Available online at www.sciencedirect.

com

Applied Thermal Engineering 28 (2008) 17981805 www.elsevier.com/locate/apthermeng

Design optimization of shell-and-tube heat exchangers

Rio de Janeiro State University (UERJ), Instituto de Qumica, Rua Sao Francisco Xavier, 524, CEP 20550-900, Rio de Janeiro, RJ, Brazil Federal University of Rio de Janeiro (UFRJ), Escola de Qumica CT, Bloco E, Ilha do Fundao, CEP 21949-900, Rio de Janeiro, RJ, Brazil Received 20 October 2006; accepted 12 November 2007 Available online 19 November 2007

Abstract This paper presents a study about the design optimization of shell-and-tube heat exchangers. The formulated problem consists of the minimization of the thermal surface area for a certain service, involving discrete decision variables. Additional constraints represent geometrical features and velocity conditions which must be complied in order to reach a more realistic solution for the process task. The optimization algorithm is based on a search along the tube count table where the established constraints and the investigated design candidates are employed to eliminate nonoptimal alternatives, thus reducing the number of rating runs executed. The performance of the algorithm and its individual components are explored through two design examples. The obtained results illustrate the capacity of the proposed approach to direct the optimization towards more eective designs, considering important limitations usually ignored in the literature. 2007 Elsevier Ltd. All rights reserved.
Keywords: Shell-and-tube heat exchanger; Design; Optimization

1. Introduction Shell-and-tube heat exchangers are the most common type of thermal equipment employed in chemical process industries. This widespread use can be justied by its versatility, robustness and reliability. Despite the technological advances of other exchanger types (e.g., plate-and-frame, spiral, lamella, etc.), shell-and-tube heat exchangers will maintain a central position in industrial activities in the next years. Due to the important role of shell-and-tube heat exchangers, a considerable number of papers has been devoted to the design optimization problem, employing different techniques, such as, numerical resolution of the stationary point equations of a nonlinear objective function [1,2], graphical analysis of the search space [3,4], simulated annealing [5], genetic algorithms [6,7], mixed integer nonlinear programming [8], systematic screening of tube count
*

Corresponding author. Tel.: +55 21 2587 7631. E-mail address: andrehc@uerj.br (A.L.H. Costa).

tables [9,10], among others. These techniques were employed according to distinct problem formulations in relation to: (i) objective function: heat transfer area or total annualized costs (i.e. capital costs of the heat exchanger and pumps/compressors associated to uid ow operating costs); (ii) constraints: heat transfer and uid ow equations, pressure drop and velocity bounds, etc.; and (iii) decision variables: selection of dierent search variables and its characterization as integer or continuous (e.g., tube diameter can be considered a xed parameter, a continuous variable or a discrete variable). In spite of the algorithmic developments applied to heat exchanger design, the complexity of the task allows some criticism of the eectiveness of optimization procedures for real industrial problems [11]. In the context of the development of new design algorithms, this paper presents an optimization procedure integrated with practical design guidelines, aiming to provide a feasible alternative in an engineering point of view. These design rules are usually ignored in the algorithms present in the literature, which restrains the eective application

1359-4311/$ - see front matter 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.applthermaleng.2007.11.009

A.L.H. Costa, E.M. Queiroz / Applied Thermal Engineering 28 (2008) 17981805

1799

Nomenclature A Aexc Areq Bc CSL CSNtp CSRbs Ds Dt F L Lbb Lmax b m NDs NDt NL NNtp NRbs Ntp Ntt Ntpp OPT Qreq Rbs Rtp heat exchanger area (m2) excess area required area (m2) bae cut subset of tube lengths subset of numbers of tube passes subset of ratios between bae spacing and shell diameter inner shell diameter (m) outer tube diameter (m) correction factor for DTLM tube length (m) shell-bundle clearance (m) maximum unsupported span (m) mass ow rate (kg/s) number of elements of SDs number of elements of SDt number of elements of CSL number of elements of CSNtp number of elements of CSRbs number of tube passes total number of tubes number of tubes per pass set of optimal solution candidates service heat load (W) ratio between bae spacing and shell diameter tube pitch ratio SDs SDt SL SNtp SRbs U v set of shell diameters set of tube diameters set of tube lengths set of numbers of tube passes set of ratios between bae spacing and shell diameter overall heat transfer coecient (W/(m2K)) uid velocity (m/s)

