group t4 - expt c3

Upload: chandnichellappa

Post on 03-Jun-2018

241 views

Category:

Documents


1 download

TRANSCRIPT

  • 8/11/2019 Group T4 - Expt C3

    1/46

    National University of Singapore

    Department of Chemical and Biomolecular Engineering

    CN3019 Chemical Engineering Process Laboratory III

    Experiment C3Feedforward and Cascade Control

    Group T4

    Ang Yan Shan (Leader) U083592W

    Chandni Chellappa (Experimenter) U083655U

    Loo Ching Choo (Data Analyst) U084262H

    Yew Kian Wei (Data Analyst) U083694R

    Khoo Kian Guan (Literature Reviewer) U083698M

    Date of Experiment: 2nd

    /6thSeptember 2011

    Demonstratorssignature:

    GRADE:

  • 8/11/2019 Group T4 - Expt C3

    2/46

    Summary

    Process control is important in ensuring the smooth functioning of industrial operations.

    Choice and design of control system is crucial in order for appropriate corrective actions to

    be taken once a deviation sets in. In this experiment, two types of control, i.e. feedforward

    control and cascade control, are studied and evaluated based on their ability to maintain the

    vessel level at set point value in a miniaturized plant system. Various deviations, e.g. set

    point change, primary and secondary disturbances, were introduced to the system and the

    corresponding response curves plotted to track level changes and controller action over time.

    The performance of the two control systems were assessed based on three criteria, i.e. the

    maximum deviation magnitude, time required for re-stabilization and integral of absolute

    error value. Based on the data collected, it is concluded that cascade control shows betterperformance in the control of level than the feedforward control for this system. This is due

    to the presence of the slave loop in the former case which allows for quick acting response.

  • 8/11/2019 Group T4 - Expt C3

    3/46

    Table of Contents

    1. Introduction ........................................................................................................................ 1

    2. Theoretical Background ..................................................................................................... 22.1. Terminology and Basic Control Equipment................................................................ 2

    2.2. Control Strategies ........................................................................................................ 3

    2.2.1. Feedback Control ................................................................................................. 3

    2.2.2. Feedforward Control ............................................................................................ 3

    2.2.3. Cascade Control ................................................................................................... 4

    2.3. Transfer function of control strategy ........................................................................... 5

    2.3.1. Transfer function of feedforward-feedback control ............................................. 5

    2.3.2. Design of cascade control .................................................................................... 6

    2.4. Methods for the Evaluation of System Response........................................................ 7

    2.4.1. Characteristics of system response curve ................................................................ 7

    2.4.2. Integral error criteria ................................................................................................ 8

    3. Experiment........................................................................................................................ 10

    3.1. Apparatus List ........................................................................................................... 10

    3.2. Experimental Procedures........................................................................................... 10

    3.2.1. Feedforward Control .......................................................................................... 10

    3.2.2. Cascade Control ................................................................................................. 12

    3.3. Safety Analysis .............................................................................................................. 14

    4. Results and Calculations ................................................................................................... 15

    4.1. Feedforward Control ................................................................................................. 15

    4.1.1. Feedforward Control with Feedforward Gain = 0 ............................................. 15

    4.1.2. Feedforward Control with Feedforward Gain = 0.2 .......................................... 17

    4.1.3. Feedforward Control with Feedforward Gain = 0.4 .......................................... 18

    4.1.4. Feedforward Control with Feedforward Gain = 0.6 .......................................... 19

    4.1.5. Feedforward Control with Feedforward Gain = 0.8 .......................................... 20

    4.1.6. Feedforward Control with Feedforward Gain = 1.0 .......................................... 21

    4.1.7. Summary of Parameters Used to Evaluate Control System Performance ......... 22

    4.2. Cascade Control ........................................................................................................ 24

    4.2.1. Set point Change Perturbation ........................................................................... 24

    4.2.2. Disturbance to SystemChange in Inflow Rate ............................................... 27

  • 8/11/2019 Group T4 - Expt C3

    4/46

    4.2.3. Disturbance to SystemChange in Outflow Rate ............................................ 29

    5. Discussion and Analysis ................................................................................................... 32

    5.1. Feedforward Control ................................................................................................. 32

    5.1.1. Block Diagram of Feedforward Control ............................................................ 32

    5.1.2. Limitations of Simple Feedback Control ........................................................... 32

    5.1.3. Limitations of Cascade Control ......................................................................... 33

    5.1.4. Circumstances that feedforward control is recommended ................................. 33

    5.1.5. Limitations of Feedforward Controller .............................................................. 34

    5.2. Cascade control ......................................................................................................... 35

    5.2.1. Block Diagram of Cascade Control ................................................................... 35

    5.2.2. Advantages of Cascade control compared with single loop feedback control .. 36

    5.2.3. Limitations of Cascade Control ......................................................................... 36

    5.2.4. Classification of Feedforward Control as Cascade Control ............................... 37

    5.3. Error Analysis ............................................................................................................... 38

    6. Conclusion ........................................................................................................................... 39

    References ................................................................................................................................ 40

    Table of Notation ..................................................................................................................... 40

  • 8/11/2019 Group T4 - Expt C3

    5/46

    Table of Figures

    Figure 1. A general block diagram for feedback control loop ................................................................ 3

    Figure 2. A general block diagram for Feedforward Control ................................................................. 4

    Figure 3. A simplified block diagram of Cascade control ...................................................................... 5Figure 4. A block diagram for a feedforward-feedback control system (Adapted from Process

    Dynamics and Control by Seborg et al) .................................................................................................. 6

    Figure 5. A block diagram for a cascade control system (Adapted from Process Dynamics and Control

    by Seborg et al) ....................................................................................................................................... 7

    Figure 6. Basic characteristics of a system performance curve. ............................................................. 8

    Figure 7. Picture of CE117 Process Trainer - (Top) Miniaturized plant set up (Bottom) Mimic control

    panel. ..................................................................................................................................................... 10

    Figure 8. Schematic layout of connections made for feedforward control experiment. ....................... 10

    Figure 9. Flowchart for feedforward control experiment. ..................................................................... 11

    Figure 10. Schematic layout of connections made for cascade control experiment. ............................ 12Figure 11. Flowchart for cascade control experiment. .......................................................................... 13

    Figure 12. Response graph for feedforward control with gain = 0. ...................................................... 15

    Figure 13. Response Graph for Feedforward Control with feedforward gain 0 (400s till 1247.9s) .... 16

    Figure 14. Response Graph for Feedforward Control with feedforward gain 0.2 ............................... 17

    Figure 15. Response Graph for Feedforward Control with feedforward gain 0.2 (400s till 799.2s) ... 17

    Figure 16. Response Graph for Feedforward Control with feedforward gain 0.4. .............................. 18

    Figure 17. Response Graph for Feedforward Control with feedforward gain 0.4 (200s till 729.5s) ... 18

    Figure 18. Response Graph for Feedforward Control with feedforward gain 0.6 ............................... 19

    Figure 19. Response Graph for Feedforward Control with feedforward gain 0.6 (200s till 729.5s) ... 19

    Figure 20. Response Graph for Feedforward Control with feedforward gain 0.8 ............................... 20Figure 21. Response Graph for Feedforward Control with feedforward gain 0.8 (200s till 729.5s) ... 20

    Figure 22. Response Graph for Feedforward Control with feedforward gain 1.0 ............................... 21

    Figure 23. Response Graph for Feedforward Control with feedforward gain 1.0 (200s till 725.9s) ... 21

    Figure 24. Response graph for (a) initial start and (b) level set point changes (From 7V to 8V and

    back to 7V). ........................................................................................................................................... 26

