You are on page 1of 9
THE EFFECT OF LEAKAGE THROUGH THE LONGITUDINAL BAFFLE ON THE PERFORMANCE OF TWO-PASS SHELL EXCHANGERS: Mothematical analysis wos developed which permits the pre the effects of leakage through the longitudinol baffle in a two-pass shell iction of T. Rozenman: and’ J. Taborek. exchanger (IEMA F-type). The leckage lowers the thermal effectiveness of the exchanger but also lowers the pressure drop. The effec leakage is expressed by two dimensionless numbers, combine all the with the assumption for the clearance between the Graphs showing the variation of flow, pressure drop, torms of these dimensionless numbers. shell ond the tion factors aro given in mote, o dimensionless plot showing the effect of conducti t of the Ro and Lk, which known operation and design porameters of the system e baffle. ‘and F correc: Further: ion through the botfle on the performance of the exchanger is given. A practical ex: ‘omple of reting an actual exchanger illustrates the use of the given here to analyze the problem of leckage in @ twor exchanger. In many cases, the use of two-pass exchangers (TEMA F-type) seems attractive to the process engineer. Two- pass shell exchangers afford true countercurrent flow and thus eliminate incfficiencies due to temperature profile overlap. Also, due to doubling of the shell-side velocity, higher heat transfer rates can be obtained. This is particularly important in case of unequal shell-and- tube flow rates. ‘The two-pass shell exchanger requires a longitudinal baffle which separates the two shell streams. Sometimes this baffle is welded to the shell. In such case no fluid Jeaks from one side of the shell to the other. However, heat is transferred from the hot side of the bafile to the cold side, thus reducing the overall efficiency of the ex- changer (thermal leakage). In other cases, the center bar fle is removable with the tube bundle, and shell fluid might leak from one side to the other, decreasing further Heat Transfer Reseutch, Inc. Alhambra, California, method pass shell the efficiency of the exchanger. This physical leakage affects the performance of the exchanger in two w: |. Decreasing the thermal efficiency of the exchi ‘The leakage causes the hot fluid to mix with the col of the two-shell passes and also decreases the overall heal j transfer coefficient. 2. Reducing the ally flowing in the baff Whistler (1) showed the effect of conduction through the baffle in a 2-2 exchanger. He provided graphs of F correction factor as a function of temperature difference! in the exchanger. Untit wow; this was the only work pul lished, discussing inefficiences in twonyassshell ex changers. { In the following, we present an analytical investigatior of the physical teakage problem. The analysis is carried out for a 2-2 exchanger and is divided into two sections, In the firstsection, the factors that affect the pressure drop and the true resultant flow are determined. The sure drop, since less fluid is actu section of the shell. ‘ R No. 118, Vol. 68 main results are presented graphically in terms of dimen- sionless parameters. In the second section the analysisis extended to solve for the heat transfer efficiency. The re- sults of the first section are used to obtain the mean tem- perature difference correction factors in the usual dimen- sionless groups of Rand S. The method derived is illustrated by solving a practical problem. EFFECT OF LONGITUDINAL BAFFLE LEAKAGE ON SHELL-SIDE FLOW AND PRESSURE DROP In.a removable tube bundle two-pass shell heat ex => changer, the longitudinal bafMle forms a sliding joint with the shell. The leskage of fluid around the sliding zone must be considered. The leakage is caused by the pres- sure difference that exists between the top and bottom sides of the bafile. Figure | shows a schematic diagram of a 2-2 exchanger, 2-2 meaning two-pass in the shell and two-pass tubes. The shell fluid enters at one end, flows through the baffled top section, turas around at the other end, and continues to the exit nozzle whichis located -at— same end as the inlet nozzle. As can be-easily-seen—— ror Figure 1. the highest pressure difference betweemthe: two sides of the baffle is at the end where:the nozzles are located. At the other end of the shell; the pressure differ ence is obviously zero. Tn a nonleaking exchanger