Академический Документы
Профессиональный Документы
Культура Документы
Laboratory Experiment
Supervisor:
Date:
Group:
Table 3: FOPDT Parameters of the CO step from 42-52% and 25-35% using calculations
CO step from 42% to 52% CO step from 25% to 35%
𝐶𝑂1 = 42% → 𝑃𝑉1 = 92.0℃ 𝐶𝑂1 = 25% → 𝑃𝑉1 = 99.3℃
𝐶𝑂2 = 52% → 𝑃𝑉2 = 88.8℃ 𝐶𝑂2 = 35% → 𝑃𝑉2 = 94.7℃
∆𝑃𝑉 = 88.8 − 92.0 = −3.2 ∆𝑃𝑉 = 94.7 − 99.3 = −4.6
∆𝐶𝑂 = 52 − 42 = 10 % ∆𝐶𝑂 = 35 − 425 = 10 %
∆𝑃𝑉 −3.2 ∆𝑃𝑉 −4,6
Process gain, 𝐾𝑝 = ∆𝐶𝑂 = 10
= −0.32 ℃/% Process gain, 𝐾𝑝 = ∆𝐶𝑂 = 10
= −0.46 ℃/%
Based from the results from Figure 1 and 2, the two sets of the FOPDT model parameters are
different for the step change of CO from 42-52% and 25-35%. All parameters except the time constant, 𝜏𝑝
are larger when on higher operating levels. This is because the dynamic process behaviour changes based
on operating levels.
2. Nature of the Process
To being able to compare the parameter of the operating level the entire plot test was separated. Below are
4 sets of data from operating level of 30%-40%, 40%-50%, 50%-60% and 60%-70%. The plot and table of
results are shown.
By comparing the parameter obtained it is determined as non-linear process due to the change of KP ,τP , θ
as we manipulated the operating level. The process variable from Figure 7 of the overall plot seems to be
decreased over time. At the lowest operating condition, the KP is found to be 0.4 while at higher condition
the value decrease to -0.229. Same change was observed for τP and θ thus it can be concluded that the
jacketed reactor is a nonlinear process. Below showcases the process model equation.
𝐾𝑝 𝑒 −𝜃𝑝 𝑠
𝐺𝑝 (𝑠) =
𝜏𝑝 𝑠 + 1
−0.319𝑒 −0.803𝑠
𝐺𝑝 (𝑠) =
1.69𝑠 + 1
3. Disturbance Model
In order to determine the model parameter for the disturbance, the Controller Output (CO) was
maintained at 42%. The value of disturbance was manipulated between 50℃-60℃ and the result were
shown below.
Figure 8 show an increase in temperature when increase the disturbance value. KP, τP, and θ were obtained
from the simulator. This parameter will be used for the Tuning and evaluation of controller design later on.
The disturbance model equation is shown below.
𝐾𝑝 𝑒 −𝜃𝑝 𝑠
𝐺𝑑 (𝑠) =
𝜏𝑝 𝑠 + 1
0.668𝑒 −0.961𝑠
𝐺𝑑 (𝑠) =
2.39𝑠 + 1
Table 9: The suggested IMC Tuning correlation for controller output step up (42% to 52%)
Model IMC Tuning Correlations
Parameters Controller Type Kc τI τD α
KP = -0.319 P only -1.57 0 0 0
τP= 1.69 PI -0.733 1.69 0 0
θ = 0.803 PID -0.960 2.09 0.324 0
Model’s R 2
= PID with Filter -0.907 2.09 0.324 1.10
0.999
Table 10: The suggested IMC Tuning correlation for disturbance step up (50°C to 60°C)
Model IMC Tuning Correlations
Parameters Controller Type Kc τI τD α
KP = 0.668 P only 0.920 0 0 0
τP= 2.39 PI 0.414 2.39 0 0
θ = 0.961 PID 0.527 2.88 0.400 0
Model’s R2 = PID with Filter 0.498 2.88 0.400 1.07
1.000
Based on model parameters obtained from the controller output step up and the dynamic
disturbance model, the dynamic disturbance model parameters are deemed not suitable for controlling the
jacketed stirred reactor. One of the reasons why this is the case is because the disturbance test data shows
a positive process gain which is the wrong action causing the valve to move in the wrong direction which
would increase the error rather than correct them. However, it would be useful if designing a feed forward
control element design which would consider high disturbance fluctuations.
