You are on page 1of 22

Chemical Product and Process

Modeling
Volume 5, Issue 1

2010

Article 6

Simulation and Control of a Commercial


Double Effect Evaporator: Tomato Juice
Praveen Yadav, Indian Institute of Technology, Kharagpur
Amiya K. Jana, Indian Institute of Technology, Kharagpur

Recommended Citation:
Yadav, Praveen and Jana, Amiya K. (2010) "Simulation and Control of a Commercial Double
Effect Evaporator: Tomato Juice," Chemical Product and Process Modeling: Vol. 5: Iss. 1,
Article 6.
DOI: 10.2202/1934-2659.1443

Brought to you by | Universiti Teknologi Malaysia


Authenticated
Download Date | 1/11/16 12:01 PM

Simulation and Control of a Commercial


Double Effect Evaporator: Tomato Juice
Praveen Yadav and Amiya K. Jana

Abstract
This work aims to present a detailed study on a commercial double-effect tomato paste
evaporation system. The modeling equations formulated for process simulation belong to
backward feeding arrangement. Open-loop process dynamics has been studied by rigorous
simulation of the model structure. In the next, three multi-loop control schemes, namely
conventional proportional integral (PI), gain-scheduled PI (GSPI) and nonlinear PI (NLPI), have
been synthesized for the sample process. Finally, several simulation experiments have been
conducted to investigate the comparative closed-loop performance based on set point tracking and
disturbance rejection.
KEYWORDS: evaporator, double-effect, tomato paste, modeling, dynamics, control

Brought to you by | Universiti Teknologi Malaysia


Authenticated
Download Date | 1/11/16 12:01 PM

Yadav and Jana: Simulation and Control of an Evaporator

1 INTRODUCTION
An evaporator is commonly used to concentrate a solution by removing a part of
the solvent in the form of vapor. It has various areas of application. In most of the
industrial and commercial applications, the multiple effect evaporator is used due
to its advantages over the single effect system. The first and most important
advantage is the economy. Multiple effect scheme evaporates more water per kg
of steam fed to the unit by reusing the vapor from one effect as the heating
medium for the next. Secondly, the heat transfer gets improved due to the viscous
effects of the products as they become more concentrated. The invention of the
multiple effect evaporators is the result of the demand of sugar industry. A
revolution in the sector of sugar industry was brought by Norbert Rillieux by the
development of multiple pan evaporation system for use in sugar refinery.
Literature reviews revealed that in the beginning, the study on the multieffect evaporators was based on steady state analysis. Subsequently, the research
attention was paid to develop the dynamic model of the evaporation system. In the
1960s, the mathematical model for a single effect evaporator was proposed by
Andersen et al. (1961) and the simulation was carried out after reduction and
linearization of that model to study the closed-loop control performance using an
analog computer. An empirical input-output model of a single concentration
evaporator with PID control application was described by Kropholler and Spikins
(1965). Andre and Ritter (1968) formulated the nonlinear model of a doubleeffect evaporator.
During 1970s, it was realized through the development of several process
models that the important behaviors of the evaporator system can easily be
described by its dynamic nature. Linear and nonlinear models of a generic
evaporator were discussed in detail by Newell and Fisher (1972). The simulation
study became easier with the development of a computer code which is capable of
simulating the steady state condition of a multiple effect evaporator. This
computing technique was brought by Bolmstedt and Jernquist (1976) which was
further supported by their publication in 1977 showing a dynamic simulator
through blocks which is capable of simulating more complex plants.
A mathematical model with a wide variety of its extension for plants of
different configurations, including the death-time arising due to circulation in
each effect and through the pipe within effects, was developed by Tonelli (1987).
In the past, the mathematical models were constructed for open-loop simulations
and for application of conventional control laws. But in the last decade, the state
space models suitable for designing the multivariable controllers and state
estimations (Newell and Lee, 1989) were reported. Cadet et al. (1999) formulated
a detailed evaporator model based on energy and mass balance with considering
semi-empirical equilibrium functions. This model was implemented in a sugar

