Experiment E.

Liquid Level Control Using Coupled Tanks

이름: 윤초록

학번: 20180422
 1. Introduction

Process control is the ability to change the process actively based on the consequences of

process monitoring in order to take a desired output. When device detecting an out-of-control

situation, the manager makes a alternation in order to put the process back into control. As the

chemical processes are generally operated in steady-state, it is important to maintain the

condition for each unit in plant. Off-control can result in the production of off-spec products,

and the decreased profitability due to insufficient consummation of raw materials and energy.

Most importantly, several deviations from the steady-state operation can eventually lead to an

explosion in chemical plant. Automatic process control can supply robustness of process in

the present of disturbance and enhanced process safety. We will use the simple liquid-level

control experiment to learn how we can design the controller and choose each parameter.

Control systems are developed to maintain process conditions at aimed values by

manipulating some process variables. A typical example of a control system is home hot

water heater. (Figure 1.)

Figure 1. Example of a Control System, Home Hot Water Heater

The control system in Figure 1. efforts to keep the temperature in the tank of heating process

at the aimed value by manipulating the fuel flow to the burner (or the electrical input to the
heater). For instance, if we desire the temperature by 50℃, we apply the system to response

smoothly and efficiently. The thermostat which is a controller compares the measured

temperature recorded in thermocouple to the aimed variable. Then we call such difference

between the two values as error. (Equation: error = set value – measured signal value) The

controller takes proper action by sending a signal to the final control element (Fuel valve)

based on the magnitude/sign of error, which manipulates input of process to lead to the set

point. The concept that we get information about the deviation of the system from its desired

state and use this for controlling the system is called feedback control. These information

about state of system is sent to controller for formation of feedback and process utilizes this

information to alter the system in some way.

In the experiment E, our purpose is comprehensive understanding on liquid level controlling

and learn how to design coupled tank systems.

2. Theoretical Background

2-1. Modeling and Mathematical Tools

Consider a system that controls the water level in the tank by connecting two pumps.
 Figure 2. Configuration of the coupled tanks system

We will use type of configuration, Quanser Inc. QUARC User Manual, (2012) in the liquid-

level control experiment. We want to maintain the water level of the tank through control

system at the desired value. According to the tank 1 of Figure, the amount of water drawn

from a pump/ the outlet volumetric flow/ the height of the tank1 (L1) can be considered as an

input/disturbance/output, respectively. For tank2, the out-flow from the tank1 is input of

tank2, and the height of the tank2 (L2) can be considered as output.

In tank 1, we set water level as L1 (t), cross sectional area as A1, outlet cross sectional area

as a 1 . Also, k v is pump power constant, g is gravitational acceleration constant, and V(t) is

input voltage. In tank 2, similarly with tank 1, we set water level as L2 (t ), cross sectional area

as A2, outlet cross sectional area as a 2 .

Assume that V0, L10, and L20 are the input voltage, water level of tank 1 and water level of

tank 2 at steady-state, respectively. Also, differences between the variables and their steady-

state values are defined as V1, L11, and L21;

※ deviation variables
L11 =L1 −L10

L21=L2−L20

V 1=V −V 0

dL10 d L20
Since the system is in a steady-state (no change over time), = =0.
dt dt

Volumetric flow rate by pump is set to k v V . Also, flow rate balance equation can be

obtained through Bernoulli's principle (the velocity of fluid at the tank outlet can be

expressed as √ 2 gh) and Taylor expansion.

dL11 −a1 √ g k V dL12 −a 2 √ g a √g

= L11 + v 1 = L21+ 1 L11
dt √2 L10 A 1 A1 dt √ 2 L20 A 2 √ 2 L10 A 1

L11 ( s ) K
We can obtain = c from above equation. In addition, L10 , L20 ,V 0 can be obtained
V 1 ( s ) τs +1

by using steady-state condition ¿).