Greek symbols d thickness (m) DP pressure drop (Pa) DTLM logarithmic mean temperature dierence (C) q uid density (kg/m3) Subscripts s shell side t tube side Superscripts k tube diameter index l shell diameter index m tube length index n number of tube passes index o bae spacing index min minimum value max maximum value

of the design solutions found. Additionally, another important aspect of the proposed optimization scheme consists in the exploration of the relations between heat exchanger performance and design modications in order to direct the search towards the optimal solution. The rest of this paper is organized as follows. Section 2 presents the adopted formulation of the optimal design problem, Section 3 describes the optimization algorithm, Section 4 illustrates the application of the algorithm in two design examples and Section 5 summarizes the nal conclusions. 2. Problem formulation 2.1. Decision variables The design of a heat exchanger involves the selection/ sizing of its main mechanical components to fulll a desired service. Each set of design parameters constitutes a candidate solution to the problem and, since there are several candidates that can execute a certain service, an optimization procedure can be applied to identify the best option, usually based on economic factors. However, the concept of a good design involves aspects that cannot be easily described in a single economic

objective function: fouling suppression, maintenance easiness, mechanical resistance, ow distribution, potential tube vibration, etc. Therefore, the set of design parameters of a shell-and-tube heat exchanger can be divided into two groups. The rst group contains parameters that are dened by the engineer, since they are, partially or entirely, connected to broader engineering aspects. The second group includes parameters that can be searched to nd the most economical design for the proposed service. Table 1 describes the design parameter classication adopted in this work. Additionally, practical design rules can be established among the searched parameters: the shell should be full

Table 1 Classication of shell-and-tube design parameters User dened parameters Fluid allocation (tube or shell) Materials (shell, tube, bae, etc.) Shell type (TEMA E, F, G, H, J, etc.) Head type (TEMA A, B, L, M, S, T, etc.) Bae type (single, double, NTIW, etc.) Tube pitch ratio (1.25, 1.33, 1.50) Tube layout pattern (30, 45, 60) Tube wall thickness Searched parameters Inner shell diameter Total number of tubes Number of tube passes Tube length Outer tube diameter Bae spacing Bae cut

1800

A.L.H. Costa, E.M. Queiroz / Applied Thermal Engineering 28 (2008) 17981805

of tubes according to a tube count table [12] and bae cut and bae spacing values should respect an established relation [13]. The rst rule implies that for a given set composed by a head and shell types, shell diameter, tube pitch and layout, tube diameter and number of tube passes; the total number of tubes is automatically xed by a tube count table. The second rule states that for a given bae spacing and shell diameter, the bae cut can be estimated automatically. Considering single-phase ow with segmental baes, this relation is represented by the following equation, obtained from a graph presented in [13]: Bc 0:16 0:20Rbs ; 1

the shell diameter and the maximum number of tube passes. 2.2.2. Velocity constraints These constraints represent uid velocity limits in order to reduce fouling and erosion problems. The tube side stream velocity must obey lower and upper bounds: vt P vmin ; t vt 6 vmax : t 7 8

The corresponding constraints for the shell side stream are vs P vmin ; s vs 6 vmax : s 9 10

