You are on page 1of 8

Available online at www.sciencedirect.

com

Chemical Engineering and Processing 47 (2008) 906913

Design and optimization of multipass heat exchangers


J.M. Ponce-Ortega a,b , M. Serna-Gonz lez a , A. Jim nez-Guti rrez b, a e e
a

Facultad de Ingeniera Qumica, Universidad Michoacana de San Nicol s de Hidalgo, Morelia, Mexico a b Departamento de Ingeniera Qumica, Instituto Tecnol gico de Celaya, Celaya, Mexico o Received 28 April 2006; received in revised form 11 February 2007; accepted 12 February 2007 Available online 23 February 2007

Abstract In this paper, a simple algorithm is developed for the design and economic optimization of multiple-pass 12 shell-and-tube heat exchangers in series. The design model is formulated using the FT design method and inequality constraints that ensure feasible and practical heat exchangers. Simple expressions are obtained for the minimum real (non-integer) number of 12 shells in series. A graphical method is also presented to develop some insight into the nature of the optimization problem. The proposed algorithm enables engineers to design optimum multipass heat exchangers quickly and easily. It is also shown how the method can be applied for optimal design of multipass process utility exchangers. 2007 Published by Elsevier B.V.
Keywords: Multipass heat exchangers; Design; Optimization

1. Introduction Multiple-pass shell-and-tube heat exchangers allow thermal expansion and easy mechanical cleaning, as well as longer owpaths for a given exchanger length. In addition, the high velocities achieved for the tube uid help increase heat transfer coefcients and reduce surface fouling [1]. The design of multipass exchanger involves the determination of the heat transfer area and number of shells in series for a specied heat duty. A typical solution to this problem involves a trial-and-error graphical method to determine the number of shells in series for values of the FT correction factor equal to or greater than 0.75 (see for instance, Kern [2] and Ahmad et al. [3]). The size of the exchanger is then found from the basic design equation. This graphical method is often tedious and time consuming. Furthermore, the FT correction factor may be difcult to compute, particularly in the steep regions of the FT charts. Fakheri [4] has presented some explicit expressions that avoid the difculties associated with the use of the FT charts. Ahmad et al. [3] introduced a new parameter, Xp , and derived simple equations to calculate the number of shells in series explicitly, in terms of the heat capacity ratio, R, the heat exchanger thermal effectiveness, P, and the parameter Xp . Thus,

Corresponding author. Tel.: +52 461 611 7575x139; fax: +52 461 611 7744. E-mail address: arturo@iqcelaya.itc.mx (A. Jim nez-Guti rrez). e e

if the values of Xp and the terminal temperatures are specied, these expressions allow a more straightforward solution for multiple-pass heat exchangers problems than the graphical method. In addition, these equations are useful for targeting and synthesis of multipass heat exchanger networks. As stated by Ahmad et al. [3] the Xp parameter must be specied by the designer such that 0 < Xp < 1. Since the Xp value of 0.9 guarantees that FT 0.75 almost over the full range of R, Ahmad et al. [3] suggest the general use of that value to simplify calculations. As an alternative approach, Shenoy [5] reported some equations to estimate Xp for different values of R. However, these approximations can predict values of Xp that yield unfeasible exchangers. To avoid this limitation, Moita et al. [6] attempted to quantify a trade-off between large heat-transfer area at high Xp and large number of shells in series at low Xp values, through the minimization of the investment cost of the exchanger. As a constraint, a lower bound on the FT correction factor of 0.75 was used. Similarly and based on the logical requirement that Ns must be integer, Ponce-Ortega et al. [7] proposed a mixed integer non-linear programming formulation for this optimization problem. As an additional constraint in that approach, an upper value for the area per shell was specied. In this paper, we present an optimization method for the design of multiple-pass shell-and-tube heat exchangers. The method is simpler and more efcient than previous algorithms. Simple analytical expressions for minimum real (non-integer) number of 12 shells in series are derived. A simple design

0255-2701/$ see front matter 2007 Published by Elsevier B.V. doi:10.1016/j.cep.2007.02.004