2-2 Transfer Function

A transfer function relates two variables (generally input and output) of a physical (real)

process.
 Figure 3. Transfer Function

Here, X(s) indicates the input variable and Y(s) indicates the output variable affected by

physical process. We can set the input and output balance with intermediate transfer function,

G(s),

Y ( s )=G(s) X (s)

Y ( s)
Finally, we can get the form of transfer function G(s), G ( s )= . We state that the transfer
X (s )

function G(s) operates on X(s) to produce Y(s).

2-3. First-Order System

The general form of the equation;

dy
τ + y=K c x (t )
dt

y is output, x(t) is the input function. The initial condition is y ( 0 )= y s=K p x ( 0 )=K p x s. ( y s

and x s are y and x values in steady-state, respectively). In this form, we can get deviation
 dy
variable expressed as X =x−x s ,Y = y− y s. Therefore, τ + y=K c x (t ) can be replaced with
dt

dY ( t )
τ +Y =K c X (t ). (At steady-state, change over time is 0) Through Laplace transform of
dt

Y ( s) K
this differential equation, we get transfer function G(s) = = c . Here, a Laplace
X ( s ) τs+1

dY ( t )
transform sY(s) of was used.
dt

Assume that a step change of X(t) with magnitude A is applied into the first-order system,

A A Kc
the Laplace transform of X(t) is . Also, we can obtain Y ( s )= and
s s τs+1

L−1 {Y ( s ) }= y ( t )=K c A {1−exp ( −tτ ) } can be obtained using an inverse Laplace transform.

Figure 4. Response of a first-order system to a step input

 2-4. Second-Order System

Second-order systems contain a second-order differential equation. There is a general form

of second-order system.

d2 y (t ) dy ( t )
A 2
+B +Cy ( t )=x (t)
dt dt

This equation relates output variable y to the input variable x.

Figure 5. Two first-order systems in series yield an overall second-order system

A second-order transfer function can arise physically whenever two first-order processes are

connected in series. Above figure shows the signal flow relation for such a process in block

diagram.

In this second-order case, the total transfer function G(s) is

U ( s ) Y ( s) Y ( s) K1 K2 Kc
G ( s )= = =G 1 ( s ) G 2 (s ). G ( s )= = because
X ( s) U ( s) X ( s ) τ 1 s+ 1 τ 2 s+1 (τ ¿¿ 1 s+1)(τ 2 s+1)¿

K1 K2
G 1 ( s )= ,G 2 ( s )= (according to transfer function of first-order system). We can
τ 1 s+1 τ 2 s +1

define K c as K 1 K 2, and equation of second order case has a standard form of

w2
G ( s )= . Depending on the value of ζ , the type of second-order system can be
s 2+ 2 wζs+ w2

classed in Table.
 Damping Coefficient Characterization of Roots of Characteristic

Response Equation
ζ >1 Overdamped Real and unequal
ζ =1 Critically damped Real and equal
0< ζ <1 Underdamped Complex conjugates (of the

form a + ib and a - ib)

Table 1. The Three Classes of Second-order Transfer Function

For underdamped case, if step-input of the second-order process is applied, the response will

be shown by below Figure.

Figure 6. Describing an underdamped second-order response

Firstly, tr is the rise time, which is the time it takes that output reaches the new steady-state

π
value first, and tp is time to first peak ( ¿, which means the time it takes that the
w √ 1−ζ 2

output reaches its first maximum value. ts is settling time, which means the time it takes that

the output starts to remain within specific range for steady-state value. The range here is ±5%
 a
of value b (sometimes ±1% of b is used in some cases). Overshoot (OS) is defined as (=
b

−2 πζ
)) value. Decay ratio (DR) means c (= exp(
−πζ
exp( 2
) 2
2 = OS ) value (c is height of
√ 1−ζ a √ 1−ζ
second peak) Period of oscillation (P) means the time between two successive peaks or

−2 π
valleys of response (exp (√ )
1−ζ 2
¿.

