ENGINEERING LABORATORY

Level Control

Grup A1
Adrián Martínez
Emili Brau
11/10/2023

Introduction
 Objective
In order to investigate the operation of a level control system, the following objectives have
been set:

1.Establishing Steady State:

First, determine the system's steady state and analyze its characteristics. This involves
identifying the equilibrium condition and pinpointing the maximum valve opening percentage
at which this equilibrium occurs, along with the corresponding water level.

2.Dynamic Behavior Analysis:

Once the steady state is identified, conduct step changes within the system. The goal is to
study its dynamic response. This involves adjusting a First Order Plus Dead Time (FOPDT)
model to fit the system's behavior. It's crucial to provide a physical explanation for how
various parameters of the model change as a result of these step changes.

3.Proportional Controller Gain Variation (KC):

Investigate the effects of altering the proportional controller gain (KC) while keeping other
variables constant. Analyze how changes in KC impact system performance and stability.

4.Integral Time Variation:

Examine the influence of modifying the integral time (Ti) while keeping other factors
unchanged. Assess how variations in Ti affect the system's response and control
performance.

Theoretical Basis

In the realm of process control, there exist two categories of variables:

First, we have the input variable, which illustrates the influence of the environment on the
process. Within this category, there are two subtypes of inputs. To begin with, we have the
manipulated inputs, which can be regulated either by an operator or through the control
system. Secondly, there are disturbances, which are inputs beyond our control.

Next, we have the output variables, which are the results affecting the environment. Process
controls are instruments employed to manage a specific parameter, such as temperature,
level, or pressure. These instruments vary in the manner in which they connect the controller
input (error) with the controller output (drive signal).

Process controls are devices employed for the regulation of specific parameters, such as
temperature, level, or pressure. These devices exhibit diversity in how they establish a
connection between the controller input (error) and the controller output (drive signal).

2
Proportional Control (P) represents a type of feedback control mechanism. It serves as the
most straightforward continuous control method available for implementation in a closed-
loop system. Its primary function is to reduce fluctuations in the process variable; however, it
may not consistently bring the system to the desired set point. Despite its advantage of
providing a quicker response time, it tends to introduce an unwanted offset or deviation from
the set point. This can be likened to the systematic error found in a calibration curve, where
a constant error persists. To mitigate this issue, it is possible to enhance the control system
by incorporating another form of control, such as I or D control.

Proportional-Integral Control (PI) is another variant of feedback control. PI control serves

to both stabilize the system, eliminating fluctuations, and guide it back to its designated set
point. While the response time of PI control is quicker than that of Integral (I) control, it
remains somewhat slower, being up to 50% less rapid than Proportional (P) control.

Proportional-Derivative Control (PD) represents a fusion of feedforward and feedback

control strategies. It functions by considering both the existing state of the process and its
anticipated conditions. Within PD control, the output is determined through a linear
combination of the error signal and its derivative. This enables proportional control to
mitigate fluctuations while allowing derivative control to forecast process errors, thereby
anticipating changes in process conditions and scrutinizing variations in errors. The primary
advantage of D controllers lies in their capacity to avert system fluctuations, effectively
eliminating any form of oscillation.

Proportional-Integral-Derivative control (PID): This form of control is the most commonly

employed because it amalgamates the merits of all control types. This encompasses a
swifter response time akin to P control, diminished offset akin to I control, and the ability to
anticipate disturbances through the measurement of error change facilitated by D control.
Nevertheless, despite the apparent suitability of the PID controller, it also carries the highest
cost. Consequently, it is reserved for applications where the precision and stability offered by
the PID controller are essential requisites for the process.

The PID control method derives its name from its three correction components, the
combined total of which forms the manipulated variable (MV). To compute the output of the
PID controller, the proportional, integral, and derivative components are added together. If
we designate the controller output as u(t), the PID algorithm's ultimate expression is
determined as follows:

[Eq. 1]
where:
- Kp is the proportional gain, a tuning parameter
- Ki is the integral gain, a tuning parameter,
- Kd is the derivative gain, a tuning parameter,
- e(t)=SP-PV(t) is the error (SP is the setpoint, and PV(t) is the process variable)
- t is the time or instantaneous time
- τ is the variable of integration (takes on values from time 0 to the present t)

3
The Ziegler-Nichols method is a control tuning technique used in the field of automatic
control theory. Its primary purpose is to determine the parameters of a proportional (P) or
proportional-integral (PI) controller in a control system to achieve desired performance. This
method is based on observing the system's response to a change in input and uses a series
of experiments to adjust the controller parameters until an optimal response is achieved,
minimizing error and avoiding undesirable oscillations in the system.

