You are on page 1of 33

Development of Mathematical Model for Mass,

Energy and Property Integration



Ting Siew Hoo
School of Chemical Engineering
Curtin University Sarawak
Miri, Malaysia
Tingsoufu@hotmail.com


Abstract With current rising cost of materials, energy, waste treatment and new safety and environmental
regulations, the driving force for further improvements towards optimization of a process has never been
stronger. Significant efforts have led to the development of systematic methods in order for mass, energy,
or property of a process to be integrated. But unless all three mass, property and energy are integrated
simultaneously; the optimal network obtained would not be able to fulfill all three mass, energy and
property constraints. To date, little work has been conducted in developing a heuristic approach to integrate
all three mass, property and energy simultaneously to obtain an optimal network design in terms of total
annual cost. Thus this research aims to address the problem by developing a heuristic approach to
simultaneously integrate mass, energy and property. It was decided that due to the complexity of
simultaneously integrating mass, energy and property, mathematical modeling was used to model the
process and solve for the optimal network in terms of total annual cost. A superstructure was developed and
two case studies applied to demonstrate the applicability of the model. Through application of the case
studies, it was proven that the model is capable of integrating simultaneously mass, energy and property to
solve for the minimum total annual cost. The model was also proved to be capable of solving for minimum
operating cost, conduct utility selection, and size coolers, heaters and heat exchangers.

Keywords-component; Simultaneous, mass, property, energy, integration, in-direct heat exchange.

I. INTRODUCTION

During the last few decades, sustainability has been a much discussed topic both in industry and
academia. Long gone are the days where the abundance of natural resources and relaxed or non-existent
environmental and safety regulations could be taken for granted. Just electricity alone, tariff rates of
industrial consumers have been increasing
[20][9][21]
. As for raw material inputs for example petrochemicals,
prices have also been increasing annually
[4].
Water which is a vital raw material for any process plant has
seen a price rise as well
[12]
. New environmental regulations for examples the Mercury and Air Toxics
Standards (MATS)
[21]
or the Carbon Tax Scheme are constantly being introduced which increase
production costs. The primary preferred method in industry to meet these challenges of increased production
cost is to boost the efficiency of the process
[11]
. This has led to the development of Process Integration to
better the efficiency of resource consumption.

Process Integration is defined by El-Halwagi (1997) as a holistic approach to process design which
emphasizes the unity of the process
[8]
. In Process Integration, there are three main categories, which are
mass, energy and property integration. Unless all three integrations are conducted simultaneously, other
combinations of Process Integrations types would have its limitations as summarized in the Table 1. But as
to date, only one paper regarding the implementation of all three mass, property and energy integration has
been published
[14]
. But in their paper, only direct heat exchange due to mixing was considered.

Among the three types of Process Integration, property integration remains the least developed.
Property integration was made possible when Shelly and El-Halwagi (2000) introduced a property
clustering technique whereby non conserved properties are allocated surrogate properties and their
relationships mapped to allow visualization of the problem
[18]
. By doing so, a stream could be described in
terms of its properties thus an integration can be done on the selected properties. The property clustering
technique has since then been developed further where the Group Contribution Method was introduced to
map materials properties thus allowing for material selection to be implemented into a design problem
[10]
.

Other techniques and application have been developed for example algebraic techniques via
component-less design
[17]
where the limitations of graphical methods to tracking only 3 properties are
overcome. This approach to property integration was employed to simultaneously synthesis a property-
based Water Reuse/Recycle and Interception networks for batch process
[6]
as well as resource conservation
network synthesis in palm oil mills
[2]
. Other recent development was the implementation of property based
automatic targeting techniques which was originally developed for mass integration
[7]
combined with the
cascade analysis techniques to design resource conservation networks
[3]
. This has allowed network design
using property integration to be conducted independently using steps very much similar to that of mass
integration.

Deng-Chun and Xiao-Fung (2011) then proposed and improved problem table (IPT) to be used in
conjunction with the composite table algorithm (CTA) to obtain targets for a property based water network
with multiple resources
[5]
. In their research, they conducted a property pinch analysis using composite
curves and proceeded with the network design. Because mass integration and property integration
methodologies are very much similar, it is then possible to simultaneously integrate property and mass
through the use of IPT and CTA. Property integration has also been further studied to obtain not only the
targets for optimal network design but also the optimal synthesis and scheduling times for a hybrid
dynamic/steady- state property integration network
[13]
. In their research, they utilized algebraic techniques
to model the property interception network and solved it using LINGO software at different time intervals to
obtain the optimal equipment scheduling time.

Research on property integration thus have been heavily focused on either property integration by
itself or combined with mass integration. Currently little work has been published regarding the inclusion of
energy integration. The closest research available was an optimization analysis of a direct recycle networks
with the simultaneous consideration of the property, mass and thermal effects
[14]
. Mathematical modeling
was used to generate viable network designs to meet all three flowrates, temperature, and vapor pressure
constraints but the scope of their research only encompasses the thermal effects of mixing streams without
recovery of the energy through non direct heat exchange. As a result, their model would only prove feasible
for systems where the temperature constraints can be met through only mixing of streams which in industry
are rare. Thus in this work, the main objective is to develop a mathematical model to address the issue of
mass, energy and property integration which incorporates not only direct heat exchange but also indirect
heat exchange in order to generate an optimal network design.



