before performing the next step. As shown in
52
Figure 5-15 below, the control valve displayed perfect movement, as the flow
changed with every step change. This was not expected as it was almost certain
that the valve was sticky.
#1 (A) PRP.FCLP20.OP 27/10/2008 12:21:00 42.96 % (Raw) *G.L. TO PRECIPITATOR 02 0 / GLIQ TO PRECIP FLOW**#2 (A) PRP.FCLP20.PV 27/10/2008 12:21:00 787.44 kl/h (Raw) *G.L. TO PRECIPITATOR 02 0 / GLIQ TO PRECIP FLOW**#3 (A) PRP.FCLP20.SP Hidden (Raw) *G.L. TO PRECIPITATOR 020 / GLIQ TO PRECIP FLOW#4 (A) PRP.FCLP20.MODE 27/10/2008 12:21:00 2.00 kl/h (Raw) *G.L. TO PRECIPITATOR 02 0 / GLIQ TO PRECIP FLOW**
50.00
10.00
1000.00
0.00
1000.00
0.00
2.00
0.00
27/10/2008 09:42:00 27/10/2008 12:21:00
New Plot Title @ 2h39m 0s
Figure 5-15: Valve stroking FCLP20
There was a delay between the time when the step change was made to when the
flow changed, which was usually 15 seconds but varied up to 1 minute. This was
very puzzling because positioners usually move very quickly in response to OP
changes. The test was then conducted again, this time it was out in the field with
the operator making the changes. The field tests confirmed the trend data, and
the belief is that the I to P converter of the actuator is faulty resulting in the
delayed transmission of the signal from the DCS. Therefore, it was recommended
that this be checked and repaired.
The results for FCLP40 were more conclusive. The test began with an OP of
42.5%.Two single unit step changes were then performed and as illustrated by the
flow in the plot below, the valve did not move for either step. The same occurred
for a 1.5% step change from 43.5% to 45%. The valve only moved when stepped
to 50%. Thus the minimum change in the OP required to effectively move the
53
valve is 8% which is unacceptable. This was to be expected, since the controllers
valve characteristic plot developed in the initial analysis did indicate a deadband
between 8% and 10%
#1 (A) PRP.FCLP40.OP 27/10/2008 11:22:31 50.00 % (Raw) *G.L. T O PRECIPITAT OR 040 / GLIQ T O PRECIP FLOW**#2 (A) PRP.FCLP40.PV 27/10/2008 11:22:31 769.91 kl/h (Raw) *G.L. T O PRECIPIT AT OR 040 / GLIQ T O PRECIP FLOW**#3 (A) PRP.FCLP40.SP Hidden (Raw) *G.L. T O PRECIPITAT OR 040 / GLIQ T O PRECIP FLOW**#4 (A) PRP.FCLP40.MODE 27/10/2008 11:22:31 0.00 kl/h (Raw) *G.L. T O PRECIPIT AT OR 040 / GLIQ T O PRECIP FLOW**
53.00
40.00
800.00
620.00
1300.00
0.00
2.00
0.00
27/10/2008 11:10:00 27/10/2008 11:22:31
New Plot Title @ 12m31s
Figure 5-16: Valve Stroking FCLP40
It turned out that the control valves for both controllers were replaced before
they went online, which was 9 days before the test was conducted. Thus it is
believed that rather than scale, it is likely that the valve for FCLP40 was not
positioned correctly when installed in the flange and as a result the disc is
continually hitting the edges. For this reason, it was recommended that the valve
alignment be checked once it has been taken offline.
5.7. Analysis of Coarse Seed Controllers
There are a total of 28 coarse seed controllers in precipitation of which four
control the level of the repulper tanks, four control the total coarse seed flow to
precipitation and the rest control the flow to either the first or second precipitator
of each bank (refer to Figure 5-2 for the layout of these controllers). Performance
of these controllers is provided below:
54
• 5 controllers were Excellent,
• 9 exhibited Fair performance
• 2 performed Ok
• 7 exhibited Poor performance, whilst
• 4 were high level controllers which basically empty the repulper tanks
when the level got too high. These were never used and were classed as
inactive.
