Estimation of the installed cost of heat exchanger networks

Estimation of the installed cost of heat exchanger networks

Internutional Elsevier Jourrlal of Production Economics, Estimation of the installed K. Suaysompol” and R.M. Woodb cost of heat exchanger a ES...

670KB Sizes 8 Downloads 51 Views

Internutional Elsevier

Jourrlal of Production

Economics,

Estimation

of the installed

K. Suaysompol”

and R.M. Woodb

cost of heat exchanger

a ESSO Standard Thailand Ltd., Bangkok, Thailand b School qf’ Chemical Enginewing & Industrial Chemistry, (Received

17 July 1992; accepted

303

29 (1993) 303-312

networks

Unicersity of’ New South Wales, Kensington,

in revised form 19 October

Austrulia

1992)

Abstract Much progress has been made in the development of synthesis methods for heat exchanger networks through the use of targeting procedures. Such advances include the development of pinch technology, the application of mathematical programming techniques and the use of the A* heuristic search procedure based on artificial intelligence by the present authors. Networks of substantially different topologies can be produced by the alternative synthesis procedures, but the methods used for cost estimation typically rely heavily on a simple area exponent calculation. As substantial capital costs are incurred for piping a simplified method is proposed which does allow for piping costs and so may be used to discriminate more accurately between alternative networks. The use of this procedure is illustrated by using some networks from published problems.

1. Introduction Targeting energy, heat transfer area and the number of shells has enabled capital cost, operating cost and the total cost (e.g. annual operating cost due to the consumption of utilities such as steam and cooling water, together with the capital charges on an annual basis) of heat exchanger networks to be predicted. As a result the trade-off between capital and operating costs can be performed prior to design, so as to determine the optimal minimum approach temperature and utility consumption (see e.g. Ref Cl]). In such work there is little uncertainty about the annual utility costs, but the predicted network capital costs are estimated from the predicted mean area of a heat exchanger unit or shell. (The target area/target number of units or shells). The exchanger costs are then usually calculated from simple equations involving an area to an exponent. If the exponent is 0.6 for example, then doubling the Correspondence to: K. Suaysompol, ESSO Standard Thailand Ltd., 1016 Rama 4 Road Bangkok 10500, Thailand.

09255273/93/$06.00

0

1993 Elsevier Science Publishers

area of a unit only increases the capital cost by 50%. Therefore the form of this relationship means that for a given total area, networks with exchangers having a wide variation in area will have lower capital costs than those with the area distributed more evenly. The actual capital cost is also influenced by the network topology as this will influence the cost of pipework. Network topology as well as the distribution of areas and shells among the expected number of units are all unknown prior to an actual design. As a result several networks need to be generated, not only for operability and other practical considerations, but also for evaluation of network costs. An alternative network cost estimation procedure is proposed here, which it is believed would be useful for reducing the number of alternatives which would otherwise need to be screened. The use of this procedure is illustrated for two literature examples, for which several networks having widely different topologies have been published.

B.V. All rights reserved.

304

K. Sutr~~~r~~pol,R.M.

~ood.!E.stimcltiorl

2. Network cost estimation model In practice, simpler heat exchanger network (HEN) designs (i.e. having fewer matches (units) and/or shells) will have a lower installed cost even if some additional heat transfer area is required. The conventional cost estimation methods assume that capital costs are determined by heat transfer area alone. Cost capacity exponents in the range 0.44-0.95 are quoted in the literature 121. A lower exponent will favour simpler networks as area will be concentrated into fewer units. However, it is an oversimplification to base capital costs on such a simple method. A more realistic cost estimation method is developed and proposed to reflect the complexity of HENS structure. This method takes piping, instrumentation, foundations, control, and other associated facilities into consideration and is based on the factor method of Guthrie 131. The cost correlation can be updated by using the Marshall and Swift (M&S) equipment cost index published monthly in Chemical Engineering. For this work the year 1988 (M&S = 852.0) was chosen. The purchased equipment cost of shell and tube exchangers as a function of area was obtained from Peters and Timmerhaus [4]. This has a cost-capacity exponent of 0.55 and is based on carbon steel costs in 1979 (M&S = 599.4). Current cost data suggests that if the costs are escalated from 1979 to 1988, relative cost comparisons will be acceptable. The purchased piping cost as a function of diameter was obtained from Holland et al. [S]. This correlation having a cost-capacity exponent of 1.33 is based on carbon steel with M&S at 1000.0. Typical pipe velocities and allowable pressure drops are used to estimate pipe diameters. For average hydrocarbon liquids, Simpson [6] recommended an optimum pipe velocity (I/) of 3.0 m/s. By using Eq. (1) with fluid physical properties from Table 1, the pipe diameters for hot and cold streams can be obtained and then selected from commercially