    Figure 25. Response graph for inflow rate change 10V8V10V ................................................ 27

    Figure 26. Response graph for inflow rate change 10V6V10V ................................................ 27

    Figure 27. Response graph for inflow rate change 10V4V10V ................................................ 28

    Figure 28. Response graph for inflow rate change 10V2V10V ................................................ 29

    Figure 29. Response graph for outflow rate change (drain valve fully open1/3 closedfullyopen) ..................................................................................................................................................... 30

    Figure 30. Response graph for outflow rate change (drain valve fully open2/3 closedfully

    open) ..................................................................................................................................................... 31

    Figure 31. A control amplifier connected to proportional valve ........................................................... 37

    Table 1. Comparison of the various feed forward gains 0, 0.2, 0.4, 0.6and 0.8 .................................. 22

    Table 2. Summary of IAE at respective feedforward gain value .......................................................... 24

    Table 3. Output variable (water level) response to setpoint changes .................................................... 26

    http://c/Users/adminNUS/Desktop/C3/Group%20T4%20-%20Expt%20C3%20(Updated).docx%23_Toc303639463http://c/Users/adminNUS/Desktop/C3/Group%20T4%20-%20Expt%20C3%20(Updated).docx%23_Toc303639465http://c/Users/adminNUS/Desktop/C3/Group%20T4%20-%20Expt%20C3%20(Updated).docx%23_Toc303639465http://c/Users/adminNUS/Desktop/C3/Group%20T4%20-%20Expt%20C3%20(Updated).docx%23_Toc303639463
  • 8/11/2019 Group T4 - Expt C3

    6/46

    1

    1. Introduction

    Control in process industries refers to the regulation of various operating parameters in a

    process, such as temperature, level, pressure and flow rates. Precise control over these

    conditions is important in ensuring that the system is running efficiently and cost-effectively

    at its optimum conditions and for product quality assurance. Also equally important, if not

    more important application of process control is in ensuring the process is running within the

    safety limit and in compliance with environmental regulatory standards. It is therefore

    important to choose the types of control system and design the corresponding parameters

    carefully with respect to the system in control to fulfill the abovementioned requirements.

    There are three main types of control systems commonly used in the industry, i.e. feedback

    control, feedforward control and cascade control. They differ in terms of the type of input

    measured, i.e. control variable, secondary variable and disturbance respectively, which has

    subsequent implications on the speed and quality of control. Each has its own pros and cons,

    which will be elaborated in later sections, and discretion has to be exercised in making a

    choice for different systems under control.

    In this experiment, the scope of investigation is focused on studying feedforward and cascade

    control. There are two main objectives in this experiment:- (1) to study the limitation of

    feedforward control and (2) to study how cascade control can improve control of level. Using

    CE117 Mimic Panel, a simple miniaturized plant system consisting of a vessel, pump and

    valves is put under control. The aim of control is to maintain the water level in the vessel

    (controlled variable) at a specified set-point value. To achieve the first objective, disturbance

    in the form of change in pump voltage was introduced to the system at various feedforward

    gain values and the response of water level and controller action was plotted over time. In thesecond part of the experiment, disturbances in the form of outflow rate and inflow rate were

    introduced to the system to study the improvement in level control due to the presence of the

    slave loop in cascade control.

  • 8/11/2019 Group T4 - Expt C3

    7/46

    2

    2. Theoretical Background

    2.1.Terminology and Basic Control Equipment

    To understand process control, it is vital for one get familiarized with the basic components

    of a control system which are mainly:

    a) Sensor: It is a detection device uses to measure variables (such as pressure,

    temperature, level etc.) and transmits signals to the comparator.

    b) Comparator: When signals are received from the sensor, a comparator compares the

    set point and the measured value and transmits the difference of values as an error to

    the controller for controllers action.

    c)

    Controller: It takes corrective action as soon as possible if the controlled variable/measured disturbances deviated from the set point.

    d) Actuator: The actuator serves as a final control element. Depending on the controller

    output, the control elements can be control valves, cooling rate, energy input and etc.

    Another three important terms that for analysing control systems are the following:

    e)

    Manipulated variables: The process variables that the controller adjusts to keep the

    controlled variables at their desired values. These include the input flow rate, fuel

    flow rate and etc.

    f) Disturbance variables: The variables that disturbed the system which is undesirable.

    Disturbance variables may be known or unknown variables. The presence of

    disturbances causes the fluctuation of the system response and thus the need for

    automatic control system.

    g) Controlled variables: The variable that the controller controlled. They are usually

    output variables of a process.

  • 8/11/2019 Group T4 - Expt C3

    8/46

    3

    2.2.

    Control Strategies

    2.2.1. Feedback Control

    Feedback is the process of measuring the deviation of controlled variable from the set point

    and utilizing the measured information to exert control on the controlled variable itself.

    A simplified block diagram as shown in

    Figure 1below shows a typical feedback control system. Measured deviation, which is the

    difference between output and set point, is shown as E in the figure. The deviation is then

    transmitted to controller and the controller responds to deviation by controlling actuator. The

    process is corrected and thus output returns to set point eventually.

    Figure 1. A general block diagram for feedback control loop

    2.2.2. Feedforward Control

    Feedforward control, on the other hand, is the process of measuring disturbance and taking

    action based on the measured disturbance to ensure the disturbance does not upset the system

    response. Output is not measured in this strategy.

    A simple block diagram depicting how feedforward control takes action on disturbance isshown in

    Set PointController Process

    Disturbance

    Sensor

    Actuator

    -

    E Output

  • 8/11/2019 Group T4 - Expt C3

    9/46

    4

    Figure 2.Disturbance is directly measured by sensor and the result is transmitted to

    feedforward controller. An accurate process model is then used by controller to control

    actuator which in effect cancels out or reduces the impact of disturbance on system response.

    As disturbance can be effectively removed, perfect control is theoretically possible for

    feedforward control.

    Figure 2. A general block diagram for Feedforward Control

    However, feedforward control is usually used together with feedback control, as a process is

    subjected to known and unknown disturbances. Feedforward control serves to enhance the

    control performance by eliminating known disturbance while feedback corrects unknown

    disturbances.

    2.2.3.

    Cascade Control

    Cascade control consists of two or more nested feedback loops with the secondary loop(s) is

    placed within the primary loop. The primary loop is often called the master loop as it controls

    the output as well as the set point of the secondary loop. The secondary loop, on the other

    hand, is named slave loop. The use of slave loop(s) is to enhance a system performance by

    fast correction of unknown disturbances presented in the system. Therefore, slave loop(s)

    responds faster than master loop

    Disturbance

    Sensor

    ProcessActuator

    Feedforward

    Controller

    Manipulated

    variable

    Controlled

    Variable

  • 8/11/2019 Group T4 - Expt C3

    10/46

    5

    Figure 3below shows a simplified block diagram of cascade control. As can be seen in the

    figure, output from process 1 serves as the controlled variable of slave loop while output from

    process 2 serves as the controlled variable of master loop. Note that the deviation of the

    master loop is the set point for process 2. Any disturbances introduced into the system will be

    quickly corrected by slave controller and hence the output is less disturbed.

    Figure 3. A simplified block diagram of Cascade control

    2.3.