The purpose of the total flow controllers is to ensure that the available flow is
evenly distributed to the first 2 precipitators of the two banks it supplies
(precipitators 10, 20, 40 and 30). The controller first obtains the flows from the
lines going out of the repulper tank, then adds them together to get the total seed
flow which then becomes the input to the controller. The setpoint is set by the
operator and each total flow controller tries to control this total flow to setpoint.
The controller output is then split 4 ways and becomes the setpoint for the four
controllers on its line. Thus there is no single control valve for total flow
controllers. The block diagram in Figure 5-17 illustrates the block diagram of one
total flow controller, FCR22U1, which controls the total flow of the controllers
on the line out of repulper 22 (Refer to Figure 5-2).
K=1 K=0.25
?RATIO
CONTROLSP
FCR22U1
FCR22P10
FCR22P40
FCR22P30
FCR22P20
∑RATIO
CONTROLSP
FCR22U1
K=1 K=0.25K=1 K=0.25
?RATIO
CONTROLSP
FCR22U1
FCR22P10
FCR22P40
FCR22P30
FCR22P20
∑RATIO
CONTROLSP
FCR22U1
K=1 K=0.25
Figure 5-17: Simplified Block Diagram of FCR22U1
55
This controller underwent retuning in order to improve performance. Figure 5-18
shows the response before retuning. It is evident that the response is very
oscillatory, with amplitude of around 50kl/hr and a period of about 9 to 10
minutes. ISE indicated an OK performance rating.
#1 (R) prp.fc r22u1.op 2/09/2008 04:59:25 46.37 % (Raw) *ST SEED EX 44R22 CONT ROL /**#2 (R) prp.fc r22u1.pv 2/09/2008 04:59:25 1110.17 kl/h (Raw) *ST SEED EX 44R22 CONT ROL /**#3 (R) prp.fc r22u1.sp 2/09/2008 04:59:25 1050.00 kl/h (Raw) *ST SEED EX 44R22 CONT ROL /**
110.00
-10.00
1500.00
500.00
1500.00
500.00
2/09/2008 04:59:08 2/09/2008 08:59:08
New Plot Title @ 4h0m 0s
Figure 5-18: FCR22U1 before retuning
The problem with total flow controllers is that if one of its secondary control
loops is performing bad, then that will effect its own performance and as a result
the performance of all the other controllers it is controlling. Thus, it was unclear
whether the oscillations were due to FCR22U1 or bad tuning on one of the
secondary control loops. There was however an event that did confirm that it was
FCR22U1 that needed tuning. The operators put the secondary control loops
FCR22P10 to 40 all in auto in order to stop the oscillations in these loops. As
shown in Figure 6-19, when this occurred, the oscillations in the total flow
ceased. This was an important finding as it meant that the control loops don’t
oscillate when it is not cascaded with FCR22U1, thus indicating that the primary
controller was too aggressive.
56
#1 (R) prp.fc r22u1.op 30/08/2008 20:26:42 53.74 % (Raw) *ST SEED EX 44R22 CONT ROL /**#2 (R) prp.fc r22u1.pv 30/08/2008 20:26:43 1104.32 kl /h (Raw) *ST SEED EX 44R22 CONT ROL /**#3 (R) prp.fc r22u1.sp 30/08/2008 20:26:43 1050.00 kl/h (Raw) *ST SEED EX 44R22 CONT ROL /**#4 (R) prp.fc r22u1.mode 30/08/2008 20:26:43 2.00 kl /h (Raw) *ST SEED EX 44R22 CONT ROL /**#5 (R) prp.fc r22p20.mode 30/08/2008 20:26:43 1.00 kl/h (Raw) *44R22 T O PRECIP 20 / SEED T O PRECIP FLOW**#6 (R) prp.fc r22p10.mode 30/08/2008 20:26:43 1.00 kl/h (Raw) *44R22 T O PRECIP 10 / SEED T O PRECIP FLOW**#7 (R) prp.fc r22p30.mode 30/08/2008 20:26:43 1.00 kl/h (Raw) *44R22 T O PRECIP 30 / SEED T O PRECIP FLOW**#8 (R) prp.fc r22p40.mode 30/08/2008 20:26:43 1.00 kl/h (Raw) *44R22 T O PRECIP 40 / SEED T O PRECIP FLOW**
110.00
0.00
1200.00
500.00
1200.00
500.00
2.00
0.00
2.00
0.00
2.00
0.00
2.00
0.00
2.00
0.00
30/08/2008 11:09:04 31/08/2008 11:09:04
New Plot Title @ 1d0h0m 0s
Figure 5-19: When the secondary controllers were taken out of cascade mode, the oscillations ceased.