II. PROBLEM STATEMENT

The problem addressed in this research is defined as:

Provided are sets of process sources NSOURCE which consists of mass streams that may be reused/
recycled or discharged. Each source is designated a flowrate

where the subscript


i
indicates the process
source. Each source is then characterized by their temperature T
i
, compositions C
i
, and properties

.

A set of fresh source NFRESH is also provided at a cost of Cost
r
to supplement the process source in
the formation of an optimal network. The subscript
r
indicates the fresh source. The fresh sources are
designated by their Temperature T
r
, composition C
r
, and properties

.

A set of utilities designated NUTILITY are provided which are streams that are required in order to
cool or heat streams. The subscript
k
indicates the type of utility. Each utility is then characterized by their
temperature T
k
, and composition C
k
. The utility is purchased at a cost; COST
k
to supplement the heating and
cooling needs of the network.

A set of sinks henceforth referred to as NSINKS are process units that can accept the flowrate of the
sources. The sinks are then further categorized into three types of sinks. NSINKpro which represents the
process sinks, NSINKutil which represents the utility sinks and NSINKenv which represents the
environment sinks.

For NSINKpro, each is required to fulfill either a fixed or ranged flowrate F
j
, composition C
j
,
temperature T
j,
or property

constraint. The subscript


j
indicates the sink. However for NSINKenv, only
some may be required to fulfil composition C
j
, temperature T
j
or property

constraints. For NSINKutil,


only the composition constraints C
j
need to be fulfilled.

For property, mixing rules are required to define the splitting and mixing of different properties.
Such mixing rules are defined by Equation 1.

(1)

) and

are property operators that govern the properties of mixture

and stream

.
The subscript
p
refers to the type of property while

refers to the fractional contribution of stream towards


the total mixtures

, flowrate.

As for temperature, mixing rules are also required in order to define the splitting and mixing of each
stream. The mixing rule for temperature is as shown in Equation 2.

(2)

F, denotes flowrate while Cp denotes the specific heat capacity, T denotes the temperature and T
0

denotes the reference temperature. The subscript indicates the stream in which the variable represents.
Equation 2 shows that temperature of the mixture stream is determined based on the contribution of heat
energy from each stream.

In the event that the use of fresh and process sources are unable to meet the temperature demands of
the sinks, heat exchange equipments would thus be required. The formula that governs the energy exchange
between the hot stream and the cold stream is shown in Equation 3.

)

The subscript
Hot
refers to the hot stream while the subscript
Cold
refers to the cold stream. Subscript
a

represents after heat exchange and subscript
b
represents before heat exchange.

The objective is thus to synthesize a network that provides the optimal allocation of resources while
satisfying the flowrate, composition, temperature, and property constraints of the sinks at minimum TAC.





III. OUTLINE OF THE MODEL

NSINK, NSOURCE, NUTILITY, NFRESH, NMIX represent the sets for sinks, process sources,
utilities, fresh sources and mixers. To denote each process source, sink, utilities, fresh sources and mixers,
the following indices i, j, k, r, and m are used respectively. As for sinks, they are further categorized into
three types of sinks which are process sinks j
Pro
designated as the set NSINKpro, environmental sinks, j
Env
as
set NSINKenv and utilities sinks, j
Util
as set NSINKutil. Each type of sink is given an additional subscript
behind the j indices as specified in the nomenclature section. As for the mixers they are further categorized
into two types which are process mixers m
Pro
designated as set NMIXpro, and energy integration mixers
m
Ener
designated as set NMIXener.

It can be noted that each process sink is designated an energy integration node (EIN) for which the
proposed model would determine it to be a heat exchanger, heater, cooler or disabled. Upstream each EINs
are two mixers. Ones to the left of each EIN are the mixers designated NMIXpro and function to meet the
sinks mass and property demands while mixers on the top of each EIN are for the energy exchange streams
used to meet the temperature constraints (if required) of each sink named NMIXener. Each fresh source is
connected to individual splitters and then each split stream connected to the NMIXpro and NMIXener. The
same is done to process sources. Each process source is connected to individual splitters but the only
difference is that split streams from process sources splitters are connected not only to the NMIXpro and
NMIXener but also to NSINKenv. For utility streams, they are connected to their designated splitters and
the split streams connected only to the NMIXener and NSINKener.

From each EIN, there are now two outputs, one from the NMIXpro and the other from the
NMIXener. The streams from the NMIXener are then sent to their respective environmental sinks
NSINKenv or utility sinks NSINKutil based on their compositions. As for the streams from the NMIXpro,
they are then connected to their respective process sinks, NSINKpro and must meet the mass, property and
thermal constraints to be deemed feasible.

To model the temperatures of each stream, the specific heat capacity of each stream are required.
The molar fractions of components in each stream are then tracked and the specific heat capacity determined
based on the molar fraction of components within the stream. With the specific heat capacity, input
flowrates and initial temperatures of the fresh source and process source known, the temperature of any
stream after mixing can then be determined by tracking the heat energy in each stream.