Brought to you by | Universiti Teknologi Malaysia


Authenticated
Download Date | 1/11/16 12:01 PM

Chemical Product and Process Modeling, Vol. 5 [2010], Iss. 1, Art. 6

plant and showed satisfactory results. Recently, a phenomenological, stationary


and dynamic model of a multi-effect evaporator was presented by Miranda and
Simpson (2005) for simulation and control purposes.
In the recent years, energy conservation is a big issue for the research and
industrial organizations. Thus the latest research efforts by the scientists and
engineers working in industrial organizations are towards more efficient use of
energy. Balkan et al. (2005) did the performance evaluation of a triple-effect
evaporator with forward feeding using exergy analysis. Kaya and Sarac (2007)
developed a model for a multiple effect evaporator and performed energy
analysis. There are several types of feeding patterns for the evaporator systems
such as forward feeding, backward feeding, mixed feeding and parallel feeding.
Each operation was investigated by the authors with and without pre-heating
arrangements. The effect of pre-heating on evaporation process was investigated
from the point of energy economy. Mohanty and Khanam (2007) developed a
simplified model based on the principles of process integration for the analysis of
multiple effect evaporator systems taking into account the variation in physicothermal properties as well as boiling point rise. It included new concepts of stream
analysis, temperature path and internal heat exchange for the formulation of the
model equations.
Modeling and online control is very relevant for food concentrates, mainly
because of its influence on product quality and also on energy consumption
(Cadet et al., 1999). Although, a significant progress has been made on modeling,
there are limited papers dealt with the control of multi-effect evaporation systems
(e.g., Runyon et al., 1991; Kam and Tad, 2000). It is with this intention that the
present work has been undertaken.
In this paper, a systematic study is conducted on a commercial doubleeffect tomato paste evaporator. The dynamic process model, consisting of mass
balance, energy balance and empirical correlations, is presented by the
differential-algebraic equations. The simulation of the model structure is
performed for open-loop process dynamics. For closed-loop study, three multiloop control strategies, namely conventional proportional integral (PI), gainscheduled PI (GSPI) and nonlinear PI (NLPI), have been synthesized. Finally, a
comparative control performance is addressed on the sample process. The
contribution of this paper is the comparison of the three control schemes.
2 THE PROCESS
The example process as shown in Figure 1 is a double-effect evaporator with
backward feeding arrangement used for tomato concentrate. The two effects are
numbered from left to right as Tank1 and Tank2, respectively. The raw juice
having flow rate F , concentration X f and temperature T f enters Tank2, and the
DOI: 10.2202/1934-2659.1443

Brought to you by | Universiti Teknologi Malaysia


Authenticated
Download Date | 1/11/16 12:01 PM

Yadav and Jana: Simulation and Control of an Evaporator

steam with flow rate S and temperature T S enters Tank1. The mass holdup in the
two tanks are defined as M 1 and M 2 . V 1 and V 2 are the vapor flow rates from
the overhead of two tanks with temperature T 1 and T 2 , respectively. P1 and P 2
are the product flow rates from the two effects with product concentration X p
and X 2 , and temperature T 1 and T 2 , respectively. The steady state and parameter
values are listed in Table 1 (Runyon et al., 1991).

Figure 1 Schematic of a double-effect evaporator.


Process model
The dynamic model for the sample process is derived for tomato concentrate
based on the study of Runyon et al. (1991), and Miranda and Simpson (2005).
This model is also validated (Runyon et al., 1991) for a different set of design and
parametric variables. An evaporation process involves mass and heat transfer. The
tomato juice is assumed as a binary solution of water and soluble solids, both
considered inert in a chemical sense. The macroscopical evaporator model
consisted of a set of differential-algebraic equations (DAEs) has been constructed
based on conservative laws and empirical relationships. It should be noted that

Brought to you by | Universiti Teknologi Malaysia


Authenticated
Download Date | 1/11/16 12:01 PM

Chemical Product and Process Modeling, Vol. 5 [2010], Iss. 1, Art. 6

only the juice phase is considered for modeling. The assumptions involved in the
formulation of model are listed below.
Negligible heat losses to the surroundings
Homogeneous composition and temperature inside each evaporator
Variable liquid holdup and negligible vapor holdup
Overhead vapors considered as pure steam
Latent heat of vaporization or condensation varied with temperature
No boiling point elevation of the solution
Table 1 Steady state and parameter values.

