PERFORMANCE OF LIQUID LEVEL TANK CONTROL OF LINEAR

CONTROLLERS
1. INTRODUCTION

1.1 Description of the Liquid Tank System

Over the last two decades, the control of liquid tank level systems has attracted the attention of many researchers
worldwide. It is one of the most challenging benchmark control problems due to its nonlinear and non-minimum phase
characteristics. The control objective in a tank system is to maintain the liquid's desired level when there is an inflow and
outflow of water out of the tank, respectively. The tank system is a single-input and single-output system (SISO) with
flow rate as input and water level as the output. Even though the tank system is simple from a construction point of view,
there are a lot of control challenges owing to the following characteristics.
Fig.1.1 depicts a basic representation of a typical liquid-level system.

Figure 1.1 liquid-level system examples

Figure 1.2 representation of typical liquid level system

1.2 Description of Liquid Tank Experimental Setup

1.2.1 Unit description
The RT 674 Flow/Level Control Demonstration Unit is a tabletop unit that deals with the following control tasks.
 1. Liquid level control
2. Flow rate control
3. Liquid level/flow rate cascade control
Water is used as the operating medium.
The control task is performed by two electronic industrial controllers.
These can be configured as P, Pl, or PID controllers.

1.2.2 Unit components and functions

1. Water tank
2. Pump
3. Float flow meter with electronic signal transmitter
4. Control valve (electromagnetic proportional valve)
5. Control panel with process diagram Figure 1.3 process unit
6. Flow rate controller FIC
7. Liquid level controller LIC
8. Adjusting valve for flow rate
9. Liquid level tank
10. Level sensor
11. Adjusting valve for drainage from liquid level tank
12. Selector switch for control mode
13. Switch for pump
14. Signal jacks for the control signal y2 from flow rate controller FIC
15. Control button for external/internal control signal
16. Button pressed (red mark):
➢ External control signal Button released: (no mark) > Internal
controllers
17. Flow rate controller FIC (controller 2)
18. Signal jacks for control signal y1 from liquid level controller
(corresponds to input signal w2 on flow rate controller FIC for
Figure 1.4 control panel
cascade control)
19. Signal jacks for the flow rate signal x2 from the flow rate sensor
20. Liquid level controller LIC (controller 1)
21. Signal jacks for the liquid level signal x1 from the liquid level
sensor

1.2.3 Process diagram

A (9) Liquid level tank

B (10) Liquid level sensor
C (7) Liquid level controller LIC
D (6) Flow rate controller FIC
E (3) Float flow meter
F (3) Flow rate sensor
G (4) Control valve
H (8) Adjusting valve flow rate Figure 1.5 process diagram
I (11) Adjusting valve for drainage from the liquid level tank.
K (2) Pump
L (1) Water tank

1.3 Objectives
The objectives are as follows
• Design and implement different advanced controllers for the level control to maintain the desired liquid level in
the tank liquid level system.
• To pursue a comparative study among all the developed controller techniques to choose the best controller based
on tracking capability performance for the liquid level control.
• To validate all results from the simulation (using MATLAB) and then through real-time experimentation on a
tank liquid level setting

2. DYNAMICS MODELLING OF TANK SYSTEM

The mathematical modeling of the liquid level in the tank is obtained using Mass balance. The scheme of the tank system
is shown in Figure 2.1. Water flows to the tank at a flow rate of Qin(lit/hr.), and the outlet flow is Qout(lit/hr). The cross-
section area of the tank is A(cm2). The height of the water level in the tank is represented by h(cm), which is controlled by
adjusting the pump's flow rate. Assuming the density of the inlet and outlet flow
rate is constant, the tank has a uniform cross-sectional area.
A material balance around the SISO tank gives.

(accumulation of mass in the tank) = (Massin) –(Massout)

𝜌𝑑𝑉
= 𝜌(𝑄𝑖𝑛 − 𝑄𝑜𝑢𝑡)
𝑑𝑡
𝐴𝑑ℎ
= 𝑄𝑖𝑛 − 𝑄𝑜𝑢𝑡
𝑑𝑡

The inlet flow rate Qin is a function of time. The outlet flow rate Qout is modeled
as a nonlinear function of the liquid tank by using Torricelli’s law

𝑄𝑜𝑢𝑡 = 𝐶𝑑 . 𝑎. √2𝑔ℎ

𝑄𝑜𝑢𝑡 = 𝐶𝑑 . 𝑎. √2𝑔ℎ ; 𝑙𝑒𝑡𝑠 𝑎𝑠𝑠𝑢𝑚𝑒 𝐶𝑑 = 1 Figure 2.1 Schematic diagram of

Liquid level tank
𝑄𝑜𝑢𝑡 = 𝑎. √2𝑔ℎ
𝑑ℎ
𝐴 = 𝑄𝑖𝑛 − 𝑎√2𝑔ℎ
𝑑𝑡
On performing Taylor’s series expansion for the above equation, one can obtain the linear mathematical model for the
tank system. At the equilibrium point for constant water level, the derivatives are zero.