In summary, we can find a, b, c in above Figure, and express these terms with a, b, c.

π a −πζ
t p=
w √ 1−ζ 2
OS= =exp
b (
√1−ζ 2 )
c −2 πζ −2 π
2
DR=( OS ) = =exp
b √1−ζ 2 ( ) P=exp
( √1−ζ 2 )

2-5. Process Control

There is a block diagram of a typical feedback controller below.

Figure 7. Block diagram for feedback control

Proportional-integral-derivative (PID) control is the primary type of feedback control.

 PI control (proportional-integral control) includes proportional, integral mode. PID control

includes proportional, integral, derivative mode.

R(t), a set value, is a aimed value of a certain variable as a reference. L { R ( t ) } =R ( s)

Y(t) is the output value of the system. L { Y ( t ) }=Y (s )

e(t) is an error function. e (t )=R ( t )−Y (t ). L { e ( t ) }=E ( s ) =R ( s )−Y (s).

U(t) is the input function. Assume that Ú is in the reference state. U ' ( t )=U ( t )−Ú .

Deviation function of input is U’(t). L {U ' ( t ) }=U ' ( s)

t
' ' 1 ¿ ¿ de ( t )
In PID control, we can express U ( t ) as U ( t )=U ( t )−Ú =K p {e (t ) + ∫ e ( t ) d t +τ D }.
τ1 0 dt

Kp
Kp, K i= , K d =K p τ D
τI

U' (s ) 1
When the transfer function C(s) can be expressed as C ( s ) =
E (s )
=K p 1+ +τ s .
τIs D ( )
In proportional control, the output of the controller is adjusted in proportion to the error

value. In addition, by changing the value of the controller gain, it is possible to control how

sensitive the output changes with the error, and by adjusting the sign of Kp, the output can be

increased or decreased as the error signal increases.

Integral control allows the controller’s output to integrate the error signal over the entire
¿
time. It is controlled by the value. Integral control continuously controls the system until e (t )
¿
congerges to 0 because the e (t ) value is not 0 when the output value is not in the steady

state. Therefore, integral control is widely used because it has the advantage of reducing the

offset so that the output is in a steady state.

 Derivative control is a control that predicts the future trend through the rate of change of

the error signal. Because of its role, it is also called anticipatory control. However, if the error

value is kept constant at a value other than 0, the output of the controller cannot be adjusted

because the rate of change of the error is 0. Therefore, it is used together with other control

methods.

When the feedforward control is used together with the feedback control, the error generated

in the modeling process and the error caused by the output of the process can be reduced, so

that the system can be better controlled.

2-6. Modeling of the Closed loop system with feedback controller

The closed loop system of the tank controlled by the feedback controller can be expressed as

the following block diagram.

Figure 8. Closed loop system

C(s) and P(s) are the transfer functions of the controller and the tank, respectively. In this

system, Y ( s )=U ' ( s ) P ( s )=C ( s ) P ( s ) E ( s )=C ( s ) P ( s ) { R ( s )−Y ( s ) } Thus, total transfer function

Y (s ) C ( s ) P ( s)
G(s) can be expressed as G ( s )= = .
R ( s ) 1+C ( s ) P ( s )
 In order to obtain G(s), C(s) and P(s) should be calculated.

U' (s ) K
For PI controller, Kd = 0. Then, C ( s ) = =K p + i is calculated. Because tank have first-
E (s ) s

L11 ( s ) K
order system formation in this system, P ( s )= = c can be obtained.
V 1 ( s ) τs+1

V 1 ( s): the deviation function of the input of tank 1

L11 ( s): the deviation function of the output.

Since C(s) and P(s) obtained, we can obtain G(s) which has a standard form of

w2
, (It is form of transfer function of the second order system)
s 2+ 2 wζs+ w2

3. Scheme of experiment

Figure 9. Overall Scheme of Experiment

 4. Experimental Method

▷ Tank pressure sensor calibration

- Fill the distilled water as high as two-thirds of the vessel.

