Professional Documents
Culture Documents
Flexible Heat Exchanger Networks
Flexible Heat Exchanger Networks
Flexible Heat
temperatures are dis- temperature difference
turbed by D and all defined in Equation (4)
target temperatures are Ti Internal temperature i
fixed, controlled C) ∆Ti Change in internal
CP Heat capacity flowrates
Exchanger
CPC Heat capacity flowrates temperature Ti
for the cold stream TSi Supply temperature of
CPH Heat capacity flowrates stream i
for the hot stream TTi Target temperature of
D Disturbed parameters stream i
Networks
(in the example, start U Overall heat transfer
temperatures are dis- coefficient
turbed by D and all <UA>j Contingency for
target temperatures are exchanger j
When designing a 1
D TS1 = 30°C
1 3 C
C
TT1 = 80°C
network of integrated C
TT2 = 40°C
conditions Figure 1. Heat exchanger networks should be designed for flexibility in operating
parameters such as supply temperatures, heat capacity flowrates, and heat transfer
coefficients, while maintaining controlled parameters such as target temperatures
T
Tc Td Disturbed parameter D Cooler C
he ever important aim of energy 2
efficiency in the chemical process Figure 2. A single heat exchanger Controlled parameter C Heater H
industries (CPI) is brought closer with hot Stream 1 and cold Stream 2
to its target by the practice of en-
ergy recovery in heat exchanger net- approach is developed, automated and sive response of internal and target
works (HENs). When heat exchang- illustrated on an example problem. network temperatures to changes
ers are designed to work together to The approach is applied through in supply temperatures, heat capac-
exchange heat between hot and cold two steps. The first one is developing ity flowrates and effective values of
streams, the required utilities are ef- a strategy for the choice of the HEN’s heat transfer coefficients for the heat
fectively reduced. The HEN designer base case, upon which sensitivity ta- exchangers. Moreover, combined ac-
usually applies the assumptions of bles are generated. The second step is tions can be considered using sensi-
fixed operating parameters at nominal using sensitivity tables to reach target tivity tables where tradeoff between
conditions for given specifications of a temperatures, and hence, target utili- energy, capital cost and flexibility can
process. But practically, a HEN should ties of the alternative cases. In the be established. The new mode of opera-
remain operable under variations in end, we can have a HEN that realizes tion can be considered as a deviation
operating conditions without losing maximum energy savings under dif- from the base case, so it represents
stream temperature targets and, at ferent operating conditions — in other the disturbances in supply and target
the same time, maintain economically words, a HEN with flexible and opti- temperatures, and even heat capacity
optimal energy integration. mum properties. flowrates. Disturbances may occur for
This article illustrates how to use short or long periods where, certain
sensitivity tables to design an opti- Design of optimal, flexible HEN parameters have to be fixed. These pa-
mum, flexible HEN for a multi-period The sensitivity tables approach. rameters are called controlled parame-
process that is feasible for N periods The use of sensitivity tables replaces ters (C) and disturbed parameters (D).
of operations. The sensitivity tables rigorous simulation to study the pas- The period for which the disturbances
32 Chemical Engineering www.che.com April 2011
History of HEN Synthesis and Sensitivity Tables
H
EN synthesis is one of the most extensively studied problems in the N periods of operation achieving minimum cost.
CPI process design, given the importance of determining the A systematic methodology in the design of HENs under multiple
energy costs for a process and improving energy recovery. periods of operation is presented by Verheyen and Zhang [12].
The first systematic method to consider targets for energy recov- The model presented a superstructure-based on mixed integer
ery was the thermodynamic approach of the pinch concept, intro- nonlinear programming (MINLP) model, which minimizes the
duced during the 1970s. Mathematical programming, stochastic total annualized cost containing heat exchanger area cost and
optimization approaches and hybrid methods developed between utility costs. The model is based on the superstructure by Yee and
the two have also been approached for optimal HEN design. Fur- Grossmann [13, 14], which was formulated for multiple periods
man and Sahinidis [1] reported that over 400 papers have been by Aaltola [15].
