Você está na página 1de 34

NANYANG TECHNOLOGICAL UNIVERSITY

SCHOOL OF CHEMICAL AND BIOMEDICAL ENGINEERING DIVISON OF CHEMICAL AND BIOMOLECULAR ENGINEERING

CH3702 Formal Lab Report


Experiment CE3 PROCESS CONTROL Prepared by:
Lau Jia Sheng U0920916H GP10

Date of Experiment:
22 March 2012

Experiment CE3 Process Control

Lau Jia Sheng U0920916H

TABLE OF CONTENT

1. Aim 2. Abstract 3. Principles


3.1 Linear Model 3.2 Model Identification 3.3 On/Off Controller 3.4 PID Controller 3.5 Closed-loop Performance

Page 3 Page 3 Page 4

4. Equipments and Materials 5. Procedures


5.1 Experiment A (On/Off Control) 5.2 Experiment B (Model Identification for Control using Inlet Flowrate) 5.3 Experiment C (PID control using Inlet Flowrate)

Page 11 Page 12

6. Results
6.1 Experiment A (On/Off Control) 6.2 Experiment B (Model Identification for Control using Inlet Flowrate) 6.3 Experiment C (PID Control using Inlet Flowrate)

Page 14

7. Discussion
7.1 Questions 7.2 Sources of errors and Possible Improvements to the experiment

Page 29

8. Conclusion

Page 34

Experiment CE3 Process Control

Lau Jia Sheng U0920916H

1. AIM
To study the on/off control and the dynamics of the level of a tank, derive a model for it using step-response analysis, design a Proportional-Integral-Derivative (PID) controller using Internal Model Control (IMC) method for level control, as well as study the effect of changes in P, I and D terms of the PID controller on closed-loop performance.

2. ABSTRACT
The equipment described in Section 4: Equipments and Materials was used in this experiment to study the dynamics of the level of a tank. Using the flowrates and tank level change with time, a model can be derived and controllers such as the PID controller can be designed. Also, by varying certain parameters, setpoint tracking and disturbance rejection analysis can be done. Performance of controllers with different parameters can then be evaluated through the determination of various performance measures. At the end of this experiment, one should be able to understand the on/off controller works and also how to design a PID controller using the IMC control method.

Experiment CE3 Process Control

Lau Jia Sheng U0920916H

3. PRINCIPLES
3.1 Linear Models Automatic control is normally used in process industries for tracking setpoint changes and rejecting disturbances. Commonly used controllers such as the on/off and PID controllers can be used to control the level of a tank. As illustrated in Figure 1, the inlet and outlet flowrates can be varied to control the tank level.

Figure 1. Controlling tank level via Inlet and Outlet Flow

By doing a mass balance, the system used in the experiment for both cases is:
(1)

Where h is the tank level, A is the cross-sectional area, Fout is the outlet flowrate and Fin is the inlet flowrate. For level control using inlet flowrate (Figure 1a), the outlet flowrate follows the power law and is dependent on the pressure head in the tank.
(2)

Substitute equation (2) into equation (1),


(3)

Experiment CE3 Process Control

Lau Jia Sheng U0920916H

Linearizing equation (3) at the nominal operating points of h and F in and then taking laplace transforms, the linear model between h and Fin is:
(4)

where

and

are the deviation variables,

and

are the level and inlet flowrate at and

nominal operating point. The deviation variables can be expressed as .

Equation (4) is a first-order process model (Refer to section 3.2 for model identification). The gain and time constant of the process is and respectively.

As for level control using outlet flowrate (Figure 1b), the outlet flowrate can be modeled to vary linearly with pump speed:
(5)

where Ps is the speed of pump and c is a proportionality constant. Similarly, by substituting equation (5) into equation (1) and performing both linearization and Laplace transform, the linear model for level control using outlet flowrate is:
(6)

where , and are the deviation variables, , and are the level, pump speed and inlet flowrate at nominal operating point. The deviation variables can be expressed as , and .