Figure 9: P only controller studies using set point tracking with suggested IMC tuning correlation by
controller output step up (42% to 52%)
Figure 10: PI controller studies using set point tracking with suggested IMC tuning correlation by
controller output step up (42% to 52%)
Figure 11: PID controller studies using set point tracking with suggested IMC tuning correlation by
controller output step up (42% to 52%)
Figure 12: P-only controller studies using disturbance rejection with suggested IMC tuning correlation by
controller output step up (42% to 52%)
Figure 13: PI controller studies using disturbance rejection with suggested IMC tuning correlation by
controller output step up (42% to 52%)
Figure 14: PID controller studies using disturbance rejection with suggested IMC tuning correlation by
controller output step up (42% to 52%)
Figure 15: PI controller studies using set point tracking with suggested Cohen and Coon by controller
output step up (42% to 52%)
Figure 16: PI controller studies using set point tracking with suggested Ziegler-Nichols by controller
output step up (42% to 52%)
Figure 17: PI controller studies using disturbance rejection with Cohen and Coon by controller output step
up (42% to 52%)
Figure 18: PI controller studies using disturbance rejection with Ziegler-Nichols by controller output step
up (42% to 52%)
Based on Figure 9, the graphs show that the use of a P-only controller creates an offset of about -
2.1% which would be not ideal. As the process is nonlinear, the offset may become larger which is
undesirable. In Figure 12, the P-only controller fails the disturbance rejection test as it was unable to
compensate for the change and created a larger offset. For the PI controller in Figure 10, the offset was
removed by the integral action however it increased the time for the process to settle. However, when
subjugated to disturbances the settling time significantly increased but was able to compensate compared
to the P-only controller. The PID controller in Figure 11 showed similar performance to the PI controller
however it has frequent fluctuations in the controller output or noise which can cause wear and tear towards
the controller valve. Therefore, based on observation of the response of P, PI and PID controller towards
set point tracking and disturbances rejection, the ideal controller choice would be PI. Among the PI
controllers, 3 tuning correlations were selected which are IMC, Cohen and Coon and Ziegler-Nichols to
select the best. The 4 performance criterions were checked for each of the tuning. The criterions are
minimum offset, quick settling time, peak overshoot ratio and decay ratio. All the performance criterions
were tabulated in Table 12 for set point tracking and Table 13 for disturbance rejection. The graphical
values obtained to calculate the performance criterions were tabulated in Table 11. A sample of
performance criterion calculation is shown below for Figure 9. The values for performance criterion were
obtained via graphical analysis however as it is difficult to determine, however the values of the
performance criterions obtained are within ±5% error of margin. Based on the data obtained, the IMC
tuned PI controller has the best performance compared the Cohen and Coon and Ziegler-Nichols. It has no
overshoot ratio, decay ratio and offset however it has a larger settling time when used in set point tracking
and disturbance rejection. As this process is nonlinear, a suggestion which can improve the reliability of
the controller is to take the average process gain, time constant and dead time at multiple controller output
(CO) step up changes to ensure that the process is stable regardless of the change in disturbance or set point.
Furthermore, the use on a feed forward control loop would help mitigate the effects of large disturbances
acting on the process as well.
PVsettled (°C) 93 95 95 94 95 95 95
PVsetpoint (°C) 95 95 95 95 95 95 95
tsettled (mins) 21.0 33.9 64.4 21.0 17.4 15.5 15.8
tstart (mins) 14.8 13.2 41.9 12.6 5.9 9.9 8.9
A ( change in set point) 3 3 3 3 3 3 3
Table 12: Performance criterion for IMC, Cohen and Coon and Ziegler-Nichols tuning correlation for set
point tracking
Performance Criterion IMC Cohen and Coon Ziegler-Nichols
Minimum Offset (%) -2.1 0.0 0.0 -1.1 0.0 0.0 0.0
±5%
Settling Time (mins) 6.2 20.7 22.5 8.4 11.5 5.6 6.9
Overshoot Ratio (%) 0.0 0.0 0.0 0.0 53.3 40.0 20.0
±10%
Decay Ratio (%) 0.0 0.0 0.0 0.0 30 15.5 0.0
±25%
Table 13: Performance criterion for IMC, Cohen and Coon and Ziegler-Nichols tuning correlation for
disturbance rejection
Performance Criterion IMC Cohen and Coon Ziegler