Term

Abbreviation (unit)

Value

Tank1 mass holdup

M10 (kg)

2268

Tank2 mass holdup

M20 (kg)

2268

Input feed flow rate

F0 (kg/hr)

26103

Input steam flow rate

S0 (kg/hr)

11023

Tank1 liquid product flow rate

P10 (kg/hr)

5006

Tank2 liquid product flow rate

P20 (kg/hr)

14887

Vapor flow rate from Tank1

V10 (kg/hr)

9932

Vapor flow rate from Tank2

V20 (kg/hr)

11165

Feed composition

Xf0 (kg/kg)

0.05

Tank1 composition

Xp0 (kg/kg)

0.2607

Tank2 composition

X20 (kg/kg)

0.0874

Steam temperature

Ts0 ( C)

115.7

Feed temperature

Tf0 ( C)

85.0

Temperature in Tank1

T10 (0C)

74.7

Temperature in Tank2

T20 ( C)

Heat transfer area of Tank1

A1 (m )

Heat transfer area of Tank2

A2 (m2)

Overall heat transfer coefficient for Tank1


Overall heat transfer coefficient for Tank2

52.0

102
412
20

5826

20

2453

U1 (kJ/hr.m . C)
U2 (kJ/hr.m . C)

DOI: 10.2202/1934-2659.1443

Brought to you by | Universiti Teknologi Malaysia


Authenticated
Download Date | 1/11/16 12:01 PM

Yadav and Jana: Simulation and Control of an Evaporator

Total mass balance


First effect: dM 1 = P2 P1 V 1
dt

(1)

Second effect: dM 2 = F P2 V 2
dt

(2)

Component (solids) mass balance


First effect:

d (M 1 X p)
= P 2 X 2 P1 X p
dt

(3)

M1

dX p
dM 1
+Xp
= P 2 X 2 P1 X p
dt
dt

(4)

M1

dX p
dM 1
= P2 X 2 P1 X p X p
dt
dt

(5)

Substituting equation (1),


M1

dX p
= P2 X 2 P1 X p X p ( P2 P1 V 1)
dt

M1

dX p
= P2 X 2 P1 X p X p P2 + X p P1 + X p V 1
dt

dX p P2 ( X 2 X p ) + X p V 1
=
dt
M1

Second effect:

d ( M 2 X 2)
= F X f P2 X 2
dt

Simplifying and rearranging, finally we get


dX 2 = F ( X f X 2) + X 2V 2
dt
M2

(6)

(7)

(8)

(9)

(10)

Brought to you by | Universiti Teknologi Malaysia


Authenticated
Download Date | 1/11/16 12:01 PM

Chemical Product and Process Modeling, Vol. 5 [2010], Iss. 1, Art. 6

Energy balance

The steam flow rate to the first effect is obtained through energy balance on the
first effect heat exchanger as:

S (T S ) = U 1 A1 (T S T 1)

(11)

where, is the latent heat. That means,


( )
S = U 1 A1 T S T 1
(T S )

(12)

Similarly, the vapor flow rate to the second effect is derived from the energy
balance on the second effect heat exchanger as:
V 1 (T 1) = U 2 A2 (T 1 T 2)

(13)

( )
V 1 = U 2 A2 T 1 T 2
(T 1)

(14)

In the following, the energy balance equations are derived.


d [ M 1 h(T 1 , X p )]

First effect:

M1

M1

dt

dh(T 1 , X p )
dt

dh(T 1 , X p )
dt

= P2 h(T 2 , X 2) + S (T S ) P1 h(T 1 , X p ) V 1 H (T 1) (15)

+ h(T 1 , X P ) dM 1 = P2 h(T 2 , X 2) + S (T S )
dt
P1 h(T 1 , X p ) V 1 H (T 1)

(16)

