Escolar Documentos
Profissional Documentos
Cultura Documentos
EXCHANGER
By
Khalid A. Bakhurji
August, 2015
The goal of this thesis is to create the experimental setup and the necessary
instrumentation for the temperature control of a double-pipe heat exchanger. The heat
State University, Long Beach, has been used in the past for temperature control
experiments, in which the manipulated variable was the incoming water flow. In the
control experiments performed for the current work, the manipulated variable is the
incoming steam pressure. For this purpose cascade control has been used, with an inner
loop controlling the steam pressure setpoint, and an outer loop controlling the water
outlet temperature. Despite major fluctuations in the steam pressure supply and
inappropriate sizing of a control valve, the experimental results show that the designed
cascade control can track the temperature setpoint and reject load-induced disturbances
satisfactorily. System parameters are obtained via experimental modeling through open-
loop experiments. A LabView-based data acquisition and control program has been
1
written and implemented for acquiring and processing sensors outputs, and for providing
control commands to the final control elements. Several controller parameter tuning
methods, such as IMC, ITAE, Cohen-Coon, and Ziegler-Nichols, were implemented and
2
CASCADE CONTROL FOR A STEAM-HEATED DOUBLE-PIPE HEAT
EXCHANGER
A THESIS
In Partial Fulfillment
Committee Members:
College Designee:
By Khalid A. Bakhurji
August, 2015
ProQuest Number: 1596967
In the unlikely event that the author did not send a complete manuscript
and there are missing pages, these will be noted. Also, if material had to be removed,
a note will indicate the deletion.
ProQuest 1596967
Published by ProQuest LLC (2015). Copyright of the Dissertation is held by the Author.
ProQuest LLC.
789 East Eisenhower Parkway
P.O. Box 1346
Ann Arbor, MI 48106 - 1346
Copyright 2015
Khalid A. Bakhurji
As a believer, I start these lines by thanking God. Who gave me the mind,
Special thanks go to Dr. Jang and Dr. Chassiakos for supervising and correcting
this thesis. Dr. Jang provided me with the needed knowledge and skills in LabView and
inspired me during the tuning stage. Dr. Chassiakos participated with his ideas during the
analysis stage and was there during the whole work progress for support whenever it is
needed. Thanks also to Abdullah Al Qahtani for providing me with needed documents
and loved ones who contributed either directly or indirectly to our achievement. I would
like to dedicate this work to my parents who were and still are my biggest supporter.
They have always encouraged me and pushed me to climb the highest academic ladder
since I was little. I would also like to pass my gratefulness to my great friend Berenice
Morales for assisting me during the writing and reviewing stage of this work.
Finally, I would like to dedicate this work to Ana Flores, a special person and a
friend for her support and encouragement during the development of this paper.
iii
TABLE OF CONTENTS
Page
CHAPTER
3. AUTOMATIC CONTROL......................................................................................8
iv
CHAPTER Page
v
CHAPTER Page
9 CONCLUSION ......................................................................................................72
APPENDICIES ..................................................................................................................73
REFERENCES ..................................................................................................................87
vi
LIST OF TABLES
TABLES Page
12. PID IMC Tuning Parameters for the Slave Controller ................................................51
13. IMC Tuning Parameters for the Master Controller (4 GPM Load). ............................54
14. PI Tuning Values for the Cascade Control Master Loop .............................................64
vii
LIST OF FIGURES
FIGURES Page
1. Control function block for the heat exchanger using combined cascade and
feedforward control .........................................................................................................2
10. Steam Pressure control valve and the pressure transmitter ..........................................28
13. Steam pressure cycling under fully open bypass valve ..............................................32
14. Steam pressure cycling for valve openings between 5%-30% ...................................33
viii
FIGURES Page
18. Open-loop function block test for the double-pipe heat exchanger ............................41
21. Water Temperature setpoint tracking using PI controller under different tuning
values ..........................................................................................................................46
23. Cascade control function block for the double-pipe heat exchanger ...........................48
25. Open-loop process reaction curve for the cascade slave loop .....................................50
26. Slave controller closed loop test result for setpoint tracking .......................................52
27. Open-loop control function block for the cascade master loop ...................................52
28. Open-loop process reaction curve for the cascade master loop (4GPM load) .............54
29. Open-loop process reaction curve for the cascade master loop (2.5 GPM load) .........55
30. Heat exchanger cascade control setpoint tracking using IMC tuning..........................58
31. Heat Exchanger cascade control setpoint tracking for optimum tuning ......................58
32. Heat Exchanger cascade control disturbance rejection test to load variation ..............59
38. Graph representation of the performance indices for the tuning methods ...................68
ix
FIGURES Page
39. Summery graph of the average indices for the tuning methods...................................69
41. Model 9052 Double-Pipe heat exchanger process flow diagram (PFD)
[Duplicate of figure 6] ................................................................................................75
42. Scale and flask filled with steam condensate during the test. ......................................77
x
CHAPTER 1
LITERATURE REVIEW
1.1 Introduction
Heat exchangers have been the focus of studies for many years, in both the
control theory and mechanical design fields. In the control theory field, they are the main
examples in many text books in terms of chemical process models and control strategies.
Thus in this work, applying process model approximation and applying control strategies
to a double-pipe heat exchanger will be the main focus rather than test simulation.
The double-pipe heat exchanger used in this thesis and previous work is located at
the California State University, Long Beach (CSULB) Chemical Engineering department
laboratory. Originally, the heat exchanger was operated using a simple PID feedback
control where the manipulated variable is the opening of the cold water flow control
valve, and the process variable is the water outlet temperature. The heat exchanger
control strategy was further studied and enhanced later in 2014. Basic PID control,
cascade control and feedforward control were tested. It was found that a combination of
cascade control and feedforward control is the best control strategy for the double-pipe
performances. Figure 1 shows the control function block of the combined control
strategies applied to the heat exchanger. Figure 2 shows the system response to
1
combined control strategies. The control algorithm was tested for system response to
FIGURE 1. Control function block for the heat exchanger using combined cascade and
feedforward control [1].
The objective of the double-pipe heat exchanger control in the previous study was
to control the cold water outlet temperature by manipulating the cold water flowrate
through the cold water incoming line control valve. However in industrial applications,
heat exchangers are controlled through tight controlling of the hot stream, being the
saturated steam pressure in the double-pipe heat exchanger under study. The cold water,
being the heat exchanger load, would vary throughout the operation of the exchanger
2
FIGURE 2. Previous system response to the combined control in terms of setpoint and
disturbance rejection [1].
Different from the previous study, the objective of this work is to control the cold
water outlet temperature of the double-pipe heat exchanger through manipulating the
saturated steam pressure control valve. The system should have good setpoint tracking
indices defined for the purposes of this thesis. The indices can be understood as a robust
exchanger.
3
The major challenge in this double-pipe heat exchanger is identified as the
saturated steam pressure fluctuation (disturbance). The steam pressure goes into cycles
of an average of 3.5 minutes per cycle, and varying amplitude between 4 PSIG and 12
PSIG with fully open control valve. Since we do not have control over the boiler
generating steam in this system, the disturbance acting on the steam pressure has to be
rejected through control. Varying back pressure of the steam pressure network is also a
concern. In some cases with a fully closed steam pressure valve, the pressure can reach
up to 6 PSIG.
The second major challenge in controlling this heat exchanger is the oversized
steam pressure control valve. Compared to the maximum saturated steam flow required
Drawings related to the heat exchanger (Technovate, model 9052) internals and
materials details were not found at the search stage of this thesis. This was also a
challenge. Without an internal look to the piping internal design, it is hard to distinguish
potential disturbances that appears intermittently during the operation of the heat
exchanger. Only a Process Flow Diagram (PFD) were found related to the heat
exchanger.
4
CHAPTER 2
Heat exchangers are equipment used to transfer thermal energy from one medium
assisting the flow of heat from one medium and transfer it to another. Since their
invention in 1855, modern heat radiators used in automobile engines are an essential
application of heat exchangers [2]. In this process, the heat is transferred from the engine
refrigerators and air conditioning. On a wide industrial scale, heat exchangers are used in
chemical processing plants, petroleum refineries, food processing and power plants.
both mediums, hot and cold, are flowing through separate channels while exchanging
thermal energy across the wall separating the flow paths. In comparison, Regenerative
heat exchangers have single flow channels where both mediums alternately pass through
5
Over 50% of all installed heat exchangers are of recuperative type, specifically
recuperative heat exchangers and widely used in small scale and large scale industries
[5]. In contrast, double-pipe heat exchanger, which is a tubular type exchanger, is the
simplest exchanger form used in the industry. They are known for their cheap
construction and maintenance, yet, known for their low efficiency and large foot print.
