Applied Thermal Engineering 70 (2014) 369e379

Contents lists available at ScienceDirect

Applied Thermal Engineering

journal homepage: www.elsevier.com/locate/apthermeng

Comparison between pinch analysis and bridge analysis to retrofit

the heat exchanger network of a kraft pulp mill
J.C. Bonhivers a, E. Svensson b, T. Berntsson b, P.R. Stuart a, *
a
NSERC Design Engineering Chair in Process Integration, Department of Chemical Engineering, Ecole Polytechnique de Montreal, Canada
b €teborg, Sweden
Heat and Power Technology, Chalmers University of Technology, Go

h i g h l i g h t s g r a p h i c a l a b s t r a c t

Reducing the heat consumption im-

plies decreasing the flow-rate of heat
cascaded through the HEN.

A bridge between cooler and heater
is necessary to achieve heat savings
in an existing HEN.

Bridge analysis provides more infor-
mation for network retrofit than
pinch analysis does.

a r t i c l e i n f o a b s t r a c t

Article history: Pinch analysis is based on the hot, cold and grand composite curves and is the most commonly-used
Received 30 January 2014 approach to identify strategies for reducing energy consumption by heat exchanger network retrofit.
Accepted 19 April 2014 This method was originally developed for the synthesis of new networks, and there remain certain
Available online 14 May 2014
difficulties for its application to improve existing networks. The advanced composite curves have been
developed for retrofit situations specifically, and use data about existing heat exchangers to provide more
Keywords:
information about the modifications necessary to achieve heat savings. Bridge analysis, which is based on
Process integration
the energy transfer diagram, is a new method and enumerates the sets of heat transfer modifications
Heat-exchanger network
Retrofit
necessary to save energy. In this paper, the grand composite curve, the advanced composite curves and
Energy savings the energy transfer diagram have been constructed for analysis of the heat exchanger network of a kraft
pulp mill. Links between these methods are made explicit; then results are discussed and compared. It is
shown that the information provided by these approaches is consistent; however, the level of detail
progressively increases from the grand composite curve to the advanced composite curves until the
energy transfer diagram. Fundamentally, reducing the energy consumption implies decreasing the flow
rate of heat cascaded through the network from the hot utility until the environment. As a consequence,
any heat savings solution includes network modifications bridging coolers to heaters. Traditional pinch
analysis does not provide information about the network modifications required after removal of cross-
pinch transfers, while the advanced composite curves indicate the heat savings potential attainable
through modifications of few existing heat exchanger units. Bridge analysis provides more detail about
heat savings modifications, which bridge existing heaters and coolers, than traditional pinch analysis and
the advanced composite curves do.
© 2014 Elsevier Ltd. All rights reserved.

* Corresponding author.
E-mail address: paul.stuart@polymtl.ca (P. Stuart).