Figure 1: Ziegler-Nichols method

The Cohen-Coon method, also known as the Cohen-Coon PID method, is an approach
used in automatic control to adjust the parameters of a proportional-integral-derivative (PID)
controller in a control system. Unlike the Ziegler-Nichols method, the Cohen-Coon method
doesn't rely on step response experiments but uses steady-state oscillation response
experiments to determine the optimal values of the proportional (P), integral (I), and
derivative (D) coefficients of the PID controller. This is done with the aim of improving the
performance and stability of the control system.

Figure 2: Cohen-Coon method

Oscillations

4
Undesirable oscillations may occur in the process output variable when the proportional,
integral, and derivative control parameters are not correctly configured. These oscillations
can be characterized by five parameters: rise time, peak time, overshoot ratio, decay ratio,
and period.

● Rise time (tr) signifies the duration it takes for the system to attain its steady state for
the initial time, excluding any dead time.

● Peak time, (tp) corresponds to the amount of time until the first peak is reached
(without including dead time).

● Overshoot ratio, (OS): corresponds to the exceed between the first oscillation and
the steady state level in relation to the steady state variation.

● Decay ratio, (DR): is the size of the successive peaks.

● Period, (P): duration between peaks in an oscillation.

Figure 3: System responding to the oscillations [1]

.

Experimental Part

5
Equipment and material:

For this experiment, the level control equipment was used. This can be seen below in figure
4.

Figure 4: Control Level P&ID

Reagents:

Water

Experimental procedure:

Before starting with the experimentation it is necessary to open the air valve and turn on
both the PLC and the computer.

The first step is to obtain the loading and unloading curves of the tank T-101. Begin making
sure that the tank is empty and that the valves V-104, V-105, V-106 V-102, V-103 are closed
and that the valve V-101 is open. To record the values for the first loading curve, set the VC-
101 pneumatic valve to an open setting of 10% through the PLC touch screen which must be
in manual mode. This procedure will be repeated for increments of 10% until the valve is
100% opened. In order to record the unloading curves, when the tank is completely filled,
start the level probe measuring and open the V-104 valve. This procedure is repeated 3
times.

The tank width and length dimensions are measured with measuring tape. The values
obtained are 31cm x 31cm. The height value is determined with the maximum value the
probe reads before stopping automatically in the loading curves. This value is 44 cm. Once
the volume of the tank is known, the values of the volumetric flow rates of the different
loading curves are found, and thus the maximum valve opening percentage to achieve a
steady state.

6
The next step is to manually start the measuring of the height values with the discharge
valve open, setting the percentage of the pneumatic pump opening to 10% and once the
liquid level stabilizes, without stopping the measurement, change the open setting to 20%
and let it stabilize again. This procedure is repeated once more, but in this case starting with
a 20% open setting and changing it to 30%.

For the next part, we proceed to study the P, PI and PID controllers. In order to do so the
PLC is set to automatic. This study is done by varying the values of Kc, Ti and Td.

First, for the controller P we study the effect of the Kc. The height setpoint limit is set to 15
cm and the values Ti = 100 and Td = 0 are fixed. The Kc is set to 0,2 and with the discharge
valve open the cycle is started. Once the height has stabilized the values are plotted. The
procedure is repeated with a Kc = 5 and Kc = 50.

Next, the controller PI is studied, and therefore the effect of the Ti value. The height setpoint
limit is kept the same. For the first measure, both Kc and Ti are set to 50 while Td = 0. After
that Kc is fixed at 0,2 and two measures are taken at Ti = 1 and Ti = 25.

Finally the controller PID is studied, varying the Td value. The height setpoint limit is kept the
same and the values Kc = 0,2 and Ti = 0,7 are fixed. The Td is set to 0,1 and then 10. Two
more measures are taken at Kc = 5, Ti = 50, Td = 10 and Kc = 5, Ti = 5, Td = 1.

To conclude the experiment, a study is conducted to investigate how various disturbances

can impact the valve opening regulation for controlling the incoming flow rate. Two separate
experiments are carried out using two interconnected tanks. Initially, data measurement
begins, and one of the tanks is filled. Once the fluid has reached a stable state, the sensor
surface of the corresponding tank (where the introduction is taking place) is covered for one
minute. Afterward, the process continues normally until it stabilizes again.