where Bc is the bae cut and Rbs is the ratio between bae spacing and shell diameter, variable employed to represent the bae spacing in the optimization algorithm. Finally, these concepts allow the representation of each candidate exchanger in the optimization algorithm by ve decision variables: outer tube diameter (Dt), inner shell diameter (Ds), tube length (L), number of tube passes (Ntp) and the ratio between bae spacing and shell diameter (Rbs). The variables Dt, Ds, L and Ntp are assumed to be discrete variables owing to their nature and standard commercial alternatives. Due to the algorithm structure, the variable Rbs is also discretized, however without loss of generality (if needed a ner grid can always be used). The integer alternatives of the decision variables are organized in the following sets: SDt, SDs, SL, SNtp and SRbs, respectively. 2.2. Constraints 2.2.1. Geometric constraints The shell-and-tube heat exchanger candidate must respect a series of geometric constraints, involving the following rules [13]: the ratio between tube length and shell diameter must be between 3 and 15; the ratio between bae spacing and shell diameter must be between 0.2 and 1.0; the bae spacing cannot be lower than 50 mm; the bae spacing must obey the maximum unsupported span; there is an upper bound on the number of tube passes for a given shell diameter, minimizing the gap in the tube bundle due to multiple passes. These constraints can be represented by the following mathematical expressions: 3 6 L=Ds 6 5; 0:2 6 Rbs 6 1:0; Rbs P 0:050=Ds ; Rbs 6 0:5Lmax =Ds ; b N tp 6 f Ds ; 2 3 4 5 6

2.2.3. Service constraints This kind of constraint represents the thermal and hydraulic demands of the desired service. The thermal performance of the exchanger must fulll the required overdesign, represented by an excess area, Aexc, which must be larger than a prescribed value, Amin : exc Aexc A Areq P Amin ; exc Areq 11

where A is the actual area of the heat exchanger and Areq is the required area for the service. The required area is calculated by Areq Qreq ; U DT LM F 12

where Qreq is the service heat load, U is the overall heat transfer coecient, DTLM is the logarithmic mean temperature dierence for the equivalent counterow arrangement and F is its correction factor. The hydraulic requirements of the service are represented by upper bounds on the pressure drop of both streams: DP t 6 DP max ; t DP s 6 DP max ; s 13 14

where DPt and DP max are the pressure drop and its maxit mum value for the tube side stream and DPs and DP max s are the corresponding values for the shell side stream. 2.3. Objective function The optimization procedure aims to reduce the capital costs of the heat exchanger represented by its heat transfer surface. 2.4. Thermal and hydraulic evaluations The structure of the algorithm involves a set of heat transfer and uid ow equations organized in an independent rating code. This external code is responsible to evaluate the required area and pressure drops of the solution candidates, as presented in Fig. 1. This approach gives

where Lmax is the maximum unsupported span, a linear b function of the tube diameter and f is the relation between

A.L.H. Costa, E.M. Queiroz / Applied Thermal Engineering 28 (2008) 17981805

1801

Dt , Ds , L Ntp , Rbs

Rating Code

Areq Pt , Ps

Lmin 3Dl ; s L
max

15 16

15Dl : s

Fig. 1. Heat exchanger rating scheme.

These bounds determine the elements L(m) (m = 1, . . . , NL with L(m) < L(m+1)) of the subset of tube lengths, CSL & SL: CSL fLm jL1 P Lmin ^ LNL 6 Lmax g: 17

exibility to the solution procedure, since the user can employ the optimization algorithm connected to any desired rating code. 3. Optimization algorithm 3.1. Scope The optimization algorithm is suited to the design of shell-and-tube heat exchangers with single-phase uids owing in a unique shell. Before the optimization, the user must dene the values of the parameters indicated in the left column of Table 1. The algorithm is based on an oriented search along the tube count table. If a feasible design is found, it is saved for the nal determination of the alternative with the smallest area at the end of the search. In the algorithm, an outer loop passes through the tube diameter set and another loop moves along the shell diameter set. For a certain shell diameter, the algorithm seeks to identify the smallest tube length which can fulll the service. Thus, in the analysis of a certain tube length, the number of tube passes and bafe spacing are manipulated in order to nd a feasible alternative. If there is no feasible design, the search moves to a larger tube length. If all values of tube lengths were investigated and no feasible design was found, the shell diameter is increased and the process is repeated. After the examination of all shell diameters, the tube diameter is increased and the search goes on. If the last tube diameter was analyzed, the algorithm stops. A fundamental aspect of the search is the analysis of the constraints and the investigated congurations to reduce the number of rating runs. This procedure is conducted using the denitions of variable bounds, feasibility tests and fathoming. 3.2. Variable bounds The geometric and velocity constraints can be organized in a set of decision variable limits. For a given pair of tube k k and shell diameters, Dt (k 2 {1, . . . , NDt} with Dt < k1 l l l1 Dt and Ds (l 2 {1, . . . , NDs} with Ds < Ds , the constraints allow the establishment of bounds on the other decision variables. These bounds are related to the current k values of Dt and Dl , guiding the search inside the feasible s region and contracting the sets of integer alternatives. 3.2.1. Tube length Constraints (2) dene bounds on the exchanger tube length:

3.2.2. Number of tube passes Constraints (7) and (8) determine bounds on the number of tubes per pass (Ntpp, equivalent to the ratio between the total number of tubes and the number of tube passes): N min tpp N max tpp mt =qt =vmax t pDt 2dt =4 mt =qt =vmin t pDt 2dt =4
k 2 k 2

; ;

18 19

where mt and qt are the mass ow rate and density of the tube side stream and dt is the tube thickness. Since the shell diameter is previously known, these parameters can be employed to determine, in the tube count table, bounds on the number of tube passes, N min 1 and N max 1 . tp tp Additionally, constraint (6) also imposes a second upper bound: N max 2 f Dl : tp s 20

The resultant bounds must consider both groups of constraints: N min N min 1 ; tp tp N max tp minfN max 1 ; N max 2 g: tp tp
n n

21 22
n1

Thus, the elements N tp (n = 1, . . . , NNtp with N tp < N tp of the subset CSNtp & SNtp are given by CSNtp fN tp jN tp P N min ^ N tp tp
n 1 NNtp

6 N max g: tp

23

3.2.3. Ratio between bae spacing and shell diameter Constraints (9) and (10) are involved in the following bounds, related to the shell side cross-ow area: 2 3 ! 6 7 1 ms =qs =vmax min s 7 Rbs 1 6 4 5 Dl ; l k Ds Dt Lbb k k s Lbb Rtp Dt Dt Ltp;eff ! 6 7 1 ms =qs =vmin s 7 Rmax 1 6 bs 4 5 Dl ; l k Ds Dt Lbb k k s Lbb Rtp Dt Dt Ltp;eff 25 where ms and qs are the mass ow rate and density of the shell side stream, Rtp is the tube pitch ratio, Lbb is the k shell-bundle clearance and Ltp,e is equal to Rtp Dt for 2 3 24

1802

A.L.H. Costa, E.M. Queiroz / Applied Thermal Engineering 28 (2008) 17981805

30 and 90 layouts and is equal to 0:707Rtp Dt for 45 layout. Additionally, constraints (4) and (5) impose a second pair of bounds: Rmin 2 0:050=Dl ; bs s Rmax 2 bs 0:5Lmax =Dl : b s The resultant bounds are Rmin maxfRmin 1 ; Rmin 2 g; bs bs bs Rmax minfRmax 1 ; Rmax 2 g: bs bs bs The elements Rbs (o = 1, . . . , NRbs with Rbs < Rbs subset CSRbs & SRbs are given by CSRbs fRbs jRbs P Rmin ^ Rbs bs
o 1 NRbs o o o1

of same shell diameter, if both of them were not thermally feasible, none other conguration with this shell diameter will need to be analyzed and the search must be conducted to the next shell diameter. 3.3.2. Tests H2, T2a and T2b For given values of tube diameter, shell diameter and k tube length, Dt , Dl and L(m), the conguration with the s lowest number of tube passes and the largest bae spacing, 1 NR N tp and Rbs bs , corresponds to the lowest pressure drops. Thus, if this conguration is not hydraulically feasible, the search must be immediately directed to a larger shell diameter (further increases of the tube length with this shell diameter would imply in even larger pressure drops). This test is identied as Test H2. If Test H2 is positive, the Tests T2a and T2b are applied. The Test T2a involves the conguration with a single tube 1 1 pass and the smallest bae spacing, N tp and Rbs 1 (if N tp > 1 then the test result is assumed negative, without rating). The Test T2b is applied to the conguration with the highest number of tube passes and the smallest bae NN NN 1 spacing, N tp tp and Rbs (if N tp tp 1 then the test result is assumed negative, without rating). Since, the algorithm assumes that the highest value of excess area involves one of these two congurations, if both tests are negative then the search must be directed to the next tube length. 3.3.3. Tests H3 and T3 k Considering a certain design candidate, Dt Dl Lm s n o N tp Rbs , the hydraulic and thermal feasibility verications correspond to the Tests H3 and T3. 3.4. Fathoming