Thus, they are often used for educational purposes [5], [6].
single processing unit, or combinations of processing units, used for the conversion of
raw materials into finished products. Maintaining this process or processes commonly
involve control. The control of chemical process can be either manual in which human
interference is required all the time, or automatic in which a set of control algorithms,
centralized processing unit, field instrumentation and final control elements are all
integrated to form what we call these days Process Control Systems (PCS).
Industrial chemical processes in recent years have grown globally. The industry
have developed sophisticated processes, some have set high product quality index, many
have energy consumption obligations laws and others have strict safety standards. Most
organizations try to improve productivity, improve quality and reduce cost. All these
factors are challenges raised in the past two decades and demand advance control
processing plant. Being an energy processing unit, many industries economically rely on
6
reliable control of heat and energy. In addition, heat exchangers play a role in larger
process to supply quality stream temperature, mass flow or pressure. Therefore, heat
exchangers control has been the center of many control system studies and examples [8].
There are many algorithms contributing to the control of a simple heat exchanger
depending on the cold and hot streams used. In general, there are four main control
control solutions with good modeling, estimation and disturbance identification can result
in excellent control [8]. However in some cases, modeling a heat exchanger for the
purpose of control can be a difficult task because of noise and instabilities present in the
In recent years, many studies have been concentrating on the mechanical design
improving efficiency. Other factors such as heat transfer and material heat capacity are
In this thesis, a double-pipe heat exchanger is put under test to apply one of the
known control algorithms to achieve dynamic robust control. The heat exchanger uses
saturated steam (hot stream) generated from a boiler to heat up flowing cold water (load).
By manipulating the steam pressure control valve under different loads, the objective of
the control algorithm is to maintain the cold water at a desired set point temperature
7
CHAPTER 3
AUTOMATIC CONTROL
3.1 Background
automation growth. To a certain degree, it can be said that the first feedback application
was invented during the ancient Greek and Arab worlds. The feedback control at that
time was in the form of float valve regulator to control water clocks, oil lamps and water
tanks level. In the third century BC, Ktesibios and Philon were known names in the
application of feedback control. Heron of Alexandria was also active in the control field
in the first century AD. The Arabs also contributed to the development of feedback
control. Names such as Al-Jazari and Ibn al-Sa-ati continued the development of the
water clock during 1203-1206 [9]. In modern days, almost all machines used by humans
employ the concept of feedback control. Operating these machines accurately and
reliably would not be possible without feedback control. Since the ancient Greek times
and up to current days, control schemes have been represented using block diagram
sensing element, a transducer, a controller, the process, and a final control element or
8
actuator. Figure 3 shows the arrangement of flow control architecture including typical
control elements.
Sensors are a core element of any control system. A sensor measures the status of
the variable to be controlled such as pressure, flow, temperature or level. This process
can either be continuous or sampled, such as in the case with analyzers. Depending on
produce a raw signal that needs to be converted to a standardized electrical signal such as
For example, traditional pressure sensors sense the spring movement caused by an
electrical single by the pressure transducer. In the modern digital world, sensors and
transducers are combined together in one unit to form what we call now transmitter.
9
The measured state of the variable to be controlled is sent to a controller or what
we called logic solver. A controller works as the brain or the decision maker in any
control scheme. The controller receives the status of the process variable to be controlled
and makes a decision to be sent in the form of an electrical signal to the final control
element.
Final control elements in the chemical process industry are typically a control
valve or an actuator. In a simple traditional control loop, the logic solver sends a decision
command to either open or close a control valve. Control valves are an important
element in any control scheme, and come in different shapes, sizes and design purposes.
Sizing of control valves for the purpose of this thesis will be discussed in chapter 5.2.3.
In simple words, a feedback control system measures the state of the controlled
variable output (process variable) and feeds it back to the input. It compares the fed back
signal to a reference signal (set point) to calculate the error. The objective of the control
in this scenario is to minimize the error and match the reference signal. Figure 4 shows a
In figure 4, the sensor measures the process variable or control variable (CV) and
feeds it back to the input. The CV is compared to the setpoint to produce the error (e).
The control block contains the control algorithm i.e. PID, PI or simple gain. An action
signal (c) is produced based on the control algorithm to the final control element or
actuator. The actuator changes the status of the manipulated variable (u). The process at
this point is adjusted based on the manipulated variable while typically disturbances are
10
acting on the process. The objective of the feedback control is to maintain the (CV) at
Since its first commercial appearance in the 1930s, PID or (Proportional, Integral,
Derivative) control algorithm has remained the first choice for process control designers
implementations, many control software developers and DCS manufacturers offer PID
feedback control modes. Features of those PID packages vary between developers. Yet,
With reference to figure 4, the objective of the control is to reduce the error (e) to
(1)
11
The aim of the proportional gain is to keep the error (e) in proportion to the
( ) ( )+ (2)
Where,
: is the error.
: is the steady-state bias value (It can be adjusted based on process operating
range).
Based on the above definition the term direct action mode of the controller
means that the controller output increases as the CV increases. Similarly, the term
reverse action mode of the controller means that the controller action will decrease as
The purpose of the integral action in a PID controller is to make the rate of change
of the controller output proportional to the error to eliminate offset. This can be
( ) ( ). (3)
Where,
seconds).
12
Integral control action to eliminate error offset is rarely used by itself since the
control action does not respond immediately until the error persists for sometime, unlike
proportional control where the control action responds immediately to error. Therefore,
integral control action is usually combined with proportional control action in the form of
1
( ) + ( ( ) + ( ) )
!
(4)
The anticipatory action or derivative action refers to the situation where changes
in error, typically small, can propagate through the system and eventually becomes big.
In this situation, the control action is proportional to the rate of changes of error [12].
( )
( ) + ( ( ) + " )
(5)
Where,
": is the derivative time or rate time (measured usually in minutes or seconds).
The three control modes can be combined together to form the PID controller.
Still a lot of chemical processes are using PI only, which is common and sufficient.
13
There are many features associated with modern PID controllers developed by
different manufacturer. Indeed, industrial PID controllers found inside DCSs and PLCs
are governed by the International Electrotechnical Commission (IEC 61131). Part of the
IEC 61131 mandates safety and minimum feature design requirements for programming
As a general rule in this thesis, PI control is used for temperature control and PID
3.5.1 General
values. This methodology is based on how the process model is obtained. In this thesis,
and in a lot of chemical processes, the empirical model approximation method is used.
This method involves disturbing the operation of the process for a period of time to
collect data. Thus, this method is hardly used in larger facilities or complicated processes
limited, to Internal Model control (IMC), Cohen-Coon (C-C), Integral Time Absolute
Error (ITAE) and Ziegler-Nichols (Z-N) method. Generally, these are well-known tuning
methods used in a variety of control applications. These tuning rules are calculated from
the process model. As will be seen in Chapter 8.3, the approximate process model is
based on fitting a first order system plus time-delay (FOPTD) to generate the process
reaction curve. Each tuning method has its own characteristics and it is applied in the
14
field. The theory behind each tuning method will not be discussed in this introduction,
3.5.2 IMC
Tuning PID controllers based on IMC was developed initially by Rivera and
Morari [7]. In this work, Loop-Pro tuning software package is used to develop PID
tuning values based on IMC method. The IMC in Loop-Pro relies on the system closed
loop time constant deduced from the model. Loop-Pro uses equation (6) to calculate PI
1
#
# $%# + &
(6)
Where,
The closed loop time constant is calculated by Loop-Pro based on the controlling
table 1.
15
TABLE 1. IMC Closed-Loop Time Constant Calculations
A desirable process response for some chemical processes is the quarter decay
ratio. Cohen-Coon developed a PID tuning rule in which the process response is
somewhat oscillatory. This tuning rule is more suitable for processes models with 0.1 <
+- . < 1.0. Table 2 shows the C-C tuning rule for PI and PID controllers.
,
,
%
1 1 30 + 3( )
2 3 40.9 + ( )7 % 8 :
% 2 % %
# 9 + 20( )
PI --
%
1 4 1 % 32 + 6( ) 4
2 3 4 + ( )7 % 8 : % 8 :
% 3 4 % %
# 13 + 8( ) 11 + 2( )
PID
16
3.5.4 Ziegler-Nichols (Z-N)
Similar to C-C tuning method, Ziegler and Nichols developed a tuning rule to
achieve a quarter decay ratio process response. In addition, the tuning rule works well
with processes that have 0.1 < + . < 1.0 . Table 3 shows the Z-N tuning rule for PI
,
- ,
0.9
( ) 3.33 %
%
PI
#
--
1.2
( ) 2.0 % 0.5 %
# %
PID
ITAE tuning rule is based on optimizing absolute error that appears for long time
where the purpose is to decrease integral-time. Again, the ratio 0.1 < + . < 1.0 is
,
-,
important when implementing this tuning rule. Table 4 shows the ITAE tuning rule for
PI and PID controllers. The developers of this tuning rule distinguished tuning for
setpoint tracking and tuning for disturbance rejection. The table shows the tuning rule for
17
TABLE 4. ITAE Tuning Rule for PI and PID Setpoint Tracking
0.586
( )!.?@A %
% B1.03 0.165 ( )C
PI --
#
0.965 %
( )!.DEE % 0.308 ( )!.?G?