    Transfer function of control strategy

    2.3.1. Transfer function of feedforward-feedback control

    FromFigure 4below, the transfer function for feedforward-feedback control can be

    developed. Firstly, the closed loop transfer function for disturbance change, sDsY

    , is derived

    as follow:

    mpvc

    pvftd

    GGGG

    GGGGG

    sD

    sY

    1 Equation (1)

    Since perfect control is desired in feedforward control, i.e. controlled variable remains at set

    point regardless of any upsets in the system. Hence, the deviation of output Y(s) is zero for

    perfect control. D(s) is not 0 because it presents in the system and therefore the following

    conclusion is reached by substituting Y=0 into equation 1:

    0 pvftd GGGGG

    Disturbance 2Disturbance 1

    Master

    Controller

    Process

    2

    Set Point 1

    OutputSlave

    Controller

    Process

    1

    Primary Loop

    Secondary Loop

    Sensor 2

    Sensor 1

    Actuator

    ++-

    -+

    Set

    Point

    2

  • 8/11/2019 Group T4 - Expt C3

    11/46

    6

    Rearranging,

    pvt

    df

    GGG

    GG Equation (2)

    Usually, lead-lag units (Equation (3)). are used to provide reasonable approximations to

    ideal feedforward controllers Feedforward gain, Kfand the controllers time constants, 1and

    2,are parameters that can be tuned. Note that a feedforward gain of 0 will simply bring the

    control system back to simple feedback control.

    1

    1

    2

    1

    s

    s

    KG ff

    Equation (3)

    Figure 4. A block diagram for a feedforward-feedback control system (Adapted from Process

    Dynamics and Control by Seborg et al)

    2.3.2. Design of cascade control

    The closed loop transfer function for disturbance changes are derived based on the block

    diagram shown inFigure 5:

    12121222

    21

    2

    1

    1 mppvccmpvc

    dp

    GGGGGGGGGG

    GG

    sD

    sY

    Equation (4)

  • 8/11/2019 Group T4 - Expt C3

    12/46

    7

    12121222

    2221

    1

    1

    1

    1

    mppvccmpvc

    mpvcd

    GGGGGGGGGG

    GGGGG

    sD

    sY

    Equation (5)

    The characteristic equation is therefore:

    01 12121222 mppvccmpvc GGGGGGGGGG Equation (6)

    If the secondary loop is removed, i.e. Gc2=1 and Gm2=0, the characteristic equation reduces to

    simple feedback control.

    Figure 5. A block diagram for a cascade control system (Adapted from Process Dynamics

    and Control by Seborg et al)

    2.4.

    Methods for the Evaluation of System Response

    There are different types of methods that can be exploited to examine the performance of

    system. This involves basic studying of the response curve such as time to reach the steady

    state and overshoot of the system response, or using integral error criteria that analyzes the

    magnitude of error that is introduced to the system.

    2.4.1.

    Characteristics of system response curve

    Km1 Gc1

    MasterController

    Gc2

    SlaveController

    Gv

    ControlValve

    Gp2 Gp1

    Gd2 Gd1

    Gm2

    Gm1

    +-

    ++

    +-

    ++

    Ysp1 1~spY 2

    ~spY E1 E2

    Ym2

    Ym1

    P Y2 Y1

    D2 D1

  • 8/11/2019 Group T4 - Expt C3

    13/46

    8

    A controller performance can be examined by looking at a few obvious criteria from a system

    performance curve. These characteristics include time to reach first maximum, overshoot,

    decay ratio, rise time and settling time.

    First characteristic is the time to reach first maximum which is labeled as tpinFigure 6.The

    second characteristic is overshoot, which is the ratio of a and b, as labeled in Figure 6.

    Thirdly, decay ratio which shows how fast the fluctuation of the system settles is defined as

    the ratio of c and a. These three concepts are prominent for oscillatory response.

    Another two important concepts that are used to characterize the performance of the control

    system are rise time and settling time, denotes trand tsrespectively in figure below. Rise time

    is the time where the system response reaches the set point if the system whereas settling time

    is the time required for the response to settle down within a fixed percentage (for e.g. 5% as

    shown in the figure below).

    Figure 6. Basic characteristics of a system performance curve.

    2.4.2. Integral error criteria

    To achieve the best tuning of the controller, it can be difficult just to judge based on the

    abovementioned characteristics of the response curve. In such case, integral error criteria can

    be used to determine the best tuning of the controller. There are a few types of criteria that

    can be utilized which include integral of the absolute value of error (IAE), Integral of time

    weighted absolute error (ITAE), integral of squared error (ISE) and integral of time weighted

    squared error. Their definitions are as follow:

  • 8/11/2019 Group T4 - Expt C3

    14/46

    9

    The four integral error methods described above penalize the error differently. The first

    criterion IAE simply integrates the overall error in the system without any penalization of

    error. It gives response in between ISE and ITAE, with reasonable intial response and smaller

    amplitude oscillation.

    In ITAE, however, the error is penalized by a factor of length of time. The time factor ensures

    that the error will not persist for a long time in the system. The system settles quicker but

    with sluggish initial response which is good to avoid sustain oscillation

    On the other hand, ISE penalizes errors that are large, since the errors are squared. In such

    case, tuning from ISE ensures small fluctuation of the system but it might result in sustained

    low amplitude oscillation.

  • 8/11/2019 Group T4 - Expt C3

    15/46

    10

    3. Experiment

    3.1.Apparatus List

    1 x CE117 process trainer ()

    1 x CRT monitor

    Stirrer, fan and heater were not used throughout the whole process of this experiment

    Figure 7. Picture of CE117 Process Trainer - (Top) Miniaturized plant set up (Bottom) Mimic

    control panel.

    3.2.

    Experimental Procedures

    3.2.1. Feedforward Control

    Figure 8. Schematic layout of connections made for feedforward control experiment.

  • 8/11/2019 Group T4 - Expt C3

    16/46

    11

    1. The computer connected to the system was turned on.

    2. Folder CE117 was opened and C3 Feedfoward Control.ict was loaded

    3.

    The connections of the mimic panel were made as shown inFigure 8.All miscellaneous

    switches were placed to manual.

    4. Pump 2 was set to external, and drain valve was opened while bypass valve closed.

    5.

    Feedforward gain was left at 0, PID block was set as follows:

    Proportional = 5, Integral = 0.2. Derivative = 0

    N.B. For the whole experiment, controller has no derivative action.

    6. Start the process and allow the system to stabilize for about 5 minutes.

    7. A disturbance was introduced by increasing the pump voltage to 10 V. The reaction of

    the system was observed as it stabilized

    8.

    Increase the pump voltage back to 7 V. The system was allowed to stabilize.

    9. Repeat step 6 - 8 were repeated for feedfoward gains of 0.2 0.4, 0.6, 0.8 and 1.

    Figure 9. Flowchart for feedforward control experiment.

    Open the PID controller block. Set proportional gain to 5, integral to 0.2,

    leaving derivative at 0

    Set level setpoint and pump voltage at 7 V. Start the process and allow itto run till stabilization is reached (typically in about 5 minutes).

    Increase pump voltage to 10V. Observe the changes in controller action

    and response curve. Wait for system to stabilize again.

    Decrease the voltage back to 7 V and observe the response again. Allow

    the s stem to settle.

    Repeat this process for feedfoward gains of 0, 0.2, 0.4, 0.6, 0.8 and 1.0.

  • 8/11/2019 Group T4 - Expt C3

    17/46

    12

    3.2.2. Cascade Control

    Figure 10. Schematic layout of connections made for cascade control experiment.

    1. From the same CE117 folder, C3 Cascade Control.ict was loaded and the mimic panel

    was readjusted as shown inFigure 10.

    2. The proportional valve was set to 10 V (fully open), while the reference level was set at 7

    V.