= P2 h(T 2 , X 2) + S (T S ) P1 h(T 1 , X p )
dM 1
V 1 H (T 1) h(T 1 , X p )
dt

(17)

Substituting equation (1),

DOI: 10.2202/1934-2659.1443

Brought to you by | Universiti Teknologi Malaysia


Authenticated
Download Date | 1/11/16 12:01 PM

Yadav and Jana: Simulation and Control of an Evaporator

M1

dh(T 1 , X p )
dt

= P2 [h(T 2 , X 2) h(T 1 , X p )] + S (T S ) V 1[ H (T 1) h(T 1 , X p )] (18)

Using equation (11),


M1

dh(T 1 , X p )
dt

= P2 [h(T 2 , X 2) h(T 1 , X p )] + U 1 A1 (T S T 1) V 1[ H (T 1) h(T 1 , X p )]


------------ (19)

This gives,
dh(T 1 , X p )
dt

P 2 [h(T 2 , X 2) h(T 1 , X p )] + U 1 A1 (T S T 1) V 1[ H (T 1) h(T 1 , X p )]


M1
-----------(20)

Second effect:
d [ M 2 h(T 2 , X 2)]
= Fh(T f , X f ) + V 1 (T 1) P 2 h(T 2 , X 2) V 2 H (T 2)
dt

(21)

Simplifying and rearranging, finally we have


dh(T 2 , X 2) F [h(T f , X f ) h(T 2 , X 2)] + U 2 A2 (T 1 T 2) V 2 [ H (T 2) h(T 2 , X 2)]
=
dt
M2
-----------(22)
Empirical correlations

The enthalpy of the product (tomato juice) is represented as (Heldman and Singh,
1981):
h(T , X ) = (4.177 2.506 X ) T

(23)

The pure solvent vapor (steam) enthalpy is obtained using a polynomial


regression equation of values from the steam tables as:

Brought to you by | Universiti Teknologi Malaysia


Authenticated
Download Date | 1/11/16 12:01 PM

Chemical Product and Process Modeling, Vol. 5 [2010], Iss. 1, Art. 6

H (T ) = 2495.0 + 1.958 T 0.002128 T 2

(24)

For the condensate streams, the pure solvent liquid enthalpy is also found from
the steam tables as:
h(T ) = 4.177 T

(25)

The latent heat of vaporization can be computed as:

(T ) = H (T ) h(T ) = 2495.0 2.219 T 0.002128 T 2

(26)

Using the above correlations, the energy balance equations ((20) and (22)) have
the following final forms:
dT 1 P2 (4.177 2.506 X 2)(T 2 T 1) U 2 A2 (T 1 T 2) + U 1 A1 (T S T 1)
=
dt
M 1 (4.177 2.506 X p )

(27)

dT 2 = F (4.177 2.506 X f )(T f T 2) + U 2 A2 (T 1 T 2) + V 2 [4.177 T 2 H (T 2)] (28)


dt
M 2 (4.177 2.506 X 2)
Therefore, the final form of the model includes equations (1), (2), (8), (10), (12),
(14) and (25)-(28).
3 APPLICATION OF CONTROL THEORY
3.1 Control objectives

The control objectives for an evaporation system are selected taking into account
the product specifications, operational constraints and cost considerations. For the
concerned process, the primary objective is to maintain the product solids
concentration (or product viscosity) at its desired value. In order to achieve the
desired product quality in presence of disturbance and uncertainty, several
additional control schemes, as mentioned below, need to employ with the
evaporator.
To prevent the overflow or drying out of evaporator tubes, liquid mass holdup
should be controlled.
To avoid the product degradation or damage, temperature must be
maintained at the desired value.

DOI: 10.2202/1934-2659.1443

Brought to you by | Universiti Teknologi Malaysia


Authenticated
Download Date | 1/11/16 12:01 PM

Yadav and Jana: Simulation and Control of an Evaporator

To reduce the steam consumption, steam economy should be maximized.