Next, the introduction is stopped, and the tanks are allowed to completely drain. Once they
are completely empty, the first tank is refilled. However, in contrast to the previous case,
after it has stabilized, the drain valve of the second tank is opened. The equipment is then
allowed to stabilize again before it is stopped.

Results and Discussion

7
 Steady state

The first objective is the determination of the maximum flow rate and opening fraction to
reach steady state. To determine this we first obtain the loading and unloading curves.

Graph 1: Height vs. Time for every valve % opening and unloading

Once we have obtained these values we can find the values of flow rate for both loading and
unloading using the equation:

dV l∗w∗dH
Eq.1 =
dt dt

dV
Where is the variation of volume with respect to the variation of time, l is the length of
dt
dH
the tank, w is the width of the tank and is the variation of height with respect to the
dt
variation of time. We have measured the l and w with measuring tape. We know that when
dH
filling up the tank, the is constant as the volumetric flow being pumped to the tank is
dt
dV dV
constant, thus can be considered constant. We calculate the average values of only
dt dt
for loading with 10%, 20% and 30% since at higher opening percentages there is turbulent
dV
flow and a lot of “noise”. The calculated values of are shown in the table below:
dt

Table 1: Average volumetric flow rate for each valve % opening

Valve opening (%) Average Volumetric Flow (m3/s)

10 0,00025889

20 0,00046382

8
 30 0,00058846

In order to obtain a steady state, the following equation must be fulfilled:

dV out
Eq.2 dV ∈ ¿ = ¿
dt dt

Which basically means that:

Eq.3 Q∈¿ Q out

The problem is that unlike Qin, Qout is not linear, so in order to better visualize which valve
opening % flow rate better adapts to the Qout we graph the average inflow calculated rates
on top of the Qout and obtain the following graph:

Graph 2: Volumetric flow rate vs. Time for unloading and average 10, 20 and 30 % valve opening

From looking at Graph 2 it can be deduced that the open percentages of the pneumatic
pump that better adapt to the unloading volumetric flow are 10% and 20%.

Step response

Once the graph 2 information has been analyzed, the next step is to approximate the step
response changes with the FOPDT model, allowing us to obtain and calculate the
parameters of Kc, Ti and Td for P, PI, PD and PID with which the PLC will work with when
on automatic mode.

First we did the step change from 20% to 30% and the tank quickly reached 44 cm and the
pneumatic pump automatically stopped. Thus we change the step response from 10% to
20% valve opening and obtain the following graph where we apply the Ziegler-Nichols
method:

9
 Graph 3: Height vs. Time for step response from 10% to 20% valve opening

From the graph 3 we can obtain the following variables:

Table 2: Ziegler-Nichols variables

M 0,2
K 9,5
L 0,3
T 2
R 0,5
NL 1,425

These variable are necessary to determine the experimental values of the proportional
Controller Gain Variation (Kc), the Integral Time Variation (Ti) and the Derivative Time
Variation (Td) by using the Cohen and Coon method:

10
 Table 3: Kc, Ti and Td values found find the Cohen and Coon method

Controller Kc Ti Td

P 0,1474 - -

PI 0,1281 0,7613 -

PD 0,1789 - 0,0762

PID 0,1924 0,6951 0,1062

Controller study

For the controller study the experimental values of Kc, Td, and Ti obtained in table 3 will be
tested and varied to see what the optimal value for these variables is.

First we study the P controller and obtain 3 graphs for the 3 values of Kc: 0,2, 5 and 50
shown below in graph 4.

Graph 4: Height vs. Time for P controller

It can be observed that in this case the higher the value of Kc, the faster it stabilizes and
reaches the “equilibrium”. At lower values such as 0,2, not only does it take longer to fill up,
but it also overshoots the 0,15 m mark by quite a larger distance than the rest of Kc values.
Nevertheless, at lower values of Kc, when it reaches the equilibrium, it is much more stable,
as can be seen at Kc = 0,2 where the height value barely changes from time 1200 s. If we
look at Kc = 5 and Kc = 50, when it reaches the 0,15 m mark, the height values are not as
stable.

11
Next we study the PI controller and obtain 2 graphs for the 2 values of Ti = 1 and 25 while
keeping Kc constant at 0,2 shown below in graph 5.

Graph 5: Height vs. Time for PI controller

It can be observed that at a lower Ti value the height increases quicker, however it also
overshoots more which leads to a bigger oscillation and it doesn’t stabilize. In order to
choose the best value of Ti we have to balance a low value for less oscillations and a bigger
value for a faster height rise value.