- Open the Tank Calibration file.

- Offset 1 potentiometer calibration: With an empty tank, adjust the Tank 1 Offset:

potentiometer to set V_L1 to 0 Volts.

- Gain 1 potentiometer calibration: Fill tank 1 up to 25cm, and adjust the "Tank 1 Gain"

potentiometer to set V_L1 between 4.0 to 4.2 Volts.

▷ Step input response

- Fill the distilled water as high as two-thirds of the vessel.

- Open the step_input simulink file, and set the parameter to 7 V.

- Open the saved file, and extract the data.

▷ PI-controller

- Fill the distilled water as high as two-thirds of the vessel.

- Open the Tank1_FFPI and set the set-point and feedforward gain to 9cm and 2.39,

respectively.

- Set Ki and Kp as 0.

- Set Ki as 0 and, vary Kp from 2 to 10 with an increment of 2.

- Set Kp as 0 and vary Ki from 2 to 10 with an increment of 2.

 - Set the proper values of Ki, and Kp at least 5 times, in order to satisfy the proposed control

conditions

- Save all the experimental data.

▷ Data plotting

Open the simulink file, then the assigned data will be saved as format of "mat".

and then, follow the procedure below.

In the case of step input experiment,

X=importdata('stepinput.mat');

plot(X(1,:), X(2,:));

5. Discussion

1) Derive the ODE equation model of the tank 1 system.

We will use type of configuration, Quanser Inc. QUARC User Manual, (2012) in the liquid-

level control experiment. (We will only use tank 1.) We want to maintain the water level of

the tank through control system at the desired value. According to the tank 1 of Figure, the

amount of water drawn from a pump/ the outlet volumetric flow/ the height of the tank1 (L1)

can be considered as an input/disturbance/output, respectively. For tank2, the out-flow from

the tank1 is input of tank2, and the height of the tank2 (L2) can be considered as output.
 Figure 10. Configuration of the coupled tanks system (Quanser Inc. QUARC User Manual,
(2012))

The dynamic modelling of tank 1 is

dL1 (t) 1
= (−a1 √ 2 g L1 ( t ) + k v V ( t ))
dt A1

The dynamic modelling of tank 2 is

dL2 (t) 1
= (−a2 √ 2 g L2 ( t )+ a1 √2 g L1 ( t ))
dt A2

We can set this equation by constructing the mass balance assuming constant fluid density.

In tank 1, we set water level as L1 (t), cross sectional area as A1, outlet cross sectional area

as a 1 . Also, k v is pump power constant, g is gravitational acceleration constant, and V(t) is

input voltage. In tank 2, similarly with tank 1, we set water level as L2 (t ), cross sectional area

as A2, outlet cross sectional area as a 2 .

Firstly, we can express volumetric flow rate from pump as k v V (t ). Then we can utilize

1 2 1
Bernoulli equation, which is ρ v 1 + P1 + ρg h1=P2 + ρ v 22 + ρg h2.
2 2
 Figure 11. Bernoulli equation

With Bernoulli equation, we can express v1(t) (outlet flowing velocity) with variable that is

mentioned previously. Based on means of Bernoulli's principle (the velocity of fluid at the

tank outlet can be expressed as √ 2 gh), we can get v1(t).

Because balance equation which is (accumulation = Input - Output + Generation) is

established, we can get tank 1 volumetric flow rate equation; (generation is 0)

d L1 ( t )
A1 =( inflow¿ pump )−( out flow through outlet )=k v V ( t )−a1 v 1 ( t )=k v V ( t )−a1 √ 2 g L1 (t )
dt

(with Bernoulli's principle)

dL1 ( t ) 1
= {−a 1 √ 2 g L1 ( t ) +k v V ( t ) } ⋯ Tank L1 equation
dt A1

Similarly, we can get tank 2 volumetric flow rate equation;

d L2 ( t )
A2 =( inflow¿ pump )−( out flow through outlet ) ¿ a 1 v 1(t)−a 2 v2 ( t )=a1 √ 2 g L1 (t)−a2 √ 2 g L2 (t )
dt