As for energy exchange at the EIN, the stream that exits the NMIXpro into the EIN would undergo
energy exchange with the stream that exits the NMIXener into the EIN. To identify correct match of
temperatures to ensure no temperature cross occur, constraints are set based on the inlet and the outlet
temperatures of the streams entering and exiting the EIN.

To determine the cost of individual heaters, coolers and heat exchangers for the TAC optimization
objective function, sizing of the equipment was required. To calculate for area, the duty, log mean
temperature difference and overall heat transfer coefficient are required. To obtain the duty Equation 4 was
used. The mass flowrates, specific heat capacities, inlet outlet temperatures, and overall heat transfer
coefficients of the streams going into each EIN are evaluated using the proposed model.

(4)

Q denotes the heat duty, Cp denotes the specific heat capacity and denotes temperature
difference.



IV. MODEL FORMULATION

A. Fresh Feed Overall Mass Balance NFRESH.

Fresh sources are split and sent to process mixers and environmental sinks as shown in Equation 5.




F
r
denotes the total fresh source flowrate,

denotes the flowrate of fresh source entering the


NMIXpro, and

denotes the flowrate of fresh source entering NSINKenv.




B. Process Source Overall Mass Balance NSOURCE.

Process sources are split and sent to process mixers, energy integration mixers and environmental
sinks as shown in Equation 6.




F
i
denotes the total flowrate of process source while

denotes the flowrate of process source


into NMIXpro,

denotes the flowrate of process source into NMIXener and

denotes the flowrate


of process source into NSINKenv.


C. Utilities Overall Mass Balance NUTI LI TY.

Utilities are split and sent to energy integration mixers and utility sinks as shown in Equation 7.




F
k
denotes the total flowrate of utility,

denotes the flowrate of utility entering NMIXener,

denotes the flowrate of utility entering NSINKutil.







D. Overall Mass Balance at NMI Xpro

The total flowrates to the inlet of any NMIXpro mixers are from the fresh and process sources as
shown in Equation 8.

denotes the total flowrate entering NMIXpro.



As for the outputs from NMIXpro, they are sent to the EIN and then to their designated process
sinks. Since no other mixing or splitting of streams occur downstream of the NMIXpro, the flowrate that
exits the NMIXpro must equal to the input flowrate into each sink as shown in Equation 9.

denotes the total flowrate entering NSINKpro.



For Equation 9,

.


E. Overall Mass Balance at NMI Xener.

The total flowrates to the inlet of any NMIXener mixers are from the process source and utility as
shown in Equation 10.

denotes the total flowrate entering NMIXener.


For the output from NMIXener, they are sent to the EIN and then to the environment sink
NSINKenv or the utility sinks NSINKutil. Thus the output from the NMIXener must equal to the flowrate
entering the EIN and also equal to the sum of the flowrate that enter the environment or utility sinks as
shown in Equation 11.

denotes the flowrate out from NMIXener entering NSINKenv while

denotes
the flowrate out from NMIXener entering NSINKutil.


F. Overall Mass Balance at NSI NKpro.

Since each process sink is designated an EIN, the flowrates to the EINs NMIXpro must equal the
flowrates into each designated NSINKpro. The balance is shown in Equation 12.



For Equation 12,

.


G. Overall Mass Balance at NSI NKutil.


The total flowrate for the utility sink,

comes only from the utility source,

, and the
EIN output from the NMIXeners,

. Equation 13 shows the balance.


denotes the total flowrate into each utility sink.




H. Overall Mass Balance at NSI NKenv.

The total flowrate to the environment sink comes from the fresh source, the process source and the
NMIXener outputs as shown in Equation 14.

denotes the total flow to the environment sink NSINKenv.




I . Composition Balance at NMI Xpro.

To convert the mass flowrate balance to composition balance, the mass fraction of the desired
composition is multiplied with the mass flowrate to yield the flowrate of the desired component. The
symbol C is used to denote mass fraction and the subscript denotes the stream which the mass fraction is
referring to.

Thus for the composition balance of NMIXpro input, the equation is shown in Equation 15.

denotes the composition of the input stream into NMIXpro while

denotes the composition


of the fresh source and

denotes the composition of the process source.



As for the output of NMIXpro that enters through the EIN to the process sinks, the balance is shown
in Equation 16.

denotes the composition of the flow entering NSINKpro.



For Equation 16,

.


J . Composition Balance at NSI NKpro.

The composition balance of NSINKpro is shown in Equation 17.



For Equation 17,

.


K. Composition Balance at NSI NKenv.

The composition balance at NSINKenv is shown in Equation 18.

denotes the composition of the flow entering NSINKenv while

denotes the composition


of the flow from NMIXener.
L. Composition Balance at NSI NKutil

The composition balance at NSINKutil is shown in Equation 19.

denotes the composition of the flow entering NSINKutil while

denotes the composition of


the flow from utility source entering NSINKutil.


M. Composition Balance at NMI Xener.

The composition balance of the input into NMIXener is shown in Equation 20.



The composition balance of the output of NMIXener that passes through the EIN and enters either
the environmental sink or the utility sink is shown in Equation 21.