J.M. Ponce-Ortega et al. / Chemical Engineering and Processing 47 (2008) 906913 Table 1 Basic design equations Description Energy balance hot uid Energy balance cold uid Design equation Thermal effectiveness Heat capacity ratio log-mean temperature log-mean temperature FT correction factor Effective P1,2 FT correction factor Effective P1,2 Equations Q = CPh (Thi Tho ) Q = CPc (Tco Tci ) Q = UAFT TLM Tco Tci P= Thi Tci Thi Tho CPc = R= CPh Tco Tci (Thi Tco ) (Tho Tci ) TLM = for R = 1 ln((Thi Tco )/(Tho Tci )) TLM = Thi Tco = Tho Tci for R = 1 R2 + 1 ln[(1 P1,2 )/(1 RP1,2 )] FT = (R 1) ln ((2 P1,2 (R + 1 R2 + 1))/(2 P1,2 (R + 1 + R2 + 1))) 1/Ns 1/Ns 1 1 RP 1 RP P1,2 = 1 R for R = 1 1P 1P 2P1,2 FT = for R = 1 (1 P1,2 ) ln ((2 P1,2 (2 2))/(2 P1,2 (2 + 2))) P P1,2 = for R = 1 P Ns P + Ns

907

(1) (2) (3) (4) (5) (6) (7) for R = 1 (8) (9) (10) (11)

algorithm is developed taking advantage of the fact that Ns can only take integer values. It should be noted that the design model uses directly the FT correction factor instead of the Xp parameter. First, we consider the problem in which the heat capacity ow rates of both uids are known and one wishes to nd the optimal number of shells in series. Then, we consider the problem of nding both the optimal number of shells in series and the optimal mass ow rate of a process utility exchanger for a given heat duty. 2. Mathematical model and constraints In the design of multiple-pass shell-and-tube heat exchangers, the major items are the total heat transfer area and the number of shells in series. The basic equations for such problems have been reported elsewhere [2,8], and are summarized in Table 1. An examination of Eq. (8) from Table 1 shows that when P1,2 = 2/(R + 1 + (R2 + 1)1/2 ), a division by zero will

occur in the logarithm term of the denominator. Also, when 2 P1,2 (R + 1 + (R2 + 1)1/2 ) < 0, the argument of the logarithm will be negative. Thus, these conditions impose limitations on the values one can assign to P1,2 . Specically, for any value of R, the maximum theoretically attainable value of P1,2 is given by (see Ahmad et al. [3] and Smith [1]): Pmax = 2 1 + R + R2 + 1 (12)

Fig. 1 shows the relationship between the FT correction factor and P1,2 for different values of R. It shows how FT is a monotonically decreasing function of P1,2 . Note that, in the limit, where P1,2 Pmax , the correction factor FT approaches zero. It can be seen from Eq. (3) that, in this limiting case, the area of the heat exchanger is innite. It can also be seen from Fig. 1 that FT = 1 as P1,2 0. Thus, the minimum area requirement A = Amin = Q/(U TLM ) is obtained when P1,2 0. It follows from this analysis that the region, where Eq. (8) has positive

Fig. 1. FT correction factor for shell-and-tube heat exchangers as a function of P1,2 for different values of R.

908

J.M. Ponce-Ortega et al. / Chemical Engineering and Processing 47 (2008) 906913

Therefore, we conclude that the physically allowable range of the number of 12 shells in series, Ns , satises: Ns, min < Ns < (19)

Fig. 2. Relationship between R and P1,2 for R = 0.8333 and P = 0.878.

and, therefore, physical solutions for the FT correction factor is: 0 < P1,2 < Pmax (13)

Eqs. (12) and (13) also apply to Eq. (10) when R = 1. Also, for all real multiple-pass shell-and-tube heat exchangers the range of the heat-transfer area satises: U Q <A< TLM (14)

Eqs. (9) and (11) can be rewritten to yield expressions for the real (non-integer) number of 12 shells in series in terms of R, P and P1,2 (see Ahmad et al. [3]): Ns = and Ns = P 1P 1 P1,2 P1,2 , R=1 (16) ln[(1 RP)/(1 P)] , ln((1 RP1,2 )/(1 P1,2 )) R=1 (15)