3.2 Degrees of freedom analysis

The evaporator model presented earlier includes fifteen independent variables


[ F , P1 , P2 , M 1 , M 2 , T f , T S , T 1 , T 2 , S , V 1 , V 2 , X f , X p and X 2] and eight
independent equations ((1), (2), (8), (10), (12), (14), (27) and (28)). Among these
variables, the inputs, namely T f , TS , X f , S and F are specified by the external
world. Obviously, the degrees of freedom is seven. In order to have a completely
determined process, the number of its degrees of freedom should be zero. For this
purpose and to meet the control objectives, five control pairs have been selected
and two input variables, X f and T f , can be treated as known disturbances.
3.3 Selection of control pairs

For the example tomato paste evaporator, the followings are selected as controlled
variables: (i) final product concentration ( X p ), (ii) temperature of the inlet steam
( TS ), (iii) temperature of the second effect ( T 2 ), (iv) liquid mass holdup in first
effect ( M 1 ), and (v) liquid mass holdup in second effect ( M 2 ). In order to
regulate the process variables, the corresponding manipulated variables are
chosen as the product flow rate from first effect ( P1 ), the steam flow rate ( S ), the
vapor flow rate from second effect ( V 2 ), the product flow rate from second effect
( P2 ), and the feed flow rate ( F ). Figure 1 as well as Table 2 includes all these
control configurations.
Table 2 Control pairings and controllers used.

Controlled variable
Xp
TS

Manipulated variable
P1
S

Controller type
PI, NLPI and GSPI
PI, NLPI and GSPI

T2
M1
M2

V2
P2
F

PI, NLPI and GSPI


P-only
P-only

3.4 Control equations

In this paper, a comparative control study is presented. For this, three types of
multi-loop controllers, namely PI, NLPI and GSPI, have been designed in this
9

Brought to you by | Universiti Teknologi Malaysia


Authenticated
Download Date | 1/11/16 12:01 PM

Chemical Product and Process Modeling, Vol. 5 [2010], Iss. 1, Art. 6

section for the example evaporator. The ISE performance criterion has been used
in selecting the controller tuning parameter values and they are reported in Table
3.

3.4.1 Proportional integral (PI) controller


The general form of a single-loop PI control law is given by the following
expression:

U = U 0 + K [(Y sp Y ) +

(Y

sp

Y )dt ]

(29)

Here, K is the proportional gain, the integral time constant, U 0 the bias signal
and Y sp the set point value of Y .

Parameter
Kf

Table 3 Tuning parameters values.


P or PI
NLPI
a = 3.0

1000 hr -1
0

GSPI

KV2 , KV20

1250 kg/hr. C

1250 kg/hr. C

1550 kg/hr.0C

KS , KS0

15 kg/hr.0C

15 kg/hr.0C

15 kg/hr.0C

KP1, KP10

30000 kg/hr

20000 kg/hr

20000 kg/hr

-1

KP2

800 hr

V 2

0.03 hr

0.01 hr

0.01 hr

0.35 hr

0.52 hr

0.52 hr

P1

0.05 hr

0.01 hr

0.01 hr

3.4.2 Nonlinear proportional integral (NLPI) controller


Different forms of nonlinear PI (and PID) controller are available in scientific
literature. It can be developed by adding higher-order terms of the error signal and
integral of the error to the control law. Also, there is a way to make the controller
parameters functions of the error (Cheung and Luyben, 1980) or parameter
scheduling (Rugh, 1987) or both (Jutan, 1989). Actually, there is no particular
control scheme that has become standard in the literature.

DOI: 10.2202/1934-2659.1443

Brought to you by | Universiti Teknologi Malaysia


Authenticated
Download Date | 1/11/16 12:01 PM

10

Yadav and Jana: Simulation and Control of an Evaporator

Here, the following nonlinear PI control scheme is applied on the


evaporator. The proportional gain of the NLPI is a function of the absolute value
of the error.

U = U 0 + K [(Y sp Y ) +

(Y

sp

Y )dt ]

with K = K 0 (1 + a Y sp Y )

(29)
(30)

where, K 0 is the initial fixed gain. It is obvious that if there is no integral term,
the controller output is effectively proportional to the square of the error.