(with Bernoulli's principle)

 dL2 ( t ) 1
= {a 1 √ 2 g L1 (t )−a2 √ 2 g L2 (t ) } ⋯ Tank L2 equation
dt A2

Assume that V0, L10, and L20 are the input voltage, water level of tank 1 and water level of

tank 2 at steady-state, respectively. Also, differences between the variables and their steady-

state values are defined as V1, L11, and L21;

※ deviation variables

L11 =L1 −L10

L21=L2−L20

V 1=V −V 0

dL10 d L20
Since the system is in a steady-state (no change over time), = =0.
dt dt

With Tank L1 equation and Tank L2 equation, we can get expression of L10 and L20 with terms

of V0.

k 2v V 20
Tank L1 equation → 0=k v V 0−a1 √ 2 g L10 → L10= 2
2 a1 g

a21 k 2v V 20
Tank L2 equation → 0=a1 √2 g L10−a2 √ 2 g L20 → L20= L 10 =
a22 2a 22 g

Suppose that there is a change of input voltage from V0 to V and this change leads to

change of L1(t) and L2(t) over time. Deviation function can be defined as

V 1=V −V 0 , L11 =L1−L10 , L21=L2−L20

 dL1 d L11 d L10 d L11 dL2 d L21
so = + = , = , V =V 1+V 0 .
dt dt dt dt dt dt

In Taylor series expansion,

1 '' 1
y=f ( x )=f ( x 0 ) + f ' ( x 0 )( x− x0 ) + f ( x 0 ) ( x−x 0 )2+ ⋯ f n ( x0 ) ( x−x 0 ) n+ ⋯ ≈ f ( x 0 ) + f ' ( x0 ) ( x−x 0 )
2! n!

When x is very small, it is √ x+ a ≅ √ a+1 ! ( 2 1√ a ) x+2 !( 4−1√ a ) x + ⋯ (with Taylor series

3
2

expansion at x 0=0 )

1
so√ 2 g L1= √2 g ( L11 + L10) =√ 2 g ×
( √ L10+1 ! ( ) L +⋯
2 √ L10 11 )
≅ √2 g L10 + √g L
11
√2 L10

g
√ 2 g L2 ≅ √ 2 g L20+ √2 L L21.
√ 20

Substitute it to Tank L1 equation and Tank L2 equation,

Tank L1 equation

dL1 ( t ) d L11 1 a √g 1 −a √ g
→
dt
=
dt
=
A1 {
−a1 √2 g L10 +k v V 0 − 1
√ 2 L10
L11 + k v V 1 = ( 1
A1 √ 2 L10 11 v 1}
L +k V ). (

0=k v V 0−a1 √2 g L10 (steady−state )¿

dL11 −a1 √ g k
= L11 + v V 1 ⋯ Equation L11
dt A 1 √ 2 L10 A1

Tank L2 equation
 dL2 ( t ) d L21 1 a √ g a2 √ g 1 a 1 √ g a2 √ g
→
dt
=
dt
=
A2 {
a1 √2 g L10−a2 √ 2 g L20+ 1 −
√ 2 L10 √2 L20
L21 = − L
} (
A2 √2 L10 √ 2 L20 21 )
(0 = a 1 √ 2 g L10−a 2 √ 2 g L20 ¿

dL21 a1 √ g a2 √ g
dt
=
( −
A 2 √2 L10 A2 √2 L20 )
L21 ⋯ Equation L21

L11(t) and V1(t) in Equation L11 can be changed to L11(s) and V1(s) by Laplace transform.