http://dx.doi.org/10.1016/j.applthermaleng.2014.04.052
1359-4311/© 2014 Elsevier Ltd. All rights reserved.
370 J. Bonhivers et al. / Applied Thermal Engineering 70 (2014) 369e379
Abbreviations and conventions
ACC advanced composite curves
ACLC actual cooling load curve
AHLC actual heat load curve
AILC actual internal load curve
crx receptor of cooling system Cx (environment)
csx supplier of cooling system Cx
CUC cold utility curve
Cx cooling system x
Ey internal heat exchanger y
ery receptor of internal heat exchanger Ey
esy supplier of internal heat exchanger Ey
ECLC extreme cooling load curve
EHLC extreme heat load curve
EILC extreme internal load curve
1. Introduction
Increasing energy costs and environmental concerns provide
motivation for reducing the energy use in industry. Heat exchanger
networks (HENs) play a significant role in industrial process energy
systems. A high degree of heat recovery in the HEN is important for
overall energy efficiency of the plant. By retrofitting existing HENs
for improved heat recovery, utility consumption and thereby also
energy costs are reduced.
Methods for energy analysis of existing heat exchanger net-
works (HEN retrofit) involve the identification and evaluation of
inefficiencies in the current network, heat-saving modifications to
reduce these inefficiencies, and the selection of the most promising
modifications, i.e., the most profitable ones with acceptable oper-
ability. Methods for HEN retrofit can be broadly categorized into
optimization-based approaches and insight-based approaches. The
optimization-based methods for HEN retrofit can, in turn, be
divided into deterministic and stochastic methods. The determin-
istic methods range from early developments by, e.g., Ciric and
Floudas [11] and Yee and Grossmann [27] to more recent contri-
butions by, e.g., Bjork and Nordman [5], Kralj [14], Nguyen et al. [18]
and Zhang and Rangaiah [28]. One risk of using deterministic
models is that they can be trapped in local optima. The stochastic
methods (see, e.g., Athier et al. [2]; Bochenek and Jezowski [6];
Rezaei and Shafiei [21]; Wang et al. [26] are more likely to find the
global optimum for HEN retrofits, but usually at the expense of long
computation times even for small problems. The optimization-
based methods are highly complex, and evaluation of the quality
of solution may be difficult in practice considering possible trap-
ping in local optimum and inevitable model and parameter un-
certainties. Therefore, in practice, the insight-based approaches
such as pinch analysis [16] are still the most widely used for in-
dustrial applications e also for retrofit (see, e.g., Asante and Zhu [1];
Tjoe and Linnhoff [23]). In pinch analysis, the insights from
graphical tools such as composite curves are used to calculate en-
ergy targets. Heuristics are then applied for network design to
achieve these targets.
The first complete method based on pinch technology for HEN
retrofit targeting and design was presented by Tjoe and Linnhoff
[23]. Their main point in the retrofit design is to eliminate heat
transfer across the pinch by identifying and modifying heat ex-
changers in the existing HEN that exchange heat through the pinch.
While the aim of the retrofit design method of Tjoe and Linnhoff
[23] is to minimize the investment costs for a certain energy saving,
EMAT exchanger minimum approach temperature
Env environment
ETD energy transfer diagram
FRAM future resource-adapted mill, Swedish programme
GCC grand composite curve
Hz heater z
hrz receptor of heater Hz
hsz supplier of heater Hz (hot utility steam)
HEN heat exchanger network
HRAT heat recovery approach temperature
HU hot utility
HUC hot utility curve
TCLC theoretical cooling load curve
THLC theoretical heat load curve
TILC theoretical internal load curve
the economic assumptions are largely based on grassroots design
conditions.
In order to enable a more accurate representation of the costs
involved in HEN retrofit, including costs such as costs for piping,
pressure drops and area extensions of existing heat exchanger, the
Matrix method [10] was developed. Also the Matrix method, which
is based on the concepts of pinch analysis, aims at eliminating pinch
violations. However, some relaxation of this requirement is
allowed. The approach is iterative and based on heuristics.
Different approaches have been proposed in order to simplify or
relax the HEN retrofit problem. For example, the so-called Path
analysis was proposed by van Reisen et al. [24] as a way to reduce
the complexity of the problem by decomposition and prescreening
of the HEN before solution. An extension of the Path analysis called
Structural targeting was presented a few years later [25].
Lakshmanan and Ban ~ ares-Alc
antara [15] proposed a visualiza-
tion tool called the retrofit thermodynamic diagram (RTD). The RTD
is a modification of the conventional stream representation grid
used in pinch analysis and allows for a graphical visualization of
possible retrofit options.
The concept of network pinch was introduced by Asante and Zhu
[1] based on the observation that heat recovery in a HEN is ther-
modynamically constrained by the network topology. The network
pinch, unlike the process pinch, can be altered by modifications such
as relocations or introductions of heat exchangers, or re-piping. The
two-stage approach proposed by Asante and Zhu [1] makes use of
the knowledge about the network pinch by starting with the iden-
tification of a modification that eliminates the network pinch. The
approach relies on mathematical programming for the final opti-
mization of the network design parameters. Hybrid methods
combining pinch analysis and optimization have been proposed also
by others (see, e.g., Briones and Kokossis [9]; Smith et al. [22]).
The main advantages of insight-based methods such as pinch
analysis are their simplicity, their graphical representation, and the
possibility of the design engineer to interact and influence the so-
lution process. However, difficulties in data extraction, practical
targeting and redesign of the network are still encountered using
pinch analysis, not the least in retrofit situations.
In retrofitting, the existing equipment constrains the opportu-
nities for cost-efficient integration. Consequently information
about the existing heat exchanger network has to be included in the
analysis. The advanced composite curves [19] which are based on
the classical pinch curves, include information about the actual
placement of heaters and coolers in the existing HEN. The advanced
 J. Bonhivers et al. / Applied Thermal Engineering 70 (2014) 369e379
composite curves provide an illustration of the current energy
demand as well as the potential savings that would result from a
reduced temperature driving force, but they also illustrate the
location of heaters and coolers in the current network as well as
their theoretical upper and lower temperatures.
A new insight-based method that has the advantages of pinch
analysis in terms of user interaction and graphical visualization
tools, and also overcomes some of the problems with data extrac-
tion and representation of the existing network has recently been
proposed by Bonhivers et al. [7,8]. To reduce the hot utility con-
sumption in a HEN, the flow rate of heat transferred from the
heating utility, through the existing heat exchangers and ultimately
rejected to the environment must be reduced. This progressive
transfer of heat in the existing heat exchanges is explicitly illus-
trated in the so-called energy transfer diagram (ETD) [8], and the