published on the subject over the last 40 years. Gundersen and Aguilera and Nasini [17] proposed a mixed-integer linear-
Naess [2] and Ježowski [3, 4] have also contributed thorough programming (MILP) formulation for testing the flexibility of the
reviews on HEN synthesis. HEN for flowrate variation, and later Aguilera and Nasini [16]
Generally, HEN synthesis takes place under the assumptions of introduced a flexibility test for the HEN with non-overlapping inlet
fixed operating parameters at nominal conditions for given spec- temperature variations. Tantimuratha and others [18] proposed a
ifications of a process. So, when operating conditions change, screening and targeting process for the HEN design with flexibility
as they almost certainly do, stream temperature and energy ef- consideration in both grassroots and retrofit cases. The screening
ficiency targets can easily fall out of reach. Reviews of research stage is based on the screening models of Briones and Kokossis
into flexibility and operability can be found in Furman and Sa- [19–21], and it considers both economic and flexibility aspects
hinidis [1]. Marselle and others [5] defined resilience for heat prior to network development. Konukman and others [22] intro-
exchanger networks and stated other properties of resilience. duced simultaneous flexibility targeting and the synthesis of the
They proposed a heuristic design method for structurally resilient minimum utility HEN.
networks with respect to inlet parameter variations. Swaney and Kotjabasakis and Linnhoff [23] introduced “sensitivity tables”,
Grossmann [6] introduced a flexibility index, which defines the which simply correlate the response of the network temperatures
maximum parameter range that can be achieved for a feasible to changes of input streams’ temperatures or heat capacity flow-
operation. Grossmann and Floudas [7–11] introduced an active rates and heat transfer rates (UA) of network exchangers. Sen-
set strategy for the automated solution of the flexibility test and sitivity tables can be utilized to find the necessary corrections in
the flexibility index of Swaney and Grossmann [6] and intro- order to make a nominal design sufficiently flexible, and for mak-
duced a systematic procedure for synthesizing flexible heat ex- ing decisions for the trade-offs between cost effectiveness and
changer networks for multiperiod operation. It is assumed that, in flexibility of the design. The aim is to design an HEN that is both
general, different values are specified for the flowrates and inlet optimized and flexible. This is the approach that has been auto-
and outlet temperatures of the streams for N periods of opera- mated and applied here for the purpose of designing a flexible
tion. The objective is synthesizing a network that is feasible for HEN for multiperiod processes. ❏
Equations (7) and (8) are linear with Three types of sensitivity tables can 2. T(CP) sensitivity tables: These ta-
respect to temperature and nonlin- be constructed, where they simply bles correlate the response of the
ear with respect to both heat capacity correlate the response of the network network temperatures according to
flowrates and UAs. For the same heat temperatures to changes of input changes in heat capacity flowrate
exchanger in Figure 2, knowing two of stream temperatures (TS), heat capac- for a specific stream. Because Equa-
the four temperatures (Ta, Tb, Tc, Td) ity flowrates (CP) and the UA of the tions (7) and (8) are non linear with
and knowing UA, CPH and CPC, the networked exchangers. The informa- respect to CP, the CP sensitivity
other two temperatures can be deter- tion needed are the base case stream tables are constructed for different
mined by solving Equations (7) and data and the network structure. The levels of CP variation in a specific
(8). This way of formulation is effective different types of sensitivity tables are stream. Interpolation between these
for each exchanger in any feasible net- as follows: values is possible.