N. Property Balance at NMI Xpro.

For property balance, unlike composition balance, not all properties mix and split linearly. Thus
properties have to be converted into a linear form in order for it to be tracked for all streams within the
system. This is achieved through the use of different property operators for different properties as. The
symbol

denotes the property operators and the type of property is denoted by the subscript.

The property balance of the NMIXpro input is shown in Equation 22.

demotes the property


p
of stream NMIXpro while

denotes the property


p
of the fresh
source entering NMIXpro and

denotes the property


p
of the process source entering NMIXpro.

As for the NMIXpro output property balance, the balance is shown in Equation 23.

denotes the property


p
of the flow entering NSINKpro.

For Equation 23,

.






O. Property Balance at NSI NKpro.

The output from the NMIXpro passes through the EIN and enters the NSINKpro without any
subsequent mixing or splitting. Thus the property of the stream that exits the NSINKpro must equal to the
property that enters the NSINKpro as shown in Equation24.



From Equation 24,

.


P. Energy Balance for NMI Xpro.

As for the energy balance, stream temperatures are tracked based on the amount of heat energy in
each stream using Equation 5. To evaluate the heat of a stream, the specific heat capacity is denoted using
Cp. As for temperature, all streams are evaluated based on the difference from T
0
which equals 273.15
Kelvin. The streams temperatures are denoted by the symbol T. As for the subscripts, a dictates the
temperature of the stream AFTER heat exchange and b as BEFORE heat exchange.
The energy balance for the input into NMIXpro is then defined in Equation 25.

and

denotes the specific heat capacity and temperature of the stream entering
NMIXpro.

and

denotes the specific heat capacity and temperature of the fresh source entering
NMIXpro.

and

denotes the specific heat capacity and temperature of the process source entering
NMIXpro.

As for the output temperature from NMIXpro, it is dependent on the energy exchange (if any) that
occurs at the EIN. Further explanation on the mathematical model of the energy exchange would be
described in section S.



Q. Energy Balance for NSI NKenv.

The energy balance input for the environmental sink is modeled as Equation 26.

and

denotes the specific heat capacity and temperature of the stream entering
NSINKenv.

and

denotes the specific heat capacity and temperature of the stream entering
the NMIXener.


R. Energy Balance for NMI Xener.

The energy balance input for NMIXener mixers as shown in Equation 27.

and

denotes the specific heat capacity and temperature of the NUTILITY streams.

The output temperature from NMIXene is also dependent on the energy exchange (if any) that
occurs at the EIN. Further explanation on the mathematical model of the energy exchange would be
described in section S.


S. Energy Exchange at the EI N.

To determine the output temperature of the stream from the EIN, an energy balance between the
outputs of the NMIXpro and NMIXener is required. Equation 28 shows the mathematical model of the
balance.

denotes the temperature of the stream from NMIXener entering the EIN while


denotes the temperature of the stream from NMIXener exiting the EIN after energy exchange (if any).

denotes the temperature of the stream from the NMIXpro exiting the EIN after energy
exchange (if any) while

denotes the temperature of the stream from the NMIXpro entering the
EIN.

Since the stream from NMIXpro is the same as the stream that enters NSINKpro as stated in
Equation 12, Equation 29 must then hold true as the temperature of the stream from NMIXpro AFTER heat
exchange must equal to the temperature of the stream entering the process sinks.

denotes the temperature of the stream entering NSINKpro.



Equation 28 shows that if LINGO designates 0 flow entering NMIXener,

, then

must equal 0 as

and

are never 0. This means that no heat is exchanged as the


temperature of NMIXpro entering the EIN comes out the same. When

, it shows that the


stream undergoes heating while if

, it shows that the stream undergoes cooling. As to


whether the EIN is a cooler/heater or heat exchanger would depend on the stream entering the NMIXener. If
the source is from NSOURCE, then the EIN is considered as a heat exchanger. If the source for the
NMIXener is from NUTILITY, then the EIN is considered as a heater/cooler.
The main challenge when LINGO is left to evaluate if the EIN is a cooler, heater or heat exchanger
is when trying to determine the existence of temperature crosses. Temperature crosses are when heat is
being transferred from a cooler source to a hotter source which is impossible as stated by the 2
nd
law of
thermodynamics. The 2
nd
law of thermodynamics state that heat will spontaneously flow only from a hot
region to a cold region
[1]
. To determine if the stream exiting NMIXpro is the cold stream or the hot stream,
the difference between

and

are used. If

, the stream undergoes heating


thus it must be a cold stream and the stream from NMIXener must be a hot stream. If

,
the stream undergoes cooling thus it must be a hot stream and the stream from NMIXener must be a cold
stream. In order to prevent temperature crosses, the inlet and outlet temperatures for the hot stream must
always be higher than the inlet and outlet temperatures for the cold stream. This is where the use of the
logical function IF is used as shown in Equation 30 and 31. Assuming within the energy exchange unit, the
flow is counter current, Equation 30 evaluates the temperature match between the streams from NMIXpro
entering the EIN with the streams from the NMIXener exiting the EIN. Equation 31 evaluates the
temperature match between the streams NMIXener entering the EIN with the streams from the NMIXpro
exciting the EIN. The evaluated variable is then given the symbols B
1
and B
2
as shown in Equation 30 and
31.