As shown in Section 3, the specication of Ns is sufcient to x P1,2 and A. Thus, the bounds given by conditions (13) and (14) cannot be treated as independent constraints since only explicit bounds on Ns are necessary. The above analysis shows that an increase on the FT value leads to a decrease of the exchanger area A and P1,2 , which increases Ns . It should be noted that, for any value of R, the lowest value of FT corresponds to Ns = Ns,min . This value increases with the number of shells. Fig. 1 shows that there is a point FT = FT that divides the , the curves have curves into two regions. When 0 < FT < FT steep slopes. This implies that a heat exchanger design located in this region will not be able to handle uctuations in process conditions. On the other hand, in the region FT FT 1, the dependence of FT on the thermal effectiveness becomes very weak when P1,2 moves close to 0. Thus, a proper design should be such that the FT curves provide values of the correction factor above FT . It is convenient, therefore, to nd the point FT = FT that divides the graphical correlations of FT into two different regions. According to Shah and Skiepko [9] it is a common industrial practice to use a FT value of 0.8. Below this point, as may be seen in Fig. 1, FT decreases sharply because it approaches a vertical slope at given values of R and P1,2 . In addition, this value avoids the condition of temperature cross in each shell [2]. Thus, for designing feasible multiple-pass shell-and-tube heat exchangers, the following constraint must be satised (see Shah and Skiepko [9]): FT 0.8 3. Degrees of freedom and design algorithm The heat exchanger model has 15 variables and eight independent equations (Eqs. (1)(5), (6) or (7), (8) or (10) and (9) or (11), as needed). For design problems, the input parameters are the inlet and outlet temperatures of the hot (or cold) uid, the inlet temperature of the cold (or hot) uid, and the heat capacity ow rates of both uids. We assume that the overall heat transfer coefcient is also specied. Hence, there are nine remaining variables in the heat exchanger model [Q, Tco (or Tho ), A, TLM , P, R, FT , P1,2 and Ns ], providing one degree of freedom. We can rst notice that Eqs. (1), (2), (4), (5) and (6) or (7) allow a sequential calculation of Q, Tco , P, R, and TLM , respectively. Thus, these ve variables provide the rst set to be computed. The second subset is composed of three equations (Eqs. (3), (8) or (10) and (9) or (11), as appropriate) and four variables (A, FT , P1,2 , and Ns ). A or Ns can be used as the best design variable; Ns is selected by convenience because it can take only integer values. Such a selection provides a sequential solution for the remaining variables. (20)

For given values of R and P, the relation between Ns and P1,2 may be generated from Eqs. (15) or (16). Fig. 2 shows one such graph for R = 0.8333 and P = 0.878. One can notice that Ns /P1,2 < 0 (i.e. Ns decreases when P1,2 increases). One can also observe there are two asymptotes. The rst one is vertical, and it denes the limit of innite number of 12 shells needed at P1,2 0. The second asymptote is horizontal (Ns,min in Fig. 2) and is obtained when P1,2 Pmax . Ns,min represents the minimum real (non-integer) number of 12 shells in series theoretically needed for multiple-pass shell-and-tube heat exchangers. One can obtain similar results, for any values of R and P, for either limiting case of P1,2 Pmax or P1,2 0. The minimum asymptotic values of Ns are found from the combination of Eqs. (12), (15) or (16), and the condition P1,2 = Pmax :
Ns, min = ln((1 RP)/(1 P)) , ln((1 R + R2 + 1)/(R 1 + R2 + 1)) R=1

(17)

and Ns, min =

2 2

P 1P

R=1

(18)