During linearization by Taylor Series expansion, the higher order terms are neglected because these are exceedingly
small, and some parameters of the tank system are not known perfectly.
1
√ℎ = √ℎ𝑠 + (ℎ − ℎ𝑠)
2√ℎ𝑠

𝑑ℎ 1
𝐴 = 𝑄𝑖𝑛 − 𝑎√2𝑔 (√ℎ𝑠 + (ℎ − ℎ𝑠))
𝑑𝑡 2√ℎ𝑠

AT steady state h=hs

 𝑑ℎ𝑠
𝐴 = 𝑄𝑖𝑛 − 𝑎√2𝑔(√ℎ𝑠)
𝑑𝑡
Subtracting the above two equations and rearranging, we get

𝑑ℎ 𝑎√2𝑔
𝐴 = 𝑄𝑖𝑛 − 𝐻
𝑑𝑡 2√ℎ𝑠

𝑎√2𝑔 1
=
2√ℎ𝑠 𝑅

𝑑𝐻
𝑅𝐴 + 𝐻 = 𝑅 ⋅ 𝑄𝑖𝑛
𝑑𝑡
(𝜏𝑠 + 1)𝐻(𝑠) = 𝑅 ⋅ 𝑄𝑖𝑛(𝑠)

The system's Transfer function in the forward path can then be found analytically or experimentally. If the flow Q i is
assumed to be laminar, then the transfer function of the liquid tank, as outlined above, will be linear and can be expressed
as

𝐻(𝑠) 𝑅
𝐺𝑝(𝑠) = =
⋅ 𝑄𝑖𝑛(𝑠) (𝜏𝑠 + 1)
𝑘 = 𝑅 ;𝜏 = 𝐴𝑅

The above is the first-order Transfer function for the liquid tank. The inflow rate is controlled by adjusting the voltage
applied to the pump motor. The pressure transducer senses the liquid level using a bubbler method that produces a
proportional voltage. An error signal is generated by comparing this voltage with the voltage corresponding to the set
reference level of the liquid. This error further acts as an input signal to the selected controller to initiate the appropriate
controlling action to keep the liquid level in the tank at the set value.

Table 2.1 Parameters of the system

parameters value Units

Area of liquid tank A 117.376 cm2
Height H 46 cm
hydraulic resistance R 0.10622 s/cm2
outlet area of the tank a 4.08 cm2

The final transfer function of tank level is evaluated by using the values of the parameters shown in Table 2.1 as follows.

𝐻(𝑠) 0.10622
=
𝑄𝑖(𝑠) 12.3𝑠 + 1

3. PID CONTROLLER

PID control is composed of three types of controllers (P, I, and D), as shown in Figure 3.1
 P controller

e(t) control signal.

desired state

I controller
M
+

D controller

feedback signal

measured state

Figure 3.1 Schematic of PID Controller

Proportional Control Integral control Derivative control

The proportional controller reduces Integral control ultimately drives The derivative controller acts upon the
the error between the process output the offset to zero produced by the derivative of the error, so it is most active
(measure value MV) and the set proportional control. Can cause when the error changes rapidly. The
point (SP) but cannot eliminate it. oscillations derivative controller is worked to decrease
oscillation in the output process
𝑡 𝑑
𝑷𝒐𝒖𝒕 = 𝑲𝒑. 𝒆(𝒕)
𝐼𝑜𝑢𝑡 = 𝐾𝑖. ∫ 𝑒(𝑡) 𝐷𝑜𝑢𝑡 = 𝐾𝑑 . 𝑒(𝑡)
0
𝑑𝑥
4.TUNNING OF THE CONTROLLER

4.1 Ziegler-Nicholus

The Ziegler-Nichole (Z-N) setting has been widely used as a benchmark for
evaluating different tuning methods and control strategies. Ziegler-Nichole
(Z-N) is applied in a closed-loop system for tuning a PID controller as
follows:

1. Reduce the (integral time) and (derivative time) to zero and use
only the P-controller.
2. Increase Kp until oscillations occur at critical value Kp=Kcu)
3. Evaluate (ultimate gain Kcu) (ultimate period Pu sec) as shown in
Figure 4.1

The PID-controller parameters are now specified as shown in Table 4.1

Figure 4.1 Determination of Kcu
Table 4.1 Ziegler-Nichols setting based on the continuous cycling method

Ziegler-Nichols Proportional gain Kc integral time Ƭi derivative time Ƭd

P controller Kcu/2 ---- ----
PI controller Kcu/2.2 Pu/1.2 ----
PID controller Kcu/1.7 Pu/2 Pu/8

4.2 Cohen-Coon
The Cohen and Coon tuning method is applied in the graphical
construction as shown in Figure 4.2 method reduces the process reaction
curve to first order with the transport lag model given by the equation.