@IF


then

]
else




@IF


then

]
else

are binary values.



B
1
and B
2
cannot be 0 as division by 0 would be encountered during the sizing of the equipment. But
when the EIN node is disabled, there is a possibility that B
1
and B
2
equals to 0. Thus binary values are used.
In both Equation 30, and 31,

are binary values which determine if each individual EIN is enabled or


disabled. When

=1 means that the EIN is enabled while

=0 means that the EIN is disabled. M


being any positive value would then allow B
1
and B
2
to equal 0 when

= 0 as subtraction between

with

and

with

would equal to any number lower than 0. Thus the right


hand side equation must be lower than zero in order for LINGO to accept

or

to be 0. When

= 1, Equation 26 and 27 ensures that for the EIN, the hot streams
input and output is always higher than the cold streams input and output.


T. Heater, Cooler, Heat Exchanger Sizing.

To optimize the network based on TAC, any heaters, coolers and heat exchangers have to be sized to
determine their costs. Through sizing, the area required to achieve the desired energy exchange is obtained
and using area, the cost of the unit can be determined. Equation 32 is used to obtain the area of each heater,
cooler or heat exchanger.

|
|

)
|
|

denotes the area of the heat exchange equipment for a particular NSINKpro while U denotes
the overall heat transfer coefficient.

(

denotes the overall heat transfer coefficient if process source is being used to heat or cool while

denotes the overall heat transfer coefficient if a particular utility is being used to heat of cool.

From Equation 32 a situation arises where when B
1
= B
2
, then the term (

) in Equation 33 would
equal 0 thus a division by 0 error is present. Thus Equation 34 is used to prevent cases where B
1
= B
2
.

|

| ( (

))



Equation 34 ensures that when the EIN is enabled, B
1
can never equal B
2
as the absolute value
obtained by subtraction between the two would yield a number greater than 0.


U. Cooler, Heater, Heat Exchanger Cost Estimation

To estimate the Cost of each heater, cooler and heat exchanger, the area is multiplied by the cost per
unit area (CPUA) of a heat exchange
[16]
. The total cost is then converted to the Equivalent Annual Cost
(EAC) based on a cost of capital (COC) of 5% and an expected lifetime of 20 years. The formula is shown
in Equation 35.


(


EAC denotes the equivalent annual cost while COC denotes the cost of capital, CPUA denotes cost
per unit area and y denotes the expected lifetime.

V. Specific Heat Capacity Calculation.

Unless the specific heat capacity of each stream is given, it needs to be evaluated. The specific heat
capacity of each stream is tracked based on the mole fraction of components in the stream. Equation 36
represents the Cp model for NFRESH, NSOURCE, NMIXpro, NSINKenv, and NSINKpro streams.
Equation 37 represents the Cp model for NMIXener streams.

denotes the specific heat capacity of NFRESH, NSOURCE, NMIXpro, NSINKenv, and
NSINKpro streams while

denotes the mole fraction of the contributing component of NFRESH,


NSOURCE, NMIXpro, NSINKenv, and NSINKpro streams.

denotes the specific heat capacity of the


pure contributing component.

denotes the mole fraction of the contributing utility component to the NMIXener streams
while

denotes the specific heat capacity of the pure utility.



To convert the mass fractions into mole fraction, the mass flowrates of each stream or a basis mass
flowrate is multiplied with the mass fraction of each component in the stream to determine the mass
flowrate of each component. By dividing the mass flowrate of each component by its molar mass, the mole
flowrate of each component in a stream can be determined. With the mole flowrates, it would then be
possible to determine the mole fraction of each component in a stream.


W. Mass Balance Constraints

Flowrate constraints are imposed on NSOURCE and NMIXpro or NSINKpro. Since the flowrates of
NMIXpro must equal to NSINKpro as stated in Equation 12, only the constraints of NMIXpro is required to
be specified. Equation 38 and 39 represents the mass balance constraints for NSOURCE and NMIXpro.

(39)

The subscript
s
indicates that the value is set by the user base on data obtained from case studies


X. Component Balance Constraints

Component constraints are set for the fresh source, process source and the sink. The sink constraints
are set for two main purposes. One reason is to ensure that the input into each process sink meets the
component constraints set and the other reason is to ensure that the different utilities and process source do
not mix. Since the component flow going into NSINKpro is the same as the component flow entering
NMIXpro as shown in Equation 17, the constraint is set at NMIXpro. The fresh source, utilities, and process
source sinks are set by Equation 40, 41, 42, 43, 44, and 45. Equation 46 and 47 governs the component
constraints for streams entering NSINKpro while Equation 48, 49, 50, and 51 ensures that the different
utilities and process source do not mix and enters the correct environmental or utility sink.

denotes the concentration of component q in fresh sources.