26 27

28 29 ) of the 30

6 Rmax g: bs

It is important to mention that constraints (3) are previously included in the delimitation of the set SRbs. 3.3. Feasibility tests The service constraints are veried through the application of rating runs to proposed design alternatives. A hydraulic test checks if the pressure drops of a certain candidate are smaller than the allowable specications (a positive test result means that the heat exchanger alternative is hydraulically feasible). A thermal test checks if the exchanger area is larger than the required area, according to the prescribed overdesign (in this case, a positive test result means that the exchanger alternative is thermally feasible). A proper exploration of these feasibility tests can eliminate from the search large sets of candidate alternatives. 3.3.1. Tests H1, T1a and T1b Considering a certain pair of tube and shell diameters, k Dt and Dl , the design alternative with the lowest pressure s drops is composed by the shortest tube length, the lowest number of tube passes and the largest bae spacing: L(1), 1 NR N tp and Rbs bs . Thus, if this alternative is not hydraulically feasible, it is not necessary to continue the analysis of other alternatives with the same shell diameter and the search must be directed to the next shell diameter in the tube count table (note that the search follows, for each investigated tube diameter, an increasing sequence of shell diameters). This feasibility test will be identied as Test H1. If Test H1 is positive, then more two tests, T1a and T1b, are applied to investigate the thermal feasibility. The Test T1a involves a conguration composed by the longest tube length, a single tube pass and the smallest bae spacing: 1 1 1 L(NL), N tp and Rbs (if N tp > 1 then the test result is assumed negative, without rating). The Test T1b is applied to a conguration composed by the longest tube length, the highest number of tube passes and the smallest bae NN NN 1 spacing: L(NL), N tp tp and Rbs (if N tp tp 1 then the test result is assumed negative, without rating). Since, the algorithm assumes that one of these two congurations presents the highest value of excess area of the candidates

Let Amin be the lowest area of a feasible design alternak tive already identied during the search and let Dt and l Ds be the values of the tube and shell diameters explored at the moment. The tube length related to Amin is Lmax; Amin pDt N tt
k

31

where Ntt is the total number of tubes indicated in the tube NN k count table associated to the values of Dt , Dl and N tp tp . s Since design alternatives with tube length longer than Lmax,* will present higher area than the best candidate already found, this parameter can be employed to a further contraction of the subset CSL: CSL fLm jL1 P 3Dl ^ LNL s 6 minf15Dl ; Lmax; gg: s 32

In this case, if the contracted subset CSL becomes empty because the shortest tube length in the original subset CSL is larger than Lmax,*, then it is not possible to nd a heat exchanger with area smaller than Amin related to the shell diameter considered. Additionally, since any further increase of the shell diameter imply in the same situation,

A.L.H. Costa, E.M. Queiroz / Applied Thermal Engineering 28 (2008) 17981805

1803

the search must be directed by the external loop to a dierent tube diameter. 3.5. Algorithm steps Step 1: Start Select the values of the xed design parameters (see Table 1). Establish the sets of the decision variables: SDt, SDs, SL, SNtp and SRbs. Initialize the set of feasible candidates to the optimal design: OPT = ;. Initialize the tube diameter index: set k = 0. Step 2: Update tube diameter If k = NDt then stop, otherwise increase the value of the tube diameter investigated: k k + 1 and initialize the shell diameter index: l = 0. Step 3: Update shell diameter If l = NDs then go to Step 2, otherwise increase the value of the shell diameter investigated: l l + 1 and initialize the tube length index: m = 0. Step 4: Identication of variables bounds (4.1) Update the bounds Lmin and Lmax to determine the corresponding subset of tube lengths, CSL. If CSL = ; then go to Step 3. (4.2) Update the bounds N min and N max to determine tp tp the corresponding subset of number of tube passes, CSNtp. If CSNtp = ; then go to Step 3. (4.3) Update the bounds Rmin and Rmax to determine the bs bs corresponding subset related to bae spacing values, CSRbs. If CSRbs = ; then go to Step 3. Step 5: Fathoming If the set OPT is not empty, apply the fathoming procedure, updating the subset of tube lengths, CSL. If CSL = ; then go to Step 2. Step 6: Application of Tests H1, T1a and T1b (6.1) Apply Test H1. If Test H1 is negative then go to Step 3. (6.2) Apply Tests T1a and T1b. If both tests are negative then go to Step 3. (6.3) If Test T1a is negative then contract the subset: CSNtp CSNtp {1}. (6.4) If Test T1b is negative then contract the subset: CSNtp {1}. Step 7: Update tube length If m = NL then go to Step 3, otherwise increase the value of the tube length investigated: m m + 1 and initialize the number of tube passes and bae spacing indices: n = NNpt and o = 1. Step 8: Application of Tests H2, T2a and T2b (8.1) Apply Test H2. If Test H2 is negative then go to Step 3.