Equation (6) is an integrating process model, and c/A is the slope of the integrator. F in is acting as the disturbance in this case.

3.2 Model Identification A model of the process can be found through step response analysis to eliminate the need to calculate unknown parameters (a, b, c and A) in derived models (Equation 4 and 6).

Experiment CE3 Process Control

Lau Jia Sheng U0920916H

In practice, processes can be modeled as a first-order process with time delay:


(7)

where y is the output of the process, u is the input of the process, K is the gain, is the time constant and is the time delay of the process. Figure 2a shows the response of y due to a step change in u for a process with K = 10, = 10 and = 100. In this case, the transfer function is:
(8)

Figure 2. Response of y for Step change in u

The gain of the process, K, can be estimated as

, can be estimated by measuring the time

where there is no change in y after a change in u. The time constant can be calculated using the 63.2% method. At 63.2% of y, time value is ( + ). As shown on Figure 2a by the red line, ( + ) = 110. Since = 10, = 110 = 100. Processes can also be modeled as an integrating process such as:
(9)

where is the time delay and K can be estimated by the slope of y as shown in Figure 2b.

Experiment CE3 Process Control

Lau Jia Sheng U0920916H

3.3 On/Off controller An on/off controller is a simple controller that can keep y between the upper and lower limit, y U and yL. With on and off mode, u is set to the maximum u U and uL respectively. Correspondingly, y will reach its max value - yU and minimum value yL. The on/off controller is also known as bang-bang controllers because the input continuously changes between its maximum and minimum values. A common used for such controller is in the house-hold controller where cooler is switched on when the temperature rise above a desired value, and switches off when temperature is low.

Figure 3. On/Off Controller

The closed-loop response for the process in (9) is shown in Figure 3. Due to the inertia and time delay of the system, y is allowed to exceed its upper and lower limits. Therefore it is important to choose the limits conservatively so that the actual upper limit is within the allowable upper limit (e.g. yU < actual allowable upper limit) There are 2 types of on/off controller used in this experiment, namely: 1) Level switch 2) Differential level switch There is a slight difference between level and differential level switch. For level switch, the difference between the upper and lower limits of level are fixed (fixed deadband), whereas in differential level switch, both upper and lower limits can be changed independently (variable deadband).

Experiment CE3 Process Control

Lau Jia Sheng U0920916H

3.4 PID Controller The most widely used controller is the PID controller. It can be represented as:
(10)

where P, I and D stands for Proportional gain, Integral time and Derivative time respectively. By defining the error as e(t) = r(t) y(t), where r(t) is the setpoint for desired y(t) value, we can compute u(t) as:
(11)

For this experiment, the Internal Model Control (IMC) tuning procedure will be used to find the parameters P, I and D. IMC tuning parameters are as shown:
Process First-order process: Integrating process: P I min{ , 4(c + )} 4(c + )
Table 1. Recommended tuning parameters for PID controller (IMC)

D 0 0

In Table 1, c is the closed-loop time constant, where a smaller value corresponds to a faster response. The D term is often set to zero for first order and integrating processes. For higher order processes, D is non-zero. In some cases where proportional gain P cannot be specified directly, proportional band would be used instead. It is defined as the inverse of P:
(12)

3.5 Closed-loop Performance To compare the performance of different controllers, various metrics are needed. For on/off controllers, oscillation period and amplitude are used as the measures. A well performing on.off controller will have oscillation amplitude as small as possible and oscillation as large as possible. The following are more general performance measures:

Experiment CE3 Process Control

Lau Jia Sheng U0920916H

1) Rise time (Tr): the time it takes for the output to first reach 90% of its final value, which is required to be small. 2) Settling time (Ts): time after which the output remains within e% of its final value (typically e = 5), which is usually required to be small. 3) Overshoot: the peak value divided by the final value, which should typically be 1.2 (20%) or less. 4) Decay ratio: the ratio of the second and first peaks, which should typically be 0.3 or less. 5) Steady-state offset: the difference between the final and the desired final value, which is required to be small.