of’ the installed

available

cost 01‘ HEN

pipe sizes.

Dk = 2 x lo6

CPI, 71fk vkcp,

(&HSuCS}.

(1)

Based on typical oil refinery situations, the average on-site equipment spacing between two heat exchanger units in different services is 7 m. The actual pipe length (L) is approximately four times the on-site equipment spacing. This length accounts for the average piping required by each exchanger (which is additional to other pipe runs associated with process streams). The purchased equipment costs, as obtained above, together with the installation factors for heat exchangers (F,) proposed by Guthrie [3] are used to estimate the installed equipment costs. The details of these installation factors using a modular approach are listed in Table 2. The installed network cost (C”) correlation based on number of shells which seeks to reflect network complexity is presented in

Eq. (2).

M&S + O.O4F, ~ (LiD!‘33 + Lj,~‘“‘) i 1000.0 >

1 ,

(2)

Where Aij and Nij denote the total heat transfer area and number of shells of the match between hot stream i and cold streamj satisfying the following conditions (10 < (Aij/Nij) < 500}, and (25
Table 1 Typical fluid physical

properties

[7]

Fluid

Paraflins Oils Heat transfer fluid Cooling water

800 800 800

1000

2239 2880 2880 4176

K. Suaysompol,

Table 2 Field installation

costs (Guthrie

R.M.

Wood/Estimation

qf the installed

Heat transfer exchanger

items are based on carbon

equipment

shell & tube heat

1.oo

FOB Equipment Concrete Steel Instruments Electrical Insulation Paint Total material (M) Erection & Setting (L) Total direct M&L factor Freight, insurance, taxes, engineering Overhead or field expense Total indirect factor Total module factor (exci. piping) Total module factor (incl. piping) for individual

305

c?fHEN

[3])

Details of cost break down

a Factors

cost

0.05 0.03 0.10 0.02 0.05 1.25 0.63 1.88 0.08 0.95 1.03 2.91 3.31 steel material.

pipe-lengths for hot stream i and cold stream for matches between these j respectively, streams (Aij> 0). The value of F, excluding piping is 2.91 as suggested by Guthrie [3]. Holland et al. [S] recommended that the installation factor for piping (F,) equals 13.0 times the purchased piping cost, which is consistent with IChemE [S] cost data. The conventional installed equipment cost correlations based on units or shells (C: or Cf) could be obtained from Eq. (2) with F, set to zero and F, equal to 3.37.

3. Network case studies 3.1. Case Study I This is a four stream problem and four different designs are shown in Figs. l-4. These network designs are shown in the grid format for heat exchanger networks. The streams being cooled (hot streams) are at the top of the diagrams and flow from left to right, with the cold streams flowing from right to left. Each exchanger (El, E2,. . . .) achieves heat transfer from a hot stream to a cold stream. Coolers

(Cl, c2,. . . .) exchange heat from hot streams to cold utilities (e.g. cooling water). Heaters (Hl, H2,.... (not required here)) transfer heat from a hot utility (e.g. steam) to a cold stream. Temperatures are shown for all hot and cold streams on the grid diagrams and the exchanger loads Q are shown beneath the exchangers and coolers: Q = CPAT,

(3)

Three designs [9, lo] were obtained using Algorithmic Design Methods (ADMs) based on mathematical programming and the fourth [l l] by the Flexible Pinch Design Method (FPDM) as incorporated into the computer program FLEXNET. The ADMs use network superstructures which incorporate many alternative designs. Mathematical programming techniques then seek an optimal design structure from the superstructures and optimise the design parameters. Sometimes the ADMs produce networks having an unusual structure such as that of Fig. 2. Although this is the simplest network in having the fewest exchanger units (four), this network requires the complex bypassing and mixing structure for hot stream 1. The FPDM uses a step by step

K. Suuysompol,

306

R.M. WoodlEstinzcltion

of the installrd cost of HEN

1.CO

20.00

I.CO

l.GO

a

00 d

‘,

r-

10.00

2 L-l

2400

Fig. I. Algorithmic

design using DICOPT+

+ for case study

I (Vishwanathan

and Grossmann

[9]).