J.M. Ponce-Ortega et al. / Chemical Engineering and Processing 47 (2008) 906913

909

It is important to notice that the design has to satisfy the constraints given by Eqs. (19) and (20). Thus, not all integer values of Ns yield feasible designs. However, one can also notice that Ns,min sets a lower limit on meaningful values of Ns . Consequently, the design algorithm can be simplied and initialized by choosing an integer value for Ns equal to or higher than Ns,min . Then, FT is calculated from Eqs. (8) or (10). A search over Ns is conducted until Eq. (20) is satised. Therefore, the following algorithm provides the initial feasible design for the optimization problem. (1) Set input parameters: inlet and outlet temperatures of hot uid, inlet temperature of cold uid, heat capacity ow rates of both the hot uid and the cold uid, and the overall heat transfer coefcient. (2) Calculate Q, Tco , P, R, and TLM from Eqs. (1), (2), (4), (5) and (6) or (7). (3) Calculate Ns,min from Eqs. (17) or (18). (4) Choose the smallest integer value for Ns satisfying Ns > Ns,min to ensure that the initial design specication is meaningful. (5) Calculate P1,2 from Eqs. (9) or (11) for the current value of Ns . (6) Find the FT correction factor from Eqs. (8) or (10). (7) Check whether the condition given by Eq. (20) is satised. If so, a feasible solution has been detected; calculate the heat exchanger area from Eq. (3). Otherwise, set Ns = Ns + 1 and go to step 5. 4. Formulation of the optimization problem The optimization problem uses the minimization of the investment cost as the objective function. The optimal design should satisfy constraints (19) and (20), with Ns integer. The investment cost for heat exchangers may be expressed as a function of the equipment area through a power law expression. For multiple-pass heat exchangers, the investment cost can be estimated by (see for instance Smith [1]): C = a + bNs A Ns
c

Fig. 3. The investment cost C as a function of the number of shells in series Ns for example 1. The asterisk denotes the location of the unconstrained optimum solution.

reect the economies of scale generally observed in chemical process equipment. It should be noted that for these fractional values of c this optimization process is different to that of minimizing the total area; the minimum area would be obtained in the limit when Ns . The solution to this optimization problem is rst shown through a graphical method. The idea is that the knowledge gained through this method will be helpful for the numerical solution of the optimization problem. 4.1. Graphical method Example 1, taken from Moita et al. [6] and shown in Table 2, is used to illustrate the graphical method. Fig. 3 shows the investment cost as a function of Ns . The curve was obtained by direct solution of the model equations, for several values of Ns . The constraint imposed by the correction factor FT is shown as a dashed line. The feasibility region lies on the right hand side of the FT constraint. For nonlinear cost functions with 0 < c < 1 (which reects the cases of practical interest), Eq. (21) has a unique global minimum in the feasible region given by condition (19). The following analysis shows this property. When Ns Ns,min , FT 0, and A as discussed in Section 2. Thus, in this limiting case, the objective function goes to innite. When Ns , the opposite occurs: FT 1, and A has the smallest value. Hence, in this limit, the function objective will also approach innity. There-

(21)

where a, b, and c are cost law constants which depend on construction materials, pressure rating and the type of exchanger. Values for c are typically between 0.6 and 0.9 [1,10,11], which
Table 2 Data for the examples Example Thi ( C) Tho ( C) Tci ( C) Tco ( C) Q (kW) U (kW/m2 K) a b c Ex1 410 110 0 360 2000 0.1 8600 670 0.83 Ex2 500 270 40 195 2000 0.1 0 7000 0.65 Ex3

Ex4 570 150 50 150 2000 0.1 0 7000 0.65

Ex5 570 150 0 200 2000 0.1 0 7000 0.65

Ex6 400 320 120 330 2,000 0.1 40,000 2,300 1.0

500 130 40 180 2000 0.1 0 7000 0.65

910

J.M. Ponce-Ortega et al. / Chemical Engineering and Processing 47 (2008) 906913

(1) Specify Thi , Tho , Tci , CPc , CPh , U, a, b, and c. (2) Use the design algorithm to generate an initial feasible design. (3) Compute the investment cost C associated with the initial design using Eq. (21). (4) Fix Ns = Ns + 1 and generate the corresponding feasible design. (5) Evaluate the cost function of the new design. If a better objective function is found, go to step 4; otherwise, the former design is the optimum solution. It can be noted that the algorithm is based on three basic features. First, the number of 12 shells in series must be integer. Second, the optimum integer solution must be at or to the right hand side of the FT constraint. And third, the objective function is unimodal. 4.3. Optimization of multipass coolers The analysis shown above is extended for optimization problems of multipass coolers (or heaters). In the previous analysis, the heat capacity ow rates of both uids are xed; now the problem is to nd the optimal mass ow rate or outlet temperature for the cold (hot) utility as well. Therefore, this problem is formulated so as to minimize the total annual cost (TAC): TAC = Af a + bNs A Ns
c

Fig. 4. Investment cost C as a function of the number of shells in series Ns for example 1. Dark circles indicate integer values of number of shells in series.