    3. The 2 PID blocks were set as follows:

    Master Slave

    P 20 1

    I 1 1

    D 0 0

    4. The software was started after opening the drain valve fully and switching pump 2 to

    External. The system was left to stabilize.To investigate the effect of change in level set point,

    5. The level set point was increased by 1 V and the system was allowed to stabilize.

    6. The level set point was returned to 7 V and the system was allowed to stabilize.

    To investigate the effect of change in inflow rate,

    7. The system was started again and allowed to reach steady state at 10 V

    8 The valve was changed to 8 V and the system was allowed to stabilize.

    9. The valve was changed back to 10 V and the system was allowed to stabilize.

    10. Repeat Step 7-9 for 4.0V, 6.V0 and 8.0V step change.

  • 8/11/2019 Group T4 - Expt C3

    18/46

    13

    To investigate the effect of change in outflow rate,

    11. The drain valve was 1/3 closed. Allow the system to stabilize.

    12. The drain valve was reopened fully and the system was allowed to stabilize again.

    13. Repeat Step 1112 for the case when drain valve is 2/3 closed.

    Load C3 Cascade Control.ict from folder CE117

    Set the proportional valve block to 10 V. Set the reference level to 7V.

    Go to the 2 PID blocks. For the master controller set P to 20 and I to

    1; for the slave, P=I=1. For both blocks leave D at 0.

    Start the process. Allow the system to settle (within a few seconds).

    Increase reference level by 1 V and observe the response. Allow

    about 30s for system to settle.

    Reduce the level back to 7 V and observe the change in response curve.

    Adjust the valve settings for 2.0V, 4.0V, 6.0V, 8.0V.Observe

    systems response to changes in inflow rate.

    Manually adjust the drain valve (1/3 and 2/3 closed). Observe

    systems response to changes in outflow rate.

    Figure 11. Flowchart for cascade control experiment.

  • 8/11/2019 Group T4 - Expt C3

    19/46

    14

    3.3. Safety Analysis

    1) Many electrical connections are involved in this experiment. It is therefore important to

    keep the workspace dry especially where the power points are to avoid electric shocks.

    Within the design of the experiment, set appropriate flow rates to ensure that the water in the

    vessel does not overflow. Open the drain valves in the process vessel to prevent incident

    whereby the pump is accidentally switched on without setting the control system in place. In

    the event that water spillage did occur, switch off all electric points and inform laboratory

    safety officer immediately.

    2) Ensure air vent of vessel is opened to prevent pressure build up in system, which could

    potentially rupture the vessel.

    3) Switch off the heater control which is not required for this experiment, but could

    potentially pose thermal hazard if kept running.

    4) Ensure that water is circulating in the system while the pump is switched on, otherwise the

    pump could overheat and spoil, posing mechanical hazard.

  • 8/11/2019 Group T4 - Expt C3

    20/46

    15

    4. Results and Calculations

    4.1.Feedforward Control

    In the first part of the experiment, responses of the system to disturbance were investigated inthe feedforward control. Feedforward gains were set to 0, 0.2, 0.4, 0.6, 0.8 and 1.0. For each

    feedforward gain, the system was allowed to reach steady state before disturbance was

    introduced into the system. The pump voltage, which was the disturbance in this experiment,

    was increased from 7V to 10V and back to 7V again. The responses of how the controller

    reacts to disturbance and makes corrective action to the actuator signal were observed.

    4.1.1. Feedforward Control with Feedforward Gain = 0

    Initially, the tank was being filled up to the desired set-point of 7V. As shown in Figure 12,

    there was a gradual increase in water level until it overshot which peaked at about 70.3s.

    Between 70.3s till 181.5s, the system was approaching the desired set point by decreasing the

    water level. At this point, the controller reacted to the correct the action and hence explained

    the increase in the controller voltage. At about 181.5s, the system stabilized.

    Figure 12. Response graph for feedforward control with gain = 0.

    -5

    0

    5

    10

    15

    20

    25

    30

    35

    40

    45

    0 200 400 600 800 1000 1200 1400V

    olt(V)

    Time (s)

    Setpoint Control Level

    Start up - for

    system toreach steady

    state

    Pump voltage

    increased

    from 7V to

    10V

    Pump voltage

    decreasedfrom 10V to

    7V

  • 8/11/2019 Group T4 - Expt C3

    21/46

    16

    Figure 13. Response Graph for Feedforward Control with feedforward gain 0 (400s till

    1247.9s)

    Pump voltage increased from 7V to 10V at 426s

    The increase in pump voltage caused greater flow of water into the vessel and hence

    deviation of water level from the set point as shown in Figure 13. The PI feedforward

    controller reacted by sending signal to the valve to reduce the inflow of water into the vessel.

    The drop in the control voltage was due to the controller action to eliminate the offset.

    Pump voltage decreased from 10V to 7V at 804.7s

    The decrease in pump voltage, which meant decrease in pump speed, resulted in drop in the

    water level. The deviation of water level from the set point causes the PI controller to

    increase the valve voltage and hence bring the water level back to the set point level of 7.0V.

    As shown inFigure 13,an overshoot in the control voltage from 3.38V to 4.504V explained

    the oscillatory mode as the controller tried to bring the water level back to the set point.

    0

    1

    2

    3

    4

    5

    6

    7

    8

    400 600 800 1000 1200 1400

    Volt(V)

    Time (s)

    Setpoint Control Level

  • 8/11/2019 Group T4 - Expt C3

    22/46

    17

    4.1.2. Feedforward Control with Feedforward Gain = 0.2

    Figure 14. Response Graph for Feedforward Control with feedforward gain 0.2

    Figure 15. Response Graph for Feedforward Control with feedforward gain 0.2 (400s till

    799.2s)

    Pump voltage increased from 7V to 10V at 416s

    Pump voltage decreased from 10V to 7V at 712.1s

    -5

    0

    5

    10

    15

    20

    25

    30

    35

    40

    0 200 400 600 800

    Volt(V)

    Time (s)

    Setpoint Control Level

    Start up - forsystem to reach

    steady state

    Pump voltage

    increasedfrom 7V to

    10V

    Pump

    voltage

    decreased

    from 10V

    to 7V

    0

    1

    2

    3

    4

    56

    7

    8

    400 500 600 700 800

    Volt(V)

    Time (s)

    Setpoint Control Level

  • 8/11/2019 Group T4 - Expt C3

    23/46

    18

    4.1.3. Feedforward Control with Feedforward Gain = 0.4

    Figure 16. Response Graph for Feedforward Control with feedforward gain 0.4.

    Figure 17. Response Graph for Feedforward Control with feedforward gain 0.4 (200s till

    729.5s)

    Pump voltage increased from 7V to 10V at 230.5s

    Pump voltage decreased from 10V to 7V at 484.2s

    -5

    0

    5

    10

    15

    20

    25

    30

    35

    40

    0 200 400 600 800

    Volt(V)

    Time (s)

    Setpoint Control Level

    Start up -

    for systemto reach

    steady state

    Pump voltage

    increased from7V to 10V

    Pump voltage

    decreased from10V to 7V

    0

    1

    2

    34

    5

    6

    7

    8

    200 300 400 500 600 700 800

    Vol

    t(V)

    Time (s)

    Setpoint Control Level

  • 8/11/2019 Group T4 - Expt C3

    24/46

    19

    4.1.4. Feedforward Control with Feedforward Gain = 0.6

    Figure 18. Response Graph for Feedforward Control with feedforward gain 0.6

    Figure 19. Response Graph for Feedforward Control with feedforward gain 0.6 (200s till

    729.5s)

    Pump voltage increased from 7V to 10V at 291.8s

    Pump voltage decreased from 10V to 7V at 545.5s

    -5

    0

    5

    10

    15

    20

    25

    30

    35

    40

    0 100 200 300 400 500 600 700 800

    Volt(V)

    Time (s)