Cl’ kW/C

82 GO

30.00

400

1.GO

15.00

1.60

20.00

no

140 I.GO

(.

--

-

10.00

2x0

Fig. 2. Algorithmic

design using MAGNETS

for case study 1 (Yee and Grossmann

method based on heuristic rules for network development, and is guided by the A* heuristic search procedure. All designs achieved the same energy recovery of 4.7 MW which corresponds to the net-

[lo]).

work Heat Recovery Approach Temperature (HRAT) of 56°C (i.e. a network cannot be designed in which all the exchangers have minexceeding temperatures approach imum HRAT). The matches are for counter-current

K. Suaysompol,

R.M.

Wood/Estimation

of the installed

cost of HEN

307 Cl’ klI/C

I.60

15.00

l.GO

20.00

.lO.OO 2400

Fig. 3. Algorthmic

design using MINLP

for case study 1 (Yee and Grossman

[lo]).

h

CI’ kW/C

kH/m2C

1 .GO

30.00

I .GO

15.00

1.60

20.00

I .GO

40.00 2400

Fig. 4. Flexible pinch design using FLEXNET

for case study 1 (Suaysompol

exchangers and each unit corresponds to a single heat exchanger shell. Details of the designs are summarised in Table 3 which presents the network number of units (IV:), and shells (Nr), heat transfer area (AN), cost based

and Wood

[l 11).

on units (C:) and cost based on Eq. (2) (CN) as well as the minimum temperature difference between a hot stream and a cold stream at either the hot or cold side of an exchanger (EMAT). (The parameters in Eq. (2) were

308

K. Suaysompol,

Table 3 Design and cost comparisons Case

EMAT

R.M.

Wood!Estimution

of‘ the installed

cost of‘ HEN

for case study I

Unit

Target

1

7

c Unit Unit Shell Shell mz %, k$ “/o k$ %

5.6 4 0 4 0 471 0 698 0

scaled so as to give similar network costs (CN) to those obtained from the data used in the original publications.) The target items (see e.g. Linnhoff and Ahmad [l] for the network can be calculated from the stream data (i.e. inlet and outlet temperatures, mass flowrates and heat capacities) and these targets are independent of any designs. E.g. the number of units target is one less than the sum of the hot, cold and utility streams in the network. As seen from Table 3 the three algorithmic designs have values for the Exchanger Minimum Approach Temperature (EMAT) which are below HRAT. However FLEXNET chose a design for which HRAT = EMAT and this design is the same as that obtained from the Pinch Design Method (PDM) [12]. Although there are variations in the numbers of units and network areas, the network costs (Cr) based on shells (units) and areas are all virtually the same. However when piping costs are also included (CN) the simplest design has the lowest cost and there is now probably a significant difference, between the cost of case 2 and that of the design produced by FLEXNET. However the piping costs for the designs in Fig. 1 and 2 do not contain items for the control equipment required to regulate the split streams. Furthermore Eq. (2) only contains costs for piping connected to exchanger units and does not allow for the part of stream

ADM 3 2.7 5 +I 5 +1 524 +9 722 +3 782 +2

ADM 4

ADM

1.8

2.6

4 0 4 0

5

593 + 23 720 +3 764 0

+1

FPDM

5.6 6 +Z

5

6

+1 562 + 17 729 +4 802 +5

f2 475 +1 730 +4 820 f7

1, which is shown in Fig. 2 as having no direct contact to an exchanger.