insight of these so-called heat transfer bridges is captured in the
method of bridge analysis [7].
While bridge analysis is a promising approach for energy ret-
rofits of process plants, pinch analysis is still the most used
approach in industrial applications. Consequently, there is a clear
need to compare the concepts and results of these two methods.
In this article, a case study of the heat exchanger network of a
typical Scandinavian Kraft pulp mill forms the basis of such a
comparison.
1.1. Objective
The aim of this paper is to illustrate and critically analyze the
similarities and differences between bridge analysis and the pinch-
based approach to HEN retrofit based on a case study of a typical
Scandinavian Kraft pulp mill. More specifically, one objective of the
paper is to describe the links between the basic pinch curves, the
advanced composite curves (ACC) and the energy transfer diagram
(ETD), both in general terms and as applied to the heat exchanger
network of the pulp mill. A second objective is to compare the re-
sults for the retrofit analysis of the pulp mill obtained using the
pinch and bridge approaches; this involves the identification of
heat-saving bridges and their correspondence to previously iden-
tified pinch violations.
2. Principles of compared methods for HEN retrofit
HEN retrofit opportunities can be investigated if a preliminary
analysis of the industrial energy system shows that there is an in-
vestment opportunity for heat savings. This part presents the
principles of the three approaches for network retrofit that are
compared in the paper, i.e. classical pinch analysis, advanced
composite curves and bridge analysis.
2.1. Classical pinch analysis
The reader is assumed to be familiar with pinch analysis con-
cepts and visualization tools (see e.g., Linnhoff and Hindmarsh [16],
Smith [17], and Kemp [13] for comprehensive material on this
topic). The principle of basic pinch-based retrofit methods is to
identify modifications that remove pinch violations. Cross-pinch
heat transfers, heaters below the pinch and coolers above the
pinch in the actual network are identified and removed; then the
network is rebalanced below and above the pinch.
2.2. Advanced composite curves
The major difference between the traditional pinch curves and
the advanced composite curves is that information about the
existing network is included in the latter. The advanced composite
371
curves show in a single illustration the potential savings that would
result from a reduced temperature difference, the actual locations
of heaters and coolers in the network before retrofit, the theoretical
upper and lower temperatures of each, and the excess heat, both
directly available and potential.
The method uses four composite curves above the pinch and
four below [19]. Only the construction of the curves above the
pinch is explained; the curves below the pinch are constructed in
the same way. The four curves above the pinch are called the Hot
Utility Curve (HUC), the Theoretical Heat Load Curve (THLC), the
Actual Heat Load Curve (AHLC), and the Extreme Heat Load Curve
(EHLC). The corresponding curves below the pinch are the Cold
Utility Curve (CUC), the Theoretical Cooling Load Curve (TCLC), the
Actual Cooling Load Curve (ACLC), and the Extreme Cooling Load
Curve (ECLC).
The HUC is a composite curve of the utility streams in the existing
heaters plotted at real temperatures. Correspondingly, the AHLC is a
composite curve of the parts of the process streams that are heated
by utility in the existing heaters. The EHLC shows the temperatures
at which heat would be supplied to the process if the heat exchange
was carried out so that all the external heat was supplied at the
highest possible temperature and with the same heat requirement as
for the existing process. This corresponds to the part of the cold
composite curve at the highest temperature. The THLC shows the
lowest possible temperatures at which external heat could be sup-
plied for the same heat requirement as for the existing process if
allowing an exchanger minimum approach temperature difference
(EMAT) smaller than the Heat Recovery Approach Temperature
(HRAT) corresponding to the current network.
Correspondingly, the TCLC shows the highest possible temper-
atures at which heat would be extracted if all thermodynamically
possible measures for Energy Conservation had been used with a
given EMAT. Consequently, a certain amount of cooling demand in
the TCLC can be matched with a certain amount of heat demand in
the THLC. This amount of transfer corresponds to the heat savings
resulting from reducing the minimum temperature between the
composite curves from HRAT to EMAT. In summary, the curves
above the pinch show the loads and temperature levels of the
utilities in heaters (HUC), the process streams in the actual (AHLC),
the lowest (THLC) and the highest (EHLC) possible temperature
levels for heaters; the curves below the pinch show the loads and
temperature levels of the utilities in coolers (CUC), the process
streams in the actual (ACLC), the lowest (ECLC), and the highest
(TCLC) possible temperature levels for coolers.
The advanced composite curves provide an estimation of heat-
saving potential of economically feasible projects before detailed
design calculations. Releasing heaters placed low in temperature
and coolers placed high in temperature is usually easier and less
expensive than releasing other heaters or coolers, because fewer
heat exchanger modifications and less exchange area are required.
Consequently, the number of modifications and the required heat
exchanger area of a HEN retrofit ought to be smaller when the AHLC
and the ACLC are close. The advanced composite curves also indi-
cate the temperature levels at which usable excess heat, that is,
heat below the pinch, can be extracted and used, e.g., for eco-cyclic
industrial parks, district heating or cooling, or heat upgrading.
2.3. Bridge analysis
Energy is conserved and degraded through heat exchanges and
process operations. To reduce the heat consumption in an existing
HEN, it is necessary to reduce the flow rate of heat cascaded from
the heating utilities to the environment by modifications in the
heat exchangers. It has been shown that heat savings modifications
must include a bridge between heat outlets and hot utility users
372 J. Bonhivers et al. / Applied Thermal Engineering 70 (2014) 369e379
[7,8]. Bridge analysis has been developed for HEN retrofit and in-
cludes the following: the description of the heat transfer modifi-
cations necessary to save energy which are termed “bridge”, an
energy transfer diagram which allows identifying the bridges, a
method to enumerate them, and a network table for easy
evaluation.
The terminology and conventions of the method are here pre-
sented. Heat is transferred in a heat exchanger from a supplier to a
receptor. Supplier of cooling system Cx, internal (processeprocess)
heat exchanger Ey, and heater Hz are noted csx ; esy and hsz ; respec-
tively. Receptor of cooling system Cx, internal (processeprocess)
heat exchanger Ey and heater Hz are noted crx ; ery and hrz ; respec-
tively. A match is a couple composed of a supplier and receptor. For
example, the match corresponding to heater Hz is noted hsz hrz :
A match is a couple composed of a supplier and a receptor. A set
of new matches connecting heat outlets to hot utility users is
termed “bridge”. A prime bridge cannot be decomposed into other
bridges and has one of the two following formats:

s r
cx hz
n or o modifications systematically, without taking into account paths or
csx ery1 ; esy1 ery2 ; …; esyn1 eryn ; esyn hrz
where x is the label of a cooler C, z is the label of a heater H, and
{y1,…, yn} is a subset of n distinct labels of processeprocess ex-
changers E.
In the first format, heat that was sent to the environment before
retrofit is directly used for the receptor of a heater. In the second
format heat that was sent to the environment before retrofit is used
for the receptor of an internal heat exchanger; its supplier before
retrofit which is now partially or totally released can be used for
another receptor, and so until reaching the receptor of a heater. No
heat can be saved without including a bridge. Bridges can be
identified with an algorithm, a grid diagram, an energy transfer
diagram, or a network table.
The energy transfer diagram presents the flow rate of cascaded
heat through each actual cooler, heater, and internal heat
exchanger in function of temperature. Each curve corresponds to a
heat exchanger unit; all the curves are summated (stacked); the
total flow rate is presented by the “network curve”. All bridges can
be identified with this diagram.
A bridge, which connects a cooler supplier csx (heat outlet) to a
heater receptor hrz through a set of matches, is necessary to reduce
the hot utility consumption. Fig. 1 shows the meaning of a bridge.
Cooler supplier csx ; which is heat sent to the environment before
retrofit, is used to exchanger receptor ery ; the corresponding sup-
plier esy is released and used to heater receptor hrz : As a
Fig. 1. Necessity of a bridge to achieve heat savings e bridge fcsx ery ; esy hrz g is shown, which results in translating down the flow rate of cascaded heat by S and in S heat savings.
consequence, the flow rate of heat cascaded from the hot utility
until the environment is reduced.
Evaluation of the heat savings capacity of a bridge corresponds to
a small linear programming problem (LP). It is important to
remember that bridge analysis identifies heat transfer modifications.
The corresponding network design, including eventual stream
splitting, can then be identified by engineering analysis or mathe-
matical programming. The optimization approach implies the use of
binary variables to take into account design constraints. A formula-
tion based on a transportation model for heat exchanges, such as the
model developed in Refs. [20,4,18], can be used, and solved by mixed
integer linear programming (MILP). The decision variables are
limited to the streams involved in bridge modifications.
It is important to make the distinction between a bridge and a
heater-cooler path. A bridge is a set of heat transfer modifications. A
heater-cooler path is a structural characteristic in a heat exchanger
network; this is not a set of modifications. Methods for HEN retrofit
have been developed using the concept of path, e.g. Path-Analysis
[24] and Network Pinch Approach [1]. The presence of a path in
the HEN before or after retrofit is not necessary for achieving heat
savings. In bridge analysis, it is proposed to identify heat savings
any other structural characteristics of the network. As a conse-
quence, bridge modifications occurring along a path and resulting
in profit, if there are, also are identified.
3. Case study
A model of the average Scandinavian kraft pulp mill has been
developed in the framework of the Swedish national research
programme “Future Resource-Adapted Pulp Mill” [12]. The aim of
the programme is to investigate how the pulp and paper produc-
tion can have less environmental impact. A central part of FRAM is
to assess the energy-saving potential of the model mills. Details
about model mills can be found in FRAM [12]. The mill was created
to represent a typical Scandinavian market Kraft pulp mill in the
year 2000 with regard to equipment, size and efficiency. It produces
1000 ADt/d of bleached market pulp from softwood and has a total
steam consumption of 193 MW. Our case study is the bleached
market pulp mill with high water usage, which has a process water
consumption of 21 ton per air dried ton.
A detailed energy analysis of this mill is presented in Ref. [3],
where results from pinch analysis, change in equipment, and use of
an innovative system to reduce the steam consumption at the black
liquor multiple-effect evaporation, called “process-integrated
evaporation”, are compared. Based on the hot and cold streams
extracted from the FRAM typical Scandinavian kraft pulp mill with
high water usage (ibid.), a heat exchanger network has been
 J. Bonhivers et al. / Applied Thermal Engineering 70 (2014) 369e379
Fig. 2. Grid diagram of FRAM kraft pulp mill.
designed. The corresponding grid diagram is presented on Fig. 2; it
includes 40 units, i.e. 13 coolers, 19 heaters, and 8 internal heat
exchangers. The individual contribution to minimal temperature
difference from each stream is also indicated. The hot utility con-
sumption before retrofit is equal to 147 MW, after exclusion of
46 MW of high pressure steam at 450  C used for power generation
and soot blowing.
The grand composite curves, advanced composite curves and
energy transfer diagram of the actual network mill are first pre-
sented; then results from the three methods for HEN retrofit will be
discussed.
4. Comparison between the diagrams of these approaches for
HEN retrofit
This part presents the GCC, the ACC, and the ETD corresponding
to the case study, as well as the links between them.
373
4.1. Links between the grand composite curves (GCC) and the
energy transfer diagram (ETD)
The GCC translated until reaching the actual consumption re-
sults in the ETD network curve; reciprocally the network curve
translated until touching the temperature axis results in the GCC.
The GCC (with individual minimum temperature differences) cor-
responding to the mill HEN is presented on Fig. 3. The GCC shows
the flow rate of heat cascaded through the network at minimal
consumption, i.e. 113 MW. Two pinch points are present, located at
60  C and 105  C. The actual heat demand is greater than the
minimum heat consumption by 34 MW.
Energy is conserved, and if not converted into another form of
energy, is progressively degraded until environment temperature
through process operations and heat exchangers. The energy
transfer diagram of a complete process shows the flow rate of heat
cascaded through each actual heat exchanger and process opera-
tion from the hot utility until the environment. In this paper, only
374 J. Bonhivers et al. / Applied Thermal Engineering 70 (2014) 369e379
Fig. 3. grand composite curve and network curve of FRAM kraft pulp mill (with in-
dividual DTmin).
the part corresponding to the HEN is presented in order to explain
its links with the pinch analysis curves. The network curve repre-
sents the total flow rate of heat cascaded through the actual HEN; it
results from the sum of all energy transfer curves of the heat ex-
changers (Fig. 3). The ETD is normally represented with tempera-
ture on the X-axis and heat flow rate on the Y-axis. This corresponds
to the common graphical representation of a parameter Y that is a