fore, a minimum investment cost must exist at some intermediate value of Ns . It should be noted that the mathematical behavior for linear cost functions is such that the objective function decreases monotonically with A, providing the trivial and not useful optimum solution at the limit A Amin for Ns (which is equivalent to the problem of minimizing the total area). To avoid this impractical result, it is needed in these cases to include an upper limit lower than 1 for the FT correction factor in the model formulation. It is apparent from Fig. 3 that the minimum investment cost is obtained by using a number of 12 shells in series of approximately 6.25. The minimum value of C is D 109,037. However, feasible multiple-pass shell-and-tube heat exchangers must provide FT correction factors greater than 0.8, with an integer number of 12 shells in series. Therefore, the optimization problem is to nd the best integer value of Ns , with the constraints given by Eqs. (19) and (20), that yields the minimum investment. This situation is shown in Fig. 4 for example 1, where circles indicate the acceptable integer values of Ns . The optimal integer solution is Ns = 6, for a minimum investment cost of D 109,064. In this case, restricting Ns to integer values has only increased the investment cost by D 27. In general, depending on the economic data and process conditions, the relaxed optimum solution may lie on either side of the FT constraint. However the best integer solution is always at or to the right hand side of the FT constraint. These observations allow the development of a simple numerical optimization algorithm. 4.2. Numerical algorithm The optimization algorithm is based on a single variable search, in this case over the number of shells in series. The design algorithm discussed above provides a starting feasible design. Then, the design process is simply repeated for higher integer values of Ns until a minimum investment cost is detected. The optimization procedure is summarized as follows:

+ HY Cu mu

(22)

where Af is the annualization factor for capital investment, mu is the owrate of the process utility with a unit cost Cu , and HY is the annual operating time. We outline below the analysis of a multipass cooler that uses cooling water as the process utility. The hot uid inlet and outlet temperatures as well as the heat duty are usually xed by the process, and the water inlet temperature Tci is determined by site conditions. Thus, the cooling water owrate can be calculated from: mu = Q CPu (Tco Tci ) (23)

This problem has two degrees of freedom, and two suitable optimization variables for this case are Ns and Tco . To minimize the total annual cost, an iterative procedure is used. The basic idea of this algorithm is to choose the water outlet temperature, Tco , and calculate the corresponding optimal value of Ns using the single-variable optimization algorithm of Section 4.2; then, for the assumed value of Tco , the cooling water ow rate and the total annual cost can be determined from Eqs. (23) and (22), respectively. This procedure can be applied for varying water outlet temperatures to determine the minimum total annual cost. As shown in Fig. 5, for nonlinear cost functions with 0 < c < 1, the total annual cost is unimodal with respect to the water outlet temperature. We use the golden section method to determine the optimal value of Tco . Many designers [2] limit the maximum water outlet temperature to 49 C because of the rapid deposit formation of mineral salts over the heat transfer surface at greater temper-

J.M. Ponce-Ortega et al. / Chemical Engineering and Processing 47 (2008) 906913 Table 3 Results obtained with the proposed optimization algorithm Example Ex1 Ex2 Ex3 Ex4 Ex5-a Ex5-b Ex6 Ns 6 1 2 1 2 1 2 Xp 0.717 0.720 0.719 0.915 0.6736 0.9520 0.646 FT 0.907 0.909 0.929 0.814 0.937 0.661 0.952 A (m2 ) 289.888 82.807 118.747 110.159 87.571 124.205 169.683

911

C (kD ) 109.06 123.55 199.05 148.73 163.30 160.80 430.27

Fig. 5. Finding the optimum outlet temperature for cooling water in multipass coolers for 0 < c < 1. Each point implies a local optimum solution for the optimum number of shells (solution to example of Table 4).