−a 1 √ g k
s L11 ( s )−L11 ( 0 )= L11 ( s)+ v V 1 (s) ( L11 ( 0 )=0 ¿
A1 √2 L10 A1

A 1 √ 2 L10 k √ 2 L10
{ a1 √ g }
s+1 L11 ( s ) = v
a1 √ g
V 1 ( s)

k v √ 2 L10
L11 ( s ) a1 √ g k A √2 L10 k √ 2 L10
→
V 1 (s )
=
A 1 √ 2 L10
= c τ= 1
τs+ 1 (
a1 √ g
, kc= v
a1 √ g )
⋯ Equation Laplace L21

{ a1 √ g }
s+1

When we set the input V as a step input, V −V 0 will be A ( constant ) so

A kc 1 1
A
V 1 ( s ) becomes L { A }= and
s
L11 ( s )=
s τs+ 1
=k c A −
s
s+
1
τ
{ }
{
L−1 { L11 ( s ) }=L11 ( t )=k c A 1−exp ( −tτ )}(τ = Aa√√2gL
1

1
10
, k c=
k v √ 2 L10
a1 √ g
)

2) Find the definition of the Laplace transform and how to apply the transform to

solve an ODE.
 Definition of the Laplace transform: Laplace transform of function f(t) is defined in F(s)

∞
F ( s ) =L { f ( t ) } =∫ e−st f ( t ) dt (if it converges) (Laplace transform of f(t) for (t ≥ 0))
0

Inverse Laplace transform of F(s) to f(t)

L−1 { F ( s ) }=f (t )

The purpose of Laplace transform is to solve the ordinary differential equation (ODE) with

ease.

Figure 12. Scheme of Laplace transform (example)

As long as ODE can be modified by Laplace transform, the solution of ODE in the Laplace

domain can be obtained by simple algebraic manipulation. The inverse Laplace transform

eventually gives the solution of ODE.

L {α f 1 ( t ) + β f 2 ( t ) }=α F 1 ( s ) + β F 2 ( s ) , L−1 { α F 1 ( s ) + β F 2 ( s ) }=α f 1 ( t ) + β f 2 ( t )

By definition of linear transform (Linear transform should satisfy properties of both

additivity and homogeneity), Laplace transform is linear transform.

 There are examples of Laplace transforms of some basic functions.

Figure 13. Transforms of Some Basic functions

Also, Laplace transform of a derivative is here;

L { y (n ) ( t ) }=s n Y ( s )−sn −1 y ( 0 )−sn−2 y ' ( 0 ) −⋯− y n−1 (0)

Laplace transform can be used to solve ODEs because the differential terms can be

expressed as polynomials when we know the initial condition for the differential term.

3) Derive the first peak time (tp), Overshoot (OS) and Decay ratio (DR) in second order

system.

Figure 14. Underdamped second order system

 Y ( s) U ( s) Y ( s ) k1 k2 w2
G ( s )= = =G 1 ( s ) G2 ( s )= = , ( 0<ξ <1 )
X (s ) X (s ) U (s ) ( τ 1 s+ 1 )( τ 2 s+1 ) s2 +2 wζs +w 2

1 { } 1 w2
Step input X(t) = constant (value: 1) → X ( s )=
s( )
L 1 = , Y ( s )= 2
s s (s +2 wζs+ w2 )

w2 1 s +2 wζ
Y ( s )= = − 2 2
s ( s + 2 wζs+ w ) s s +2 wζs +w
2 2

2 2
w √ 1−ζ −wζi w √ 1−ζ + wζi
1
¿ − { 2 w √1−ζ 2
+
∗1
2 w √ 1−ζ 2
s s +wζ −w √ 1−ζ 2 i s+wζ + w √1−ζ 2 i
∗1
}
β−αi β+ αi
1
¿ −
s {
2β
∗1
+
2β
∗1
s+ α− βi s +α + βi } 2
, ( α =ζw , β=w √ 1−ζ )

¿)