Figure 4. Measures for closed-loop performance

Both the rise time and settling time are used to measure the speed of response. On the other hand, the overshoot, decay ratio and steady-state offset are used to measure the quality of response. Besides addressing the output response, y(t), through the 5 measures mentioned above, one should consider the magnitude of the manipulated input (control signal, u), which usually should be as small and smooth as possible. Smoothness can be determined by calculating the total variation:

(13)

Experiment CE3 Process Control

Lau Jia Sheng U0920916H

Total variation equation measures the total up and down movement of u, which should be small. Closed-loop performance is measured for setpoint changes as well as disturbance changes. During the disturbance changes, the reference value for u(t) is changed to 0 from 1.

10

Experiment CE3 Process Control

Lau Jia Sheng U0920916H

4. EQUIPMENTS AND MATERIALS


The experimental setup is shown in Figure 5 below. It consists of a clear cylindrical tank with an inner and outer cylinder. Water flows into the space (annulus) in between the outer and inner cylinders from the bottom of the tank and water level is measured with the level scale pasted on the outer cylindrical wall. The drain valve is located at the bottom of the tank. Both level switch and differential level switch are mounted on the lid and extends into the annulus. The depth of the switches into the annulus can be adjusted at the lid.

Figure 5. Schematics of experimental setup

The experimental setup is connected to a computer and operated through the interface of software PCT40. Data collection and process monitoring is also available through the PCT40.

11

Experiment CE3 Process Control

Lau Jia Sheng U0920916H

5. PROCEDURES
5.1 Experiment A (On/Off Control) 1. The software PCT40 was launched and experiment 1 was loaded. Equipment was connected as shown in the schematic diagram (Figure 4). Make sure SOL1 is connected. Inlet flow meter valve was open fully. 2. The level switch was positioned near the centre of the tank (around 150mm mark). 3. In PCT40, the control mode was switched from manual to level switch. Following the adjustment, the performance was observed and recorded. 4. Subsequently, a disturbance was introduced to the system by opening the SOL2 valve through the PCT40 software. The performance was observed and recorded. Disturbance rejection analysis was then done. 5. SOL2 valve was switched off. Process is left to achieve steady state. 6. Now in the PCT40, control mode was set to differential switch and the performance was observed and recorded. 7. Similar to step 4, a disturbance was introduced to the system by opening the SOL2 valve through the PCT40 software. The performance was observed and recorded. Disturbance rejection analysis was then done 8. The variable deadband was adjusted by shifting the relative of the two sensors and performance analysis was done. Repeat Step 6 to 8, 2 more times for different deadband.

5.2 Experiment B (Model Identification for Control using Inlet Flowrate) 1. The software PCT40 was launched and experiment 2 was loaded. Equipment was connected as shown in the schematic diagram. Make sure PSV is connected. Inlet flow meter valve was open fully. 2. In the PCT40 interface, the proportioning solenoid valve (PSV) was set to 30% open. The tank was allowed to reach about half full before the opening of drain valve. Drain valve was used to achieve steady-state in the system. (e.g. Steady-state is considered achieved when tank level did not fluctuate more then 0.2mm.) 3. A step change of approximately 5% was introduced in PSV. (e.g. Setting PSV from 30% to 40%). The process is allowed time to reach steady-state.