3.4.3 Gain-scheduled proportional integral (GSPI) controller


A gain-scheduled PI law (Bequette, 2003) has been designed for the
representative process. The GSPI has the following form:

U = U 0 + K (Y )[(Y sp Y ) +

(Y

sp

Y )dt ]

(31)

The controller gain, K , is varied aiming to keep K K P constant, which then


keeps the stability margin constant. When the process gain is characterized as a
function of the scheduling variable, K P (Y ) , then the controller gain can be
scheduled as:

K (Y ) =

K (Y0 ) K P (Y 0)
K P (Y )

(32)

The gain of the GSPI scheme used has the following forms:
(i) When Y > Y0 ,
1 Y0
K (Y ) = K 0
1Y

(33)

where, K P (Y ) = 1 Y and K (Y 0) = K 0 .
(ii) When Y < Y0
K (Y ) = K 0

(34)
11

Brought to you by | Universiti Teknologi Malaysia


Authenticated
Download Date | 1/11/16 12:01 PM

Chemical Product and Process Modeling, Vol. 5 [2010], Iss. 1, Art. 6

Here, Y 0 represents the nominal operating value of Y (= Y sp ). It is worthy to


mention that this is a one-way approach. Since the process gain increases with
lower purity, maintaining a constant controller gain speeds up the response when
the product is less pure.
4 RESULTS AND DISCUSSION

Matlab codes have been developed to generate simulation results. The 4th order
Runge-Kutta method is used to solve the differential equations contained in the
model. In the subsequent discussion, the open-loop followed by the closed-loop
evaporator performance is presented.

DOI: 10.2202/1934-2659.1443

Brought to you by | Universiti Teknologi Malaysia


Authenticated
Download Date | 1/11/16 12:01 PM

12

Yadav and Jana: Simulation and Control of an Evaporator

Figure 2 Effect of a pulse input change in Tank2 product flow rate (changed from
14887 to 15200 kg/hr at time = 5 hr and then from 15200 to 14887 kg/hr at time
=10 hr).
4.1 Open-loop performance

Figure 2 illustrates the effect of Tank2 product flow rate on the main product
composition. Two consecutive step changes have been introduced in the product
flow rate (step increase: 14887 15200 kg/hr at time =5 hr; step decrease: 15200

13

Brought to you by | Universiti Teknologi Malaysia


Authenticated
Download Date | 1/11/16 12:01 PM

Chemical Product and Process Modeling, Vol. 5 [2010], Iss. 1, Art. 6

14887 kg/hr at time =10 hr). As a consequence, the final product purity gets
disturbed and the process attains a new steady state against a change. This result
shows the interactive behavior of the process variables and it confirms that the
sample evaporator is open-loop stable.
4.2 Comparative closed-loop performance
4.2.1 Disturbance rejection

The comparative regulatory performances are depicted in Figures 3 and 4


introducing a pulse input change in feed concentration (step increase: 0.05
0.0525 kg/kg at time = 5 hr; step decrease: 0.0525 0.05 kg/kg at time = 15 hr)
and feed temperature (step increase: 85 900C at time = 5 hr; step decrease: 90
850C at time = 15 hr), respectively. Both the figures include the control
performance in terms of final product composition and liquid mass holdup in
Tank1. However, Tables 4 and 5 record the integral square error (ISE) for all
control loops. It is obvious from the simulation results that the NLPI and GSPI
provide better performance over the classical PI controller. Due to the moderate
nonlinearity of the process, the nonlinear controllers outperform the PI.
Table 4 ISE values corresponding to Figure 3.
M1
M2
Xp
T2
Controller
PI
NLPI
GSPI

2.16x10-6
4.93x10-7
4.97x10-7

0.0052

7.23x10-4

5.76x10-7
1.73x10-7
6.89x10-8

Table 5 ISE values corresponding to Figure 4.