The subscript
Util
refers to the utility components being tracked from NUTILITY.

denotes the
concentration of component
Util
in fresh source.

denotes the concentration of component q in process sources.


denotes the concentration of component


Util
in process sources.

denotes the concentration of component q in utility sources.


denotes the concentration of component


Util
in utility sources.

denotes the minimum concentration of component q of the NSINKpro streams


while

denotes the maximum concentration of the NSINKpro streams.

denotes the
concentration of component q of NMIXpro streams.

denotes the concentration of component Util in the NMIXpro streams.


denotes the concentration of component q in the NSINKenv streams.


denotes the concentration of component Util in the NSINKutil streams.


denotes the concentration of component Util in the NSINKenv streams.


denotes the concentration of component q in the NSINKenv streams.



By setting up the constraints for the sinks as shown in equation 48, 49, 50, and 51, streams from
NUTILITY would not be able to mix with streams from NSOURCE or NFRESH as there would be no sink
available for the resultant mixed stream which contains both utility components
Util
and process components
q
to flow into.


Y. Property Balance Constraints

The property constraints are set on NSOURCE, NFRESH, NSINKpro, NMIXpro and NSINKenv.
From Equation 23, the property at NSINKpro is the same as the property at NMIXpro, thus the property
constraint for NSINKpro can be set at NMIXpro. Equation 52 shows the constraints for NSOURCE and
NFRESH. Equation 53 shows the constraints for NMIXpro and NSINKpro while Equation 54 shows the
constraints for NSINKenv.

denotes the property


p
of stream d while the superscript MIN represents the
minimum constraint and the superscript MAX denotes the maximum constraint.

denotes the property p for streams.







Z. Energy Balance Constraints

The energy constraints are set as temperature constraints for NSOURCE, NFRESH, NUTILITY, and
NSINKpro. Based on Equation 29, the temperature of the stream from NMIXpro that exits the EIN equals
the temperature of the process sink NSINKpro. Thus the temperature constraints for NSINKpro can be set
as the temperature of the stream from NMIXpro which exits the EIN. Equation 55 depicts the temperature
constraints for NSOURCE, NFRESH, and NUTILITY while Equation 56 depicts the temperature
constraints for NSINKpro.

denotes the temperature for the streams.




AA. Objective Functions

The objective function is determined by the desired optimized variable and in most cases; TAC is
desired to be minimized. TAC involves not only the OC, but also the Equivalent Annual Cost of the heaters,
coolers and heat exchangers. Piping costs may be included in the TAC if the data is available.

It is however not uncommon to find cases where the desired optimized variable is the OC which
includes only the material cost of NFRESH and NUTILITY and the treatment costs of NSINKutil and
NSINKenv.

Equation 58 models the minimization of the TAC while Equation 59 models the minimization of
OC, and Equation 60 models the minimization of TAC plus piping costs.

)









The symbol Cost denotes the material cost while the subscript denotes which stream to cost is
referring to. Tcost however symbolizes the treatment cost while the subscript denotes the stream in
reference. As for piping costs, the symbol used is Pcost while the subscript indicates at which point in the
system does the pipeline starts and the ends at.





V. CASE STUDIES

In order to illustrate the applicability of the proposed model, the proposed model was applied to two
case studies. The first case study was based on a vinyl acetate monomer plant modified by Tan et al. (2012)

[19]
. This case study was selected as the simplicity of the system allows for an easy understanding of the
application of the proposed model. The TAC and the OC were evaluated and the results compared to those
obtained by Tan et al. (2012).

The second case study was adopted from Kheireddine Houssein, et al, (2011) which is a phenol
production plant utilizing the decomposition of cumene hydroperoxide (CHP) that forms phenol and acetone
in the presence of sulphuric acid
[14]
. This case study was selected to demonstrate that the proposed model is
applicable for larger systems as well as non-component based property constrains. The TAC and OC+PCost
were evaluated and the results compared to those obtained by Kheireddine Houssein, et al, (2011).


A. Case Study 1

A case study based on a vinyl acetate monomer plant modified by Tan et al. (2012) is adopted and
the model implemented to illustrate the approach proposed
[19]
. The data for the fresh and process source are
tabulated in Table 2 while the sinks data are tabulated in Table 3.

This case study consists of 84 variables with 120 constraints and is solved within 3 minute using a
notebook (2.3 GHz, 4GB RAM). A total of 650837 iterations were required before the global optimum was
obtained. The optimal solution is shown in Figure 2.

Based on the results, with mass, property and energy integration, the solution presents a minimum
TAC of $48.08 million/yr. $47.92 million/yr comes from the fresh feed costs while $161,640/yr from utility
and $2,700/yr from the equipment costs. Compared to the $62.7 million/yr obtained through Tan et al.
(2012) method, there seems to be a large difference, but this is due to the method in which they calculated
utility cost which is different. In their paper, utility cost was calculated based on the duty. But in this model,
the utility cost is calculated using flowrate of the utility itself. If Tan et al.s (2012) utility cost was
calculated with the same method used in this model, their TAC would add up to only $48.21 million/yr.
$48.165 million/yr from fresh feed costs and $48,900/yr for utility costs. The fresh feed flow obtained using
this model is 9584 kg/h while from the original case study was 9633 kg/h.