work. Once the overall system of equa- 1. T(TS) sensitivity tables: These ta- 3. T(UA) sensitivity tables: These ta-
tions is solved for each exchanger in bles correlate the response of the bles need to be constructed for each
the base case, new network tempera- network temperatures to unit varia- individual heat exchanger in the net-
tures can be determined when changes tion of input stream temperatures. work. These tables correlate the re-
occur in supply temperatures, heat ca- One table is enough to describe the sponse of the network temperatures
pacity flowrates and effective UAs. responses to changes in all stream to changes of the effective exchanger
Constructing sensitivity tables. supply temperatures. (in terms of UA). Because Equations
34 Chemical Engineering www.che.com April 2011
Table 3. Example T(TS) Sensitivity Table Table 4. The T(CP) Sensitivity Table
(for each 1°C change in Ts) for the Example Problem
∆Ti, ºC Stream 1 Stream 2 Stream 3 Stream 4 ∆Ti, ºC ∆CP1 ∆CP2 ∆CP3 ∆CP4 ∆CP5
∆T1 .794 0.0 0.0 0.0 –33.34 % –33.34 % –33.34 % –21.01 % –100 %
∆T2 .693 0.0 0.0 .127 ∆T1 –13.959 0.0 0.0 0.0 29.499
∆T3 .283 0.0 .634 .098 ∆T2 –17.912 0.0 0.0 .19951 25.756
∆T4 .077 .081 0.0 ∆T3 –23.384 0.0 11.134 .128 10.656
.841
∆T4 –1.667 –9.932 0.0 22.702 2.908
∆T5 .469 0.0 .341 .106
∆T5 –20.420 0.019 17.840 .1545 17.490
∆T6 .094 0.0 0.0 .695
∆T6 –2.0051 0.199 0.0 2.782 2.920
∆T7 .033 .38 0.0 .043 ∆T7 –.6772 –37.586 0.0 8.537 –.328
∆T8 .930 0.0 0.0 0.0 ∆T8 –2.400 0.0 0.0 0.0 10
Table 5. T (UA) Sensitivity Table for Exchanger No.1 of the Example Problem
% Change in UA from the base case
∆Ti, ºC + 100 % +80 % +60 % +40 % +20 % 0% –20% –40 % –60 % –80 % –100 %
∆T1 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
∆T2 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
∆T3 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
∆T4 –4.808 –4.25 –3.57 –2.7 –1.56 0.0 2.29 5.93 12.61 28.97 131
∆T5 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
∆T6 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
∆T7 4.27 3.72 3.03 2.16 1.019 0.0 –2.83 –6.47 –13.15 –29.51 –131.6
∆T8 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
∆T9 4.07 3.601 3.023 2.288 1.324 0.0 –1.93 –4.993 –10.64 –24.44 –110.6
Table 6. T (UA) Sensitivity Table for Exchanger No.2 of the Example Problem
% Change in UA from the base case
∆Ti, ºC + 100 % +80 % +60 % +40 % +20 % 0% –20% –40 % –60 % –80 % –100 %
∆T1 –2.026 –1.905 –1.709 –1.388 –0.863 0.0 1.433 3.842 7.979 15.366 29.5
∆T2 –1.769 –1.663 –1.492 –1.212 –0.754 0.0 1.252 3.354 6.967 13.416 25.76
∆T3 –0.683 –0.64 –0.569 –0.454 –0.265 0.0 .561 1.427 2.916 5.572 10.66
∆T4 –0.189 –0.177 –0.157 –0.126 –0.07 0.0 .151 .388 .794 1.52 2.91
∆T5 –1.181 –1.11 –0.993 –0.8033 –0.492 0.0 .868 2.295 4.746 17.49
9.12
∆T6 –0.413 –0.401 –0.38 –0.346 –0.290 0.0 –0.048 .2071 .645 1.426 2.92
∆T7
–0.564 –0.563 –0.562 –0.5596 –0.556 0.0 –0.538 –0.520 –0.489 –0.328
–0.434
∆T8 9.135 8.592 7.71 6.262 3.895 0.0 –6.460 –17.32 –35.97 –69.27 –133
(7) and (8) are non linear with respect Conclusion ample 1 in Ref. 8, a process that has
to UA, the temperatures responses The sensitivity tables approach can be three modes of operation (defined as
are evaluated at different values of used to design optimal flexible HEN Periods 1 through 3). The conditions of
UA percent of every effective heat ex- for multiperiods process. The tables the process changes periodically over
changer of the base case network. are constructed by detecting the varia- the year. Each period differs from other
Using sensitivity tables. Once sen- tions of the base case temperatures periods in supply, target temperatures
sitivity tables have been constructed due to disturbances in streams supply and in heat capacity flowrates. The
for the base case, the variation of any temperatures, heat capacity flowrates problem data are given in Table 1.