12

Experiment CE3 Process Control

Lau Jia Sheng U0920916H

4. The step-response method is then used to identify a first order model of the process.

5.3 Experiment C (PID control using Inlet flowrate) 1. Similar to Experiment B, PCT40 was launched and experiment 2 was loaded. Equipment was connected as shown in the schematic diagram. Make sure PSV is connected. Inlet flow meter valve was open fully. 2. In the PCT40 interface, the proportioning solenoid valve (PSV) was set to 30% open. The tank was allowed to reach about half full before the opening of drain valve. Drain valve was used to achieve steady-state in the system. 3. For Experiment C, due to time constraints, it is assumed from previous experimental runs that K = 0.1006, = 10 sec and = 695 sec. c is approximated to be half of and the PID parameters are calculated. 4. Setpoint value was set around 110 mm. In the PCT40 interface, proportional band (PB) was set to computed value, I = 0 and D = 0. The control mode was then set to automatic. 5. When process became steady, setpoint is increased by 10 mm and performance is observed and recorded. 6. Subsequently, process is allowed to reach steady state before setpoint is decreased by 10 mm back to the original setpoint. 7. SOL2 valve was opened to introduce a disturbance to the process. Performance is then observed and recorded. 8. Next, like in Step 5, setpoint value was then set around 110 mm. In the PCT40 interface, proportional band (PB) and Integrative time (I) was set to computed value, and D = 0. The control mode was then set to automatic. 9. Repeat Steps 5 to 7. 10. For the last part of the experiment, Step 8 was repeated. This time, instead of introducing a increase of setpoint by 10 mm, small increments of 2 mm was made 5 times at 10 seconds interval. 11. Process is allowed to reach steady state and performance is observed and recorded.

13

Experiment CE3 Process Control

Lau Jia Sheng U0920916H

6. RESULTS
6.1 Experiment A (On/Off Control)

Graph of Level VS Time (Overall)


172.00

171.50

171.00

Level (mm)

170.50

SOL2 opened

170.00

169.50

169.00

168.50 80 130 180 Time (s) 230 280 330

Graph 1. Graph of Level vs Time (Level switch)

At t = 95s to t = 119s, the process has achieved steady state and level was not fluctuating more than 0.2mm. At t = 120s, SOL2 valve was opened and a disturbance was introduced to the system.

14

Experiment CE3 Process Control

Lau Jia Sheng U0920916H

Graph of Level VS Time (Normal Operation)


169.44 169.42 169.40 169.38 Lvel (mm) 169.36 169.34 169.32 169.30 169.28 169.26 169.24 90 95 100 Oscillation Period 105 Time (s) 110 115 120 Oscillation Amplitude SOL2 opened

Graph 2. Graph of Level vs Time (Level switch) Normal Operations

Graph of Level VS Time (Disturbance Rejection)


172.00 171.50 171.00 Settling Time Level (mm) 170.50 170.00 169.50 169.00 168.50 100 120 140 160 180 200 Time (s) 220 240 260 280 300

Graph 3. Graph of Level vs Time (Level switch) Disturbance Rejection

15

Experiment CE3 Process Control

Lau Jia Sheng U0920916H

Graph of Level VS Time (Overall)


240

235

230

SOL2 opened

Level (mm)

225 SOL2 closed 220

215

210

205 0 100 200 300 Time (s) 400 500 600 700

Graph 4. Graph of Level vs Time (Differential switch) - Run 1 (deadband 30mm)

At t = 0s to t = 125s, Differential switch was turned on. At t = 126s, SOL2 valve was opened and a disturbance was introduced to the system. At t = 562s, SOL2 valve was closed.

16

Experiment CE3 Process Control

Lau Jia Sheng U0920916H

Graph of Level VS Time (Normal Operation)


237.52 237.50 237.48 237.46 Level (mm) 237.44 237.42 237.40 237.38 237.36 237.34 237.32 237.30 0 20 40 60 Time (s) 80 100 120 140 Oscillation Period Oscillation Amplitude SOL2 opened

Graph 5. Graph of Level vs Time (Differential switch) - Run 1 (deadband 30mm) Normal Operation

Graph of Level VS Time (Disturbance Rejection)


240.00 235.00 230.00 Level (mm) 225.00 220.00 215.00 210.00 205.00 0 Settling Time 100 200 300 Time (s) 400 500 600 700

Graph 6. Graph of Level vs Time (Differential switch) - Run 1 (deadband 30mm) Disturbance Rejection