    Setpoint Control Level

    Start up - for

    system to reach

    steady state

    Pump voltage

    increased from7V to 10V

    Pump voltage

    decreased from10V to 7V

    0

    1

    2

    3

    4

    5

    6

    7

    8

    200 300 400 500 600 700 800

    Volt(V)

    Time (s)

    Setpoint Control Level

  • 8/11/2019 Group T4 - Expt C3

    25/46

    20

    4.1.5. Feedforward Control with Feedforward Gain = 0.8

    Figure 20. Response Graph for Feedforward Control with feedforward gain 0.8

    Figure 21. Response Graph for Feedforward Control with feedforward gain 0.8 (200s till729.5s)

    Pump voltage increased from 7V to 10V at 391.8s

    Pump voltage decreased from 10V to 7V at 542.8s

    -5

    0

    5

    10

    15

    20

    25

    30

    35

    0 100 200 300 400 500 600 700 800

    Volt(V)

    Time (s)

    Setpoint Control Level

    Start up - for

    system to reach

    steady state

    Pump voltage

    increased

    from 7V to10V

    Pump voltage

    decreased from

    10V to 7V

    0

    1

    2

    3

    4

    5

    6

    7

    8

    200 300 400 500 600 700 800

    Volt(V)

    Time (s)

    Setpoint Control Level

  • 8/11/2019 Group T4 - Expt C3

    26/46

    21

    4.1.6. Feedforward Control with Feedforward Gain = 1.0

    Figure 22. Response Graph for Feedforward Control with feedforward gain 1.0

    Figure 23. Response Graph for Feedforward Control with feedforward gain 1.0 (200s till

    725.9s)

    Pump voltage increased from 7V to 10V at 295.3s

    Pump voltage decreased from 10V to 7V at 513.6s

  • 8/11/2019 Group T4 - Expt C3

    27/46

    22

    4.1.7. Summary of Parameters Used to Evaluate Control System Performance

    The control system performance was evaluated based on three criteria, i.e. time taken for

    stabilization, maximum deviation from set point and the integral of absolute error (IAE).

    Time taken for stabilization is the time required for the system to return to the set point value,

    7.0V. It measures how efficient the system reacts and takes corrective action. By analyzing

    maximum deviation from set point and the integral of absolute error (IAE), the accuracy of

    corrective actions taken can be observed. The parameters used to evaluate the first two

    criteria are summarized inTable 1.

    Table 1. Comparison of the various feed forward gains 0, 0.2, 0.4, 0.6and 0.8

    Feed

    forwardgain= 0

    Feed

    forwardgain= 0.2

    Feed

    forwardgain= 0.4

    Feed

    forwardgain= 0.6

    Feed

    forwardgain= 0.8

    Feed

    forwardgain=1.0

    Pump Voltage Increased from 7V to 10V

    Time at which

    change took

    place (s)426.0 416.0 230.5 291.8 319.8 295.3

    Time when

    level stabilized

    back to

    setpoint (s)

    488.3 493.0 286.8 347.9 374.7 347.8

    Time taken forstabilization (s)

    62.3 77.0 56.3 56.1 54.9 52.5

    Maximum

    deviation of

    level from

    setpoint (V)

    +0.117 +0.031 +0.061 +0.139 -0.246 -0.334

    Pump Voltage Decreased from 10V to 7V

    Time at which

    change took

    place (s)

    804.7 712.1 484.2 545.5 542.8 513.6

    Time whenlevel stabilized

    back to

    setpoint (s)

    863.0 792.4 561.2 616.3 600.3 572.6

    Time taken for

    stabilization (s)58.3 80.3 77.0 70.8 57.5 59.0

    Maximum

    deviation of

    level from

    setpoint (V)

    -0.130 -0.032 +0.058 +0.114 +0.218 +0.313

  • 8/11/2019 Group T4 - Expt C3

    28/46

    23

    As observed from Figure 15 (system with feedback gain 0.2) and Figure 17 (system with

    feedback gain 0.4), the response obtained was apparently better with less fluctuation in the

    water level. The maximum deviation is relatively smaller compared to other systems with

    greater feedforward gains value as shown in Table 1.The results suggested that the control

    was under compensating for the disturbance.

    For systems with feedforward gain 0, 0.2, 0.4, 0.6, water level increased when a step up of

    the pump voltage from 7.0V to 10.0V as shown in Table 1. However, for system with

    feedforward gains 0.8 and 1.0, it was found that the water level decreased despite the step up

    of pump voltage. The observation could be due to an over compensation of increased pump

    voltage by the feedforward controller whereby it over reduced the valve voltage for greater

    feedforward gains of 0.8 and 1.0. It also showed that feedback control will bring the water

    level back to the desired set point level regardless of the extent or source of disturbance.

    When the feedback gain increased from 0.2 to 1.0, it was observed that maximum deviation

    of level from the setpoint increased. The feedforward controller was changing from under

    compensating to overcompensating the disturbance by adding too much of the disturbance

    signal to the actuation signal.

    It was also observed that the controller for feedforward gain 0 took corrective action onlyafter deviation in water level had taken place. It is because the controller is a feedback

    controller which will only take corrective action after the disturbance had been introduced. In

    contrast, for systems with feedforward gain of 0.2, 0.4, 0.6, 0.8 and 1.0, the controller took

    immediate corrective action when there was step up and step down of pump voltage. This

    was evident by the sharp drop in control in Figure 14,Figure 16,Figure 18,Figure 20and

    Figure 22.This illustrate that feedforward control reacts more effectively in arresting

    disturbance.

    Generally, the time taken for system stabilization decreased from system with feedforward

    gain 0.2 to 1.0 for both disturbances introduced from step up and step down of pump voltage

    as displayed inTable 1.

    To calculate integral of the absolute value of the error (IAE), the area of square under the

    curve are summed using excel since data point were collected at small time interval of 0.1s,

    rendering the approximation to integral of curve a valid estimation. IAE value is the

    difference area between the level curve and setpoint curve.

  • 8/11/2019 Group T4 - Expt C3

    29/46

    24

    Table 2. Summary of IAE at respective feedforward gain value

    Gain Total Run Time e(t) IAE

    0 1247.8 2325.845 232.5845

    0.2 799.2 2050.563 205.0563

    0.4 725.9 2031.896 203.1896

    0.6 725.9 2221.697 222.1697

    0.8 725.9 2207.177 220.7177

    1.0 725.9 2141.617 214.1617

    As shown in

    To calculate integral of the absolute value of the error (IAE), the area of square under the

    curve are summed using excel since data point were collected at small time interval of 0.1s,

    rendering the approximation to integral of curve a valid estimation. IAE value is the

    difference area between the level curve and setpoint curve.

    Table 2,system with feedforward gain 0.2 has the smallest integral of the absolute value of

    the error. Although it took longer time to reach setpoint after disturbances, it has least

    deviation from the setpoint overall. As discussed in the previous section that the time taken

    for stabilization decreased from feedforward gain 0.2 to 1.0, however, the maximum

    deviation from setpoint after disturbances introduced increased. From the analysis, it could be

    concluded system with feedback gain 0.2 is the optimum in this experiment.

    4.2.Cascade Control

    In the second part of the experiment, the system is subjected to 3 types of manipulation, i.e.set point change, disturbance in the form of inflow rate and disturbance on the form of

    outflow rate. The master controller adjusts the vessel level and provides the setpoint for the

    slave controller which then adjusts the flow of water in the process loop.