atures. We use this value as a constraint. A similar application of the algorithm can be used for heaters with no phase change. 5. Illustrative examples Example 1 is now used to show the numerical applications of the optimization algorithm. 5.1. Initial design The constraint for the correction factor is taken as FT 0.8. The thermal effectiveness is 0.878, and the heat capacity ratio is 0.8333, from Eqs. (4) and (5). The minimum real (non-integer) number of 12 shells in series may be obtained by solving Eq. (17) as Ns,min = 3.062. Thus, the design algorithm begins with Ns = 4. For this value, the correction factor is found from Eq. (8) to be 0.7594. Therefore, the number of shells in series must be increased by 1. When Ns = 5, the correction factor provides the feasible value of 0.8599. From Eqs. (3) and (21), the corresponding heat exchange area and cost values are A = 305.7 m2 , and C = D 110,375. 5.2. Optimization We now increase the initial integer value of the number of shells in series to search for a better design; thus, for Ns = 6 the following results are obtained: FT = 0.9066, A = 289.9 m2 , and C = D 109,064. Since C is lower than the initial design, the search continues. For Ns = 7: FT = 0.9329, A = 281.7 m2 , and C = D 109,312. C is greater than the previous solution; therefore, the optimum number of shells in series is 6, with an objective function value of D 109,064. Table 3 shows the results obtained for this example. Moita et al. [6] reported an optimum number of shells in series of 5, with an investment cost value of D 110,375. We have shown that this is not the optimum solution to this problem. Although the difference in the objective function value is small, it can be established for theoretical purposes that the optimization algorithm by Moita et al. [6] fails to provide the global optimum solution. The algorithm of this work provides therefore a better (and simpler) design tool.

A sensitivity analysis with respect to the change in the number of shells in series can be easily conducted. For this example, the investment cost C is within 1.2% of the optimum so long as Ns lies between 5 and 7. Therefore, we can notice that the C versus Ns curve is fairly at in the neighborhood of the minimum investment cost. Although the C values are very close for those two values of Ns , it is observed that the heat-transfer area required is rather different (5.45% higher for Ns = 5 and 2.83% lower for Ns = 7). The results to the examples Ex2Ex4 shown in Table 2, also discussed by Moita et al. [6], are given in Table 3. The results are identical to the ones obtained by Moita et al. [6], but with the use of a simpler and easier to implement optimization method. For examples Ex1Ex4, one can notice that the implementation of the constraint FT 0.8 does not affect the optimal solution (see Table 3). This is because the FT constraint lies, in all cases, at the left hand side of the optimal constrained designs. For example Ex5, however, the optimization procedure gives different solutions for the constrained and unconstrained problems. These solutions are given in Table 3, where Ex5-a and Ex5-b represent the solutions for the constrained and unconstrained optimization problem, respectively. The results show that case Ex5-b has one less shell, which gives a 1.55% saving in the capital cost and a FT value of 0.661. Note, therefore, that if the optimization problem is solved without the constraint on the FT value, the solution would probably be close in economic terms, but would not meet practical design criteria. Example 6, taken from Moita et al. [6], serves also to illustrate the application of the model to a case with a linear cost function. In this case, the problem is unbounded with the optimal solution A = Amin when Ns = , as described above. The reported solution to this problem was obtained by setting the limit for FT,max of 0.96 (which corresponds to Xp,min = 0.6), as suggested by Moita et al. [6] The model therefore nds the closest solution to this limit as the optimal solution, which corresponds to 2 shells with an FT factor of 0.952 (see Table 3). To illustrate the application of the optimization procedure of multipass coolers, we consider the problem data given in Table 4; the basic problem consists of cooling a stream from 93 to 32 C using cooling water at an assumed temperature of 29 C. The outlet cooling water temperature and the number of shells were used as the search variables. The results from the application of the algorithm are reported in Table 5. The search for the optimal solution is displayed in Fig. 5. It should be noticed that each search point reects a local optimum solution for the optimal number of shells, obtained through a direct application

912 Table 4 Data for the cooler problem Variable Thi Tho ( C) Tci ( C) Q (kW) U (kW/m2 K) a b c Af (1/y) Cu (D /kg) HY (h/y) ( C)

J.M. Ponce-Ortega et al. / Chemical Engineering and Processing 47 (2008) 906913

Acknowledgement
Value 93 32 29 348.167 0.0443 0 5.405 0.99 1 3.97 107 8400

Partial nancial support from the Consejo de Investigaci n o Cientca of the UMSNH, M xico, is gratefully acknowledged. e Appendix A. Nomenclature