M1
M2
Xp
T2
Controller
PI
NLPI
GSPI

7.25x10-8
1.65x10-8
1.65x10-8

2.11x10-4

0.0017

0.0011
1.73x10-4
2.33x10-4

DOI: 10.2202/1934-2659.1443

Brought to you by | Universiti Teknologi Malaysia


Authenticated
Download Date | 1/11/16 12:01 PM

TS
0.2505
0.2030
0.0762

TS
0.0044
0.0026
0.0024

14

Yadav and Jana: Simulation and Control of an Evaporator

Figure 3 Effect of a pulse input change in feed concentration (changed from 0.05
to 0.0525 kg/kg at time = 5 hr and then from 0.0525 to 0.05 kg/kg at time = 15
hr).

15

Brought to you by | Universiti Teknologi Malaysia


Authenticated
Download Date | 1/11/16 12:01 PM

Chemical Product and Process Modeling, Vol. 5 [2010], Iss. 1, Art. 6

Figure 4 Effect of a pulse input change in feed temperature (changed from 85 to


900C at time = 5 hr and then from 90 to 850C at time = 15 hr).

DOI: 10.2202/1934-2659.1443

Brought to you by | Universiti Teknologi Malaysia


Authenticated
Download Date | 1/11/16 12:01 PM

16

Yadav and Jana: Simulation and Control of an Evaporator

4.2.2 Set point tracking

Figure 5 displays a comparative servo performance between PI, NLPI and GSPI
control algorithms against a pulse change in set point value of the final product
composition (step increase: 0.2607 0.27 kg/kg at time = 5 hr; step decrease:
0.27 0.2607 kg/kg at time = 15 hr). This simulation experiment confirms that
the PI controller shows relatively poor performance compared to other two PIs.
The performance in terms of ISE values is analyzed in Table 6. Overall, the GSPI
showed best performance due to its ability of timely changing the gain.
Table 6 ISE values corresponding to Figure 5.
M1
M2
Xp
T2
Controller
PI
NLPI
GSPI

-5

2.77x10
2.22x10-5
2.25x10-5

0.01917

0.0090

-6

2.59x10
2.09x10-6
1.03x10-6

TS
0.1866
0.1381
0.0568

5 CONCLUSIONS

This article presents the open-loop as well as closed-loop operation of a


commercial double-effect tomato paste evaporation system. The model structure
developed based on backward feeding approach consists of differential algebraic
equations. Open-loop simulation shows that the sample evaporator is a stable
process. Three multi-loop control strategies, namely traditional PI, NLPI and
GSPI, have been synthesized to investigate the comparative performance.
Simulation experiments based on set point tracking and disturbance rejection
show that the classical PI controller provides relatively poor performance over the
NLPI and GSPI algorithms. The quantitative performance analysis has also been
included by computing ISE values.

17

Brought to you by | Universiti Teknologi Malaysia


Authenticated
Download Date | 1/11/16 12:01 PM

Chemical Product and Process Modeling, Vol. 5 [2010], Iss. 1, Art. 6

Figure 5 Effect of a pulse set point change in final product composition (changed
from 0.2607 to 0.27 kg/kg at time = 5 hr and then from 0.27 to 0.2607 kg/kg at
time = 15 hr).
NOMENCLATURE

A
F
h(T)

Heat transfer surface area (m2)


Feed flow rate (kg/hr)
Enthalpy of pure liquid solvent (condensate) at temperature T (kJ/kg)

DOI: 10.2202/1934-2659.1443

Brought to you by | Universiti Teknologi Malaysia


Authenticated
Download Date | 1/11/16 12:01 PM

18

Yadav and Jana: Simulation and Control of an Evaporator

H(T)
K
M
P
S
t
T
U
V
X

Enthalpy of pure vapor solvent (steam) at temperature T (kJ/kg)


Proportional gain
Mass holdup of liquid product (kg)
Liquid product flow rate (kg/hr)
Steam flow rate (kg/hr)
Time (hr)
Temperature (0C)
Overall heat transfer coefficient (kJ/hr.m2.0C) or manipulated variable
Vapor flow rate (kg/hr)
Mass fraction (kg solids/kg stream)
Integral time constant (hr)