17

Experiment CE3 Process Control

Lau Jia Sheng U0920916H

Graph of Level VS Time (Overall)


235

230

225 Level (mm) SOL2 opened 220

215

SOL2 closed

210

205 0 50 100 150 200 250 300 350 400 450 Time (s)

Graph 7. Graph of Level vs Time (Differential switch) - Run 2(deadband 20mm)

At t = 0s to t = 81s, Differential switch was turned on. At t = 82s, SOL2 valve was opened and a disturbance was introduced to the system. At t = 381s, SOL2 valve was closed

18

Experiment CE3 Process Control

Lau Jia Sheng U0920916H

Graph of Level VS Time (Normal Operation)


229.18 229.16 229.14 229.12 Level (mm) 229.10 229.08 229.06 229.04 229.02 229.00 228.98 0 10 20 30 40 Time (s) 50 60 70 80 90 Oscillation Amplitude SOL2 opened

Graph 8. Graph of Level vs Time (Differential switch) - Run 2(deadband 20mm) Normal Operation

Graph of Level VS Time (Disturbance Rejection)


235.00 230.00 225.00 Level (mm) 220.00 215.00 210.00 205.00 50 Settling Time 100 150 200 Time (s) 250 300 350 400

Graph 9. Graph of Level vs Time (Differential switch) - Run 2(deadband 20mm) Disturbance Rejection

19

Experiment CE3 Process Control

Lau Jia Sheng U0920916H

Graph of Level VS Time (Overall)


222

220

218

SOL2 opened

Level (mm)

216

214

SOL2 closed

212

210

208 0 50 100 150 Time (s) 200 250 300 350

Graph 10. Graph of Level vs Time (Differential switch) - Run 3(deadband 10mm)

At t = 0s to t = 99s, Differential switch was turned on. At t = 100s, SOL2 valve was opened and a disturbance was introduced to the system. At t = 311s, SOL2 valve was closed.

20

Experiment CE3 Process Control

Lau Jia Sheng U0920916H

Graph of Level VS Time (Normal Operation)


221.28 221.26 221.24 221.22 Level (mm) 221.20 221.18 221.16 221.14 221.12 221.10 221.08 0 10 20 30 40 50 Time (s) 60 70 80 90 100 Oscillation Period Oscillation Amplitude SOL2 opened

Graph 11. Graph of Level vs Time (Differential switch) - Run 3(deadband 10mm) Normal Operation

Graph of Level VS Time (Disturbance Rejection)


222.00 220.00 218.00 Level (mm) 216.00 214.00 212.00 210.00 Settling Time 208.00 50 100 150 200 Time (s) 250 300 350

Graph 12. Graph of Level vs Time (Differential switch) - Run 3(deadband 10mm) Disturbance Rejection

21

Experiment CE3 Process Control

Lau Jia Sheng U0920916H

Normal Operation: Oscillation period (s) ^ 4 36 * 18 Oscillation amplitude (mm) 0.14 0.14 0.14 0.14

Level Switch Differential level switch (Run 1) Differential level switch (Run 2) Differential level switch (Run 3)
Table 2. Data Table for Normal Operation
^

The oscillation period for all runs are the estimated average oscillation period. *The system did not present any sign of oscillation

Disturbance Rejection Analysis: Settling Time (s) & 140 - 120 = 20 178 - 126 = 52 120 - 82 = 38 124 - 100 = 24

Level Switch Differential level switch (Run 1) Differential level switch (Run 2) Differential level switch (Run 3)
&

Table 3. Data Table for Disturbance Rejection Analysis

The settling time is taken as the time after which the level oscillations with constant amplitude and period (after introduction of disturbance).