function of another parameter X e in this case, the heat flow rate is
a function of temperature. However, the network curve of ETD on
Fig. 3 has been plotted according to the pinch analysis convention
of having heat flow rate on the X-axis and temperature on the Y-axis
to facilitate the comparison.
4.2. Advanced composite curves of the mill HEN
The usual approach to evaluate the advanced composite curves
when no individual temperatures are specified is the following. First,
a Heat Recovery Approach Temperature (HRAT) is evaluated from
the actual hot utility consumption and actual stream temperatures;
then the system is divided into two parts, below and above the
resulting pinch. A common Exchanger Minimum Approach Tem-
perature (EMAT), which is smaller than the HRAT, is specified; the
hot streams above the pinch are down-shifted and the cold streams
below the pinch are up-shifted by the value of EMAT. The EHLC, ECLC,
THLC, and TCLC are finally evaluated from the flow rate of cascaded
heat. In order to consider individual temperature differences, a
slightly different approach is used here. Firstly, streams are shifted
according to their individual contribution to temperature difference,
which are indicated on Fig. 2. Secondly, a HRAT is evaluated from the
actual hot utility consumption. Then the network is divided into two
parts, below and above the resulting pinch. Finally the EHLC, ECLC,
THLC, and TCLC are evaluated from the flow rate of cascaded heat
without any EMAT. It has been decided to use from the beginning the
curves that consider individual temperature differences, i.e. before
evaluating the HRAT and not after. This is due to the fact that the
curves that consider individual temperature differences can
completely differ from the original ones. If the original curves were
used to evaluate the HRAT and the pinch and if only then individual
temperature differences were considered, difficulties may appear;
e.g., it might be impossible to use all the hot streams above the pinch
or all the cold streams below. Consequently, adjusted curves should
be used from the beginning.
Fig. 4 presents the resulting advanced composite curves with
individual minimum temperature differences. The HRAT is equal to
28  C, and the resulting pinch is located at 32  Ce60  C. The mini-
mum sum of THLC and TCLC indicates the total potential for heat
savings, which is equal to 34 MW. The part where the AHLC overlaps
ACLC indicates potential for heat savings opportunity simply by
replacing a heater and a cooler by a processeprocess heat exchanger
at “much lower” investment cost”; this corresponds to prime
bridges that include only one match, which has the format fcsx hrz g:
More precisely, the minimum of (ACLC þ AHLC) is equal to the heat
savings capacity resulting from replacing a heater and a cooler by a
processeprocess heat exchanger. The part where the AHLC and the
ACLC are close indicates potential for heat savings opportunity at
“lower investment cost” than if they are far apart; this corresponds
to bridges that include a small number of matches. A positive and
greater temperature difference between AHLC and ACLC indicates
that heat savings projects could require increased investment cost,
for these would imply more HEN modifications; these projects
correspond to bridges including a greater number of matches.
The area between the AHLC and HUC (i.e. AHLCeHUC) shows
the flow rate of heat cascaded through actual heaters; the area
between the ACLC and CUC (i.e. ACLCeCUC) shows the flow rate of
heat cascaded through actual coolers. (THLCeHUC) and
(TCLCeCUC) indicate the flow rate of heat cascaded through
heaters and coolers if these were placed at lowest and highest
temperature, respectively. (EHLCeHUC) and (ECLCeCUC) indicate
the flow rate of heat cascaded through heaters and coolers if these
were placed at highest and lowest temperature, respectively.
4.2.1. Links between the advanced composite curves and the energy
transfer diagram
To complete the description, three supplementary advanced
composite curves corresponding to internal heat exchanges are
Fig. 4. Advanced composite curves for the HEN of FRAM kraft pulp mill.
 J. Bonhivers et al. / Applied Thermal Engineering 70 (2014) 369e379
proposed in this article: AILC, EILC, and TILC. The Actual Internal
Load Curve AILC represents the flow rate of heat cascaded through
the actual internal exchangers. The Extreme Internal Load Curve
EILC represents the flow rate of heat that would be cascaded
through internal exchanges if heaters and coolers were placed as
far away from the pinch temperature as possible (which corre-
sponds to the EHLC and EHLC). The Theoretical Internal Load Curve
TILC represents the flow rate of heat that would be cascaded
through internal exchanges if heaters and coolers were placed as
close to the pinch temperature as possible (which corresponds to
the THLC and TCLC). Notice that the TILC is null at the pinch.
The sum of AHLC, ACLC, and AILC (Fig. 5), the sum of EHLC, ECLC,
and EILC, and the sum of THLC, TCLC, and TILC (Fig. 6) result all in
the network curve. The links between the grand composite curve,
advanced composite curves, and energy transfer diagram are
summarized in Table 1.
4.3. Energy transfer diagram
In the next part, the energy transfer diagram is shown according
to the common convention regarding function of variable; flow rate
of cascaded heat which is a function of temperature is represented
on the vertical ordinate axis; temperature variable is represented
on the horizontal abscissa axis. Only the ETD corresponding to the
HEN is presented in order to show its links with GCC and ACC's.
However, the ETD can also include the individual process opera-
tions. If heat is not converted into another form of energy, then heat
is conserved and progressively degraded from the hot utility until
the environment; consequently the sum of the process operation
and HEN parts results in a rectangle shape. Note that the minimum
of flow rate of cascaded heat through the HEN corresponds to the
maximum flow rate of process operation part. The minimum of
flow rate of cascaded heat through the HEN corresponds to the heat
savings capacity by HEN retrofit.
The ETD is composed of heat transfer curves. Each curve repre-
sents the flow rate of heat cascaded through an actual cooler, heater,
or internal exchanger. Several arrangements of heat transfer curves
in the diagram are possible. On Fig. 7a, curves of heaters are placed
first at the bottom according to increasing cold-end temperature;
Fig. 5. Link between actual load curves and network curve.
375
then curves of other units (internal heat exchangers and coolers) are
placed according to increasing hot-end temperature of supplier.
With this arrangement, upward arrows correspond to new criss-
cross matches, and downward arrows correspond to thermody-
namically favorable new matches. Fig. 7b presents a simplified ETD,
with heat transfer curves merged into main groups. As noted pre-
viously, the sum of cooler heat transfer curves results in ACLC; the
sum of heater heat transfer curves results in AHLC; the sum of heat
transfer curves of internal exchanger results in AILC. Fig. 7c presents