% B0.796 0.147 ( )C
PID
#
18
CHAPTER 4
4.1 Introduction
The double-pipe heat exchanger under test in this thesis is the simplest form of
heat exchangers. It contains a set of two concentric tubes, in which one stream flows in
the outer pipe and the other stream flows inside the inner pipe. In this system, the cold
water flows in the outer pipe and the hot steam flows in the inner pipe. The cold water is
supplied from a utility source, however, the steam is supplied from a campus boiler,
where no control over the steam quality is possible. The objective of this heat exchanger
is to raise or lower the cold water temperature by manipulating the steam pressure valve.
Figure 5 shows the actual double-pipe heat exchanger. The exchanger contains 6
loops. Each loop is made of different materials for the purpose of testing. Only the
stainless steel loop is used (Tube-3), and the rest were isolated during all tests in this
thesis.
There are two flow scenarios in this heat exchanger, either co-current flow or
usually the best heat transfer efficiency achievable. In the countercurrent flow scenario,
the cold water flows from one end of the exchanger and the saturated steam flows from
the opposite end. Hence, the flow direction of water and steam are opposite to each other
19
FIGURE 5. The double-pipe heat exchanger under test.
The overall system consists of the data acquisition unit, software, the process,
temperature sensors, flow meter, cold water flow control valve, saturated steam pressure
control valve and pressure transmitter. Figure 6 shows the process flow diagram of the
heat exchanger which contains most of the system components apart from the controller
and software.
With reference to figure 6, the following are the main components of the system.
module, which can be seen in Figure 7. The FP-1601 data acquisition unit consists of the
network module which utilizes 10/100 Ethernet connection to connect with the
input/output (I/O) modules. Alongside the network module, the date acquisition unit
consists of one (1) thermocouple module, one (1) analogue output module and one (1)
20
analogue input module. All I/O modules contain 8 channels. Each FP-1601 network
module can support up to nine (9) modules. The FP-1601 network module connects
CV-1
V-18
CV-2
FIGURE 6. Model 9052 double-pipe heat exchanger process flow diagram (PFD) [1].
interface used in many automation industries. LabView version 8.2 is used in this
system.
21
FIGURE 7. National Instrument FP-1601 data acquisition and communication unit.
4.2.2 Process
The double pipe heat exchanger contains 6 tubes. Each acts as a double-pipe heat
exchanger by itself. However, only tube-3 and tube-1 is used in this study with reference
Manufacturer Technovate
Model 9052
Material Stainless Steel
Under normal operating conditions, the following are the valve arrangements:
Manual isolation valves v-1, v-5, v-11, v-12, v-15 and v-16 should be open.
22
Manual isolation valves v-2, v-3, v-4, v-6, v-7, v-8, v-9, v-10, v-13 and v-14
should be closed.
v-17 and v-18 are 3-way valves. The flow direction indication on the physical
valves should be (\) for valve v-17 and (/) for valve v-18. In this arrangement, the cold
water will go to the drain through the gauge flow meter and the steam will go directly to
CV-1 and CV-2 are the saturated steam pressure control valve and cold water
The above valve arrangement is for countercurrent flow scenario which is the
focus in this study. In this arrangement, the objective of the heat exchanger control is to
maintain the cold water at some setpoint temperature by manipulating the steam pressure
valve CV-1 position at either constant or varying cold water loads. The load can be
elements to measure temperature at both streams at different points. In this study only
FT in figure 6 is the cold water flow meter or transmitter. The flow transmitter
measures the cold water flowrate entering the heat exchanger. It sends a 4-20mA signal
to the data acquisition unit. Table 6 shows the specifications of the flow meter.
23
TABLE 6. Flow Meter Specification [15]
Manufactured by Krohne
Model OPTIFLUX 1010
Type Electromagnetic sensing element
Temperature range (-25) 120 C
Analogue signal 4 20 mA
Calibration range 0 10 gallon/min
CV-2 in figure 6 represents the cold water flow control valve. Table 7 lists the
specifications of the control valve. The control valve is equipped with a hydraulic
24
actuator that receives a 4-20mA signal from the output analogue module calibrated to 0-
100% valve opening. The actuator driving the valve body has a slow action time i.e. 60s
to open and 30s to close. This presents a control challenge if the disturbance is high or if
Manufacturer Eurotherm
Model VB-7223-268-4-2
Type Globe valve
Characteristics Equal %
Pipe size 0.5 inch
Flow size (Cv) 1.3
Maximum inlet pressure 35 psig
Fluids max. temperature -6 to 138 Co
Actuator Hydraulic EA81-11006
Safe Position Normally Closed. Increase signal to open
No load stroke at 21 Co 60 second extend. 30 second retract
CV-1 in figure 6 represents the cold water flow control valve. Table 8 lists the
specifications of the control valve. The control valve is equipped with a hydraulic
actuator that receives a 4-20mA signal from the output analogue module calibrated to 0-
100% valve opening. The actuator driving the valve body has a slow action time i.e. 60s
25
to open and 30s to close. This will affect control strategy if the disturbance is high or if
In order to control the steam pressure valve, the heat exchanger is equipped with
pressure sensor located downstream the control valve. The sensor measures the steam
pressure to the inlet of the heat exchanger. Table 9 lists the specification of the pressure
sensor.
26
TABLE 8. Steam Pressure Control Valve Specifications [16]
Manufacturer Eurotherm
Model VB-7263-000-4-4
Type Globe valve
Characteristics Modified linear
Pipe size 0.5 inch
Flow size (Cv) 4.4
Maximum inlet pressure 100 psig
Fluids max. temperature -6 to 171 Co
Manufactured by WIKA
Model IS-20
Calibration range 0 30 psig
Analogue signal 4 20 mA
Manufactured by WIKA
27
FIGURE 10. Steam Pressure control valve and the pressure transmitter.
28
CHAPTER 5
5.1 Introduction
There are several methods used to identify a chemical process. Heat exchangers,
because of their importance in many industries, have been studied in the past and are used
considered a thermal dynamic problem. These types of problems usually involve thermal
lag. A thermal lag, (often called time-delay, dead-time or transport delay), is a delay
which occurs in the energy distribution between the hot medium and cold medium.
Resistance to heat flow is the primary cause of this delay in such systems. Thus, the
process and inspect the main elements in the system. Disturbances such as steam
pressure variation or water flowrate variation could be potential factors for instability
under certain control strategies. In addition, the system reaction curve to step change in
the open-loop mode plays a major role in studying the process behavior. In order to
identify the process parameters, experimental data will be fit to a first order model using
step test.
29
5.2 Process Disturbances
Previous work on this system provided experimental data of the water pressure
and temperature behavior. Figure 11 shows the change in source water flowrate. In this
test, the cold water flow control valve was left fully open for around 8 minutes and the
flowrate was recorded. The flowrate fluctuation varies between 2.8-2.9 GPM with 0.1%
deviation [1].
thermal process, the flow variation is a small disturbance to the overall all problem. In
addition, a simple aggressive PI control if applied to the water flow valve can eliminate
Further, the cold water in this heat exchanger is considered the heat exchanger
load and the control strategy of the overall heat exchanger should be able to handle an
30
induced water load disturbance. In this thesis, the heat exchanger water load will be
The previous work has also determined the variation of source water temperature.
Figure 12 shows the test results of the variation of source water temperature over a period
of 8 min. Under no control and fully open control valve, the temperature fluctuates
between a peak value of 26.75 oC and a low value of 26.62 oC. This is not enough
process problems where the system is expected to have high thermal capacity. In fact,
the water temperature disturbance is negligible when comparing it to the source water
flowrate disturbance.
FIGURE 12. Fluctuation of source water temperature over a period of time [1].
The biggest disturbance in this system is due to the steam pressure cycling. The
pressure in the steam supply, instead of having a constant value, can vary by as much as
50% within 2 minutes. A test was performed to investigate the behavior of the steam
31
pressure and steam control valve. While the steam control valve is fully closed and the
bypass valve is fully open for a period of time, figure 13 shows the cycling of the steam
pressure. The data were taken for a duration of around 15 min and recorded a peak
pressure of 12 psig and a low pressure of 8 psig. Operating the bypass valve is not
typically part of the procedure, however, it gives an indication of the severity of the
pressure variation. In fact, this test confirms data that were obtained in previous work.