22

Experiment CE3 Process Control

Lau Jia Sheng U0920916H

6.2 Experiment B (Model Identification for Control using Inlet Flowrate)

Graph of Level VS Time (Model Identification)


160.00 150.00 140.00 Level (mm) 130.00 120.00 110.00 100.00 90.00 0 200 400 600 800 Time (s) 1000 1200 1400 1600

Graph 13. Graph of Level vs Time (Model Identification)

Using Figure 2, the system can be identified as a First-Order Process model. Initial value of level Initial value of PSV Initial Flowrate Final value of level Final value of PSV Final Flowrate K : 99.27 mm : 30 % : 759 cm3/min : 156.08 mm : 40 % : 875 cm3/min : 0.04897 min/cm2 : 0.1 min : 5.533 min

23

Experiment CE3 Process Control

Lau Jia Sheng U0920916H

Taking average,

is the time delay of the process the period of time when output was not changed after PSV is changed. From the below table: Time (s) 134 135 136 137 138 139 140 141 Level (mm) 99.12 99.27 99.12 99.12 99.27 99.27 99.27 99.41 Flowrate (cm3/min) 756 756 765 778 793 803 817 832 PSV (%) 30 40 40 40 40 40 40 40

Table 4. Extracted data table for time delay determination

= 140 134 = 6 seconds is determined using the 63.2% method. At y = [0.632 * (156.08 99.27)] + 99.27 = 135.17 mm, = 472 - 140 = 332 seconds = 5.533 min

Apply IMC tuning parameters for first-order process, Assuming c = 0.5 = 0.5*(332) = 166 s = 2.767 min,

min

24

Experiment CE3 Process Control

Lau Jia Sheng U0920916H

6.3 Experiment C (PID Control using Inlet Flowrate) Run 1: PB = PB*, I =0, D = 0

Graph of Level VS Time (Setpoint Tracking)


126.00 124.00 122.00 120.00 Level (mm) 118.00 116.00 114.00 112.00 110.00 108.00 106.00 Steady State 0 200 Setpoint adjusted to 120mm 323 400 600 Time (s)
Graph 14. Graph of Level vs Time (Expt C Run 1) - Setpoint Tracking

Setpoint adjusted to 110mm Setpoint adjusted to 110mm

800

1000

1200

Sample Calculation (Using region within dotted line)

where u is the control signal: flowrate.

25

Experiment CE3 Process Control

Lau Jia Sheng U0920916H

Graph of Level VS Time (Disturbance Rejection)


111.00 110.00 109.00 Level (mm) 108.00 107.00 106.00 105.00 104.00 0 50 100 150 200 250 Time (s)
Graph 15. Graph of Level vs Time (Expt C Run 1) - Disturbance Rejection

Steady State

Disturbance introduced at 54 seconds

b a Settling Time

300

350

400

450

500

Sample Calculation

where u is the control signal: flowrate.

26

Experiment CE3 Process Control

Lau Jia Sheng U0920916H

Run 2: PB = PB*, I =I*, D = 0

Graph of Level VS Time (Setpoint Tracking)


130.00 125.00 120.00 Level (mm) 115.00 110.00 105.00 100.00 95.00 90.00 0 100 200 300 400 500 Time (s)
Graph 16. Graph of Level vs Time (Expt C Run 2) - Setpoint Tracking

600

700

800

900

1000

Graph of Level VS Time (Disturbance Rejection)


111.00 110.00 109.00 108.00 Level (mm) 107.00 106.00 105.00 104.00 103.00 102.00 101.00 100.00 0 500 1000 1500 Time (s)
Graph 17. Graph of Level vs Time (Expt C Run 2) - Disturbance Rejection

2000

2500

3000

27

Experiment CE3 Process Control

Lau Jia Sheng U0920916H

Run 3: PB = PB*, I =I*, D = 0 (Stepwise increment of setpoint)

Graph of Level VS Time (Setpoint Tracking)


124.00 122.00 120.00 Level (mm) 118.00 116.00 114.00 112.00 110.00 0 100 200 300 Time (s)
Graph 18. Graph of Level vs Time (Expt C Run 3) - Setpoint Tracking