two examples of bridge for the mill HEN. Arrow A shows the bridge
fcs7 hr1 g; arrows B show the bridge fcs13 er7 ; es7 er8 ; es8 hr5 g:
Fig. 8 shows the effect of bridge modification on the energy
transfer diagram: the network curve is down-translated by the value
of heat savings. The total flow rate of cascaded heat between the hot
utility and the environment is reduced by the value of heat savings.
A tabular presentation is proposed for bridge identification and
evaluation. In network table resulting from first stream decompo-
sition rows correspond to suppliers, columns correspond to re-
ceptors, and cells correspond to matches. Fig. 9 presents the
network table corresponding to the case study. Match heat transfer
capacities which are indicated in MW provide information about
bridge heat savings capacity. For example, the heat savings capacity
of prime bridge fcs1 hr1 g is equal to 1.6 MW and the heat savings
capacity of prime bridge fcs1 er1 ; es1 hr2 g is equal to 1.1 MW, i.e. min
(9.0, 1.1). Column er8 corresponds to chips heating from 7 to 77  C;
its supplier before retrofit es8 is steam from second flash at the
digester. Because chips can be heated in practice only with steam
from the first or second flash, i.e. es7 or es8 ; heat transfer capacities of
all the other suppliers are null. Column hr2 corresponds to building
heating. It is supposed that building can be heated only with hot
utility steam in practice; therefore heat transfer capacities of pro-
cess suppliers are null in this column. Receptors of heaters from
H11 to H19 can be heated only by hot utility steam; as a conse-
quence, these heaters cannot be included in any bridge; these are
not shown in network table of Fig. 9. Cells above the diagonal of the
internal exchanger area correspond to new thermodynamically
unfavorable matches (these are also called criss-cross matches).
Notice that suppliers and receptors in this network table can be
decomposed; this results in new rows and columns, and leads to
network table from second decomposition which can be used for
exact evaluation of bridge heat savings capacity and for network
design (cf. paper about concepts of bridge analysis).
5. Results from pinch analysis and bridge analysis
5.1. Preliminary comment
As shown in Fig. 7c, the internal heat exchangers E7 and E8 are
important with regard to heat saving opportunities in the mill. In-
ternal heat exchanger E7 refers to the production of hot water with
steam from the second flash at the digester. Internal heat exchanger
E8 refers to chips heating with steam resulting from the first flash at
the continuous digester; i.e., chips are heated from 7  C to 77  C by
direct contact with steam at 130  C. This heat recovery with large
temperature difference is well-known in the kraft industry for
having a big impact on the energy consumption. In principle, for
heating purposes, chips could be preheated by a heat supplier at
much lower temperature. However, direct contact with steam al-
lows impregnation of chips with water, which is beneficial for the
cooking, and most modifications of E8 are difficult in practice. The
ETD of Fig. 7 clearly shows that modifying the chips heating, i.e. heat
exchanger E8, would be beneficial. Hot water can be produced with
heat sources other than steam from second flash in heat exchanger
E7, e.g. the cooling of digester bottom cs11 ; the stripper condenser cs12
or steam from smelt dissolver cs13 ; which are sent to the
376 J. Bonhivers et al. / Applied Thermal Engineering 70 (2014) 369e379
Fig. 6. a: Link between theoretical load curves and network curve. b: Link between summated (stacked) theoretical load curves and network curve.
environment today. Heating water with these heat suppliers would
release all the steam from second flash, i.e. supplier es7 ; which can
then be used to preheat the chips er8 and release steam from the first
flash es8 which can then be used for heater receptors at the digester
and consequently reduce the hot utility consumption.
5.2. Pinch analysis
Results from pinch analysis are reported in the study by Axels-
son et al. [3]. Two pinch points are present, located at 60  C and
105  C. The total amount of cross-pinch transfers is equal to 34 MW.
Table 1
Links between the grand composite curve, advanced composite curves and network curve.
Curves of pinch analysis
GCC
➢ Flow rate of heat cascaded at minimum heat consumption
➢ After translation until actual consumption, GCC results in network curve.
AHLCeHUC
➢ Flow rate of heat cascaded through the actual heaters
ACLCeCUC
➢ Flow rate of heat cascaded through the actual coolers
AILC (Actual Internal Load Curve)
➢ Flow rate of heat cascaded through the actual internal exchangers
(THLCeHUC) and (TCLCeCUC)
➢ Flow rate of heat cascaded through heaters and coolers if they were placed
at lowest and highest temperature, respectively.
(EHLCeHUC) and (ECLCeCUC)
➢ Flow rate of heat cascaded through heaters and coolers if they were placed
at highest and lowest temperature, respectively.
Heat savings capacity ¼ minimum of total flow rate of cascaded heat
Total flow rate ¼ ACLC þ AHLC þ AILC
Total flow rate ¼ ECLC þ EHLC þ EILC
Total flow rate ¼ TCLC þ THLC þ TILC
ACLC overlapping AHLC
➢ Min (ACLC þ AHLC) ¼ heat savings capacity by replacement of a cooler and
a heater with one internal exchanger, at “very low” investment cost
AHLC close to ACLC
➢ Possibility of heat savings at “low“ investment cost
AHLC far above ACLC
➢ Possibility of heat savings at “higher“ investment cost
Main sets of modifications proposed to solve pinch violations,
which result in 17 MW savings, are the following.
Set 1. Hot utility steam that is currently used for hot water
production can be replaced by other heat sources available at
lower temperature, thanks to a new hot and warm water sys-
tem. Also, hot utility steam that is currently used for the wood
yard can be replaced by other sources, such as a warm effluent.
Set 2. An important part of the pinch violations that can be
solved are related to the use of black liquor flash steam at the
digester system. In the original configuration, the second flash
Energy transfer diagram
Network curve
➢ Flow rate of heat cascaded through actual HEN
➢ After translation until temperature axis, network curve results in GCC.
Area corresponding to the actual heaters (sum of heat transfer curves
of actual heaters)
Area corresponding to the actual coolers (sum of heat transfer curves
of coolers)
Area corresponding to the actual internal exchangers (sum of heat transfer
curves of actual internal exchangers)
Area corresponding to heaters and coolers if they were placed at lowest
and highest temperature, respectively (not shown in the usual ETD)
Area corresponding to heaters and coolers if they were placed at highest
and lowest temperature, respectively (not shown in the usual ETD)
Heat savings capacity ¼ minimum of total flow rate of cascaded heat
Total flow rate ¼ network curve
Possibility of bridges that include one match
Possibility of bridges that include a small number of matches
Possibility of bridges that include a higher number of matches
 J. Bonhivers et al. / Applied Thermal Engineering 70 (2014) 369e379 377