A series of tests were conducted on the steam pressure control valve to measure
the pressure cycling. Each test was conducted by leaving the control valve open at a
certain percentage and recording the steam pressure. For each test, the steam pressure
was recorded for the same amount of time i.e. 10min. Figure 14 shows a combined tests
results for valve opening between 5%-30%. The results show that the steam pressure is
always cycling under different control valve opening, yet, reaching different amplitudes
for each opening. It is not clear why the pressure is received in this condition at the heat
32
exchanger, yet, we have no control over the exchanger upstream part. This acts as a
10 8%
8 10%
6 15%
4 20%
2 25%
0 30%
1
22
43
64
85
106
127
148
169
190
211
232
253
274
295
316
337
358
379
400
421
442
463
484
505
526
547
568
589
FIGURE 14. Steam pressure cycling for valve openings between 5%-30%.
Looking closely to the results, two observations were made. The first is that the
average cycle period T for each valve opening is almost the same, around 3.5 min. This
is a very important observation which helps when deciding on the tuning parameters. In
addition, cycling process in this magnitude and period suggests that the steam pressure
The second thing to observe from figure 14, is that the control valve is oversized.
Increasing the valve opening beyond 25% results in no change to the steam pressure. In
fact, the graph shows that at 20%, 25% and 30% valve opening, the pressure is cycling
with same amplitude, but it is phase shifted. This important result leads us to calculate
the size required for the steam control valve versus the actual size.
It is worth to mention here that during many runs of different tests, the steam loop
back pressure always varies. At some runs, the back pressure will go as low as 0.2 PSIG
33
with fully closed valve. However, at other runs, the back pressure is as high as 6-7 PSIG.
This means that the control valve shut-off pressure P ranges between 11 PSIG to 6
therefore, controlling the steam pressure guarantees quality of the steam provided to the
heat exchanger. In this section, we calculate the control valve sizing required for steam
control valve.
maximum saturated steam flowrate required by the heat exchanger. This is a tricky
process since the quality of steam received from the boiler and the system back pressure
is dynamic. However, the simulation should give a good approximation. The simulation
Steam quality is 100% entering the heat exchanger at 8 PSIG and 80 oC.
Details of the simulation and results can be found in appendix B. The simulation
shows that for the above conditions, a steam flowrate of 2.0 kg-mol/h = 36 kg/h = 79.4
ANSI/ISA 75.1 is the international standard for sizing control valves. ISA sizing
equations take into account detailed factors, such as the piping geometry correction factor
for accurate results. In comparison, many control valve manufacturers approximate these
equations or rewrite them to suit a particular stream. Indeed, these calculations lead to
34
almost the same results. Equation (7) is used by Atkomatic control valve company
I
H
3J G
L
(7)
Where,
I is the maximum steam flowrate in lbs/hr. This is the simulation result = 79.4
lbs/hr.
maximum supplied pressure from boiler which was found in 6.2.2 and equal to 4 PSIA.
The result of equation (xx) indicates a required valve Cv = 2.7. However, the
installed steam valve Cv is 4.4 which is double the needed flow capacity. This explains
why the steam pressure does not increase by increasing the control valve opening beyond
25%.
In addition, the valve actuator data sheet shows that the valve open stroke time is
60s, and 30s for close stroke travel time. In the industry, this is a slow action valve.
Adding this to the fact that the steam pressure is not stable all the time, the controllability
range for the control valve is typically between 5-15%. Thus in LabView, any steam
35
control valve action is limited to maximum 30% output. This will allow less unnecessary
travel time action and reduce the valve oversizing effect. The controller output will go
At this point, we have identified three major challenges in controlling the double-
A simple test is conducted to detect the temperature operating limits of the heat
exchanger under different loads. This test is important for determining the operating
envelop of the heat exchanger, and for determining the limitations when testing any
control strategy under different setpoints. In this test, the water flowrate is set manually
at certain flowrate and the steam valve is left fully open. Since the steam pressure is
cycling, we expect the water temperature to be cycling as well. Thus, once the steam
valve is fully open, the water temperature is left to reach a steady state cycling.
The same is done again, at the same water flow, with the steam valve fully closed. When
temperature is recorded.
Table 10 summarizes the results of the test under different water flowrates. This
table will be used as the basis for the heat exchanger operating envelope and when
36
Table 10. Double-Pipe Heat Exchanger Operating Limits
2.5 48 9 30 5
1.5 58 11 38 5
One of the many methods to develop a process model is the empirical method
using process data. It is one of the simplest methods to identify a process system. The
main advantage of this method is the development of dynamic process model in real time
rather than theoretical methods that involve many equations and variables to quantify the
system. In addition, the empirical method can characterize the process and catches small
Figure 15 shows the input and output of a process model. In the identification
method using a real time dynamic process, the input u(t), output y(t) are known. If
Disturbance d(t) can be measured/identified, as it was done in section 5.2, the process
model M can be identified. This is usually done by generating the system reaction curve
Previous studies and experimental data have shown that heat exchangers can be
37
identification method. However, real systems almost never experience a first-order
Thus, adding a time-delay to the first-order model can result in a better match
between experimental data and the first-order model [19]. This empirical method is
called fitting of a first-order plus time-delay (FOPTD). The model transfer function can
be represented by (8).
O-P
M(N)
N+1
(8)
Where,
is the time from where the process starts reacting to step change and reaches
63% of its final steady-state. It is also called the process time constant.
% is the time delay term of the process model. It represents the time it takes for the
38
Since the steam pressure cycling and the steam control valve oversizing
separate process and divide the heat exchanger into two process models. As mentioned,
Figure 16 shows the two processes function blocks and the overall heat exchanger
model, where u(t) is the input to the system, i.e. the steam valve position. The steam loop
process QRS adjusts the steam pressure and the output will be the new state of stream
pressure p(t). This will be the input to what it is called now heat loop process QRT . The
new pressure state alters the water temperature and hence that is the output of the system
y(t). The combined processes QRS and QRT represent the overall heat exchanger model.
Each model is approximated by a FOPTD linear system. Each system reaction curve is
39
CHAPTER 6
In this chapter, a single loop feedback PID control is tested. The objective of this
control logic is to test a basic control strategy and observe the outcome. The goal is to
control the water temperature exiting the heat exchanger by manipulating the steam valve
actuator. Figure 17 shows the control logic of system loop PID control. The controller
should be able to minimize the error and track the desired temperature setpoint regardless
40
The PID controller increases or decreases the signal to the steam valve actuator.
The steam control valve is configured to be direct acting, i.e. the valve will open on
increasing controller signal and close on decreasing controller signal. The control valve
action will alter the steam pressure and as a result it will change the water outlet
temperature. The thermocouple measures the water outlet temperature and feeds it back
to the input.
In order to tune the PID controller to accommodate the heat exchanger dynamics,
a process reaction curve should be produced. This is done by an open-loop test and
manipulated input variable will result in a change in the process state and in a new steady
FIGURE 18. Open-loop function block test for the double-pipe heat exchanger [1].
implement the PID feedback loop algorithm. The test starts by having the heat exchanger
valve configuration as mentioned in 4.2.2. In this test, the load or water flowrate is set at
41
4GPM. This can be achieved either through the water flowrate valve controlled manually
or automatically. Since the steam pressure fluctuates, the water outlet temperature will
also fluctuate within a small range. When the water outlet temperature reaches its steady-
state i.e. fluctuations of the water temperature stay within 5% of its average value, then
a step change to the steam pressure control valve is applied. An increase in the steam
control valve will result in an increase in the water outlet temperature and vice versa.
In this test, three step changes are applied to the control valve i.e. (1) from 5% to
8%, (2) from 8% to 10% and (3) from 15% to 10%. During the test, valve position, steam
pressure and cold water outlet temperature are recorded. These data, for the three step
changes, were loaded to Loop-Pro, the software tool for tuning and model optimization.
Typically with this tool, the process model is chosen and the software calculates the
model parameters from experimental data. In this case, a FOPTD is chosen as discussed
before. Each step change test is loaded individually and a best-fit graph is produced.
Figure 19 shows the best test result which corresponds to 0.64 R-Square goodness of fit.
The goodness of fit refers to the error magnitude between the data and best fit. In the
graph, the steam valve position trend is at the bottom and the water outlet temperature
fluctuating is at the top. From the graph, Loop-Pro calculates the FOPTD model
parameters. Transfer function (9) shows the heat exchanger over all model using FOPTD
best fit.