𝐻(𝑠) 𝑅𝑒 −Ɵ.𝑆
𝐺𝑠𝑦𝑠(𝑠) = =
𝑄𝑖(𝑠) Ƭ𝑠 + 1

The experimental procedure for the (C-C) tuning method is quite simple,
and the control loop is shown in Figure 5. The controller is changed to a
manual mode when the process reaches a steady state. Then, a (3 to 5 %)
step change in the controller output is introduced. The response of the
system is called the process reaction curve. The parameters of PID-
controller tuning by using (Cohen –Coon) open loop response is shown in Table 5

Figure 4.2 process reaction curve

Table 3 PID controller parameters by (Cohen-Coon) tuning

controller Kc Ƭi Ƭd
P 1 𝜏 1
[ + ]
𝐾 𝜃 3
PI 1 𝜏 1 𝜃
[0.9 + ] 30 + 3 𝜏
𝐾 𝜃 12 𝜃( )
𝜃
9 + 20 𝜏
PID 1 4𝜏 1 𝜃 𝜃
[ + ] 1 4𝜏 1 32 + 6 𝜏 32 + 6 𝜏 4
𝐾 3𝜃 4 [ + ]Ɵ( ) Ɵ( )(Ɵ)
𝐾 3𝜃 4 13 + 8 𝜃 𝜃 𝜃
13 + 8 𝜏 11 + 2 𝜏
𝜏

Figure 4.3 Control loop for process reaction curve method

5. SIMULINK

The simulation for the liquid tank process system was created using MATLAB Simulink software, and the block diagram
is shown in Figure 6. The system simulation responses are analyzed for change in the setpoint of the liquid level in the
tank from (5 to 90%), and responses are recorded at a different controller.
 Figure 5.1 Simulink Block Diagram for Process Tank

Figure 5.2 liquid level dynamics of a cylindrical tank

This model captures the essential dynamics of a cylindrical tank, where the inflow and outflow rates determine
the liquid level. The differential equation
𝑑ℎ 𝐹𝑖 − 𝑐 ∗ √ℎ
=
𝑑𝑡 𝐴
is implemented in this Simulink diagram, showing how the various blocks correspond to the terms in the
equation.

6. EXPERIMENTAL WORK

This study focuses on developing a control system to regulate the flow rate and liquid level within a tank using a
combination of sensors and controllers. The flow rate is measured using a rotameter installed at the pump discharge, with
a hand valve positioned between the tank and receiver to adjust the outlet flow rate accordingly. Inside the tank, a level
transmitter continuously monitors the liquid level, sending signals to a PID controller that adjusts the pump speed to
maintain the desired level. Simultaneously, a flow rate sensor provides feedback to a separate flow rate controller, which
signals an actuator to ensure precise control overflow rate and liquid height. This setup occasionally operates in a cascade
control configuration to enhance control accuracy. Additionally, a drain valve in the pipe connected to the tank introduces
disturbance into the system, challenging the controllers' ability to maintain stability and performance.

While this report focuses on simulation results, seven experiments have been conducted to validate the system under
varying conditions. Detailed analyses and findings will be reported in future publications, providing comprehensive
insights into system performance and validation of control strategies.

Modelling of liquid level

This study utilizes Simulink software to analyze the system's response to a step change (25-75% input level change). A
first-order transfer function, derived from mass balance and analytical modelling, including a dead time of 1 unit, forms
the basis of the study. Using MATLAB, analytical methods such as Ziegler-Nichols (ZN) and Cohen-Coon are
programmed to tune P, PI, and PID controllers. The objective is to evaluate and compare the efficiency of these control
strategies in regulating system response under varying input conditions.

6.1 Ziger –Nicolas and Cohen coon methods

(Z-N) It is used for closed-loop response, which made the system oscillate, and the values of ultimate gain and ultimate
period are found to be as shown in Table 4

Table 04 ultimate gain and ultimate period of Z-N method

Ultimate gain Kcu Ultimate period Pu sec

9.44 1519

Cohen and Coon used the approximate model and estimated the values of PID-controller.