3.2. Case study 2 Data for a crude oil distillation preheat train has been used to generate the four networks presented in Figs. 5-8, details of which are summarised in Table 4. These designs were obtained by the following four different procedures. (1) The HEXTRAN computer program which uses the Dual Approach Temperature Design Method (DATDM). (2) The Pseudo-Pinch Design Method (PPDM). (3) FPDM (hand design) (4) FLEXNET (Computer version of the FPDM). All designs achieve the same energy recovery of 65 MW (based on HRAT of 40°C) and the target temperature for stream 2 represents the fired heater inlet temperature. This heater is an equipment item common to all designs and the cost cannot be estimated using Eq. (2) (which is not suitable for fired equipment) therefore the heater has been omitted. All designs have shell and tube (l-2) exchangers for which the minimum log mean temperature difference correction factor is 0.75. Since all designs have the same total cold utility duty for the coolers,

K. Suaysompol,

R.M. Wood/Estimation

309

of the installed cost of HEN

Cl’

h MH/ru2C

MN/C

0.0015

0.0533 0.0459

0.002

0.2157,

I

T c

23l 3(,

(, 207 (2

0.0018

0.0023

O.OOiG

O.OOOG

107 *

0.0029

0.0013 12.7

Fig. 5. Dual approach

2.2

3.7

temperature

1.4

2.1

4.2

design using HEXTRAN

11.8

8.7

4.7

I.0

9.4

for case study 2 (O’Neill et al. [13]).

T C 93 0.0533 0.0459

234 50 41

0.2.157

207 142

0.0016

0.0535 0.04G3

203 113 42

0.0023

0.2052

270 191

0.0070

0.1402

20G 199

0.0006

0.1209

0.0015 2.0 0.002

I

0.0013

12.9

Fig. 6. Pseudo

8.3

1.0

1.7

18.0

pinch design for case study 2 (O’Neill et al. [13]).

9.8

7.0

0.5

248

I I5

0.1455

I(35 107

0.2744 0.3071 0.3074

20 106 15G 213

310

K. Suaysompol,

R.M.

WoodfEstimution

of the installed

cost of‘ HEN

11 hlH/IIl2C

c-l

0.0015

L

23’ 412

207 0.002

142

I

0.001a

0.0023

206 0.0076

0.0000

ES

T

0.0029

0.0013 17.8

18.0

Fig. 7. Flexible pinch design for case study 2 (Suaysompol

1.0

12.9

6.0

0.1

and Wood [14]).

h un/n12c

0.0015

0.002

I

0.001

G

0.0029

Fig. 8. Flexible

pinch design using FLEXNET

for case study 2 (Suaysompol

and Wood

[ll]).

K. Suaysompol,

R.M. Wood/Estimation

qf the installed cost qf HEN

311

Table 4 Design and cost comparisons Case

Unit

for case study 2

Target

2

1 EMAT NGJ AN; N,N AN: AN AAN C,” AC! CN ACN

-C Unit Unit Shell Shell m2 % k$ ; %

40.0 8 0 12 0 1927 0 248 0

_

they may be fairly compared on the basis of network capital costs. As seen from Table 4, the values for EMAT are all lower than the value for HRAT of 40°C. The numbers of units (Nr) and shells (NY) vary considerably and there is a variation of f 5% in network areas (AN). For a network capital cost based on shells and area Cr, the design achieved by the PPDM is a clear leader. However when piping costs are also considered the ranking position of the simplest network (FPDM) improves very considerably. Thus although it has the highest area of all four designs, this is more than compensated by the outstanding simplicity of the design, which achieves the units target of eight.

31.2 14 +6 15 +3 2359 + 22 281 + 13 372 + 16

30.0 11 +3 12 0 2252 + 17 256 +3 326 +2

FPDM 3

FPDM 4

22.3 8 0 12 0 2476 + 28 273 + 10 321 0

9.0 10 +2 13 +l 2231 + 16 264 +6 329 f2

cost. It would be of interest to develop this procedure further with more recent cost data.