Subscripts
f
p
S
sp
0
1
2

Feed
Final product
Steam
Set point
Steady state
First effect
Second effect

REFERENCES

Andersen, J.E., Glasson, L.W., Lee, F.P., The Control of Single Concentration
Evaporator, Transactions of the Society of Instrument Technology, Vol.
13, 121 (1961).
Andre, H., Ritter, R.A., Dynamic Response of a Double-Effect Evaporator,
Canadian Journal of Chemical Engineering, Vol. 46, 259 (1968).
Balkan, F., Colak, N., Hepbasli, A., Performance Evaluation of a Triple-Effect
Evaporator with Forward Feed using Exergy Analysis, International
Journal of Energy Research, Vol. 29, 455-470 (2005).
Bequette, B.W., Process Control: Modeling, Design, and Simulation, PrenticeHall, New Delhi (2003).
Bolmstedt, U., Jernquist, A., Simulation of the Steady-state and Dynamic
Behaviour of Multiple Effect Evaporator Plants, American Institute of
Chemical Engineers Journal, Vol. 17, 1080 (1976).
Bolmstedt, U., Jernquist, A., Simulation of the Steady-state and Dynamic
Behavior of Multiple Effect Evaporation Plants, Computer Aided Design,
Vol. 9, 29-40 (1977).

19

Brought to you by | Universiti Teknologi Malaysia


Authenticated
Download Date | 1/11/16 12:01 PM

Chemical Product and Process Modeling, Vol. 5 [2010], Iss. 1, Art. 6

Cadet, C., Toure, Y., Gilles, G., Chabriat, J.P. Knowledge Modelling and Nonlinear Predictive Control of Evaporators in Cane Sugar Production Plants,
Journal of Food Engineering, Vol. 40, 59-70 (1999).
Cheung, T.F, Luyben, W.L., Nonlinear and Nonconventional Liquid Level
Controllers, Industrial and Engineering Chemistry Fundamentals, Vol.
19, 93 (1980).
Heldman, D.R., Singh, R.P., In: (2nd edn. ed.), Food Process Engineering, AVI
Publishing Company, Westport, Connecticut (1981).
Jutan, A.A., Nonlinear PI(D) Controller, Canadian Journal of Chemical
Engineering, Vol. 67, 485-493 (1989).
Kam, K.M., Tad, M.O., Simulated Control Studies of Five-Effect Evaporator
Models, Computers and Chemical Engineering, Vol. 23, 1795-1810
(2000).
Kaya, D., Sarac, H. I., Mathematical Modeling of Multiple-Effect Evaporators
and Energy Economy, Energy, Vol. 32, 1536-1542 (2007).
Kropholler, H., Spikins, D., Principles of Control for Chemical Engineers. Part
3, Chemical Process Engineering, 558-567 (1965).
Miranda, V., Simpson, R., Modeling and Simulation of an Industrial Multiple
Effect Evaporator: Tomato Concentrate, Journal of Food Engineering,
Vol. 66, 203-210 (2005).
Mohanty, B., Khanam, S., Development of an Efficient Linear Model for the
Analysis of Multiple Effect Evaporator System, International Conference
on Advances in Energy Research, 724-730 (2007).
Newell, R.B., Fisher, D.G., Model, Development, Reduction, and Experimental
Evaluation for an Evaporator, Industrial and Engineering Chemistry
Process Design and Development, Vol. 11, 213 (1972).
Newell, R.B., Lee, P.L., Applied Process Control: A case study, Prentice-Hall,
Australia (1989).
Rugh, W.J., Design of Nonlinear PID Controllers, American Institute of
Chemical Engineers Journal, Vol. 33, 1738-1742 (1987).
Runyon, C.H., Rumsey, T.R., McCarthy, K.L., Dynamic Simulation of a
Nonlinear Model of a Double Effect Evaporator, Journal of Food
Engineering, Vol. 14, 185-201 (1991).
Tonelli, S., Estudio dinmico y control de evaporadores de mltiple efecto,
Tesis de Doctorado en Ingeniera Qumica. Universidad Nacional del Sur.
Baha Blanca, Argentina (1987).

DOI: 10.2202/1934-2659.1443

Brought to you by | Universiti Teknologi Malaysia


Authenticated
Download Date | 1/11/16 12:01 PM

20

You might also like