the flow through the shell side remains constant and is equal to the entering Now. — W,,.. However, in a leaking exchangersthe-full flowin the ~~ first passage decreases along the exchanger and-reaches~-= its lowest yalue W,aLthe turningsestion.ckaenonleak— ing exchanger the overall pressure deup-A#ivis calculated according Jo the sizcam analysis metlod-describodcbys— Palen and [aborek (2), In a leaking exchanger the actual pressure drop AP, is always smaller than AP... This be- havior is illus es ‘A mutual dependency exists between flow and pres- sure, The leakage reduces the flow and, in turn. affects - the pressure deopashich-results er 2 chimge tw the eae A system of equations witl be set-up that express the i ‘tion aimong pressure, Tow, and leakaz Figure 3 illustrates a material balance in the exchanger. shows that at any point X along the exchanger, the flow in the upper and lower sections is equal but opposite in direction. ROZENMAN and TABOREK 13 LEAKING EXCHANGER = NONLEAKING EXCHANGER 1 Spee 1 second pa 2, * aaa ang ie a MoM = H (eokage), “n-Fig.2-Comparison between a noaleaking and leaking exchanger {) Pressute 0p: (b) flow. Wir + Wie = Wa + Wie (la) Wien Wie (1b) We Tsim shat the difference between the pressure drop in the exchanger with leakage and without leakage is only due to the difference in the bulk flow through the ex- assumption can be expressed as @ _ [aB\ (Waa\* ak). \dX, re where (dPfaX).,.s pressure drop at the same point X of the leaking exchanger; (dP/4X),is the calculated pressure drop at the same point X with no leakage: Wa, is flow at any point X; W,, is nitial flow rate. n <2 ‘when the flow in the shell is turbulent and n = 1 when the flow in the shell is laminar. If the Mow is in the transition region, mis between 1 and 7) No Won Poe z Ns x o Fig. 3. Material balance inthe exchanger. 14 HEAT TRANSFER—TULSA 2. It.can be shown that at any point along the exchanger, the pressure difference betwen the two sides of the bafile can be expressed as follows: denoting Pi Ph AP, HAP.) dP) (Wa “7 (G), Ga) © The velocity of the leaking stream is defined by writing a momentum balance around a unit length of the clearance. ip, As Nae 2 # AP, Wa = (2g SP! ( 6 oe) where Vis the velocity of the leaking stream at a point X along the exchanger and X/ is the leakage coefficien: and is a function of Reynolds number based on the clear- ance dimensions and flow velocity through Denoting the clearance by a, the final expression for the leakage per unit length is /GONRPS a : (&) ‘We transform Equations (3) and (5) into a dimensionless form by using the foilowing substitution: dW x ® AP | Wee x ss eee 1 Keg Equations (3) and (5) are written in the dimensiontess form as follows: ae az (42) AP, < an dx _aVGeONER) me Waa/2L K 6) , suid ie a” The expression 2L(dP/dX),» is the total calculated pres sure of the nonleaking exchanger P. (excluding the nozzles), and Equation (6) can be written as am (222) ow ae @) AP, Equations (7) and (8) express the simultaneous variation of low and pressure difference along the exchanger. The boundary conditions are as follows: (Ga) (96) at ox=0,=1wel at ox=1,1-0 AIChE SYMPOSIUM SERIES Itcan be easily seen that Equation (7) includes the co- efficient K;, As stated previously, Kris strongly de- r pendent ona Reynolds number defined with the clear- | ance dimensions and the fiow velocity as follows (a0) Fa) # Ky = B+ 1(Ne) where 8, y and m are experimentally evaluated co- efficients, the coefficient + is strongly dependent on the parameter z = {/a, which expresses the ratio of the clear- ance to length of flow path of the leaking stream. The power m is changing for different flow regimes, For laminar flow (based on leakage Reynolds number Nqq)m = — 1; for transition and turbulent case, itis between 0 and —1, { The clearance acan be assumed to haveaconstant | value along the entire exchanger. However, Mj, varies from an initial value Vj, at ane end to zero at the other | end. Hence Ng,,, changes accordingly from some value | Nyqt0 zero. Thus K,is a function of x. We define now two dimensionless terms which com- bine all the operation and design variables of a nonleak- ing exchanger with the addition of the clearance a, which | is a parameter of a leaking exchanger. _ Vea PD W,/2L Neg Ro (2a x tke W,,/2L The Ro number expresses the ratio