After consulting with the industry supervisor, it was decided to solve the problem
conceptually. The control parameters for the controller were K1=1 and T1 =1
and the ratio control effectively has a gain of 0.25 as it splits the flow evenly 4
ways. It was found that the gain should be 1 since 100% of the total seed flow
goes to the ratio controller which then splits 4 ways. Furthermore, since the
controller was a total flow controller, it was not required that the controller react
too fast to changes in the PV. This is due to the fact that the downstream loops
need time to react to the changes, thus a total flow controller should react slower
than its secondary loops.
To achieve this, T1 was increased from 1 to 3 in order to slow the controller
down and thus eliminate the oscillations. A setpoint change from 1050 to 1075
was then performed. As shown in Figure 5-20, the results from the step response
the retuning worked extremely well, tracking the setpoint with minimal error and
no oscillatory action.
57
#1 (R) prp.fc r22u1.op 2/09/2008 08:00:58 49.03 % (Raw) *ST SEED EX 44R22 CONT ROL /**#2 (R) prp.fc r22u1.pv 2/09/2008 08:00:58 1072.23 kl /h (Raw) *ST SEED EX 44R22 CONT ROL /**#3 (R) prp.fc r22u1.sp 2/09/2008 08:00:58 1050.00 kl /h (Raw) *ST SEED EX 44R22 CONT ROL /**
110.00
-10.00
1500.00
500.00
1500.00
500.00
2/09/2008 08:00:41 2/09/2008 12:00:41
Figure 5-20: Setpoint Change on retuned FCR22U1
This is further highlighted by Figure 5-21 showing the controller’s performance
the next day. It was still performing better than before with better setpoint
tracking, small long oscillations with a period of 20 minutes and amplitude of
about 5kL which is a great improvement from the 50kL amplitude before
retuning. Improvement in performance is also shown in its ISE score which went
from 7.42E+03 (OK) to 2.39E+03 (Fair). Further improvements in performance
can be achieved by improving the performance of the controller’s secondary
loops.
#1 (R) prp.fcr22u1.op 3/09/2008 15:02:05 55.29 % (Raw) *ST SEED EX 44R22 CONT ROL /**#2 (R) prp.fcr22u1.pv 3/09/2008 15:02:05 1094.79 kl/h (Raw) *ST SEED EX 44R22 CONT ROL /**#3 (R) prp.fcr22u1.sp 3/09/2008 15:02:05 1100.00 kl/h (Raw) *ST SEED EX 44R22 CONT ROL /**#4 (R) prp.fcr22u1.mode 3/09/2008 15:02:30 2.00 kl/h (Raw) *ST SEED EX 44R22 CONT ROL /**#5 (R) prp.fcr22p20.mode 3/09/2008 15:02:05 2.00 kl/h (Raw) *44R22 T O PRECIP 20 / SEED T O PRECIP FLOW**#6 (R) prp.fcr22p10.mode 3/09/2008 15:02:05 2.00 kl/h (Raw) *44R22 T O PRECIP 10 / SEED T O PRECIP FLOW**#7 (R) prp.fcr22p30.mode 3/09/2008 15:02:05 2.00 kl/h (Raw) *44R22 T O PRECIP 30 / SEED T O PRECIP FLOW**#8 (R) prp.fcr22p40.mode 3/09/2008 15:02:05 2.00 kl/h (Raw) *44R22 T O PRECIP 40 / SEED T O PRECIP FLOW**
110.00
0.00
1200.00
500.00
1200.00
500.00
2.00
0.00
2.00
0.00
2.00
0.00
2.00
0.00
2.00
0.00
3/09/2008 11:02:01 3/09/2008 15:02:01
New Plot Title @ 4h0m 0s
Figure 5-21: Retuned FCR22U1 - One Day Later.