    4.2.1. Set point Change Perturbation

    In the first part of the cascade control experiment, the response of the system to level setpoint

    changes was observed.Figure 24 (a) and (b) shows the initial startup response and the

    response to level setpoint changes after it achieved steady state respectively. During the

  • 8/11/2019 Group T4 - Expt C3

    30/46

    25

    startup, the master controller output (F2-ref) suddenly jumped to a high value because there

    was a large error in the system. However, FT2 maintained at the maximum value of about

    4.6V because the pump was already operating at its maximum speed. As the water level

    approached the setpoint, master controller output gradually decreased, and fell below 4.6V.

    Finally, when the system reached steady state (within 99% of the setpoint), FT2 matched the

    value of the master controller output at around 2.5V.

    After the steady state, when the setpoint was changed from 7V to 8V, the master controller

    started to correct the error by giving higher output. Similar but opposite action was observed

    when the setpoint was changed back to 7V. However, there was oscillation in the water level

    as well as the controller output. This is probably due to inadequate tuning in the controller

    parameters. Furthermore, the asymmetric response to step changes in opposite directions also

    implies that there is nonlinearity in the system. In addition, fromFigure 24 (b), FT2 seems to

    lag behind F2-ref for both setpoint changes. This is because the master controller detected the

    deviation from the setpoint and kicked in first. Subsequently, the master controller output was

    fed into the slave controller as a new setpoint, therefore its reponse tailed that of master

    controller.

    (a)

    -20

    0

    20

    40

    60

    80

    100

    120

    140

    0

    1

    2

    3

    4

    5

    6

    7

    8

    0 10 20 30 40 50 60 70 80 90

    Volt(V)

    Volt(V)

    Time (s)

    Setpoint Level FT2 (secondary axis) F2-ref (secondary axis)

  • 8/11/2019 Group T4 - Expt C3

    31/46

    26

    (b)

    Figure 24. Response graph for (a) initial start and (b) level set point changes (From 7V to 8V

    and back to 7V).

    The maximum offset, rise time, and settling time for water level were calculated and

    tabulated inTable 3.It can be seen that the water level has a larger maximum error when the

    setpoint was reduced from 8V to 7V. In addition, although the rise time for the setpoint

    change from 8V to 7V is shorter, the settling time is longer. This means that the controller is

    more aggressive to negative offset, but it caused oscillation and therefore gave longer settling

    time.

    Table 3. Output variable (water level) response to setpoint changes

    Maximum erro after

    rise time (V)

    Rise time

    (s)

    Settling time within 99%

    of level setpoint (s)

    Setpoint change 7V

    to 8V-0.101 10.8 22.6

    Setpoint change 8V

    to 7V+0.705 6.2 36

    -20

    -10

    0

    10

    20

    30

    40

    50

    0

    1

    2

    3

    4

    5

    6

    7

    8

    9

    75 95 115 135 155 175 195 215 235 255 275

    Volt(V)

    Volt(V)

    Time (s)

    Setpoint Level FT2 (secondary axis) F2-ref (secondary axis)

  • 8/11/2019 Group T4 - Expt C3

    32/46

    27

    4.2.2. Disturbance to SystemChange in Inflow Rate

    In the next part of the experiment, secondary disturbance in the form of inflow rate was

    introduced to the system. FromFigure 25,when the proportional valve voltage was changed

    from 10V to 8V and from 8V back to 10V, there was no significant change in the water level

    This means that the slave controller was able to give additional corrective action to counteract

    the secondary disturbance caused by the change in the inflow rate. Together with the master

    controller action, the response was fast thus the water level was not affected much. Similarly

    inFigure 26,no significant offset was observed when there was a 4V change in the inflow

    rate. However, the controller action was larger compared to 2V change in the inflow rate.

    Figure 25. Response graph for inflow rate change 10V8V10V

    Figure 26. Response graph for inflow rate change 10V6V10V

    0

    0.5

    1

    1.5

    2

    2.5

    3

    6.8

    6.85

    6.9

    6.95

    7

    7.05

    7.1

    7.15

    7.2

    100 200 300 400 500 600

    Volt(V)

    Volt(V)

    Time (s)Setpoint Level 10V to 8V

    8V to 10V FT2 (secondary axis) F2-ref (secondary axis)

    0

    0.5

    1

    1.5

    2

    2.5

    3

    6.8

    6.85

    6.9

    6.95

    7

    7.05

    7.1

    7.15

    7.2

    450 550 650 750 850 950 1050

    Volt(

    V)

    Volt(

    V)

    Time (s)

    Setpoint Level 10V to 6V

    6V to 10V FT2 (secondary axis) F2-ref (secondary axis)

  • 8/11/2019 Group T4 - Expt C3

    33/46

    28

    When the inflow rate was changed from 10V to 4V, there was a much larger offset in the

    water level as seen inFigure 27.This suggests that the disturbance was too large for the slave

    controller to handle, and therefore caused an offset in the water level. Consequently, master

    controller countered this offset by increasing its output. However, it was observed that the

    water level increased very slowly back to the setpoint. This indicates that the flow rate into

    the vessel was limited regardless of the pump speed because the valve was closed

    significantly. Furthermore, this level offset also caused the master controller output to

    decrease very slowly and was unable to return to its previous steady state value.

    Subsequently, when the inflow rate was changed back to 10V, the water level shot up beyond

    the setpoint. This is because there was a delay in the controller response caused by the

    integral action of the controller, making it unable to change its output fast enough from the

    high value, and eventually causing a large volume of water being pumped into the vessel.

    This is referred to as reset windup of the integral action. Nevertheless, the water level was

    able to return to the setpoint relatively fast after that, and the controller output could go back

    to its previous steady state value as well.

    Figure 27. Response graph for inflow rate change 10V4V10V

    0

    1

    2

    3

    4

    5

    6

    7

    8

    9

    6.5

    6.6

    6.7

    6.8

    6.9

    7

    7.1

    7.2

    7.3

    7.4

    7.5

    1000 1100 1200 1300 1400 1500 1600 1700

    Volt(V)

    Volt(V)

    Time (s)

    Setpoint Level 10V to 4V

    4V to 10V FT2 (secondary axis) F2-ref (secondary axis)

  • 8/11/2019 Group T4 - Expt C3

    34/46

    29

    When the inflow rate was further reduced by changing the proportional valve voltage from

    10V to 2V, the flow into the vessel was so small that the level kept dropping despite

    increasing controller output, as shown inFigure 28.This is because the pump was already

    operating at its maximum speed, however, because the valve opening was too small, the flow

    into the vessel was limited. Finally, the level dropped to almost zero, and the controller

    output reached a maximum value, because the integral action of the controller was put out of

    action to prevent the controller output to increase excessively.

    Figure 28. Response graph for inflow rate change 10V2V10V

    4.2.3. Disturbance to System Change in Outflow Rate

    In the last part of the experiment, primary disturbance in the form of outflow rate wasintroduced to the system. This disturbance directly affects the water level and is outside the

    direct control of the slave controller. Here, the master controller corrects for the deviation and

    provides a new set point value for the slave controller which subsequently corrects the pump

    voltage accordingly to restore the water level to that of set point.

    FromFigure 29,the water level and controller output did not change significantly when the

    drain valve was 1/3 closed. However, when the valve was opened back fully, there was a

    notable deviation of water level from the setpoint. Consequently, the master controller output

    0

    20

    40

    60

    80

    100

    120

    140

    160

    0

    1

    2

    3

    4

    5

    6

    7

    8

    1550 1600 1650 1700 1750 1800 1850 1900 1950

    Volt(V)

    Time (s)

    Setpoint Level 10V to 2V

    FT2 (secondary axis) F2-ref (secondary axis)

  • 8/11/2019 Group T4 - Expt C3

    35/46

    30

    also increased in order to pump more water into the vessel to counteract the increased

    outflow rate. It is also noted that FT2 lagged behind FT-ref, this is because the slave

    controller had to wait for the master controller to detect the error first and then feed the

    output into the slave controller.