(8.2) Apply Tests T2a and T2b. If both tests are negative then go to Step 7. (8.3) If Test T2b is negative then set n 1. Step 9: Application of Tests H3 and T3 Apply Tests H3 and T3. If both tests are positive then save the investigated conguration in the set of design k n o candidates: OPT OPT [ fDt Dl Lm N tp Rbs g and s go to Step 3. Step 10: Update the number of tube passes and bae spacing (10.1) If DP t > DP max then decrease the number of tube t passes: n n 1. (10.2) If DP s > DP max then increase the bae spacing: s o o + 1. (10.3) If Aexc < Amin then go to Step 10.4 otherwise go to exc Step 9. n 1 (10.4) If N tp > 1 and N tp 1 then set a single pass conguration: n 1 and go to Step 9 otherwise go to Step 7. 4. Results The performance of the proposed algorithm is illustrated through the analysis of the results obtained in two examples of design tasks. The rst example aims to compare the solution reached by the optimization algorithm with a previous recent literature approach, also based in a tube count table search. The objective of the second example is to characterize the behavior of the computational eort according to the dierent components of the search algorithm. The heat exchanger rating code implemented for the solution of the examples employs the following thermouidynamic equations: the BellDelaware method is used for shell side lm coecient and pressure drop evaluation [13]; for tube side lm coecient calculations, the Gnielinski correlation is utilized in turbulent ow, the Sieder and Tate correlation is employed in developing laminar ow and Nusselt number equals to 3.66 is used in established laminar ow [14]; for tube side pressure drop, the Darcy equation is used together with an expression for the pressure drop in the exchanger heads [15]. The rating code was validated against literature examples (e.g. the oil heater described in [13]). It is imposed that the tube side velocity must be between 1 m/s and 2.5 m/s and the shell side uid velocity must be between 0.3 m/s and 1 m/s [16]. The tube count table used in both examples is present in Saunders [15]. The sets of the decision variables considered are shown in Table 2. Any positive excess area is accepted for the design. 4.1. Example 1 The rst example is based on the design problem discussed in Kara and Gu raras [10], described in Table 3.

1804 Table 2 Decision variable sets Variable Dt (m) Ds (m) Set

A.L.H. Costa, E.M. Queiroz / Applied Thermal Engineering 28 (2008) 17981805

L (m) Ntp Rbs

15.88 103, 19.05 103, 25.40 103, 31.75 103 0.203, 0.254, 0.305, 0.337, 0.387, 0.438, 0.489, 0.540, 0.591, 0.635, 0.686, 0.737, 0.787, 0.838, 0.889, 0.940, 0.991, 1.067, 1.143, 1.219, 1.295, 1.372, 1.448, 1.524, 1.600, 1.676, 1.753, 1.829, 1.981, 2.134, 2.286, 2.438, 2.591, 2.743, 2.896, 3.048 1.219, 1.829, 2.438, 3.048, 3.658, 4.877, 6.096, 7.315 1, 2, 4, 6 0.20, 0.25, . . . , 0.95, 1.00

ness of 1.65 mm [15], the tube side uid velocity would be 0.64 m/s. This relatively low value is susceptible to fouling problems. These factors stress the importance to include, together with the fundamental constraints of minimum surface area and maximum pressure drop, additional constraints to guarantee a broader feasibility for the obtained exchanger. Since these constraints are inserted in the proposed algorithm, the design alternative found do not present the mentioned deciencies: the tube length to shell diameter ratio is 6.96 and the tube side uid velocity is 1.1 m/s, according to the considered range. 4.2. Example 2