β−αi −αt (βt ) i β+ αi −αt (−βt ) i

y ( t ) =1− e e − e e
2β 2β

β−αi β+ αi ( β−αi ) i ( β +αi ) i

¿ 1−e−αt { 2β
+
2β }
cos βt−e−αt
2β
+ { 2β
sin βt }
−αt e−αt α
¿ 1−e cos βt − sin βt ¿)
β

α 2
¿ 1−e−αt 1+
√ ()β
sin ( βt+ φ )

α 2
∴ y ( t )=1−e−αt 1+
√ ()
β
sin ( βt + φ ) ¿

Ⅰ) first peak time (tp)

 dy α 2 α 2
( )
dt t =t 0
=α e−α t 1+ 0

√ () β
0

√
sin ( β t 0 +φ )−β e−α t 1+
β ()
cos ( β t 0 + φ )=0

sin ( β t 0+ φ ) β nπ
→ =tan ( β t 0 +φ ) = → tan ( β t 0+ φ ) =tanφ → β t 0=nπ , t 0 =
cos ( β t 0+ φ ) α β

(φ=tan
−1
( αβ ) , tan function has π period ¿
π π
t p=t 0 ( n=1 )= =
β w √ 1−ξ 2

Ⅱ) Overshoot (OS)

−πα
α 2 α 2
a+ b= y ( t p )=1−e −α t p

√ 1+()
β
sin ( π + φ )=1+e−α t
1+
p

β √ ()
sin ( φ ) =1+ e β 2
=1+e−πξ / √ 1−ξ (

β
sin φ= 2 )
√α + β2

b=lim y ( t )=1 ∴a=e− πξ /√ 1−ξ

t→∞

a 2
OS= =α =e−πξ / √1−ξ
b
 Ⅲ) Decay rate (DR)

−3 απ
3π α 2
b+ c= y ( t 0 ( n=3 ) ) = y ( )
β
=1−3 β

√ 1+ ()
β
sin ( 3 π + φ )=1+ e−3 απ / β

β
( sin φ= 2 )
√α + β2

b=lim y ( t )=1 a=e−πξ / √1−ξ c=e−3 απ / β = e−3 πξ / √ 1−ξ 2

t→∞

c 2 πξ
DR= =exp ⁡{− }
a √ 1−ξ2

Ⅳ) Period (P)

P=t 0 ( n=i+2 )−t 0 ( n=i ) ( ∵i : peak → i+1 : valley → i+2 : peak )

2π 2π
¿ =
β w √ 1−ξ 2

4) Derive the transfer function of the closed loop system (P-controller, I-controller, PI-

controller).

Figure 15. Block diagram of the closed loop system

C(s): transfer function of the controller

 P(s): transfer function of the tank

correlation of the system :Y ( s )=P ( s ) U ( s )=C ( s ) P ( s ) { R ( s )−Y ( s ) }

→ { C ( s ) P ( s ) +1 } Y ( s )=C ( s ) P ( s ) R ( s )

Y ( s ) : output value C ( s) P ( s)
→ total transfer function of system ,G ( s )= =
R ( s ) :set value C ( s ) P ( s ) +1

In order to get G(s), we should know the C(s) and P(s).

Transfer function of the tank has formation of first order system. (By open loop system

without the controller, P(s) is determined.

Y ( s ) L11 ( s ) kc
P ( s ) :tank → P ( s )= = =
U ' ( s ) V 1 ( s ) τs+1

t
1 de (t )
U ( t )−Ú=k p {e ( t ) + ∫ e ( t ¿ ) d t ¿ + τ D }
τI 0 dt

' kp 1
Laplace → U ( s )=k p E ( s )+ E ( s )+ k p τ D sE ( s ) ( e ( 0 )=0 )
τI s

U ' ( s) 1 1 ki kp
→ C ( s )=
E ( s) [ ]
=k p 1+ × +τ D s =k p + + k d s k i= , k d =k p τ D
τI s s τI ( )