The lower TAC obtained through implementation of this model is due to the 0.51 % lower fresh feed
usage as compared to the case study. Another thing to note is that due to the lower fresh feed usage, less
waste is also generated, 4784 kg/h compared with 4833.5 kg/h. Although the utility cost based on this model
is 3.3 times higher than the original case study, the increase in utility costs is small as compared to the
savings from the reduction of fresh feed. The cost of coolers, heaters and heat exchangers is also minor
compared to the fresh feed costs as the lifetime of the coolers, heaters and heat exchangers were set to 20
years
[14]
. The results for TAC along with the results by Tan et al. (2012) with modified utility costs are
tabulated in Table 4.


The solution from implementing the model yields the same number of heaters, coolers and heat
exchangers as the original case study. Both solutions showed 1 cooler and 1 heat exchanger but with the
model, sizing of the heat exchangers showed that the cooler requires a total area of 16 m
2
while the heat
exchanger required an area of 35.1 m
2
.

For comparison, the objective function for this case study was switched to OC and the changes to the
optimum network observed. The results are shown in Table 4 and the network shown in Figure 3.

There is notable change to the network after modifying the objective function to OC. Although the
operating costs are the same as compared to the TAC objective function case, the size of the heat exchanger
and cooler are much larger in the OC case which makes sense as they have yet to be optimized. The
difference in size is due to the different temperature of the streams entering the EIN. The size of the heat
exchanger is 43.65m
2
and the cooler is 49.9m
2
. This equates to a higher TAC cost by about $2240/yr.







B. Case Study 2

This case study is to show that the model is capable of integrating larger systems as well as
incorporate non component based property constraints. The case study consists of phenol production from
the decomposition of cumene hydroperoxide (CHP) that forms phenol and acetone in the presence of
sulfuric acid. Figure 4 shows a simplified process diagram that depicts the streams present in the system.

From Figure 4, it is found that this case study contains 2 fresh sources, 3 process sinks which are
Wash101, R104, and Wash102, and 3 process sources which are from Wash101, D101, and Wash102. The
property being tracked in this case study is vapor pressure. The data for this case study was obtained from
work by Kheireddine Houssein, et al, (2011) with the lower limit temperature requirements of the Wash102
sink modified from 60 to 7 and tabulated in Table 5 and 6
[14]
. Piping costs
are also available from their work and is tabulated in Table 7. In this case study however, the heat of mixing
is omitted as well as the varying Cp of the streams in an effort to simplify the case study.

This case study consists of 166 variables with 215 constraints and is solved within 2.50 minutes
using a notebook (2.3 GHz, 4GB RAM). A total of 288631 iterations were required before the global
optimum was obtained. The optimal solution is shown in Figure 5.

Based on the results, using process source alone to achieve the modified temperature constraint for
Wash101 is not possible, thus the need for the use of utilities to heat the stream. And since heat exchange is
not required to meet the temperature requirements of Wash102 and Neutralizer R104, no heat exchange
equipment is needed for those two sinks. The solution presents a minimum TAC of $98,200/yr compared to
the $93,825/yr obtained by Kheireddine Houssein, et al, (2011). The higher TAC is expected as the
minimum temperature constraint for Wash101 was increased causing it to require the use of utilities and a
heater. As for the size of the heater, 3.04 m
2
is required. Table 8 summarizes the results for case study 2
TAC.

To compare, the OC plus piping cost for case study 2 was generated as well and shown in Figure 6.
The data is then tabulated in Table 8. The optimal value was found to be $98,007/yr. The change to the
network after modifying the objective function to OC plus piping cost is a 1.25 kg/hr drop in hot oil feed. As
the objective function was meant to optimize only the operating cost and piping cost, LINGO would
conclude that this network with a lower flowrate of hot utility is more optimal. But through a drop in the hot
oil flowrate, the temperature profile of the heater has now changed causing the size of the heat exchanger to
increase from the initial 3.04 m
2
to 10.4 m
2
. This is due to a smaller temperature driving force which leads
to the need of a larger heat transfer area to achieve the desired final temperature. The increase in area
correlates to a higher TAC by $350/yr.



VI. CONCLUSION

Through this research, a new mathematical model for the simultaneous mass, energy and property
integration has been introduced with the capabilities of more accurately optimizing the total annual cost as
heat integration has been included. Sizing becomes possible for the heaters coolers and heat exchangers
involved in the network thus their costs can be determined using the model. In addition, the proposed
approach is capable of determining the best utility or process stream to be used for heat integration thus
aiding in utility selection for a network design. The number of heaters, coolers or heat exchangers would
also be evaluated through the use of this model. An optimization based on the total annual cost and the
operating cost by formulating an Integer Non Linear Problem (INLP) has been conducted on two case
studies and the results illustrates the usefulness of the new model.



VI. RECOMMENDATIONS

After thorough analysis of the data from the two case studies, the following recommendations are
made.

Currently, the model is capable of only exchanging heat from either process sources or from utilities
as only one EIN is provided for each sink. Further research should be done in order to increase the number
of EIN for each sink in order for more possible network designs to be available for evaluation using
optimization software.

Another recommendation would be to develop the model further as to include heat of mixing as well
as varying specific heat capacities based on temperature thus allowing for further refinement on the
accuracy of the model.