between the leakage flow and the total bulk flow in the exchanger. The leakage number Lk can be interpreted as the ratio be- tween the time that it takes a fluid element to flow in the bulk of the exchanger and the time that it takes it to leak instead through the baffle. Both these numbers Roand Lk express the two alternatives enforced on a uid ele- | ment in the exchanger under the influence of pressure | difference, namely, either flow in the bulk or leak through. By substituting the term (12a) into Equation (7), we get ~~ Ge) Ge)” sual | The Reynolds number of the leaking stream at each : point along the exchanger N,,,, is defined here as 4 Naa = Neg, « QUOC) a3 ie z ( “ley No. 118, Vol. 68 This Reynolds numbers related to the dimensionless terms defined in Equation (12) according to the following expression: Ro (AP,\"" (1t\" New, = < = 15) Wir LK Ge) () a5) As noted before, K; is dependent on N,,,, by the em- pirical formula (106). Equations (13) and (8), together with the boundary conditions, are written as aml AP, ie a (wy 8) aw AP (TLV? em eeygy at x=0,1=1 and w (9a) at x=10,0-0 (9b) In urder to facilitate the numerical technique used to solve these equations, condition (96) is transformed as follows: : BEN A atl = — (2) dx a6) fa = (3) JL ooras (7) > as Equations (8) and (13) are integrated numerically. The integration is carried out by a simple finite-difference equation since the flow in a baffled exchanger is across the bundle, For each pair of Lk and Ro the correct value TTT I te resus mare, 1 SRE Fig. 4. Leakage as a function of dimensionless parameters ¥ ROZENMAN and TABOREK 1s mesa oe i, Fig. 5. Pressure drop ina leaking exchanger. of (AP./AP,) is found from condition (18) through an iterative procedure. Ateach point along the exchanger the coefficient K, is calculated from Equations (10) and as). Equations (8) and (16) are solved for three cases: turbulent flow in the shell, assuming n = 2 and lam- inar flow in the shell m = 1. A third case was taken as n= 1.5 assuming to represent some average power for the transition case. Figures 4 and 5 show the results for In Figure 4 the ratio between the flow atthe turning end of the exchanger to the total flow 1, / Waais given asa function of the dimensionless number Ro with Lk = (u)(QL)/W,,a5a parameter. Figure 5 gives the ratio between the true pressure drop in the leaking ex- changer AP, and the calculated pressure drop 4P. for no leakage conditions as a function of the same parameters. The conditions of a leaking exchanger are readily eval- uated from these graphs. Both the dimensionless num- bers Ro and Lk can be calculated from the design and operating parameters of a nonleaking exchanger, assum- ing a value for the clearance between shell and bat. The value W/W - | — W,/1¥,, expresses the fraction of the original shell-side flow that leaked through the baffle. Figures 4 a yw that an increase in the n=l ask rere wee, Hj ee eee # \ oa vastus puweD. B+ Fig. Effect of low conditions in the shell on the leakage, 16 HEAT TRANSFER—TULSA os Fa xy i : Fig. 7, Eifect of flaw conditionsin the shell on pressure drop. “clearance, density, pressure drop, and length of baffle will increase the relative icakage, whereas higher shell-side flow and higher viscosity of shell fluid will decrease the ‘relative leakage: Figures 6 and 7 compa: the effect of flow regime in the shell on the leakage through the baffle for a typical value of Lk = 10°. Ivis clearly seen that tes eiees subgigibte aciow Kotworliggs butte io purnbers the laminar ease produces the highest leaking flow; The results thus presented are independent of the heat transfer (assuming no effect of temperature on properly averaged physical properties). They will be used to determine the effect of the leakage on the tem- ae ase profs tra heat irancler proce EFFECTS OF LEAKAGE ON HEAT TRANSFER Figure 8s a schematic diagram of the exchanger. ‘Temperatures in the upper section are denoted by @ prime, and temperatures in the lower section are denoted with two primes. A differential heat balance is written, and the overall heat balance is solved by integration. The equations are-written under the following assumptions: (1) no boiling or condensation occurs in the shell; (2) 70 heat conduction through the baffle; (3) temperature of shell fluid is uniform over any cross