Ⅰ) P-controller (k i=k d =0 ¿

kc
C ( s ) =k p , P ( s ) =
τs+1

k p kc k p kc
C ( s) P (s) τs+1 k p kc k p k c +1 k 'c
G ( s )= = = = = '
C ( s ) P ( s ) +1 k pk c k p k c + τs+1 τ τ s +1
{ }
τs+ 1
+1 k p k c +1 { s+1
}
 Y ( s) k 'c τ k k
∴ G ( s) = = ' , τ'= , k 'c = p c
R ( s ) τ s+1 k p k c +1 k p k c+ 1

Ⅱ) I-controller (k p=k d=0 ¿

ki kc
C ( s) = , P ( s) =
s τs+1

2
C ( s) P ( s ) k i k c /(τ s 2+ s) ki kc k i k c/ τ (w ')
= = = =
C ( s ) P ( s ) +1 ki kc k i k c +s +τ s2 1 k i k c k i k c s 2+ 2 s w ' ξ + ( w ' )
2

τ s 2+ s
+1 2
s +2
(
2 √ ki kc τ )√ τ
s+
τ

2
Y ( s) ( w' ) 1 ki k c
∴ G ( s) = = 2
R ( s ) s + 2ζ w' s+ ( w' )2
, where ζ =
2 √ ki kc τ
, w '
=
τ √
Ⅲ) PI-controller (k ¿ ¿ d=0) ¿

ki kc
C ( s ) =k p + , P ( s )=
s τs+1

(k ¿ ¿ p s+k i) k c kp kp
C ( s ) P (s )
=
2
τ s +s
=
( k p s+ k i ) k c
=
ki kc ( ki
s+ 1 ) =
ki kc ( ki
s+1 )
{ C ( s ) P ( s ) +1 } ( k p s +k i ) k c ( k p s+ k i ) k c + s+τ s2 k i k c + ( k c k p +1 ) s+ τ s2 2 k k +1 √ k k k k
+1 s +2× c p × i c + i c
2
τ s +s √ ki kc τ √ τ τ

2
Y (s ) ( w' ) ( τ ' s +1 ) k c k p +1 ' √ k i k c ' k p
G ( s )= = , where ζ= ,w = ,τ =
R ( s ) s2 +2 ζ w' s + ( w ' ) 2 √k i k c τ √τ ki

5) Write down the procedure of the experiment and theoretical background.

This context exits above.

6) The derivative control parameter is not widely used in practical process control,

state the reason(s) why.

 This is because that using the derivative control mode is not useful for process case that

variable (PV) has a lot of noise on it. Derivative control is based on the rate of change of the

error. Then, noise is small, random, rapid changes in the PV. Thus, noise will not place the

variable actually far from the set value. However, the derivative control can be sensitive

because the rate of change of the instantaneous error cannot be that small. (Below Figure can

illustrate this situation)

Figure 16. Effect of Noise on Derivative.

In short words, since with derivative control, this mode extrapolates slope of the error, it is

highly affected by noise.

In addition, when the unexpected change of error occurs, the overshoot which is caused by

other control signals can be on the decline (the derivative control can overwhelm it).

However, this can meddle the action of other control signals overly. Since magnitude of error

is big but rate of change of error can be small, derivative control could not function alone.

6. Meaningful Questions

1) Is there reason why we use Laplace transform? Solving the first/second ODE is not that
difficult and there is a general solution as mathematical expression.

2) What can situation make damped case?

3) How we can expect that there will be underdamped cases?

4) There are modes of proportional control, integral control, and derivative control. How are

they different (not just mathematical expression)? What does the mathematical expression

mean?

5) If we obtain not appropriate result due to noise, what kind of noise can occur?

7. References

1) Chemical & Biomolecular Engineering Laboratory, CBE301, Daejeon, KAIST, p 49-62.

2) [CBE 301] EXP. E Liquid Level Control Using Coupled Tanks, Minkee Choi, Dept. of

Chemical & Biomolecular Engineering, 2021.

3) https://blog.opticontrols.com/archives/153