Further work should also be conducted in order to recycle the utilities and process streams used for
heat exchange back into other EIN or process sinks as to further improve the number of feasible networks
for the model.



REFERENCES

[1] Asimov, and Isaac. Thermodynamics. 1996. http://mysite.pratt.edu/~arch543p/readings/thermodynamics.html#1
(accessed October 7, 2012).
[2] Chen, Cheng-Liang, Jui-Yean Lee, and Danny Sum Ng Kok. "Property Integration for Resource Conservation Via Network
Synthesis in Palm Oil Mills." Chemical Engineering Journal, 2011: 207-215.
[3] Chwan, Dominic Yee Foo, Denny Sum Ng Kok, and Raymond R. "Automated Targeting Technique for Concentration and
Property Based Total Resouce Conservation Network." Computers and Chemical Engineers, 2010: 835-840.
[4] Cos., McGrawHill. Platts: September Petrochemical Prices up 7.5% on Higher Raw Material Costs. August 10, 2012.
http://www.stockhouse.com/news/usreleasesdetail.aspx?n=8634361 (accessed October 4, 2012).
[5] Deng-Chun, and Xiao-Feng. "Targeting for Conventional and Property-Based Water Network with Multiple Resources."
Ind. Eng. Chem. Res., 2011.
[6] Denny Kok Sum Ng, Dominic Chwan Yee Foo, Arwa Rabie, Mahmoud M. El-Halwagi. "Simultaneous Synthesis of
Property-Based Water Resuse/Recycle and Interception Networks for Batch Process." AIChE, 2008.
[7] El-Halwagi, M. M., and Manousiothakis. "Automatic Synthesis of Mass-exchange Networks with Single Component
Targets." Chemical Engineering Science, 1990: 2813-2831.
[8] El-Halwagi, Mahmoud, M. Pollition Prevention Through Process Integration Systematic Design Tools. United States:
Academic Press, 1997.
[9] European Commission. Energy price statistics. November 2011.
http://epp.eurostat.ec.europa.eu/statistics_explained/index.php/Energy_price_statistics#Industrial_consumers
(accessed October 1, 2012).
[10] Fadwa, Tahra Eljack, Charles Conrad Solvason, Nishanth Chemmangattvalappil, and Mario Richard Eden. "A Property
Based Approach for Simultaneous Process and Molecular Design." Chinese Journal of Chemical Engineering, 2008: 424-
434.
[11] Global Sources. "China Supplier Survey." Expect higher China prices in months ahead. January 2011.
http://www.globalsources.com/SITE/CHINA-SURVEY-HIGH-RAW-MATERIAL-COST.HTM (accessed October 5, 2012).
[12] Global Water Intelligence. "Tariff rises outstripped by inflation." Global Water Intel. September 2012.
http://www.globalwaterintel.com/archive/13/9/market-analysis/tariff-rises-outstripped-inflation.html (accessed
October 3, 2012).
[13] Grooms, Daniel, Vasiliki Kanzatzi, and Mahmoud El-Halwagi. "Optimal Synthesis and Scheduling of Hybrid
Dynamic/Steady State Property Integration Networks." Computers and Chemical Engineering, 2005: 2318-2325.
[14] Kheirddine, Houssein,, Younas Dadmohammadi, Deng Chun, Xiao Feng, and El-Halwagi Mahmoud. "Optimization of
Direct Recycle Networks with Simultaneous Consideration of Property, Mass, and Thermal Effects." Ind. Eng. Chem. Res.,
2011: 3754-3762.
[15] McGinley, Shane. Dubai water and electricity costs to rise around 15% in 2011. December 9, 2010.
http://www.arabianbusiness.com/dubai-water-electricity-costs-rise-around-15-in-2011-366557.html (accessed October
2, 2012).
[16] Peter, Max S., Klaus D. Timmerhaous, and Ronald E. West. Plant Design and Economics For Chemical Engineers.
Singapore: McGraw Hill, 2004.
[17] Qin, X., F. Gabriel, D. Harell, and M. M. El-Halwagi. "Algebraic Techniques for Property Integration Via Componentless
Design." Ind. Eng. Chem. Res., 2011: 3792-3798.
[18] Shelly, M. D., and M. El-Halwagi. "Componentless Design of Recovery and Allocation Systems: A Functionality Based
Clustering Approach." Computer Chemical Engineers, 2000: 2081-2091.
[19] Tan, Y. L., D.K.S Ng, M.M. El-Halwagi, D.C.Y. Foo, and Y. Samyudia. "Synthesis of Heat Integrated Resource Conservation
Networks." 11th International Symposium on Process Systems Engineering (Editors: I.A. Karimi and R. Srinivasan),
Computer Aided Chemical Engineering, 2012: 985-989.
[20] Tenaga Nasional Berhad. Pricing and Tariff. 2012. http://www.tnb.com.my/residential/pricing-and-tariff.html (accessed
October 1, 2012).
[21] U.S. Energy Information Administration. "Annual Energy Outlook 2012." U.S. Energy Information Administration
Independant Statics and Analysis. June 25, 2012. http://www.eia.gov/forecasts/aeo/ (accessed Oct 4, 2012)

You might also like