section; (4) equal ‘number of tubes in each pass; (5) 4 single exchanger is involved. AIChE SYMPOSIUM SERIES Heat balance in the upper section: W.C,aT' = W,C.dt (19) Rate of heat transfer in upper section: -wiCat = UM ax(T’ 1) 20) 2 Heat balance in the lower section: W..C,dT" + MC,dX(T’ - T") = WC,de" (21) aw, Mo ae Rate of transfer in lower section: W,C, dt" = ue aX(T” -1") (22) where ng is the heat transfer area per unit length of the exchanger (n, = 4/L)and W,, C,are the tubeside flow rate and specific heat, ‘The overall heat transfer balance for the exchanger is expressed as follows: (WAP) ST om = C(t — 4) where Fis an-overall thermal correction factor which in- corporates also the vatiation of the heat transfer coef- ficient along the exchanger asa result of the leakage. U. is calculated average heat transfer coefficient for no leak- age conditions. ‘Equations (19) to (23) are transformed to 2 dimension- less form with the terms R and S (see Notation). The NTU = (U,A/W,C,) of the exchanger is expressed in terms of R, S, and F. In ( I F(R—1) (23) NTU = Using the dimensionless variables defined in the previous section, one can Write the equations as follows. Heat balance in upper section: Fig. 8, Temperature designation in the two-pass shell exchanger. ea Heat transfer in upper section: WC, dt! ee Qs Heat balance in lower section: ar _ dw an = -T)=R 6 Ve a TE @ Heat transfer in lower section: BiG GE pe age i UA dx No. 118, Vol. 68 The boundary conditions are For the upper section: at x-0,T= 7 and & . ARTO (7, - 9) (28a) For the lower section: at x= 0,7" = 7; and aT” ‘dw © (E)_- EET (7, — 0) (286) The two sections have a common boundary condition at x1 rem The temperatures in the boundary conditions (28a) and (286) are made dimensionless with R, S, and (7) ~ 1). The temperature of tube Muids t' and 4” can be elim- ced from Equations (24) to (28) by taking cross acavatives. The resultant two equations can be solved simultaneously to yield 7’ and 7” as a function of x. The values of w(x) and its derivatives needed for this integra- tion are stored in the computer from the previous calcula- tion of the leakage. The correct value of the (NTU) term or the F factor is found by an iteration scheme that matches T’ and T" atx = 1.0, The calculations are aarried out for wo case 1. Shell-side resistance is controlling... The tube side resistance is negligible. For this case U(x) = Ugw x)? U(x) is the local overall heat transfer coefficient and varies along the exchanger with the 0.7 power of the local flow. 2. Tube-side contealling. For this case the leakage has no effect on overall heat transfer coefficient, and we assume that U = U, = constant Fig. 9. Feotrection factor for 20% leakage ROZENMAN and TABOREK 7 om ses rt ats = Sa Fig. 10. NTU versus effectiveness for 20% leakage. Figure 9 presents the F factor as a function of Rand S for 20% overall leakage W,/W,» = 0.8. For any vaiue of overall leakage, we can construct similar graphs to Figure 9. The 20%, leakage graph is chosen for illustra- tion purposes. It is seen that when sheil side is con- trolling, Fis less than | even for § = 0.0, This is due to a decrease in the calculated coefficient U due to the leak- age. Figure 10 shows the reduction in effectiveness as a result of the leakage. The graphs of this section together with the results of the previous section make the leakage problem in an F shell amenable to a solution. First, the exchanger is calculated assuming no leakage conditions. From the design values, two dimensionless numbers are computed, Roand Lk. Then, from Figures 4 and 5 the correspond- ing values of the true pressure drop in the exchanger and the amount of leakage through the baffle are evaluated. Knowing W,/ W,, enables the prediction of the reduced effectiveness of the leaking exchanger through graphs similar to the one given in this section. ILLUSTRATION OF THE METHOD Before solving an example problem, the results of Whistler (/) for the effect of conduction through the longitudinal plate will be brielly reviewed. The graphs given in Whistler's original paper were ina dimensional form in terms of temperatures. Taborek (3) modified Whistler's results to a dimensioniess form that made them 18 HEAT TRANSFER—TULSA ae ef ee ‘ eke Pest f b Ba ort Co (Cool Fe. Denson prac easier to use. The modified plot