0.681 OG?.AP
M(N)
11.1N + 1
(9)
42
Loop-Pro: Te mp Ope n Lopp te st (5-8%)
M o d e l: F irs t O r d e r P lu s D e a d T im e (F O P D T ) F ile N a m e : o p e n lo o p 5 -8 % . t x t
40.3
39.0
37.7
8.1
7.2
6.3
5.4
G a in (K ) = 0 . 6 8 1 , T im e C o n s t a n t (s e c ) = 1 1 . 0 9 , D e a d T im e (s e c ) = 2 9 . 6
G o o d n e s s o f F it : R - S q u a re d = 0 . 6 4 4 6 , S S E = 4 4 7 . 4
In addition, a disturbance open-loop test was performed. Under fixed steam valve
opening at 10% and steady-state water temperature around 41-43 oC, the water flowrate
were decreased from 2.7 GPM to 2.2 GPM. The water outlet temperature increased to
46-47 oC. These test data were loaded in Loop-Pro and the reaction curve shown in
figure 20 is produced. In the figure, the lower trend is the water flowrate (manipulated
variable) and the top trend is the water temperature (process variable). FOPTD is also
used to represent the water flow disturbance model, transfer function (10).
11.14 OU.EAP
M(N)
27.1N + 1
(10)
Control (IMC). Table 11 summarizes the open-loop tests for step change and
disturbance.
43
Loop-Pro: Disturba nc e te st a t 10% ope ning
M o d e l: F irs t O r d e r P lu s D e a d T im e (F O P D T ) F ile N a m e : d is t u rb a n c e 1 . t x t
48.1
44.4
40.7
3.0
2.4
1.8
1.2
G a in (K ) = - 1 1 . 1 4 , T im e C o n s t a n t ( s e c ) = 2 7 . 0 8 , D e a d T im e (s e c ) = 7 . 5 6
G o o d n e s s o f F it : R - S q u a re d = 0 . 9 1 9 1 , S S E = 3 6 6 . 8
Process
Process Process Proportional Integral action
gain /( for '0 (VW() for PI
Reaction Time Control
gain constant
/R 'R (VW()
type Delay Sensitivity
XR (VW()
PI controller controller
At this stage, the process model is identified and IMC tuning parameter is
calculated from the process model. Next, the Proportional-Integral control logic is
applied as mentioned in figure 17. There are few things to consider when conducting the
closed-loop test:
44
Since the steam control valve is oversized, it was decided to limit the controller
action to 30% of full opening. This will allow less travel time for the control actuator and
decrease the delay. Increasing the valve opening beyond 25% will not have any effects
The water flowrate can be under manual or under controlled operation. In this
test, the flow control valve was operated manually. This will introduce a very small
disturbance that can not be noticed in the overall system. In fact, this disturbance has
already been taken into account when identifying the process model during the open-loop
test.
The test was conducted under 4 GPM water flowrate. Thus, system limitations
introduced in table 10 were considered in this test when varying temperature setpoint.
It is noted that in LabView, the integral and derivative values are in minutes NOT
seconds.
The objective of the closed loop PI control test is to study the performance of the
controller in terms of setpoint tracking and disturbance rejection. Figure 21 shows the
result of the PI control setpoint tracking. During the test, different tuning PI values were
used including the IMC tuning parameters obtained from the process model. The result
shows that a simple PI control manages to keep the water temperature in track with
setpoint, however, the control is not optimum, and unacceptable water temperature
variations can be seen. This is a result of severe steam pressure disturbance which can
45
PI Temp Control under 4 GPM water flow
45
40
35
30
25
20
15
10
5
0
1
109
136
163
190
217
244
271
298
325
352
379
406
433
460
487
514
541
568
595
622
649
676
703
730
757
784
811
838
865
892
919
946
973
1000
1027
1054
1081
1108
1135
1162
1189
1216
1243
1270
1297
1324
1351
1378
1405
28
55
82
FIGURE 21. Water Temperature setpoint tracking using PI controller under different
tuning values.
disturbance. In this test, the water temperature is kept at steady state at 40 oC. Then, the
water flow control valve was manually decreased from 100% open (corresponding to 4
GPM) to 50% open (corresponding to 2.5 GPM). In terms of thermal problem, this is not
a big disturbance, however, the temperature started fluctuating and did not recover. In
120
100
80
60
40
20
0
1
16
31
46
61
76
91
106
121
136
151
166
181
196
211
226
241
256
271
286
301
316
331
346
361
376
391
406
421
436
451
466
481
496
511
526
541
556
571
586
Temp SetPoint (oC) CW Temp out (oC) Water flow valve CW Flow rate (GPM)
CASCADE CONTROL
The steam pressure cycling mentioned earlier in addition to the control valve over
sizing makes tight control of the heat exchanger very difficult to achieve using traditional
PID control. The response to a disturbance using feedback control does not begin until
the process variable diverges from the setpoint. Minimizing energy losses by tight
controlling the water temperature in terms of setpoint tracking and disturbance rejection
Since the major disturbance acting on the heat exchanger is the steam pressure
cycling, it makes sense to separate the steam pressure control loop from the rest of the
heat exchanger model and treat them individually. That is the main purpose of
introduced such that the disturbance is measured and controlled before it is sensed by the
controlled variable primary feedback loop. Cascade control offers disturbance rejection
acting on the system without the need to precisely measure that disturbance. In addition,
Figure 23 shows the control logic function block of cascade control for the
are added to the system. QRS and QRT are the steam pressure loop model and the heat loop
47
model respectively as mentioned in chapter 5.4. The overall control strategy contains
master and slave control loops. The master controller PID 2 establishes the pressure
setpoint for the slave controller PID 1. The slave controller output is received at the
steam pressure control valve actuator for open/close action. Based on the valve action,
the steam pressure either decreases or increases. This pressure is measured using the
pressure sensor and fed back to the slave control loop input. The master loop acts to
respond to the water temperature setpoint. The objective of this control strategy is to
FIGURE 23. Cascade control function block for the double-pipe heat exchanger.
must be obtained from the process model. Similar to the PI feedback control, a process
reaction curve must be obtained from an open-loop test and approximate it by FOPTD
model. However, two open-loop tests must be conducted in order to catch QRS and QRT
48
slave and master loop, respectively. A LabView function block code was written to
The slave loop open-loop test can be summarized by the control function block in
figure 24. In the slave loop, the process variable is the steam pressure and the
set to zero.
When recording data, at least two cycles must be recorded before and after a step
change.
The open-loop test was conducted for four step changes i.e. (1) from 5% to 8% of
full opening, (2) from 8% to 10% of full opening, (3) from 10% to 15% of full opening
49
and (4) from 15% to 20% of full opening. However when loading the data to Loop-Pro
for process model and tuning parameters, poor best fit to FOPTD is achieved. This is
because Loop-Pro relies on the initial steady-state condition of the process variable which
is in this case the fluctuating steam pressure. For that reason, manual interpolation must
Initial steady-state
interpolation
FIGURE 25. Open-loop process reaction curve for the cascade slave loop.
This is done by taking the average pressure of the first two cycles before the step
change and replacing the recorded steam pressure values with the average. This should
produce a steady or constant initial stead-state. LabView captures these data and
produced best fit FOPTD model. Among the tests, the 8%-10% step change test is
50
Figure 25 shows the reaction curve result of the open-loop test for the master
loop. Transfer function (11) represents the slave loop FOPTD model. Table 12 lists the
IMC tuning values in the aggressive mode, since the cascade slave loop must be always
M#@
!.EY Z [\.]^
E.DP_@
(11)
TABLE 12. PID IMC Tuning Parameters for the Slave Controller
Step change
4.72 6.97 (0.12min) 0.96 (0.02min) Aggressive
8-10%
Although the derivative action is not always recommended for fast loops such as
pressure, yet the derivative action proved during testing that it can significantly eliminate
the steam pressure cycling effect. Thus, PID controller has to be used for the inner loop
Figure 26 shows the result of the closed loop test for the slave loop. In this closed
loop test, a modified version of the IMC tuning was used i.e. 10, 12 N , a
0.6 N . IMC tuning values in table 12 will also give satisfactory results, however, these
tuning values are optimum. The figure shows excellent result for tracking steam pressure
setpoints between 7-10 psig. Generally, as the setpoint increases beyond 10 psig, the
51
steam pressure starts cycling again. However under normal operating conditions of the
heat exchanger, the required steam pressure is usually between 8-10 psig.
FIGURE 26. Slave controller closed loop test result for setpoint tracking.
The master loop open-loop test can be summarized by the control function block
in figure 27. In the master loop, the process variable is the steam pressure and the
FIGURE 27. Open-loop control function block for the cascade master loop.