Nomenclature AN A g Cb CP C, cs

D EMAT

4. Conclusions FP_ A new procedure for evaluating the capital cost of heat exchanger networks has been desribed, which includes an estimate of the extra piping costs determined by the network topology. Results from two case studies indicate that the use of this procedure can significantly change the ranking of alternative networks compared to conventional cost estimates based on area and shells alone. In both cases networks having the minimum or close to the minimum number of units have the lowest

PPDM

DATDM

F, h HRAT

HS

L M&S

Network surface area (m2), Heat exchanger surface area (m2), Network cost using Eq. (2) ($), Network cost based on shells ($), Network cost based on units ($), Heat capacity flowrate-stream mass flowrate x Heat capacity (MW/“C), Heat capacity (J/kg”C) Set of cold streams (member j and k), Pipe diameter (mm), Exchanger minimum approach temperature in the network (EMAT < HRAT) (“C), Factor for exchanger installation cost, Factor for piping installation cost, Heat transfer coefficient for stream (kW/m2’C), Heat recovery approach temperature for the utility consumption of the network (“C), Set of hot streams (member i and k), Pipe length (m), Marshall and Swift equipment cost index,

312 N

N,N Nf;

Q T V P

K. Suaysompol,

R.M.

Wood/Estimation

Number of shells or unit (one) for a heat exchanger unit (match), Number of shells in a heat exchanger network, Number of units in a heat exchanger network, Exchanger heat load (MW), Stream temperature (“C), Velocity (m/s), Density (kg/m3).

Note added in proof An alternative method for HEN synthesis including costing (recently noted by the authors) has been proposed by Zhelev and Markov [ 151.

References [II Linnhoff, B. and Ahmad, S., 1989. Supertargeting: Optimum synthesis of energy management systems. Trans. ASME. J. Energy Resour Technol., 111: 121-130. PI Remer, D.S. and Chai, L.H., 1990. Design cost factors for Scaling-up engineering equipment, Chem. Eng. Prog., 86(8): 77-82 K.M., 1969. Capital cost estimating, c31 Guthrie, Chem. Eng., 76(6): 114142. K.D., 1980. Plant c41 Peters, M.S. and Timmerhaus, Design and Economics for Chemical Engineers, Third Edition, McGraw-Hill, New York. c51 Holland, F.A., Watson F.A. and Wilkinson, J.K., 1984. Chemical Engineers Handbook, Sixth Edi-

of the instal/ed

cost of HEN

tion, R.H. Perry and D. Green McGraw-Hill, New York, 25.64-25.80. C61 Simpson, L.L., 1968. Sizing piping for process plants, Chem. Eng., 75( 13): 192-214. c71 Coulson, J.M., Richardson, J.F. and Sinnott, R.K., 1985. Chemical Engineering, 6. Pergamon Press. Oxford, England. PI Institution of Chemical Engineers, 1988. Guide to Capital Cost Estimating, Third Edition. IChemE, Rugby, England. c91 Viswanathan, J. and Grossmann, I.E., 1990. A combined penalty function and outer-approximation method for MINLP optimization, Comput. Chem. Eng., 14: 7699782. Cl01 Yee, T.F. and Grossman, I. E., 1990. Simultaneous optimization models for heat integration ~ II heat exchanger network synthesis, Comput. Chem. Eng., 14: 116551184. Cl11 Suaysompol, K. and Wood, R. M., 1991. The flexible pinch design method for heat exchanger networks: II heuristic searching guided by the A* Algorithm, Trans. IChem. E., 69: 4655470. E., 1983. The pinch Cl21 Linnhoff, B. and Hindmarsh, design method for heat exchanger networks, Chem. Eng. Sci., 38: 1175-l 188. Cl31 O’Neill, B.K., Roach, J.R., Wood, R.M. and Trivedi, K.K., 1989. Design of Heat Exchanger Networks Employing Pseudo-Pinch Concepts, 17th Aust. Chem. Eng. Conf. (Chemeca’89) pp. 441-448. Cl41 Suaysompol, K. and Wood, R.M., 1991. The flexible pinch design method for heat exchanger networks: I heuristic guidelines for free hand designs. Trans. I Chem. E., 69: 4588464. [I51 Zhelev, T. and Markov, Y., 1991. Optimal synthesis HENS when the piping costs are taken into account. Fourth World Congress of Chemical Engineering, Strategy 2000, Karlsruhe, Germany, Preprint III, Section 8, pp. 4-10.