Table 05 the PID controller parameters based on the process characteristics

ZIEGLERS NICHOLAS COHEN COON SIMULINK TUNNED

Kc 𝜏𝑖 𝜏𝑑 Kc 𝜏𝑖 𝜏𝑑 Kc 𝜏𝑖 𝜏𝑑 N

P 4.732 10.2 19
PI 2.25 0.003358 44.60 12.66 8.60 139.79
PID 5.66 0.00074 1076 44.60 6.7 0.492 64 14.82 -10.1 6.33
Simulink provides an inbuilt tuning feature that uses a compensator formula for transfer function tuning. This method
includes a filter coefficient N, which applies only to the derivative (D) controller component. The tuning results from the
Simulink inbuilt feature are compared with the analytical tuning methods. The filter coefficient N is essential for
smoothing the derivative action, enhancing the stability of the D controller. The tuning equation incorporating the filter
coefficient is as follows:

1 𝑁
𝑐𝑜𝑚𝑝𝑒𝑛𝑠𝑎𝑡𝑜𝑟 𝑓𝑜𝑟𝑚𝑢𝑙𝑎 = 𝑃 + 𝐼 + 𝐷
𝑆 1
1+𝑁
𝑆
6.2 Simulation results

This section discusses the simulation results of Ziegler-Nichols (Z-N) and Cohen-Coon (C-C) tuning methods. The
response of the P controller to a step change in the set point of liquid level from 5% to 90% shows that the P controller
with Z-N tuning performs better than with C-C tuning. Similarly, the response of the PI controller to a step change in the
set point of liquid level from 5% to 90% indicates that the PI controller with Z-N tuning has less oscillation than C-C
tuning.For the PID controller, the response to a step change in the set point of liquid level from 5% to 90% shows no
oscillations with both Z-N and C-C tuning methods. These observations suggest that Z-N tuning is more versatile and is
based on frequency response techniques, whereas C-C tuning follows a different approach.

However, it is important to note that in real cases, the response to a step change is not uniform as seen in these methods.
In reality, the system's response to different step changes varies, and the values for Kp , τi , and τd differ with varying
step changes. The inbuilt tuning method provided by Simulink is more realistic as it allows visualization of real-time
system changes and provides flexibility to adjust and observe the system behaviour dynamically. Unlike Z-N and C-C
methods, which calculate tuning parameters for the transfer function and assume uniform responses, Simulink's tuning
method offers a comprehensive view of the process, facilitating better control strategy adjustments. A detailed
comparative analysis of Ziegler-Nichols (ZN) and Cohen-Coon (CC) tuning methods for Proportional (P), Proportional-
Integral (PI), and Proportional-Integral-Derivative (PID) controllers, with Simulink results included for reference. The
performance metrics considered were rise time, settling time, overshoot, and peak value.

Figure 6.1 Z-N tunning for PI controller

Figure 6.2 Cohen-coon tunning for PI controller

Figure 6.3 Z-N tunning of P-only controller Figure 6.4 Cohen-coon of P-only controller
Figure 6.5 Z-N tunning of PID controller Figure 6.6 Cohen-coon tunning for PID controller

Figure 6.7 performance and robustness of P-only controller

 .

Figure 6.8 performance and robustness of PID CONTROLLER

Figure 6.9 Performance and robustness of PI controller

The key findings and their implications are summarized as follows:

Practical Implications:

• Cohen-Coon Tuning: Best suited for applications requiring fast response times and rapid system adjustments. It
balances performance with some trade-offs in stability, making it ideal for dynamic environments where speed is
critical.
• Ziegler-Nichols Tuning: Best for applications where stability and minimal overshoot are paramount. Its
conservative approach ensures a smooth and stable response, making it suitable for systems where overshoot and
oscillations need to be tightly controlled.

Simulink Results as Reference:

The inclusion of Simulink results provided a benchmark, demonstrating the performance of a widely used simulation tool.
While Simulink tuning methods generally produced the fastest response times, they also exhibited higher overshoot and
peak values. These results underscore the importance of selecting an appropriate tuning method based on specific
application requirements.

Conclusion Summary:

Both Ziegler-Nichols and Cohen-Coon tuning methods have their distinct advantages and trade-offs. Cohen-Coon excels
in providing faster responses and quicker stabilization, while Ziegler-Nichols offers better control over overshoot and
stability. The choice between these methods should be guided by the specific performance criteria and operational
priorities of the application in question