400

500

600

Setpoint Tracking Analysis: Overshoot PB = PB*, I = 0, D = 0 PB = PB*, I = I*, D = 0 Stepwise increment 1.034 1.038 1.009 Decay ratio 0.142 0.129 0.252 Settling time 142 170 355 Rise time 11.0 10.5 46.0 Offset 0.57 0.28 0.57 TV (u) 3090.000 2953.857 569.824

Table 5. Setpoint Tracking Analysis Table

Disturbance Rejection Analysis: Overshoot PB = PB*, I = 0, D = 0 PB = PB*, I = I*, D = 0 0.982 0.922 Decay ratio 0.151 0.281 Settling time 176 1916 Offset -3.34 -0.23 TV (u) 1002.69 1298.58

Table 6. Disturbance Rejection Analysis Table

28

Experiment CE3 Process Control

Lau Jia Sheng U0920916H

7. DISCUSSION
7.1 Questions 1. Discuss the effect of change in deadband on the performance of on/off controller. The performance of the on/off controller can be analyzed by the oscillation period and amplitude. Theoretically, when the variable deadband is reduced, a decrease in both oscillation period and amplitude would be observed, vice versa, for the same input magnitude. In Experiment A, when the variable deadband was decreased by approximately 10mm between each run of the differential level switch setup (e.g. Run 1: 30mm, Run 2: 20mm, Run 3: 10 mm), a decrease in oscillation period was observed. A lower oscillation period is undesirable because across the same amount of time, there would be more overshoot by the peaks of the output due to inertia and time delay (as explained in Section 3.3). There was no change in oscillation amplitude because the process could be at a steady state. The performance of the on/off controller can also be measured using the settling time. In this case, the settling time decreased when the deadband was reduced. This shows that a differential level switch with smaller deadband can reject disturbances more effectively than a differential level switch with a larger deadband, because it stabilizes the process back to a steady state in a shorter time. The effect of change in deadband on the performance of on/off controller cannot be concluded because it depends on the requirement of the operation. If disturbance rejection is of the highest priority, a differential level switch of a smaller deadband would be the better performing one because it can reject disturbance better. However, if the level of water in the tank is of the highest priority, a larger deadband differential level switch would be the better performing controller because the frequency where output exceeds the allowable upper and lower limit would be lesser.

29

Experiment CE3 Process Control

Lau Jia Sheng U0920916H

2. Does the step change analysis describe here yield a reasonable first-order model? Suggest an alternative method that can give a better model. Yes, the step change analysis here yields a reasonable first-order model. Graph 13 is similar to the output response of a first-order process in Figure 2, which is in the form of:

In this experiment, time constant () and time delay () were all determined using the 63.2% method. An alternative method that can give a better model for a first-order model with time delay is the Fraction Incomplete method, where two points corresponding to 35.3% (t1) and 85.3% (t2) y on the response curve are used to determine and . The correlation for finding these parameters are:
(14) (15)

At 35.3% y = 119.21, t1 = 154 s At 85.3% y = 147.73, t2 = 580 s

Model using Fraction Incomplete method: Model using 63.2% Model: The Fraction Incomplete method would give a more accurate representation of the process because 2 points instead of 1 point were used.