58
The following will cover the analysis of the secondary control loops of FCR22U1
that were tested. These controllers were as follows:
• FCR22P10 – Coarse seed flow to Precipitator 10 from R22
• FCR22P20 – Coarse seed flow to Precipitator 20 from R22
• FCR22P30 – Coarse seed flow to Precipitator 30 from R22
• FCR22P40 – Coarse seed flow to Precipitator 40 from R22
FCR22P20 and FCR22P30 both performed extremely poor in the initial analysis.
It was initially believed that the FCR22P30 performance was due to aggressive
tuning, due to the erratic controller action exhibited in Figure 5-22.
Figure 5-22: The Trend of FCR22P30 & its associated valve characteristic plot.
However, on closer inspection, evidence of stiction was present. As shown in
Figure 5-23, there are areas where the OP is ramping up or down, yet the flow
doesn’t change until the magnitude of change has reached 10%. This should be
expected when examining the valve characteristic which does strongly indicate
the presence of stiction with a massive deadband of around 10%. Stiction
probability was 93%.
59
Figure 5-23: Close up of FCR22P30. evidence of stiction is indicated in the areas between the dotted lines
FCR22P20 was a little more obvious. Although it was also very oscillatory, there
is no doubt that it was stiction determined not only by the shape of the valve
characteristic but also by its stiction probability of 93.05%
Figure 5-24: The Trend of FCR22P20 & its associated valve characteristic plot.
Furthermore, both controllers had a low pass filter on its PV reading with a 1
minute filter time constant. This is far too large for a flow controller, as it will not
#1 (R) PRP.FCR22P30.OP 27/09/2008 07:34:37 35.69 % (Raw) *44R22 T O PRECIP 30 / SEED T O PRECIP FLOW**#2 (R) PRP.FCR22P30.PV 27/09/2008 07:34:37 310.80 kl/h (Raw) *44R22 T O PRECIP 30 / SEED T O PRECIP FLOW**#3 (R) PRP.FCR22P30.SP 27/09/2008 07:35:02 315.57 kl/h (Raw) *44R22 T O PRECIP 30 / SEED T O PRECIP FLOW**#4 (R) PRP.FCR22P30.MODE 27/09/2008 07:34:27 2.00 kl/h (Raw) *44R22 T O PRECIP 30 / SEED T O PRECIP FLOW**#5 (R) PRP.YS44R22.PVN 27/09/2008 07:34:23 1.00 (Snapshot @1 Min) *44R22 ST AT US NORMALISED ST AT US**
110.00
-10.00
600.00
0.00
600.00
0.00
2.00
0.00
1.00
0.00
27/09/2008 05:34:23 27/09/2008 07:34:23
New Plot Title @ 2h0m 0s
60
only filter the noise but also part of the real value. Thus a noise test had to be
conducted on both controllers.
When testing the controllers, the flow could not go below 50kL/hr. The reason
for this is at this point the seed begins to build up against the disc and will
continue to do so until the line is blocked. This phenomenon is known as
bogging.
Very similar results were achieved for both controllers for both the noise and
stiction. Thus the test results for FCR22P30 will be analysed. As shown in Figure
5-25, the moment the filter time constant was reduced to zero, the flow reading
became very noisy.