The same LavView code in appendix D is used in this test. The test starts by
having the steam pressure at a particular steady-state setpoint and steady-state water
temperature. Subsequently, the pressure setpoint is altered and the new water
52
temperature at steady-state is recorded. The master controller open-loop test depends on
two important factors: the flowrate of the water and tight control of the slave loop. The
following are considerations when conducting the master loop open-loop test:
In LabView code, the slave controllers must be in Auto mode and the master
In LabView, the slave controller maximum output must be set to 30%. This is to
Note that in LabView, the controller tuning parameters are in minutes NOT
seconds.
The test was conducted for three step changes i.e. (1) from 7 PSIG to 9 PSIG, (2)
from 9 PSIG to 10 PSIG and (3) from 10 PSIG to 7 PSIG. All step changes were
conducted under fixed water flowrate of 4 GPM. The data from the three tests were
loaded to LoopPro to obtain the FOPTD best fit graphs. The first step change test (7-9
PSIG) produced the best goodness of fit within R2 = 0.99. Figure 28 shows the reaction
curve of the master controller open-loop test. The lower trend represents the manipulated
variable (steam pressure) and the top trend represent the process variable (water
temperature).
53
FIGURE 28. Open-loop process reaction curve for the cascade master loop (4GPM load).
Transfer function (12) represents the master loop FOPTD model. Table 13 lists
M#G
b.YA Z [c.d^
GE.!P_@
(12)
TABLE 13. IMC Tuning Parameters for the Master Controller (4 GPM Load)
Step change 1.92 25.0 (0.4 min) 0.812 (0.01 min) Aggressive
7-9 PSIG 0.23 25.0 (0.4 min) 0.812 (0.01 min) Moderate
Another open-loop test is conducted for a different water flowrate. The same
procedure is followed under 2.5 GPM. The objective of this test is to generate different
process reaction curves under different loads for the heat exchanger. Figure 29 shows the
54
open-loop test process reaction curve under 2.5 GPM load. Transfer function (13)
represent the master loop FOPTD model when it is under 2.5 GPM load.
M#G
e G.D Z [fg.h^
@?.EP_@
(13)
FIGURE 29. Open-loop process reaction curve for the cascade master loop (2.5 GPM
load).
Despite the variation in the time-delay between the two process loads, the process
gain and the time constant is approximately matching. Further, the objective of the
cascade loop is to eliminate the internal loop disturbance (steam pressure cycling) in
addition to reject outer loop disturbance (heat exchanger load) disregarding the difference
7.3.1 General
The objective of the cascade control strategy is to eliminate the steam pressure
disturbance and contain it in the slave loop before it effects the water temperature outer
55
loop. In addition, it should control the heat exchanger water temperature in terms of
At this this stage we obtained both the master and slave controller tuning
parameters and process models for QRS and QRT . To conduct the cascade closed-loop
test, the same LabView code in appendix D is used. The control function block in section
7.1 is used to implement the cascade configuration. The following are general
In LabView code, the initial pressure setpoint must be set as low as possible e.g.
0.2
In LabView code, both controllers must be set as reverse acting controller. This
does not mean the controllers should be acting in reverse practically, however, it is only
In LabView, the slave controller maximum output must be set to 30%. This is to
In LabView, the master controller maximum output must be set to 11. This is
approximately the maximum supply pressure which will be fed to the slave controller.
In LabView, the slave controller must be switched to automatic mode first. Leave
the process to stabilize for a period of time. Enter a water temperature setpoint within the
heat exchanger operating limit i.e. table 10. Then switch the master controller to
automatic mode.
In LabView code, IMC tuning parameters for the salve and master controllers can
be used or any modified tuning values. Yet, it is important to have the slave controller
56
In LabView code, the water temperature manual output indicator should be
Note that in LabView, the controller tuning parameters are in minutes NOT
seconds.
Figures 30 shows the results of the heat exchanger under cascade control in terms
of water temperature setpoint tracking. In the graph, the top trend is the temperature
setpoint along with the water outlet temperature. The lower part represents the slave loop
with the steam pressure setpoint and process variable. In this test, IMC tuning parameters
were used for both master and slave loops obtained from table 13 and 12, respectively.
The cascade control strategy with aggressive IMC tuning managed to keep the water
temperature in track with the setpoint with excellent performance. However, as the
temperature setpoint is increased (near to the heat exchanger operating limits i.e. 40 oC),
the control becomes harder. The test was conducted under 4 GPM water flowrate. In this
test PID controllers were used for the master and slave controller.
The same closed-loop test was repeated with modified tuning parameters. Figure
31 shows the result of cascade control for different setpoints. The test was also
conducted under 4 GPM water flowrate. The result shows excellent temperature tracking
between 30 oC-38 oC for considerable amount of time. Optimum aggressive control was
used for the slave controller and IMC aggressive mode tuning for the master controller.
By visually inspecting result in figure 30 and figure 31, it can be inferred that the
response is somehow better when using tighter control for the slave loop and IMC
57
Cascade Control IMC aggressive tuning
Master Loop Kc=1.92, I=0.4min, D=0.01min
45
40
35
30
25
20
15
Slave Loop Kc=4.72, I=0.12min, D=0.02min
10
5
0
17
33
49
65
81
97
1
113
129
145
161
177
193
209
225
241
257
273
289
305
321
337
353
369
385
401
417
433
449
465
481
497
513
529
545
561
577
593
609
625
641
657
673
689
705
721
737
753
769
785
Temp Set Point (oC) C.W Temp Out (oC) Steam Pressure (PSIG) Cascade Steam Setpoint (PSIG)
FIGURE 30. Heat exchanger cascade control setpoint tracking using IMC tuning.
FIGURE 31. Heat Exchanger cascade control setpoint tracking test for optimum tuning.
The other measure in evaluating the performance of the cascade control strategy is
the load disturbance rejection. The control strategy should be able to handle water
flowrate variations and maintain the water temperature in track with the setpoint. The
test was conducted while maintaining the temperature setpoint constant during the test at
58
38 oC and altering the water rate through manually manipulating the control valve
The graph can be divided into four regions. In the first region, the control valve
opening was decreased from 100% to 80% of full opening. This did not cause enough
disturbance to the process and the flowrate was still at 4 GPM. This was followed by
another valve opening step down at 50%. In this region, the water flowrate decreased
from 4 GPM to 2.5 GPM. This was enough to cause a noticeable disturbance to the water
temperature, however, the water temperature recovered. In the third region, the water
flowrate was further decreased to 1 GPM while the temperature setpoint is still fixed at
38 oC. This caused another disturbance that the system could not recover from since the
heat exchanger is operating beyond its limits according to table 10. Finally, another
disturbance was created by increasing the water flowrate from 1 GPM to 4 GPM and the
FIGURE 32. Heat Exchanger Cascade control Disturbance rejection test to load variation.
59
CHAPTER 8
8.1 Introduction
Despite approximations made during the process model development and the
interpolation made to the experimental data, the empirical method and fitting of a first-
order plus time-delay works reliably to estimate the heat exchanger models. In addition,
cascade control managed to suppress high steam pressure cycling disturbance in the
internal loop and also successfully managed to reject water load variations acting on the
heat exchanger. With the use of appropriate aggressive controllers for both master and
slave loops, the water temperature was controlled at the desired setpoint all the time. In
terms of tuning performance, a three indices were developed for the double-pipe heat
exchanger.
The deviation of the process value from the setpoint is the error signal in the
feedback loop. Further, the area between the setpoint curve and the process variable
curve is related to the energy wasted as a result of this error. Thus, measuring this area
would provide a performance index for the tuning parameters used for the cascade
variable to a setpoint step change denoted as i in figure 33. Assuming the sampling
60
rate of the data shown in the graph is 1 second, the area of the absolute error between the
desired value and the actual value in the range between [a,b] for a step change i can be
t
u
(ps+1 ps )
j jk l m k n mkH | (p)| p q r (| kk ks+1 | + | kk ks |)
v 2
s 1
(14)
Where,
t: is the number of sampled data.
| kk k|: is the absolute difference between the measured process variable and
setpoint.
A second index based on the square of the error is also used in this study
(15)
61
An index z is defined as the area j per unit time per unit step change, based on either square or
j
z |kk k k N k N nl } m
({ l) (i)
Taking the square root of the area j when calculating the square error index was
considered, to bring down the square error indices down to a comparable values to the
other indices. Yet, this will not change the results and it can be implemented in the
future.
The other performance index will be the steam volume. The main energy
consumption in the double-pipe heat exchanger comes from the boiler in the form of
steam. Ideally, each steam valve opening percentage would allow a specific steam
which in turn can be converted to mass flowrate of steam from the boiler. The less steam
volume used to control the water temperature, the less energy dissipated in the system.
Hence, the control system strategy and controller tuning values should enhance the
However, the heat exchanger does not have the required sensors to measure the
steam flowrate. Thus, an experiment was conducted to redirect the steam recycle line to
an open drain and measure the volume of the received steam in the form of condensate.