We also did a measurement with both Kc and Ti equal to 50 in order to combine both
variables and optimize the stabilization process. A very quick stabilization and low degree of
oscillation can be seen below in graph 6.

Graph 6: Height vs. Time for PI controller (Kc, Ti = 50)

12
Next we study the PID controller, obtaining the following graphs:

Graph 7: Height vs. Time for PID

In the graph 7, experimental data series obtained with different Td values are depicted, with
the blue series standing out for having a value one hundred times higher. A minimal but not
negligible variation can be observed between both series. We notice that the peak of the
orange series is slightly lower, indicating a better ability to correct the increase in flow.
Additionally, this series exhibits a higher frequency of oscillations, leading to a more effective
long-term adjustment.

Graph 8: Height vs. Time for PID

In this experiment, the following values have been used for each of the variables: Ti=50, Td=10,
Kpi=5. In graph 8, we can observe that with these parameters, there is a very significant and rapid
overshoot, causing the tank to fill completely. This is due to an imbalance between Ti and Td.

13
 Graph 9: Height vs. Time for PID Difference between using 1 tank and 2 tanks

In this experimentation, a test is carried out to allow a comparison between flow stabilization
using one or two tanks. It can be noted that with two tanks, the over impulse is lower due to
the higher flow correction capacity resulting from having a larger volume. Also, it is observed
that when working with two tanks, the oscillations are slower (lower frequency), which
ultimately leads to a more precise adjustment around the desired value.

Graph 10: Height vs. Time for Disturbance experience Blocking the sensor

The test has been conducted under the following conditions of Ti, Td, and Kpi: Ti=5, Td=1,
and Kpi=5, using two out of the three available tanks. When examining the graph, it is
evident where the disturbance occurs. There is a clear jump in the fluid level from 0.15
meters to approximately 0.5 meters. This is due to placing an object obstructing the sensor
very close to it. This causes the sensor to detect an "excessive flow" and send a signal to
the valve to close and regulate the flow. However, since this disturbance was external and

14
the fluid level was correct, it results in an over-emptying, explaining the subsequent drop in
the fluid level.

It is worth noting that the model has well-tuned parameters, allowing it to recover quickly
from the generated disturbance. Furthermore, the oscillation that occurs before and after this
disturbance is minimal.

Graph 11: Height vs. Time for Disturbance valve opened

For the second disturbance, the same system characteristics as the previous case are
maintained. Once the system has stabilized, the discharge valve of the second tank (the one
without an inlet valve) is opened. At this point, a minimum fluid level is observed because the
system must readjust the inlet flow to compensate for the fluid loss occurring. This is also
why a subsequent maximum is observed as the valve opens abruptly to restore normalcy.
Similar to the previous test, the period during which the system remains destabilized is
minimal, as it efficiently and quickly recovers, indicating optimal parameter adjustment of the
system.

15
Conclusiones

This practice aims characterizing the unit through a series of distinct experiments. In the
initial phase of the study, manual control of the system was employed. This entailed
adjusting the valve to achieve a specific percentage opening. It enabled the precise
determination of the steady state, including a near-centimeter prediction of the system's
settling height.

Subsequently, step changes were introduced, and adjustments were made to the First Order
Plus Dead Time (FOPDT) models. Graphically, it is evident that these models provide
suitable fits, making them suitable for approximating PID parameters.

As the third objective, the impact of PID parameters was investigated. A set of experiments
were designed to estimate the PID parameters in a methodical manner. The nature of the
experiments allowed for the establishment of a range of PID parameters and see which ones
are optimal for the equipment work.

To conclude the practice, a series of tests were carried out involving the induction of
disturbances in the system to evaluate its level of adjustment and its capacity to respond to
them. As a conclusion, it was determined that in the event of any disturbance in the system,
it can react in an effective and controlled manner, without experiencing an excessive impact,
since it recovers quickly and avoids further imbalances. These tests helped to identify the
optimum parameters for equipment performance.

16
Bibliografía

[1]
https://slideplayer.es/slide/3980417/
Consultado el 09/10/2023

Nomenclature

VARIABLE MEANING UNITS

Q in Inlet flow 𝑚3/𝑠
Q out Outlet flow 𝑚3/𝑠
V Volume 𝑚3
t Time Seconds
H Height m

Δ𝑢 Change in input adimensional

Kp Process Gain 𝑚/𝑠
Tp Process Time s
17
θ Dead Time s
Δ𝑦 Change in height m
(output)
Annex

Annex 1: P&ID Diagram for Level Control

18