    Figure 29. Response graph for outflow rate change (drain valve fully open1/3 closed

    fully open)

    FromFigure 30,when the drain valve was 2/3 closed, the water level increased significantly

    and there were some oscillations before it finally reached the setpoint. The water level

    increased because the outflow rate decreased when the drain valve was closed. This primary

    disturbance affected the water level directly, causing it to increase. Subsequently, the mastercontroller responded to this deviation by decreasing its output to deliver less amount of water

    into the vessel. However, there was some oscillation in the controller output. This is probably

    due to inappropriate tuning of the controllers. Again FT2 lagged behind FT-ref because the

    slave controller had to depend on the master controller to detect the error first and then feed

    the output into the slave controller. When the drain valve was opened back fully, the water

    level dropped suddenly because of the increased outflow rate, causing a large deviation from

    the setpoint. Therefore, the master controller gave greater corrective action to pump more

    water into the vessel. Although no apparent oscillation was observed, the offset was larger

    0

    0.5

    1

    1.5

    2

    2.5

    3

    3.5

    6.8

    6.85

    6.9

    6.95

    7

    7.05

    7.1

    7.15

    7.2

    50 100 150 200 250 300 350 400 450 500

    Volt(V)

    Volt(V)

    Time (s)

    Setpoint Level Fully open to 1/3 closed

    1/3 closed to fully open FT2 (secondary axis) F2-ref (secondary axis)

  • 8/11/2019 Group T4 - Expt C3

    36/46

    31

    compared to offset obtained when closing the valve. Finally, the water level increased slowly

    back to the setpoint.

    Figure 30. Response graph for outflow rate change (drain valve fully open2/3 closed

    fully open)

    From the response graphs for change in inflow and outflow rates, it can be concluded that for

    regulatory control, the cascade controller was able to perform well to correct both primary

    and secondary disturbances with small magnitude. However, as the disturbance magnitude

    increases, the controller was unable to give satisfactory control probably due to the physical

    limitation of the actuator and insufficient fine tuning of the controller.

    0

    1

    2

    3

    4

    5

    6

    7

    8

    6.5

    6.6

    6.7

    6.8

    6.9

    7

    7.1

    7.2

    425 525 625 725 825 925

    Volt(V)

    Volt(V)

    Time (s)

    Setpoint Level Fully open to 2/3 closed

    2/3 closed to fully open FT2 (secondary axis) F2-ref (secondary axis)

  • 8/11/2019 Group T4 - Expt C3

    37/46

    32

    5. Discussion and Analysis

    5.1.

    Feedforward Control

    5.1.1.

    Block Diagram of Feedforward Control

    Discussion Question 2: In what circumstances would you recommend a feedforward control

    system rather than a cascade or a single loop feedback control system? Point out any

    limitation of the feedforward control system.

    In order to answer to this question, an understanding of limitations of feedback control or

    cascade control must be first developed. This is followed by discussing how feedforward

    control can help to overcome the stated limitation for achieving a better control. Lastly,

    limitations of feedforward control will be discussed as well.

    5.1.2.

    Limitations of Simple Feedback Control

    Simple feedback loop has its own advantage such as it can handle both known and unknown

    disturbances, it is easier to implement, it requires less understanding of the process and it

    ensures the controlled variable returns to desired set point eventually. Despite the advantages

    of feedback control, there are a few intrinsic limitations that restrain the use of feedback

    control.

    Firstly, feedback controller only reacts when deviation of controlled variables occurred. This

    means that the output has already been affected before the feedback controller can take

  • 8/11/2019 Group T4 - Expt C3

    38/46

    33

    action. If this limitation is coupled with large process dead time or large process time

    constant, the action from feedback controller is usually not satisfactory which results in

    transient behavior in the system.

    The second problem of feedback controller is the presence of closed-loop instability. When

    an open loop stable process is controlled by feedback control, the transfer function change to

    a complex one and yields different poles and zeros. If there happens to have positive poles,

    the system will become unstable and therefore inoperable. The controller in this case has to

    be carefully designed to prevent instability.

    Thirdly, feedback controller is usually more costly to operate. This is because larger control

    action has to be taken and this translates to more power consumption which eventually drives

    up the operational cost of the plant.

    5.1.3. Limitations of Cascade Control

    Cascade controller can improve the performance of feedback control, since it applies multiple

    feedback controls in the systems, with the slave system(s) having a faster response as

    compared to master control loop. However, cascade control also has its own limitation.

    First of all, sometimes it is difficult to identify a measurable secondary controlled variable in

    the system that has faster response than the primary controlled variable. In such case, cascade

    control cannot be implemented.

    Secondly, cascade control requires tuning of two controllers to achieve good control. The

    process of tuning might be difficult. Also, the introduction of multiple feedback loops

    increases the complexity of the characteristic equation, which is shown as the denominator of

    equation 4 in theoretical background session. Again, closed loop stability remains to be a

    problem.

    Thirdly, from economics point of view, feedback control is generally more expensive in

    operation as frequent control actions are taken to curb deviation. This applies to cascade

    control as well since it uses more sensors, controllers and actuators.

    5.1.4. Circumstances that feedforward control is recommended

    Since limitations of feedback and cascade control are known, the circumstances where

    feedforward control should kick in can be discussed.

  • 8/11/2019 Group T4 - Expt C3

    39/46

    34

    The first circumstance is when perfect control is necessary. Feedforward control measure the

    known disturbances of the system and compensate or react to the disturbances before they

    can affect the system. Thus, the system essential ensure the output is steady and this is

    essentially a perfect control.

    On top of that, perfect control has sound performance if the known disturbances have large

    impacts on controlled variable. This can be seen in the feedforward experiment above. Pump

    voltage, which translates to pump speed, is an important disturbance that affects the level of

    the process vessel (controlled variable). Comparison of single feedback loop(Kf=0) and

    feedforward control (Kf>0) showed that the level was well maintained in the case of

    feedforward control while in feedback control large changes occurred before controller can

    react to the deviation, i.e. level increased when step change of pump voltage occurred and

    returned to set point sluggishly.

    Moreover, feedforward control should be used if the system is large in size or slow in process

    and therefore has large time delay and time constant which system response would be in

    general slowly corrected if feedback control or cascade control is used.

    Lastly, one of the significant advantages that feedforward control is that it does not affect the

    system stability. As can be seen in section 2.3.1, inclusion of feedforward control to thefeedback control system does not affect the characteristic equation of the transfer function.

    As such, feedforward control implementation is easier for its minimum impact on the system.

    5.1.5. Limitations of Feedforward Controller

    Having said that feedforward control is a theoretical perfect control system, the system is

    usually affects by unknown and immeasurable disturbances which in effect make perfect

    control hard to achieve. In addition, physical realizability as shown in the theoretical

    background is another factor that restricts the feedforward control performance.

    Furthermore, known or measurement disturbances might not account for large deviation of

    controlled variable. If unknown or immeasurable disturbances contribute to major deviation

    of the controlled variables, feedforward control installation would be less effective as

    compared to feedback control and cascade control.

  • 8/11/2019 Group T4 - Expt C3

    40/46

    35

    Finally, an accurate process model is needed for implementation of feedforward control to avoid over compensation or under compensation with

    respect to the size of disturbance. The understanding of dynamics of process might be difficult which limits the use of feedforward control.