The test details can be seen in appendix A. By varying the control valve opening and
62
The figure shows the condensate volume in GPM at the corresponding steam
valve opening. Valve openings at 3%, 5%, 8%, 10%, 15%, 20%, 25% and 30% of full
opening were tested. Because of the variations of the steam quality received in terms of
pressure and temperature, multiple data were recorded for each valve opening and a best
least squares fit linear approximation was obtained. The equation of the line can be
shown in the graph in figure 34. The test data can be further enhanced in the future by
recording more data for each valve opening. This equation will be used to estimate the
steam flowrate used during control tests. In figure 34, 1 gallon per minutes of steam
0.25
Flow Rate (GPM)
0.2
0.15
y = 0.0051x + 0.0968
0.1
0.05
0
0 5 10 15 20 25 30 35
Flow Rate (GPM) Steam Valve Opening (%)
The slave controller output in the cascade loop, which represents the valve
opening, is used to calculate the steam flowrate per sampled data (1 sample/sec). The
index is calculated by adding all volumes for the duration of the step change test and
63
dividing them over the duration of the test. With reference to figure 33, the steam
H ~m k Nl ~ l l
i~ kl p lH kl} H ~ml k kl
({ l)
In this study, four well known tuning methods were considered; Internal Model
Control (IMC), Cohen-Coon, Integral Time Absolute Error Criterion (ITAE) and Ziegler-
Nichols method. The test conducted for each tuning method is not identical, however, it
can produce a proper index of performances. Table 14 shows the controller values for
each tuning method. It is important to note that the controller tuning parameters ,
and a for the slave loop were fixed under optimum PID aggressive control. PI
controllers were used for the master loop since addition of derivative action did not
improve the overall performance as indicated in chapter 7.3. All tests were done under 4
TABLE 14. PI Tuning Values for the Cascade Control Master Loop
Figure 35, 36 and 37 show results for setpoint tracking runs using three different
64
described in chapter 7.3.2. Generally for the double-pipe heat exchanger, decreasing the
integral time of the controller causes the process variable to oscillate more vigorously.
According to the IMC tuning rule, integral time of the controller is normally set at the
time constant of the process unit. According to the valve manufacturers manual, the
steam control valve takes around 1 min for full stroke. This indicates that the time
constant of the control valve is around 1/4 min, and that appropriate integral time for the
controller would be around min (IMC tuning rule). Therefore, if lower values of
integral time are adopted in Cohen-Coon and Ziegler-Nichols tuning tests, the response
40
35
30
25
20
15
10
0
1
21
41
61
81
101
121
141
161
181
201
221
241
261
281
301
321
341
361
381
401
421
441
461
481
501
521
541
561
581
601
621
641
661
681
701
721
741
761
781
801
821
841
861
881
901
921
941
961
981
Temp SetPoint (oC) C.W temp out (oC) Steam Pressure (PSIG) Cascade SetPoint (PSIG)
Table 15 shows a summary of the calculated indices for different tuning methods.
All indices were multiplied by 100 to make the data representable and easier to read since
65
all indices are < 1. Figure 38 shows the table results in graph representation. The
average indices for each tuning method are also calculated by adding the corresponding
index for each test and dividing by the number of tests. A graph representation of the
40
35
30
25
20
15
10
0
1
12
23
34
45
56
67
78
89
100
111
122
133
144
155
166
177
188
199
210
221
232
243
254
265
276
287
298
309
320
331
342
353
364
375
386
397
408
419
430
441
452
463
474
485
496
507
518
Temp SetPoint (oC) C.W. Temp out (oC) Steam Pressure (PSIG) Cascade Setpoint (PSIG)
40
35
30
25
20
15
10
0
1
21
41
61
81
101
121
141
161
181
201
221
241
261
281
301
321
341
361
381
401
421
441
461
481
501
521
541
561
581
601
621
641
661
681
701
721
741
761
781
801
821
841
861
881
901
921
941
961
981
Temp SetPoint (oC) C.W out Temp (oC) Steam Pressure (PSIG) Cascade Setpoint (PSIG)
Average
Step Step Time Performance
Index zPvZ
Performance volumetric
zvuP Z
Tuning Test Magnitud Period Index
Index
Method (oC) e (s)
(GPM)
35-40 5 313 18.2 46.1 23.7
IMC 40-32 8 196 17.8 95.4 9.5
32-38 6 246 14.9 51.5 19.4
Average
17.0 64.3 17.5
(%)
35-38 3 201 23.2 29.4 22.6
Cohen- 38-30 8 338 13.4 58.5 5.2
Coon 30-35 5 252 21.1 38 15.3
35-38 3 164 27.8 59.2 22.4
Average
21.4 46.3 16.4
(%)
31-37 6 241 15.3 54.5 17.9
ITAE
37-35 2 210 16.5 16.1 13.7
Average
15.9 35.3 15.8
(%)
30-38 8 231 18.2 93.5 22.4
Ziegler-
38-30 8 438 12.7 54.5 4.9
Nichols
30-35 5 256 22 41.1 15.8
Average
17.6 63.0 14.4
(%)
Although test duration is different between tuning methods, the indices based on
errors can still represent the performance since they have a unit of error per second per
step change. By looking at the graph in figure 38, the square error index is high when
using the IMC and Z-N methods with a step magnitude of 8. In general as the step
magnitude is increased, the square error increases and vice versa. The ITAE method
recorded the lowest indices when the step magnitude is only 2. Comparison of the C-C
and Z-N methods when the step magnitude is 5 shows that they are similar in
67
performance. However, same step change for same method such as the C-C when the
step magnitude is 3 results in different indices, yet, the steam flowrate index is similar.
In general, the absolute error index is similar between the tuning methods, but, the square
FIGURE 38. Graph representation of the performance indices for the tuning methods.
By observation, it can be said that to achieve system response with less square
and absolute errors indices, IMC and ITAE methods can be used, however, more steam
Figure 39 shows a summary average of the tests indices. The absolute error
index and average steam flowrate index are similar between the tuning methods,
however, the square error is high in the IMC and Z-N methods. Indeed, the ITAE
68
recorded the lowest indices based on minimum error and considerably low steam flowrate
index.
70.0
64.3 63.0
60.0
46.3
50.0
35.3
40.0
30.0 21.4
17.0 17.5 16.4 15.9 15.8 17.6 14.4
20.0
10.0
0.0
IMC Cohen-Coon ITAE Ziegler-Nichols
FIGURE 39. Summery graph of the average indices for the tuning methods.
In any experimental work, the data reliability is usually questionable. For that, a
cascade closed-loop test was selected and to be regenerated. The Cohen-Coon tuning test
was selected since it has the maximum number of step changes i.e. 4 setpoints changes.
The test was regenerated with approximately exact conditions in terms of step changes
setpoints and duration of steady-state for each setpoint. Further, the indices were re-
calculated for the new test. The indices were not expected to exactly match the original
69
The result of the test can be seen in figure 40. The top graph is the original test
shown previously in figure 35. The bottom graph is the repeated C-C tuning test. By
visually inspecting both graphs, it can be concluded that both systems responses in terms
of water temperature setpoint tracking are similar. However, the first step change in the
The indices for the repeated test were also calculated and shown in table 16.
Comparing the average indices for the 4 tests conducted for the C-C method, the indices
are almost identical with small mismatch. This leads to the conclusion that experiments
Average
Step Time Performance
Index zPvZ
Tuning Step Performance Volumetric
zvuP Z
Test Period Index
Method Magnitude Steam Flow
(oC) (s)
rate GPM
35-38 3 196 22.4 22.6 19.6
Cohen- 38-30 8 340 12.7 51.4 5.4
Coon 30-35 5 251 23.3 50.5 15.2
35-38 3 167 32.2 84.8 21.7
Average
22.7 52.3 15.5
(%)
70
Cascade Control Cohen-Coon Tuning
45
40
35
30
25
20
15
10
5
0
1
22
43
64
85
106
127
148
169
190
211
232
253
274
295
316
337
358
379
400
421
442
463
484
505
526
547
568
589
610
631
652
673
694
715
736
757
778
799
820
841
862
883
904
925
946
967
988
Temp SetPoint (oC) C.W temp out (oC) Steam Pressure (PSIG) Cascade SetPoint (PSIG)
71
CHAPTER 9
CONCLUSION
exchanger under test. By using a steam pressure controller as slave loop and a
temperature controller as master loop, the cascade control was able to maintain the cold
water temperature in tack with the setpoint. In addition, the algorithm was able to
IMC tuning obtained was successful in maintain a good performance response to step
tuning methods were applied. Based on error indices defined in this thesis, it is found
that the ITAE method has the highest performance in tracking a setpoint change. Steam
flowrate index was also defined in this thesis and it is found that to maintain aggressive
fast response system IMC and ITAE should be used, however, more flowrate is required.