Table 3 Example 1: Problem data Hot stream Fluid Mass ow rate (kg/s) Inlet temperature (C) Outlet temperature (C) Fouling factor (m2 K/W) Allowable pressure drop (kPa) Stream allocation Water 13.88 67 0.000176 12 Tube side Cold stream Water 8.33 17 40 0.000176 12 Shell side

The basic design parameters selected to solve the problem correspond to a xed tubesheet type exchanger with E-shell, single segmental baes, tube thickness of 1.65 mm and triangular tube pattern with a pitch ratio of 1.25. The adopted values of the physical properties are depicted in Table 4. The optimal solution of the problem is found by the algorithm through 84 rating runs. This solution and the alternative described in the original paper are shown in Table 5. Despite eventual dierences in the problem data and rating equations, the analysis of both results allows to state some remarks. The optimal exchanger reported in [10] presents a smaller area when compared with the proposed algorithm, but this alternative presents a tube length to shell diameter ratio of 2.3, lower than the minimum recommended. This aspect may bring problems related to ow distribution. Additionally, assuming a typical tube thickTable 4 Example 1: Physical properties values Hot stream Density (kg/m ) Heat capacity (J/(kg K)) Viscosity (Pa s) Thermal conductivity (W/(m K))
3

The data of the second example are displayed in Table 6, representing a methanol cooler. The basic design parameters employed are equivalent to the previous problem. The stream physical properties are presented in Table 7. This problem is solved in a series of ve variant algorithms, in order to analyze the role of the dierent algorithmic components: (1) full version algorithm, as utilized in Example 1, (2) the procedure of contraction of the decision variable sets due to the variable bounds is not included in the algorithm, (3) the feasibility tests 1 and 2 are not utilized to accelerate the search, (4) the fathoming procedure is not applied and (5) the utilization of a premature stopping criterion where the search ends when a rst feasible design alternative is found. All test cases reach the same solution, indicated in Table 8. The number of rating runs demanded in each test case is shown in Table 9. The comparison of Variant 1 and Variants 24 shows that the auxiliary procedures involving the variable bounds, feasibility tests and fathoming allow a considerable reduction in the number of rating runs (considering

Table 6 Example 2: Problem data Hot stream Fluid Mass ow rate (kg/s) Inlet temperature (C) Outlet temperature (C) Fouling factor (m2 K/W) Allowable pressure drop (kPa) Stream allocation Methanol 30.0 90 40 0.00015 10 Shell side Cold stream Cooling water 101.4 30 40 0.00015 25 Tube side

Cold stream 996 4178 0.827 103 0.615

983 4185 0.465 103 0.654

Table 7 Example 2: Physical properties values Table 5 Example 1: Optimal heat exchangers Area (m2) Algorithm Literature 24.63 19.60 Dt (m) 31.75 103 15.88 103 Ds (m) 0.438 0.489 L (m) 3.048 1.110 Ntp 4 2 Rbs 0.25 0.40 Hot stream Density (kg/m3) Heat capacity (J/(kg K)) Viscosity (Pa s) Thermal conductivity (W/(m K)) 750 2840 0.34 103 0.190 Cold stream 995 4200 0.80 103 0.590

A.L.H. Costa, E.M. Queiroz / Applied Thermal Engineering 28 (2008) 17981805 Table 8 Example 2: Optimal heat exchanger Area (m2) Algorithm 227.4 Dt (m) 15.87 10
3