Figure 5-25: Noise Test –FCR22P30.
Thus a filter will have to be implemented. Using the low pass filter equation, it
was worked out that a 0.15 minute filter time constant should be implemented
and as shown in Figure 5.26. The filter has removed the noise from the signal,
but allowed the controller to properly manage the flow.
#1 (R) PRP.FCR22P30.OP 3/11/2008 11:41:51 105.00 % (Raw) *44R22 TO PRECIP 30 / SEED TO PRECIP FLOW**#2 (R) PRP.FCR22P30.PV 3/11/2008 11:41:51 401.41 kl/h (Raw) *44R22 TO PRECIP 30 / SE ED TO PRECIP FLOW**#3 (R) PRP.FCR22P30.SP 3/11/2008 11:41:51 522.29 kl/h (Raw) *44R22 TO PRECIP 30 / SE ED TO PRECIP FLOW**#4 (R) PRP.FCR22P30.MODE 3/11/2008 11:41:51 2.00 kl/h (Raw) *44R22 TO PRECIP 30 / SE ED TO PRECIP FLOW**#5 (R) PRP.YS44R22.PVN 3/11/2008 11:41:51 1.00 (Snapshot @1 Min) *44R22 STATU S NORMALISED STATUS**
110.00
-10.00
600.00
0.00
600.00
0.00
2.00
0.00
1.00
0.00
3/11/2008 05:21:44 4/11/2008 05:21:44
New Plot Title @ 1d0h0m0s
Time when filter removed
61
Figure 5-26: Implementing 0.15 Minute filter time constant on FCR22P30.The pink line indicates time of implentation
The next test was valve stroking. As expected, the valve was suffering from
stiction. The valve only moved for a 10% change in the output. The test was
prematurely cut short as the line began to bog. This is shown in Figure 5-27
where the OP is being increased but the flow is steadily decreasing. Despite this,
the presence of stiction was confirmed and recommendations for its repair were
made.
#1 (A) PRP.FCR22P30.OP 3/11/2008 15:10:53 53.03 % (Raw) *44R22 TO PRECIP 30 / SEED TO PRECIP FLOW**#2 (A) PRP.FCR22P30.PV 3/11/2008 15:10:53 362.97 kl/h (Raw) *44R22 TO PRECIP 30 / SE ED TO PRECIP FLOW**#3 (A) PRP.FCR22P30.SP Hidden (Raw) *44R22 TO PRECIP 30 / SEED TO PRECIP FLOW**#4 (A) PRP.FCR22P30.MODE 3/11/2008 15:10:53 0.00 kl/h (Raw) *44R22 TO PRECIP 30 / SE ED TO PRECIP FLOW**#5 (A) PRP.YS44R22.PVN 3/11/2008 15:10:53 1.00 (Snapshot @1 Min) *44R22 STATU S NORMALISED STATUS**
110.00
-10.00
600.00
0.00
600.00
0.00
2.00
0.00
1.00
0.00
3/11/2008 15:02:29 3/11/2008 15:29:09
New Plot Title @ 26m40s
• Valve only moved for 10% change
•10% Sticky
Bogging#1 (A) PRP.FCR22P30.OP 3/11/2008 15:10:53 53.03 % (Raw) *44R22 TO PRECIP 30 / SEED TO PRECIP FLOW**#2 (A) PRP.FCR22P30.PV 3/11/2008 15:10:53 362.97 kl/h (Raw) *44R22 TO PRECIP 30 / SE ED TO PRECIP FLOW**#3 (A) PRP.FCR22P30.SP Hidden (Raw) *44R22 TO PRECIP 30 / SEED TO PRECIP FLOW**#4 (A) PRP.FCR22P30.MODE 3/11/2008 15:10:53 0.00 kl/h (Raw) *44R22 TO PRECIP 30 / SE ED TO PRECIP FLOW**#5 (A) PRP.YS44R22.PVN 3/11/2008 15:10:53 1.00 (Snapshot @1 Min) *44R22 STATU S NORMALISED STATUS**
110.00
-10.00
600.00
0.00
600.00
0.00
2.00
0.00
1.00
0.00
3/11/2008 15:02:29 3/11/2008 15:29:09
New Plot Title @ 26m40s
• Valve only moved for 10% change
•10% Sticky
Bogging
Figure 5-27: Valve Stroking - FCR22P30
#1 (R) PRP.FCR22P30.OP 5/11/2008 10:51:34 46.92 % (Raw) *44R22 T O PRECIP 30 / SEED T O PRECIP FLOW**#2 (R) PRP.FCR22P30.PV 5/11/2008 10:51:34 348.62 kl/h (Raw) *44R22 T O PRECIP 30 / SEED T O PRECIP FLOW**#3 (R) PRP.FCR22P30.SP 5/11/2008 10:51:34 350.00 kl/h (Raw) *44R22 T O PRECIP 30 / SEED T O PRECIP FLOW**#4 (R) PRP.FCR22P30.MODE 5/11/2008 10:51:34 1.00 kl /h (Raw) *44R22 T O PRECIP 30 / SEED T O PRECIP FLOW**#5 (R) PRP.YS44R22.PVN 5/11/2008 10:51:34 1.00 (Snapshot @1 Min) *44R22 ST AT US NORMALISED ST AT US**