Thus, if the objective is to use the heat exchanger such to optimize the use of steam with
used.
72
APPENDICES
73
APPENDIX A
74
FIGURE 41. Model 9052 Double-Pipe heat exchanger process flow diagram (PFD)
[Duplicate of figure 6].
With reference to the heat exchanger figure 41 above, the following is the test
Manual isolation valves v-1, v-5, v-11, v-12, v-15 and v-16 should be open.
Manual isolation valves v-2, v-3, v-4, v-6, v-7, v-8, v-9, v-10, v-13 and v-14
should be closed.
v-17 and v-18 are 3-ways valves. The flow direction indication on the physical
valves should (\) for valve v-17 and (/) for valve v-18. In this arrangement, the cold water
will go to the drain through the gauge flow meter and the steam will go directly to the
recycle line.
75
Once the above valves arrangement are sat and the heat exchanger is operating,
open the steam control valve CV-1 using the manual controller to the desired opening i.e.
Close v-16 and switch the direction of the two 3-ways valves i.e. v-17 should be
(/) and v-18 should be (\). This will allow the steam to flow through the gauge flow meter
to the drain sink. The cold water will also flow to the drain sink through the other drain.
The steam will arrive to the drain in the form of condensate. Using a measuring
flask and a stop watch start filling up the flask with steam condensate. The start of this
operation and starting the stop should be synched as close as possible. The time depends
on how fast the flask is filled. For example, at 15% valve opening, the flask will be filled
Once the time is over, measure the filled flask using a scale. Also, the flask
should be measured when it is empty. The net weight of the condensate is the filled flask
Repeat the test for different valve openings. Since the steam pressure is
fluctuating, the test should be repeated multiple times at the same valve openings.
CAUTION: at valve opening => 15%, the steam pressure can go up 13 PSIG.
The received condensate at the sink drain will be extremely hot with bursts of pressure.
76
FIGURE 42. Scale and flask filled with steam condensate during the test.
to Gallons and then divided by the time used in that test and finally convert it to GPM.
Multiple data has to be recorded in order to improve accuracy of the test. Once the data
is available, plot the valve opening percentage on the x-axis against the flowrate in GPM
on the y-axis. Ad a linear best fit trend line to the graph in excel and hence, obtain the
77
FIGURE 43. Steam condensate flowrate test [Duplicate of figure 34].
78
APPENDIX B
79
PRO II is a process design simulation and observation analysis tool. It is widely
used in the chemical process industry. The objective of the test is to determine the
maximum required saturated steam flowrate for the double-pipe heat exchanger. Figure
44 shows the components of the simulation. It consists of 4 streams heat exchanger and a
In summary, the simulation were carried out under the following conditions:
Steam quality is 100% entering the heat exchanger at 8 PSIG and 80 oC.
After running the simulation, table 17 results were obtained by the simulator.
80
FIGURE 45. PRO II streams specifications.
Stream Name S1 S2 S3 S4
Description
Phase Water Water Vapor Water
Total Stream Properties
KG-
Rate MOL/HR 37.785 37.785 2.007 2.007
KG/HR 680.703 680.703 36.165 36.165
Std. Liquid Rate M3/HR 0.681 0.681 0.036 0.036
Total Adj.Liq.Vol.Rate M3/HR 0.672 0.672 0.036 0.036
Total Adj.Vap.Vol. Rate M3/HR 830.908 830.908 44.145 44.145
Temperature K 293.150 323.150 385.775 353.147
Pressure KPA 275.790 275.790 156.483 122.009
Molecular Weight 18.015 18.015 18.015 18.015
Enthalpy M*KJ/HR 0.057 0.143 0.097 0.012
KJ/KG 84.177 209.562 2695.018 334.992
Total Liquid Fraction 1.0000 1.0000 0.0000 1.0000
Reduced Temp. 0.4530 0.4994 0.5961 0.5457
Pres. 0.0125 0.0125 0.0071 0.0055
Acentric Factor 0.3449 0.3449 0.3449 0.3449
Watson K (UOPK) 8.758 8.758 8.758 8.758
Standard Liquid Density KG/M3 999.014 999.014 999.014 999.014
81
Specific Gravity 1.0000 1.0000 1.0000 1.0000
API Gravity 10.000 10.000 10.000 10.000
Total Adj. Liq. Density KG/M3 1012.423 1012.423 1012.423 1012.423
Latent Heat KJ/KG n/a n/a n/a n/a
Vapor Phase Properties
KG-
Rate MOL/HR n/a n/a 2.007 n/a
KG/HR n/a n/a 36.165 n/a
M3/HR n/a n/a 40.301 n/a
Std. Vapor Rate M3/HR n/a n/a 44.995 n/a
Adj.Vap.Vol. Rate M3/HR n/a n/a 44.145 n/a
Specific Gravity
(Air=1.0) n/a n/a 0.622 n/a
Molecular Weight n/a n/a 18.015 n/a
Enthalpy KJ/KG n/a n/a 2695.018 n/a
CP KJ/KG-K n/a n/a 2.134 n/a
Density KG/M3 n/a n/a 0.897 n/a
Adj.Vap.Density KG/M3 n/a n/a 0.819 n/a
Thermal Conductivity W/M-K n/a n/a 0.02575 n/a
Viscosity PAS n/a n/a 0.00001 n/a
Liquid Phase Properties
KG-
Rate MOL/HR 37.785 37.785 n/a 2.007
KG/HR 680.703 680.703 n/a 36.165
M3/HR 0.682 0.689 n/a 0.037
Std. Liquid Rate M3/HR 0.681 0.681 n/a 0.036
Adj.Liq.Vol.Rate M3/HR 0.672 0.672 n/a 0.036
Specific Gravity (H2O @ 60 F) 1.0000 1.0000 n/a 1.0000
Molecular Weight 18.015 18.015 n/a 18.015
Enthalpy KJ/KG 84.177 209.562 n/a 334.992
CP KJ/KG-K 4.184 4.179 n/a 4.195
Density KG/M3 998.292 988.129 n/a 971.820
Adj.Liq.Density KG/M3 1012.423 1012.423 n/a 1012.423
Surface Tension N/M 0.0724 0.0681 n/a 0.0630
Thermal Conductivity W/M-K 0.60365 0.64069 n/a 0.66713
Viscosity PAS 0.00100 0.00054 n/a 0.00035
82
APPENDIX C
83
84
APPENDIX D
85
86
REFERENCES
87
REFERENCES
[1] A. Al Qahtani, Advance control for bench-scale double-pipe heat exchanger, M.S.
thesis, California State University, Long Beach, 2014.
[2] J. Acton, T. Adams, and M. Packer, Origin of Everyday Things. New York: Sterling
Publishing Co. 2006.
[3] S. T. Branson, Heat Exchangers: Types, Design, and Applications. Hauppauge, New
York: Nova Science Publishers, 2011.
[4] G. Walker, Industrial Heat Exchangers: A Basic Guide. New York: Hemisphere Pub.
Corp, 1990.
[7] B. A. Ogunnaike and W. Harmon, Process Dynamics, Modeling, and Control. New
York: Oxford University Press, 1994.
[8] J. G. Balchen and K. I. Mumme, Process Control: Structures and Applications. New
York: Van Nostrand Reinhold, 1988.
[9] O. Mayr, The Origins of Feedback Control. Cambridge: M.I.T. Press, 1970.
[10] Spirax Sarco, Level and flow control applications, [Online]. Available from:
http://www.spiraxsarco.com/Resources/Pages/Steam-Engineering-
Tutorials/control-applications/level-and-flow-control-applications.aspx
[12] M. King, Process Control: A Practical Approach. Hoboken, N.J.: Wiley, 2011.
[13] National Instruments, PID control, Oct, 2012. White Paper. [Online]. Available
from: http://www.ni.com/white-paper/6440/en/#toc1
88
[14] National Instruments, FP-1601 user manual, April, 2003. [Online]. Available
from: http://www.ni.com/pdf/manuals/370707a.pdf
[15] Krohne, OPTIFLUX 1000 technical data sheet, Oct, 2004. [Online]. Available
from: https://www.instrumart.com/assets/OPTIFLUX_1000.pdf
[16] Eurotherm, Control valve data sheet, [Online]. Available from: http://www.
eurotherm.com/products/valves-actuators/v-series/two-way-valve-bodies/
[17] D. R. Patrick and S. W. Fardo, Industrial Process Control Systems. Atlanta, GA:
Fairmont Press, Inc, 2009.
[18] Atkomatic. Cv flow coefficients & valve sizing. [Online]. Available from: http://
fluidprocess.com/Atkomatic/pdf/Sizing.pdf
89