1805

Acknowledgement
Ds (m) 0.591 L (m) 6.096 Ntp 1 Rbs 1.00

Eduardo M. Queiroz thanks CNPq for the nancial support. References

Table 9 Example 2: Number of rating runs Variant Rating runs 1 203 2 750 3 350 4 445 5 53

only the number of elements in the decision variable sets, there are 78,336 possible design alternatives). The level of reduction associated to each algorithmic component is variable, as can be seen in Table 9. However, this aspect is dependent of the nature of the example considered. The utilization of a premature stopping criterion in Variant 5 allows a sensible decrease of the computational eort when compared with Variant 1, without prejudicial eect in the quality of the solution. According to the logic of the algorithm, in this alternative stopping criterion, since a feasible conguration is found, no more investigations are conducted in larger shells and tube diameters. Despite heuristic rules that relate better designs with smaller shells and tube diameters, in certain specic situations, this stopping criterion may imply in a suboptimal solution. 5. Conclusions This paper approaches the optimization of the design of shell-and-tube heat exchangers. The formulation of the problem seeks the minimization of the thermal surface of the equipment, for certain minimum excess area and maximum pressure drops, considering discrete decision variables. Important additional constraints, usually ignored in previous optimization schemes, are included in order to approximate the solution to the design practice. The optimization algorithm applied to the formulated problem involves a tube count table search based on a controlled path along the decision variable space. The denition of variable bounds, feasibility tests and fathoming procedures allow a sensible reduction of computational costs. The algorithm can be associated to any desired rating code for the necessary thermal and hydraulic evaluations.

[1] F.O. Jegede, G.T. Polley, Optimum heat exchanger design, Transactions of the Institution of Chemical Engineers 70 (Part A) (1992) 133 141. [2] M. Reppich, S. Zagermann, A new design method for segmentally baed heat exchangers, Computers and Chemical Engineering 19 (Suppl.) (1995) S137S142. [3] T.K. Poddar, G.T. Polley, Heat exchanger design through parameter plotting, Transactions of the Institution of Chemical Engineers 74 (Part A) (1996) 849852. [4] K. Muralikrishna, U.V. Shenoy, Heat exchanger design targets for minimum area and cost, Transactions of the Institution of Chemical Engineers 78 (Part A) (2000) 161167. [5] P.D. Chaudhuri, U.M. Diwekar, J.S. Logsdon, An automated approach for the optimal design of heat exchangers, Industrial and Engineering Chemistry Research 36 (1997) 36853693. [6] M.C. Tayal, Y. Fu, U.M. Diwekar, Optimal design of heat exchangers: A genetic algorithm framework, Industrial and Engineering Chemistry Research 38 (1999) 456467. [7] R. Selbas, O. Kizilkan, M. Reppich, A new design approach for shelland-tube heat exchanger using genetic algorithms from economic point of view, Chemical Engineering and Processing 45 (2006) 268 275. [8] F.T. Mizutani, F.L.P. Pessoa, E.M. Queiroz, S. Hauan, I.E. Grossmann, Mathematical programming model for heat-exchanger network synthesis including detailed heat-exchanger designs. 1. Shell-and-tube heat exchanger design, Industrial and Engineering Chemistry Research 42 (2003) 40094018. [9] M.A.S.S. Ravagnani, A.P. da Silva, A.L. Andrade, Detailed equipment in heat exchanger networks synthesis and optimization, Applied Thermal Engineering 23 (2003) 141151. [10] Y.A. Kara, O. Guraras, A computer program for designing of shell and-tube heat exchangers, Applied Thermal Engineering 24 (2004) 17971805. [11] K.J. Bell, On the pessimization of heat exchangers, Heat Transfer Engineering 21 (2000) 12. [12] A.E. Jones, Thermal design of the shell-and-tube, Chemical Engineering (2002) 6065. [13] E.U. Schlunder (Ed.), Heat Exchanger Design Handbook, vol. 3, Hemisphere Publishing Corporation, 1986. [14] F.P. Incropera, D.P. DeWitt, Fundamentals of Heat and Mass Transfer, fth ed., John Wiley & Sons, 2002. [15] E.A.D. Saunders, Heat Exchangers: Selection, Design & Construction, Longman Scientic & Technical, 1988. [16] J.M. Coulson, J.F. Richardson, R.K. Sinnott, Chemical Engineering, An Introduction to Chemical Engineering Design, Pergamon Press, 1989.