A ~~
Better integration of process
equipment and operations
pays off twofold, through
simplicity in structures, and
the simultaneous best use of
both energy and capital.
Bodo Linnhoff and John A. Turner,
Inperial Chemical Industries Ltd,
Heat-recovery networks:
new insights yield big savings
(1 Process design usually involves the optimization of
individual unit operations, with litle effort spent on the
‘optimal integration of overall processes as systems
In this article, we will demonstrate that overall inte-
aration can be neglected only at one’s peril. Here, we
will concentrate on heatrecovery networks as one as-
pect of total process integration. We will present a new
approach with solved examples, and describe industrial
studies carried out at ICL
In doing 50, we will attack the conventional belief of
an inevitable tradeoff between heat recovery and capi-
tal cost. Reasons will be given as to why tradeofls need
not always exist in the context of networks. The argu-
ment is borne out both by a solved example and by the
results achieved at ICL In most cases, design changes
have been found that are less expensive in both energy
and capital.
Specifying the problem
‘The heat-recovery network (or heat integration)
problem is defined by a series of hot and cold process
streams requiring either the cooling of hot streams or
the heating of cold streams. A typical specification for
such a problem is shown in Table I
The required temperature changes may be achieved
by using interchangers between process streams and/or
heaters and coolers supplied with utilities. The design
task is that of identifying the optimum network of
interchangers, heaters and coolers with respect to an-
nual operating and capital costs. The maximum heat
load handled by any particular heat-transfer unit may
be constrained by the minimum allowable temperature
approach between the hot and cold streams, AT i,
‘Two possible networks for the data given in Table I
are shown in Fig. 1. The first (Fig. 1a) is the “no-fuss,
no-complication design” and is usually the one re-
garded as having the lowest capital cost. The second
(Fig. 1b) is a network having the lowest utility usage.
Most design engineers assume that the optimal solution
rust lie between these two extremes. Later in this arti
cle, we will challenge this assumption.
‘The methods used at ICI for tackling the problem are
somewhat unconventional. They involve two phases:
data analysis and network design.
Data analysis yields targets that correspond to the
performance characteristics of the economic optizaum
network. These targets are invaluable when dealing
with complex integration problems, and give conti-
dence to an engincer’s current design or stimulate the
engineer to search for a better solution, Furthermore,
the targeting procedures give a new insight into a very
simple but fundamentally important decomposition of
the original problem into several parts. With this in-
sight, even complex industrial problems can be easily
solved by hand,
Data analysis
The cost of a heatsrecovery network, like the cost of
any process plant, is expressed in terms of annualized
capital and operating costs. Capital costs depend pri-
56 [TTHot steame
Inrange wih hor aiey
Oro ngs eteen
Nuresel values refer ohost
foat-0, oF etih
. Lowest utility usage
inarily on the number and sizes of the heat-transfer
cunits (interchangers, heaters, coolers), while operating
costs are dominated by total tility usage (both hot and
cold), Targets are therefore set for the economic optimi
zation of heat-recovery networks, based on the minimi-
zation of utility usage, number of transfer units, and
total heat-transfer area. Simultaneously, the targeting
procedures define the problem decomposition into a
number of design principles.
In Fig, 2, we will represent the heat-recovery problem
via a box through which the hot and cold streams
flow—giving up and receiving heat, respectively. Heat
taken from the utility source is labelled Qy, while heat
rejected to the utility sinks is labeled Q,. For heat bal-
ance, the difference between hot and cold process:
stream enthalpies (ie., A/7) must equal the difference
between utility heat flows ool
| Qo = Qy = AH
AH is a constant, depending on the process stream
data only. Hence, any increase in Q, must lead to a
corresponding increase in Q.. And, any saving in source
utilities must lead to a saving in sink utilities giving a
double incentive for minimizing utility usage. The cra-
cial question is: What isthe minimum possible value for
Qy in any given problem?
Minimum utility requirements
An algorithm to predict the minimum utility require-
ments has been previously published [3] and will now
bbe briefly described with reference to Example 2
(Table TD. A step-by-step description of the algorithm is
tsven in an accompanying box on p. 59.
A Cs ers ture
range dhat by the hotest 7
ss-stream temperatures. This temperature range is di-
is ture intervals (see box
for definition), as shown in Fig. 3. Hot uty heat, Q,,
is supplied (o the first interval, ‘The heat to be rejected
GRC RGR ROVER ET
Gee eee‘mpi aloe re
im ivestuh
60+ x cy ox
1 Across the pinch b. Above the pinch_c. Below the pinch
eee ne eee
ee
to the sink utility, 0, leaves the last interval. Definition
ofthe temperature intervals ensures that a minimum
ving force (AT, is maintained at all points
‘Justas Q, and Q, ie elated she orale
steam enthalpy balance, the heat flows passed on be-
‘tween intervals, Q,, are related by the interval heat bal-
ances, AH:
Qi - Q1 = aH;
Given the temperature intervals and the interval heat
balances, itis possible to calculate all intermediate heat
flows (and ultimately Q.) for any given value of Q. To
demonstrate this, let us again consider Example 2, with
the interval heat balances now specified, as given in
Fig. 4.
Column 1 in Fig. 4 shows the heat flows resulting
from a value of Q, = 0. Between Intervals 3 and 4, the
heat flow is negative (—50 units), ie. an infeasible con-
dition, (Negativity implies that heat flows against the
natural temperature gradient.) To remedy this situa
tion, an extra 50 units of heat (ie. the largest infeasible
flow) are supplied from the external source. The result-
ing heat flows are all non-negative, and are shown in
Column 2 of Fig. 4.
Column 2 yields three important items of informa-
tion, First, the minimum heat supply from external
hheat sources, necessary for feasible heat flows, is 50
units, Second, the corresponding minimum heat load
on external sinks is 60 units. Third, there is a point in
the temperature range having zero heat flow. From now
‘on, we will call this point “the pinch.”
Problem decomposition
‘Three important constraints regarding the pinch:
= Do stot transfer heat across the pinch (Fig, 5a).