of the network temperatures can be and heat transfer coefficients (through Construction of sensitivity tables
calculated by simple summation of UAs). Then, the sensitivity tables are for the problem. In order to apply the
the variations resulting from individ- used to adjust the streams’ target tem- sensitivity tables approach, we must
ual disturbances (supply stream tem- peratures to achieve target utility re- define the base case for the HEN de-
peratures and stream heat-capacity quirements as close as possible to PDM sign upon which the sensitivity tables
flowrates). Sensitivity tables for UA results for each period by adjusting will be generated.
can then be utilized for contingency- candidate downstream heat exchang- Utility targeting and selection of
candidate exchangers in order to ers’ UA contingencies. Using PDM de- the HEN’s base case. The PDM has
quantify the necessary change in ex- sign at each period can evolve a flexible been applied on the three periods of the
changers’ UA values that rectify the HEN design much easier than using process to locate the pinch points, mini-
disturbances [23]. We assume that for sensitivity tables. However, sensitiv- mum utilities consumption and mini-
each period, adjusting stream tem- ity tables are more suitable to retrofit mum number of units and to establish
peratures at the inlet of utilities can HEN to render it flexible for limited matches for each period (Figures 3, 4
be used to nearly match target util- variations in process parameters. and 5). Table 2 displays minimum utili-
ity requirements determined via the ties consumption and pinch location for
pinch design method (PDM), which Illustration of the approach each period of the problem. The base
was developed by Linnhoff and his We have applied the sensitivity tables case is chosen as the HEN containing
colleagues [24]. approach on a literature problem (Ex- the largest number of units (Period
Chemical Engineering www.che.com April 2011 35
Feature Report
Stream Stream
CP, kW/°C CP, kW/°C
210.8°C 249°C 233.6°C
7.032 229°C 210.8°C 120.8°C 10.55 249°C 177.5°C
1 2 3 4 1 4
239°C 148.9°C 259°C 170.4°C
8.441 2 1 12.66 2 1
168.6°C 96°C 214.4°C 116°C
9.144 4 3 6.096 4 3
157.2°C
151°C 106°C 106°C 257.3°C 133.8°C 126°C
12.66 H 1 3 4 10.00 1 4
Figure 8. The final network temperatures resulting after Figure 9. Network temperatures as a result of disturbances
performing UA contingencies in Period 2 for Period 3
1). By inspection of the three periods’ case values, the equations have been ∆Ti is the change in Ti due to the com-
HEN, it can be concluded that almost solved for the following cases: bined disturbances.
all equipment required in Periods 2 1. A positive unit change in the dis- By using the sensitivity tables (Ta-
and 3 can all be covered by matches in turbed temperature, one at a time. bles 3, 4) and applying Equation (9) we
Period 1. Only the cooler of Stream 2 in 2. Variable percent change in CP from can deduce the internal temperatures
Period 3 will be added to the HEN of the base case for each stream. of the network as a result of combined
Period 1. This will be the base case, and 3. Variable percent change in UA from disturbances for Period 2 (Figure 7).