    5.2.Cascade control

    5.2.1. Block Diagram of Cascade Control

    Electrical signal [V]

    Electrical signal, FT2 [V]

    Flow sensor-transmitter (Analyzer)

    Flow Control Loop

    Inner Loop

    Level Control Loop

    Primary Loop

    Level sensor-transmitter (Analyzer)

    Process

    vessel level

    Set point

    for slave

    controller,

    F2-Ref [V]

    Pump 2 Proportional valve Flow

    Electrical

    signal,

    Level sp

    [V]

    Process

    vessel level

    set-point

    Level

    sensor-

    transmitter

    (Analyzer)

    (PI)(PI)

    Electrical signal, Level [V]

    Km1 Gc1

    Master

    Controller

    Gc2

    Slave

    Controller

    GvPr

    oces

    Control

    ValveD

    Gp2 Gp1

    Gd2 Gd1

    Gm2

    Gm1

    +-

    ++

    +-

    ++

    Ysp11

    ~

    spY 2~

    spY E1 E2

    Ym2

    Ym1

    P Y2 Y1

    D2 D1

  • 8/11/2019 Group T4 - Expt C3

    41/46

    36

    Discussion Question 2: What is the advantage of cascade control compared with single loop

    feedback control? Point out the limitation of cascade control system.

    Cascade control is a general improvement of single feedback control system if feedback

    control performance is not acceptable and if feedforward control is not able to put into action

    for reasons such as disturbances are not measurable. In this case, cascade control exploits a

    secondary measurement point which can better and faster reflects the deviation of the system

    and takes the corresponding corrections.

    5.2.2. Advantages of Cascade control compared with single loop feedback control

    Firstly, as slave control always has faster response control compared to the master control,

    the performance of control on system response would generally be better. From our

    experiment, we can observe that level fluctuation in single feedback loop was larger

    compared to cascade control. This can be partly explained by the increase value in master

    loop controller gain too. Generally, as pointed by Seborg et al, if the slave controller has

    faster response than master controller, the stability of the system will be enhanced and thus

    lead to a Kc value and a more robust response.

    Others advantages are relatively minor compared to the abovementioned, but they are worth

    mentioning. For instance, cascade control is widely accepted in industry and the tuning ofcontroller is similar to single feedback loop which utilizes PID controller. Also, cascade

    control is also good for final control element that is nonlinear in nature since it utilizes

    deviation of controlled variables.

    5.2.3. Limitations of Cascade Control

    The major limitations of cascade control have been discussed in Discussion question 2 in

    feedforward control and are briefly summarized below.

    1. Secondary measurement point, i.e. second controlled variable, might be difficult to

    identify;

    2.

    Tuning of two controllers which is generally more complex to tune;

    3. Closed loop instability remains a problem; and

    4. Cascade control is more economically expensive than feedback loop.

    Discussion Question 3: Look carefully at a single loop feedback control system and explain

    why, when the valve is considered, it can be considered as a cascade control?

  • 8/11/2019 Group T4 - Expt C3

    42/46

    37

    5.2.4. Classification of Feedforward Control as Cascade Control

    Firstly, in order for a control system to be consider as a cascade control system, the following

    criteria are necessary:

    1.

    They are at least two feedback loops;

    2.

    One of the feedback loop must be placed inside of the other feedback loop;

    3. There are two controllers, two controlled variables and one manipulated variable

    A general proportional valve is shown inFigure 31below. It shows that the valve has a

    feedback control loop that is built into the valve for controlling the valve stem position. Thus

    in this case, the master control loop controls the level (first controlled variable), the

    proportional valve controls the valve stem position to control flow.

    For feedforward control case, the manipulated variables are both the valve stem position,

    therefore in this case the single feedback loop can be considered as a cascade control system.

    The speed of action of master loop and slave loop should be about the same for proportional

    valve does not improve the loop response as compare to the case where the flow rate was the

    slave controller process output.

    For cascade control, since the manipulated variable is not the same in this case which the

    master loops manipulated variables is pump speed while valve built-in controller control

    valve stem position, as such, the system can only be loosely classified as a cascade system.

    Figure 31. A control amplifier connected to proportional valve

  • 8/11/2019 Group T4 - Expt C3

    43/46

    38

    5.3. Error Analysis

    1. Immeasurable disturbances are not taken into account to the study of the results in this

    experiment. For example, bubbling in the water may cause water level readings to be

    inaccurate and mistaken to be the system response towards the input disturbances.

    2. Assumption of ideal flow within pipes and valve may not be valid and hence

    contribute to the inaccuracy of the results.

    3. Assumption of steady state was attained after 5 minutes of observation might not be

    valid due to presence of unaccounted disturbances in the system. This can be seen

    from the oscillatory behavior in the graphs. Therefore, discretion had to be exercised

    in determining whether the steady state is reached.

    4. For the second part of the cascade control experiment, the degree of opening for the

    proportional valve and drain valve was not recorded automatically. Group members

    must coordinate accordingly to note down the time when the valves were opened or

    closed. Therefore, there might be some inaccuracies between the noted timing and

    actual timing.

    5. The change in outflow rate was controlled by closing the drain valve manually. As a

    result, the disturbance introduced might be different for each person operating the

    drain valve due to different degrees and speeds of closing. Therefore, the analysis for

    the response to change in outflow rate is only in qualitative terms.

  • 8/11/2019 Group T4 - Expt C3

    44/46

    39

    6. Conclusion

    In conclusion, the feedforward and cascade control system was successfully studied by

    observing the response curve and the corresponding controller action profile in response toset point changes and disturbances introduced to the system.

    In the first part of the experiment, the performance of the feedforward control was evaluated.

    The optimum feedforward gain used to maintain the vessel level at set point value was

    determined to be 0.2. At this setting, the maximum deviation from set point value and IAE

    was the smallest, implying that the best control was achieved out of the other gain values

    used in this experiment. It was observed experimentally that feedforward control tends to

    under- or over-compensate the disturbances depending on the feedforward gain value used

    which is one of the limitations of using feedforward control.

    In the second part of the experiment, the cascade control system gave rise to rapid corrective

    action thereby improving the control of level. This can be attributed to the presence of

    slave controller nested within the master controller loop which action kicks in rapidly

    should the disturbance manifests in the secondary variable under the control of secondary

    loop. This arrests the disturbance quickly and allows the system to stabilize within a shorter

    period of time.

  • 8/11/2019 Group T4 - Expt C3

    45/46

    40

    References

    1. Dale E. Seborg, Thomas F. Edgar, Duncan A. Mellichamp. (2004) Process Dynamics and

    Control. 2nd Edition, John Wiley & Sons, Inc.2. Chiu, M.S. & Lee, D.Y. (2010). Process Dynamics and Control: CN3121 [Lecture Notes].

    Singapore: National University of Singapore, Department of Chemical and Biomolecular

    Engineering.

    3. Industrial Electronics. (n.d.) Types of Control from a Proportional Control Valve.

    Retrieved September 9, 2011 from http://www.industrial-

    electronics.com/output_devices_amplifiers_valves_relays_variable-

    frequency_drives_stepper_motors_servomotors/Types_Control_Proportional_Control_Va

    lve.html.

    Table of Notation

    Alphabetical List

    Symbol Meaning

    D Load Disturbance

    e(t) Error, i.e. deviation from set point

    E ErrorG Transfer function

    K Gain

    P Controller output signal

    U Manipulated variable

    Y Controlled variable

    Subscript Meaning

    c Feedback controller

    d Disturbance

    f Feedforward controller

    m Feedback transmitter

    p Process

    sp Set point

    t Feedforward transmitter

    v Valve

    1 First

    2 Second

    Greek Symbol List

    Symbol Meaning

    Time constant

  • 8/11/2019 Group T4 - Expt C3

    46/46