Fig. 7. a e Energy transfer diagram of mill HEN. b e Simplified energy transfer diagram with identification of curves. c e Energy transfer diagram with arrows showing bridges.

steam is used for hot water production (i.e. the heat exchanger
denoted by E7) and the first flash steam is used for chips heating
(i.e. the heat exchanger denoted by E8). However, heat for hot
water production can be supplied by surplus heat at lower
temperature. This first change releases the second flash steam
which can then be used to preheat the chips; this second
modification releases the first flash steam which can be used to
replace hot utility steam for heating of chips and white liquor at
the digester system.
5.3. Bridge analysis
On Fig. 7c, arrow A shows a bridge of first format which corre-
sponds to set 1; arrows B show a bridge of second format which
corresponds to the second set of modifications. Fig. 9 shows the
bridge modifications (circled matches) corresponding to the solu-
tion proposed by Axelsson et al. [3]. The two sets of modifications
proposed to solve pinch violations described here above are visible.
Set 1: use of unused heat sources to replace hot utility steam for
378 J. Bonhivers et al. / Applied Thermal Engineering 70 (2014) 369e379

Fig. 8. Network curve before and after HEN retrofit.

heating of water and wood yard, fcs1 hr1 ; cs8 hr3 ; cs9 hr3 ; cs13 hr4 ; cs13 hr6 g:
Set 2: use of unused heat sources to replace second flash steam for
heating of water, use of second flash steam to replace first flash
steam for chips preheating, and use of first flash steam to replace
hot utility steam for heating chips and cooking liquor,
fcs8 er7 ; cs12 er7 ; cs13 er7 ; es7 er8 ; es8 hr5 ; es8 hr7 ; es8 hr8 g: As shown in Fig. 9,
other heat savings modifications are theoretically possible; how-
ever, the practical feasibility of most of them is lower.
6. Discussion
It is important to note that the present inefficient way of heating
wood chips being fed to the digester in the Kraft pulping is a well-
known problem; neither pinch analysis nor bridge analysis are
required to identify it and formulate remedial action. However,
these methods are useful for system analysis. To reduce the energy
consumption implies decreasing the flow rate of heat cascaded
from the hot utility to the environment through the existing HEN.
As a consequence, a bridge is necessary to achieve heat savings.
In pinch analysis, removal of a cross-pinch transfer results in
releasing a receptor below the pinch and a supplier above the
pinch. Modification of the network is then necessary to rebalance
the released receptor and released supplier; this ultimately results
in a bridge linking cooler suppliers to heater receptors through a set
of matches. In reality, removing a transfer that does not cross the
pinch and then rearranging the network to rebalance the released
receptor and released supplier such that a bridge is achieved can
also result in heat savings. The importance of the pinch is related to
the Pinch Design Method which is more useful for the synthesis of
new networks than for retrofit.
Data for this study have been extracted in the framework of pinch
analysis which especially works with indirect-contact heat ex-
changes; e.g., some non-isothermal mixings which are important in
water-borne process such as Kraft pulping have been excluded.
Given this set of data, which consider minimal individual tempera-
ture differences for extracted streams, each heat savings project
identified with pinch analysis can be identified with bridge analysis,
and reciprocally. The difference is the level of detail: bridge analysis