the temperature diagram of the prob- the base case for each exchanger. Example calculation for ∆T1: If TS1
lem is shown in Figure 6. The sensitivity tables for the problem in Period 2 is 20°C lower than in the
Generation of the sensitivity ta- are listed in Tables 3–8. They repre- base case, and the TS sensitivity table
bles. Now, we have a HEN base-case sent the response of network tem- (Table 3) shows that ∆T1 is 0.794 for
design with four heat exchangers as peratures resulting from the variation each 1°C temperature disturbance in
shown in Figure 6. The internal net- listed above. TS1, then:
work temperatures (Ti generically, Synthesizing an optimal flexible ∆T1 = (0.794)(–20) + (0.0)(–20) +
or T1 through T9 in this case), heat HEN for Period 2. The disturbances (–13.95) + (0.0) = –29.83˚C
exchanged and UA for each heat ex- in Period 2 are due to variations in TS1, By comparing Figures 4 and 7, we
changer are calculated. Then, the TS2, CP1 and CP2. The disturbance in find variations in target temperatures
variables R and B for each heat ex- the different network temperatures is due to the combined effect of varia-
changer are determined using Equa- calculated as follows: tions in target temperatures from
tions (5) and (6). Then, Equations (7) ∆Ti = ∆Ti <TS1> + ∆Ti <TS2> the base case and Period 2 (Table 1)
and (8) are constructed for each heat + ∆Ti <CP1> + ∆Ti <CP2> (9) and the disturbances in the streams’
exchanger (see Figure 2). Eight equa- Where inlet temperatures and heat capacity
tions, two for each exchanger, were ∆Ti <TS1> is the change in Ti as a con- flowrates. The target temperatures in
solved for the internal temperatures sequence of D in TS1. Figure 7 need to be adjusted to those
shown in Figure 6. A value for T9 is ∆Ti <TS2> is the change in Ti as a con- shown in Figure 4 by an increase or
obtained from T7 and T8. sequence of D inTS2. decrease in each heat exchanger’s UA,
We considered that all start temper- ∆Ti <CP1> is the change in Ti as a con- but at the same time utilities need
atures are disturbed [D] and all target sequence of D in CP1. to be kept as close as possible to the
temperatures are fixed, controlled [C]. ∆Ti <CP2> is the change in Ti as a con- target minimum values predicted by
Keeping other conditions at the base sequence of D in CP2 PDM. By using the downstream paths,
Table 8. T (UA) Sensitivity Table for Exchanger No. 4 of the Example Problem
% Change in UA from the base case
∆Ti, ºC + 100 % +80 % +60 % +40 % +20 % 0% –20% –40 % –60 % –80 % –100 %
∆T1 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
∆T2 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
∆T3 –14.1 –12.24 –10.03 –7.36 –4.08 0.0 5.39 12.57 22.71 38.1 64.2
∆T4 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
∆T5 16.3 14.19 11.64 8.57 4.78 0.0 –6.15 –14.43 –26.13 –43.9 –74
∆T6 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
∆T7 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
∆T8 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
Stream
CP, kW/°C Table 9. Comparison of UA Contingences
249°C 249°C Between Sensitivity Table
249°C 238.5°C 100°C
10.55 1 4 C1 and PDM for Period 2 and Period 3
128°C
Period 2 Period 3
259°C 170.7°C
12.66 2 1 C
UAi Sensitiv- Sensitiv-
PDM PDM
160.7°C 116°C ity table ity table
6.096 4 3
246.7°C UA1 –87% – 92 % –61 % –63 %
250°C 126°C 126°C
10.0 1 4 UA2 –90 % –87 % –100 % –100%
Integration–I. Area and Energy Targeting 18. Tantimuratha L., Asteris, G., Antonopoulos, Networks–Part III. Industrial Applications.
and Modeling of Multi-Stream Exchangers. D. K. and Kokossis, A. C., A conceptual pro- Chem. Eng. Sci., 54:685–706, 1999.
Comp. and Chem. Eng., 14(10), p. 1151–1164, gramming approach for the design of flex- 22. Konukman, A. E., Camurdan, M. C. and
1990. ible HENs. Comp. and Chem. Eng., 25 (4–6): Akman, U., Simultaneous Flexibility Tar-
14. Yee, T. F., and Grossmann, I.E.,, Simultaneous 887–892, 2001. geting And Synthesis Of Minimum-Utility
Optimization Models for Heat Integration–II. 19. Briones V. and Kokossis, A.C., Hypertargets: Heat-Exchanger Networks With Superstruc-
Heat Exchanger Network Synthesis. Comp. A Conceptual Programming Approach for the ture-Based MILP, Chem. Eng. and Proc.,
and Chem. Eng., 14(10):1165–1184, 1990. Optimization of Industrial Heat Exchanger 41(6): 501–518, 2002.