Table 5 Optimal solutioncooler design Variable Tco A (m2 ) FT Ns mu (kg/h) TAC (D /y) ( C) Value 35.29 427.11 0.9947 6 55,345 2396.7

a, b, c A Af C CP CPu Cu FT
FT

of the original design algorithm for cases with one degree of freedom. The optimal values for the search variables are an outlet temperature of cooling water of 35.3 C and a number of shells of 6, which provide a total annual cost of 2396 D /yr, based on the assumed economic environment. It can be noted that, as in the cases of problems with process streams, the design algorithm for utility process exchangers with no phase changes is very effective, showing no convergence problems. 6. Conclusions The problem of designing and optimizing 12 shell-and-tube heat exchangers in series has been addressed, and a simple design and optimization algorithm has been developed based on the FT correction factor model. The optimization algorithm uses the number of shells as a convenient search variable to provide an effective design method. The procedure guarantees the global constrained optimum solution since the objective function is unimodal with respect to the number of shells in series when the coefcient c for the cost function take values between 0 and 1. A graphical method that allows a proper insight into the optimization problem has also been presented. In addition, it has been shown how a lower bound for the number of shells in series Ns as a function of R can be used to provide an initial feasible design quickly. We have also shown how the design method can be used for the optimization of multipass utility exchangers, coolers or heaters, without phase change. Because the proposed algorithms are simple and reliable, they can be used within the solution of larger industrial problems, such as the synthesis of multipass heat exchanger networks and cooling water networks.

HY mu Ns P P1,2 Q R T TAC TLM U XP

constants of heat exchanger cost function heat exchanger area (m2 ) annualization factor (1/y) heat exchanger capital cost (D ) heat capacity ow rate (kW/ C) heat capacity for utility (W/kg C) unitary cost for utility (D /kg) correction factor for logarithmic mean temperature difference minimum allowed correction factor for logarithmic mean temperature difference operation hours per year (h/y) utility mass ow rate (kg/h) number of 12 shells in series heat exchanger thermal effectiveness thermal effectiveness of each 12 shell in the series heat exchanger duty (kW) heat capacity ratio stream temperature ( C) total annual cost (D /y) logarithm mean temperature difference ( C) overall heat exchanger transfer coefcient (kW/m2 K) Ahmads parameter, dimensionless

Subscripts c cold stream h hot stream i inlet max maximum o outlet References
[1] R. Smith, Chemical Process Design, John Wiley & Sons, England, 2005. [2] D.Q. Kern, Process Heat Transfer, McGraw Hill, USA, 1950. [3] S. Ahmad, B. Linnhoff, R. Smith, Design of multipass heat exchangers: an alternative approach, ASME J. Heat Trans. 110 (1988) 304309. [4] A. Fakheri, An alternative approach for determining the correction factor and the number of shells in shell and tube heat exchangers, J. Enhanc. Heat Trans. 10 (4) (2003) 407420. [5] U.V. Shenoy, Heat Exchanger Network SynthesisProcess Optimization by Energy and Resources Analysis, Gulf Publishing Company, USA, 1995. [6] R.D. Moita, C. Fernandes, H.A. Matos, C.P. Nunes, A cost-based strategy to design multiple shell and tube heat exchangers, ASME J. Heat Trans. 126 (2004) 119130. [7] J.M. Ponce-Ortega, M. Serna-Gonz lez, L.I. Salcedo-Estrada, A. Jim neza e Guti rrez, Minimum-investment design of multiple shell and tube heat e exchangers using a MINLP formulation, Chem. Eng. Res. Des. 84 (A10) (2006) 905910.

J.M. Ponce-Ortega et al. / Chemical Engineering and Processing 47 (2008) 906913 [8] R.A. Bowman, Mean temperature difference correction in multipass exchangers, Ind. Eng. Chem. 28 (5) (1936) 541544. [9] K. Shah, T. Skiepko, Entropy generation extrema and their relationship with heat exchanger effectivenessnumber of transfer unit behavior for complex ow arrangements, J. Heat Trans. 126 (2004) 9941002.

913

[10] L.T. Biegler, I.E. Grossmmann, A.W. Westerberg, Systematic Methods of Chemical Process Design, Prentice Hall, 1997. [11] M.S. Peters, D.K. Timmerhaus, R.E. West, Plant Design and Economics for Chemical Engineers, McGraw Hill, 2003.

You might also like