Fig. 9. Network table with match heat transfer capacity (MW). Circles show bridge modifications corresponding to pinch analysis based solution proposed by Axelsson et al. [3]. Set
1 , Set 2 .
 J. Bonhivers et al. / Applied Thermal Engineering 70 (2014) 369e379
Table 2
Comparison of functionalities between methods for HEN retrofit.
Method Complexity Information about Information about
savings capacity new design
GCC e classical pinch 1 1 1 correspond to bridges including a small number of matches.
analysis
ACC's e advanced 2 2 2 However, the method uses new symbols and terminology.
composite curves
ETD e bridge analysis 3 3 3
1: lower; 2: medium; 3: higher.
provides more information about heat savings projects than pinch
analysis does. Bridge analysis works well with water-borne process
since non-isothermal mixings lead to heat degradation which cor-
responds to energy cascade, and since constraints specific to each
match are considered from the beginning (composite curve lump
data for cold and hot streams). As a consequence, data extraction is
more complete and explicit in the framework of bridge analysis.
The GCC and the network curve represent the minimal and
actual flow-rate of energy cascaded through the HEN, respectively.
The advanced composite curves provide more information about
heat-savings potential achievable in practice than the GCC does.
Larger temperature differences between AHLC and ACLC indicates
that heat savings projects could require increased investment cost
since these would imply more HEN modifications; these projects
correspond to bridges including a greater number of matches. The
user-friendliness of ACC's is increased with the diagram showing
only the actual curves (ACLC, AHLC, AILC, the sum of these, CUC and
HUC), such as on Fig. 5. This adaptation maintains their level of
information and usefulness for HEN retrofit, while increasing the
clarity. These 6 curves show the main areas of the HEN part of the
ETD, i.e. the flow rate of heat cascaded through coolers, internal
exchangers, and heaters. Table 2 presents the summary about
functionalities for the three energy use analysis methods.
7. Conclusion
The GCC, the ACC's, and the ETD are connected. The GCC shifted
until reaching the actual energy consumption results in the ETD
network curve, which is equal to the total flow rate of heat cascaded
through the HEN. The minimum of the network curve indicates the
heat-savings potential by HEN retrofit without consideration of
connection constraints. In this paper, we have suggested three new
curves as a link between the ACC's and the ETD; these represent the
flow rate cascaded through internal heat exchanges. The sum of
actual load curves, the sum of theoretical load curves, and the sum
of extreme load curves of ACC's results all in the network curve. The
advanced composite curves approach and bridge analysis are
linked. ACLC, AHLC, and AILC correspond to the sum of ETD heat
transfer curves of heaters, coolers, and internal exchangers,
respectively. In this paper, the ACC's have considered individual
temperature differences.
Information provided by the three approaches is consistent;
however, the level of detail increases from the GCC to the ACC's
until the ETD. Traditional pinch analysis recommends suppressing
exchanges crossing the pinch temperature. However, removing
transfers crossing a specific temperature, e.g. the pinch point, is not
necessary since other temperatures can be chosen, and is not suf-
ficient since the network must then be arranged in order to reba-
lance the released receptors and suppliers. Pinch analysis succeeds
to reduce the heat consumption when a bridge is built at the
network rearrangement step.
Since a bridge is necessary to reduce the flow rate of heat
cascaded through the existing exchanges from the hot utility until
379
the environment, the advanced composite curves provide more in-
formation about network rearrangement than the GCC does. Smaller
temperature differences between AHLC and ACLC indicate that heat
savings projects could require lower investment cost; these projects
The principle of bridge analysis is fundamental and simple.
Improvement in pedagogy and explanation should facilitate its use.
References
[1] N.D.K. Asante, X.X. Zhu, An automated and interactive approach for heat
exchanger network retrofit, Chem. Eng. Res. Des. 75 (1997) 349e360.
[2] G. Athier, P. Floquet, L. Pibouleau, S. Domenech, A mixed method for retro-
fitting heat-exchanger networks, Comput. Chem. Eng. 22 (1998) S505eS511.
[3] E. Axelsson, M.R. Olsson, T. Berntsson, Heat integration opportunities in
average Scandinavian Kraft pulp mills: pinch analyses of model mills, Nordic
Pulp Pap. Res. J. 21 (4) (2006) 466e475.
[4] A. Barbaro, M.J. Bagajewicz, New rigorous one-step MILP formulation for heat
exchanger network synthesis, Comput. Chem. Eng. 29 (2005) 1945e1976.
[5] K.M. Bjork, R. Nordman, Solving large-scale retrofit heat exchanger network
synthesis problems with mathematical optimization methods, Chem. Eng.
Process 44 (2005) 869e876.
[6] R. Bochenek, J.M. Jezowski, Genetic algorithms approach for retrofitting heat
exchanger network with standard heat exchangers, in: 16th European Sym-
posium on Computer Aided Process Engineering and 9th International Sym-
posium on Process Systems Engineering, vol. 21, 2006, pp. 871e876.
[7] J.-C. Bonhivers, S. Bala, P. Stuart, Bridge analysis to reduce the industrial en-
ergy requirements by heat-exchanger network retrofit: part 1-concepts, Appl.
Therm. Eng., http://dx.doi.org/10.1016/j.applthermaleng.2014.04.078
[8] J.-C. Bonhivers, M. Korbel, M. Sorin, L. Savulescu, P. Stuart, Energy transfer
diagram for improving integration of industrial systems, Appl. Therm. Eng. 63
(1) (5 February 2014) 468e479.
[9] V. Briones, A.C. Kokossis, Hypertargets: a conceptual programming approach
for the optimisation of industrial heat exchanger networks e II. Retrofit
design, Chem. Eng. Sci. 54 (1999) 541e561.
[10] A. Carlsson, P.A. Franck, T. Berntsson, Design better heat-exchanger network
retrofits, Chem. Eng. Prog. 89 (1993) 87e96.
[11] A.R. Ciric, C.A. Floudas, A mixed integer nonlinear-programming model for
retrofitting heat-exchanger networks, Ind. Eng. Chem. Res. 29 (1990)
239e251.
[12] FRAM, FRAM Final Report Model Mills, FRAM Report No. 70, STFI, Stockholm,
2005.
[13] I.C. Kemp, Pinch Analysis and Process Integration: a User's Guide to Process
Integration for the Efficient Use of Energy, ButterwortheHeinemann, Oxford,
2007.
[14] A.K. Kralj, Optimization of an industrial retrofitted heat exchanger network,
using a stage-wise model, Energy 35 (2010) 4748e4753.
[15] R. Lakshmanan, R. Ban ~ ares-Alca
ntara, A novel visualization tool for heat
exchanger network retrofit, Ind. Eng. Chem. Res. 35 (1996) 4507e4522.
[16] B. Linnhoff, E. Hindmarsh, The pinch design method for heat-exchanger
networks, Chem. Eng. Sci. 38 (1983) 745e763.
[17] R. Smith, Chemical Process Design, McGraw-Hill, N.Y., USA, 1995.
[18] D.Q. Nguyen, A. Barbaro, N. Vipanurat, M.J. Bagajewicz, All-at-once and step-
wise detailed retrofit of heat exchanger networks using an MILP model, Ind.
Eng. Chem. Res. 49 (2010) 6080e6103.
[19] R. Nordman, T. Berntsson, Use of advanced composite curves for assessing
cost-effective HEN retrofit I: theory and concepts, Appl. Therm. Eng. 29 (2009)
275e281.
[20] F. Pettersson, Synthesis of large-scale heat exchanger networks using a
sequential match reduction approach, Comput. Chem. Eng. 29 (5) (15 April
2005) 993e1007.
[21] E. Rezaei, S. Shafiei, Heat exchanger networks retrofit by coupling genetic
algorithm with NLP and ILP methods, Comput. Chem. Eng. 33 (2009)
1451e1459.
[22] R. Smith, M. Jobson, L. Chen, Recent development in the retrofit of heat
exchanger networks, Appl. Therm. Eng. 30 (2010) 2281e2289.
[23] T.N. Tjoe, B. Linnhoff, Using pinch technology for process retrofit, Chem. Eng e
New York 93 (1986) 47e60.
[24] J.L.B. van Reisen, J. Grievink, G.T. Polley, P.J.T. Verheijen, The placement of 2-
Stream and multi-stream heat-exchangers in an existing network through
path-analysis, Comput. Chem. Eng. 19 (1995) S143eS148.
[25] J.L.B. van Reisen, G.T. Polley, P.J.T. Verheijen, Structural targeting for heat
integration retrofit, Appl. Therm. Eng. 18 (1998) 283e294.
[26] Y.F. Wang, R. Smith, J.K. Kim, Heat exchanger network retrofit optimization
involving heat transfer enhancement, Appl. Therm. Eng. 43 (2012) 7e13.
[27] T.F. Yee, I.E. Grossmann, A screening and optimization approach for the
retrofit of heat-exchanger networks, Ind. Eng. Chem. Res. 30 (1991) 146e162.
[28] H.B. Zhang, G.P. Rangaiah, One-step approach for heat exchanger network
retrofitting using integrated differential evolution, Comput. Chem. Eng. 50
(2013) 92e104.