15. Aaltola J., Simultaneous synthesis of flexible Networks–I. Grassroots Design and Network 23. Kotjabasakis, E., and Linnhoff, B., Sensitivity
heat exchanger network. App. Therm. Eng., Complexity. Chem. Eng. Sci., 54:519–539, Tables for the Design of Flexible Processes
22(8), pp. 907–918, 2002. 1999. (1) – How Much Contingency in Heat Ex-
16. Aguilera, N., and Nasini, G., Flexibility Test 20. Briones V. and Kokossis, A. C. Hypertargets: changer Networks Is Cost Effective. Chem.
for Heat Exchanger Networks with Un- A Conceptual Programming Approach for the Eng. Res. Des., 64(3):197–211, 1986.
certain Flowrates. Comp. and Chem. Eng., Optimization of Industrial Heat Exchanger 24. Linnhoff, B., Townsend, D.W., and Boland,
19(9):1007–1017, 1995. Networks–II. Retrofit Design. Chem. Eng. D. “User Guide on Process Integration for
Sci., 54:541–561, 1999. the Efficient of Energy”, The Institution of
17. Aguilera, N., and Nasini, G., Flexibility Test
for Heat Exchanger Networks with Non-over- 21. Briones V. and Kokossis, A. C., Hypertargets: Chemical Engineers. Warwick Printing Com-
lapping Inlet Temperature Variations, Comp. A Conceptual Programming Approach for the pany Ltd., Great Britain, 1982.
and Chem. Eng., 20(10):1227–1240, 1996. Optimization of Industrial Heat Exchanger
The final network for Period 2 tures of network. The best achieved the flexibility problem. Indeed it does,
that is both flexible and optimized is results are as follows: <UA1> = –56 as we mentioned before the matches
shown in Figure 8. It can be seen that %, <UA2> = –100 %, <UA3> = –100%, for the three periods of operation are
streams’ temperatures before utility <UA4> = –93 %. more or less identical, with one or
use are very close to those of Figure 4. The final Network for Period 3 is more exchangers to be bypassed and
Synthesizing an optimal, flexible shown in Figure 10. In this case we the others to be adjusted for their
HEN for Period 3. The steps applied were not able to achieve the required UAs. Table 9 displays a comparison
to reach to a flexible, optimum HEN de- target temperatures using sensitivity between UA contingencies as calcu-
sign for Period 2, can be applied again tables; a deviation of 2.3–10°C from lated from PDM and sensitivity tables
for Period 3. Sensitivity tables 3–8 are PDM results was noticed (Figure 5). for Periods 2 and 3 respectively. It is
used to detect and correct disturbances clear that there is fair agreement be-
of network, due to changes in supply Comparison with PDM Design tween the contingencies’ values calcu-
temperatures, capacity flowrates and It is interesting to find out whether lated by both methods. ■
target temperatures. In this case since PDM design can offer any insight to Edited by Rebekkah Marshall
Stream 5, a branch of Stream 4, has
been canceled (CP5 = 0.0), then auto-
matically Exchanger 2 is deleted. Only Authors
the disturbance resulting from Stream Seham Ali El-Temtamy is Eman M. Gabr is a re-
5 cancellation is taken into consid- a professor of chemical engi- searcher in the Process Devel-
neering at the Egyptian Pe- opment Dept. at the Egyptian
eration when calculating the overall troleum Research Institute’s Petroleum Research Institute
disturbances’ effect. HEN for Period 3 Process Development Dept. (Email: dremangabr@hotmail.
(Email: sehamtemtamy@ com). She holds a Ph.D. in
with disturbed temperatures is shown yahoo.com). She holds a Ph.D. Chemical Engineering from
in Figure 9. The T (UA) tables are used in Chemical Engineering from Cairo University. Her exper-
Cairo University. Her exper- tise is in energy conservation.
to establish the exchangers UA contin- tise is in energy conservation
and biotechnology.
gencies to adjust the target tempera-
t%FTJHOFEBOEFOHJOFFSFE t5FTUFEVTJOHDBMJCSBUFE
to your specifications instrumentation
t"TTFNCMFEVTJOHPOMZ t4UBSUVQBTTJTUBODFBOE
high-quality components service plans available
Circle 7 on p. 58 or go to adlinks.che.com/35064-07
1-800-USA-PUMP www.buschusa.com