30

Experiment CE3 Process Control

Lau Jia Sheng U0920916H

3. When the setpoint is increased in small steps, do you observe better performance? Explain. Comparing output response in Graph 18 and Graph 16 for increment of setpoint from 110mm to 120mm for large and small step increase in setpoint respectively, we can see that increasing the setpoint in small steps reduces the overshoot of the process and the total variation value was significantly smaller (about 4 times). However, there was an increase in decay ratio, settling time, rise time and offset of the process. The process took approximately twice as long to reach steady state and the time required to first hit 90% of the final output value was quadrupled. The higher offset value was most probably due to the longer time needed to reach steady state (i.e. will eventually hit 0.28 mm if given more time) Similar to question 1 of this discussion section, defining better performance would depend on how these two settings are used. If the operation requires a lower overshoot of the output due to equipment limitations, a stepwise increment would be preferred because this would not cause unnecessary stress on the equipment. Also if the process happens to be located far downstream in a plant, a stepwise increment would be more favorable because the change in input to this process would be less drastic and smoother (lower TV(u) value), avoiding problem such as the lack of input. Therefore small step increment would be the better performing one if considering such operating conditions. However, if the process requires a fast response (i.e. achieving steady state in a short time), and is not limited by the input, a large step setpoint change would be the better performing one because the offset in such setting is lesser.

31

Experiment CE3 Process Control

Lau Jia Sheng U0920916H

7.2 Sources of errors and Possible Improvements to the experiment 1. Lack of Oscillation In experiment A, there was no oscillation observed for all runs using the level switches during normal operation. This could be due to the system achieving steady-state, meaning the draining rate is equals to the inlet flowrate. This resulted in a relatively straight line and the determination of oscillation period and amplitude was inaccurate. The drain valve should be opened to allow a larger flowrate until oscillations are clearly visible. 2. Setpoint/Parameter change in Experiment B and C At instances where changes to the setpoints/parameter were to be made, human error might be introduced. The interface of PCT40 did not allow users to effect the change only after adjustments. In other words, all parameters changes immediately after any user input. This would cause fluctuations in readings and eventually affect the calculations. To minimize such human error, values could be first prepared in Notepad or Word and then pasted over to the interface once changes are required. 3. Default values used in Experiment C Due to the lack of time, tuning parameters were not calculated from readings of Experiment B. Results from previous lab groups were used instead. This would cause the PID controller to be overtuned or undertuned resulting in instability of the system. As seen in Table 7 below, the Proportional Gain used is about half of the experimental one. This would result in a less sensitive controller because a larger change in input is required to change the output by a same amount. Also from the table, it can be seen that the Integral term is twice of the value calculated from the experimental result. A larger integral term would result in a sluggish system as seen in disturbance rejection in Graph 17. It would also reduce the offset of the system. Experimental 0.04897 0.1 5.533 2.767 39.410 5.533 0 0.025374 Other group 0.1006 0.167 11.583 5.792 19.325 11.583 0 0.052

K (min/cm ) (min) (min) c (min) P I D PB

Table 7. Comparison of tuning parameters

32

Experiment CE3 Process Control

Lau Jia Sheng U0920916H

Experimental readings therefore should be used for PID controller tuning for better performance. 4. Limitation of PSV/SOL1 When trying to introduce a step change to the system, limitations arose in the PSV and SOL1 valves. It is difficult to find any valves that can immediately open to the new value due to water pressure. This resulted in a smooth change in input instead of a step change, affecting the output response. Because input change is assumed to be step-wise, the model derived from experimental values could be less accurate.

33

Experiment CE3 Process Control

Lau Jia Sheng U0920916H

8. CONCLUSION
The aim of studying the on/off control and the dynamics of the level of a tank, deriving a model for it using step-response analysis, designing a Proportional-Integral-Derivative (PID) controller using Internal Model Control (IMC) method for level control, as well as studying the effect of changes in P, I and D terms of the PID controller on closed-loop performance was achieved. After carrying out the various experiments using the experimental setup, output responses in the form of Tank level against Time graphs were obtained. Through the graphs and data, setpoint tracking and disturbance rejection analysis were done. It was found that the use of different controllers and the use of different tuning parameters would affect the behavior and performance of the system. It is therefore important for us to understand the requirement of each operation so that the right controller and tuning parameters are chosen. Although anomalies exist in the results and calculated values do not match theoretical values, this experiment was successful in showing us the dynamic behavior of the level of a tank and the response of a PID controller. It is also important for us to realize that when carrying out the experiment, one must observe good experimental techniques to minimize experimental anomalies.

34

Você também pode gostar