110.00
-10.00
600.00
0.00
600.00
0.00
2.00
0.00
1.00
0.00
5/11/2008 09:51:14 5/11/2008 11:51:14
New Plot Title @ 2h0m 0s
62
6. Conclusion
The internship has been a great learning experience for the intern. Over the
course of the internship, the intern was exposed to the applied world of
engineering design and research, allowing for the development of many skills in
graphics development, analysing controller performance analysis and working on
the Honeywell DCS to name a few. Furthermore, the intern was also able to
obtain a tremendous understanding of the precipitation process at Wagerup, how
it operates, how it’s controlled and the purpose of certain control schemes.
The knowledge obtained, the skills developed and the confidence gained through
the conducting the projects and working with other engineers would no doubt
beneficial for a successful career as a control engineer.
6.1. Progress and Future Work
Overall, a total of 106 control loop have been analysed thus far, making up 90 %
of the controllers needed to be analysed for the project. Of those analysed so far,
and of those poor performance was evident on over 65% of them whether that
be tuning, valve or instrument issues.
A report is currently being compiled on the results of the analysis and testing
which will be handed in at the end of the project. Maintenance has been planned
to fix the problem loops.
A great proportion of the loops analysed with less than excellent performance
was due to valve stiction. The most common cause of stiction is scale build up in
the lines. To prevent scale, the lines are meant to be caustic washed every 28 days.
However, evidence suggests that this is not the case, and in some cases, the lines
are not washed effectively.
In order to maintain the performance of the control loops and the health of the
valves, it’s essential that a regular maintenance schedule is developed and strictly
met. The maintenance schedule should also be developed to monitor and
maintain the performance of the control loops. The technical department has
63
access to automatic performance analysis software LoopScout, which scans the
controller and provides a performance report card similar to the ones produced
for this project. Controllers could be scanned in bulk every month and a report
on each one could be provided to OC3 in order to carry out the action required.
6.2. Problems Faced
There were many issues faced during the course of this project
• Precipitators offline
• Lines out of circuit
• Lightning strikes (causing plant downtime)
• Line Scaling
64
BIBLIOGRAPHY
• M.A.A. Shoukat Choudhury, et al. (2003) Modelling Valve Stiction.
University of Alberta – Department of Chemicals and Materials
Engineering, Canada; University College London – Department of
Electronic and Electrical Engineering, London
• .M.A.A. Shoukat Choudhury, et al. (2005) A Simple test to Confirm Control
Valve Stiction. University of Alberta– Department of Chemicals and
Materials Engineering, Canada.
• Alcoa in Australia (2008) Alco World Alumina, Australia.