of Whistler's original graph is shown in Figure 11. The MTD correction factor Fis given in terms of nondimensional parameter. The abscissa includes the factor a which expresses the ratio between the product of the surface area of the baffle Ay and the coefficient of heat transfer through it A, to the product (.1)(U,) of the exchanger. The value of f, con- sists of the convection heat transfer on the two sides of the bafile plus a conduction term. Denoting the film co- efficient along the baffle as fy, we can write 1 2m Fe kG (29) where the first term is the sum of the film resistance on the two sides, d, is the thickness of the baffle, and K. is the conductivity of the baffle material. A good ap- proximation for the value of 4 isto assume hy ‘Where h, is the shell-side coeffi written as jent Equation (29) can be 4 IZ dy e+ (30) noe & a For a given exchanger « can be calculated, and hence the correction factor for conduction through the baffle F can be read of Figure 11. ‘An analysis is now made that will illustrate the use of the method developed in this paper to predict the effect, of leakage in a two-pass shell exchanger. EXAMPLE A two-pass shell exchanger is designed to cool the bottoms ofa gasoline stabilizer. The bottoms enter at 400°F and are supposed to leave for storage at 75°F. Cooling medium is gasoline entering the exchanger at 40°F. The exchanger was first designed neglecting any effect of leakage. The following are the design param- eters of the exchanger: ] AIChE SYMPOSIUM SERIES 4 Tube fig Shel id Gi) isi) Rate inves 40,000 ve Tenpetre in 00" j Temperature 73°F. Spastic heat. 08 8tu/(hr)(e9 HIF) icon Li Ozeentipsse Deny . s0bbjeu.tt Prewwre drop 3 bin = Heat nonstercoaicient 130 Bew/(beisa ICR) — 230 Total teat duty Q = 7380 M E1.0./br Total heat transfer area A = 1,035 sq Overal heat transfer costfcient Uy = 80 Btu-/(he.(se.fr)(°F) ‘Shall 10.17.25 0. Eecive length 16 The effects of conduction of heat across the baffieand | leakage of shell fluid are to be investigated, Thermal Leakage The vaiues for Rand Sare R = 2.0,S Equation (30), we calculate fy: a i, ~ 130 * 1,600 G45. Using | or hy = 55 Btu./(hr.)(6a. CF) Ay = (L)(D) = 23 sq. ft. and hence, «is caleulated as — (Av) a) 263). gy gi5 (Ay) 1.935780) In order to use the graphical form of the thermal ef- ficiency, we calculate the following parameters: Je = (1.94)(015) = 0.0291 baffle is F = 0.93. 5.65 “The thermal efficiency with conduction through the | Physical Leakage (around the longitudinal baffle) In order to find the effect of the physical leakage, the dimensionless numbers Ro and Lk have to be calculated. | _@CRePAP) 5, | a Wap. — * ~ WPL | | | Substituting the design values of the exchanger, one can determine the Roand Lk numbers: R, = ()-3.05(10)? Lk = 3.9(10)~* No information is available on the clearance between the wo, 118, Vol. 68 TRELE 17-LEAKAGE RESULTS FOR TYPICAL VALUES OF CLEARANCES shell and the bafile. Under ideal conditions, no clearance {austs and, hence, for the ideal condition a is equal to ero. However, in actual operation there will exist a small Gearance that will cause shell fluid leakage. We will ex- mine the effect on this exchanger of some typical values for the clearancea, Table 1 shows the results of the calculation of the leakage for the assumed values of the earance. Itis evident from the calculations that the leakage is creasing With the increase in the clearance. A clearance »,001 in. causes 9% of the total shell flow rate to leak tncough the baffle. This leakage increases to 63% ifa stearance of 0.01 in. is assumed. Furthermore, an ex- pected clearance could have been compensated for by adding an extra [5% heat transfer area (F factor = 0.85). No such provision can be made for higher value of the clearance. ‘A calculation was also made to find the temperature profiles along the exchanger in the case where the leak- age is neglected (NTU = 1.71). Figure 12 shows the temperatuge profiles for several values of the clearance a. {tcan be noticed that the effectiveness of the lower p: of the lowsr pass of the exchanger is reduced with the merease in.clearance. Asa matter of fact, at 0.01-in, clearance, the tube fluid in the lower pass is heated by the shell fluid rather than being cooled, - Fig. 12, Temperature along the exchanger for NTU = 171 =" ROZENMAN and TABOREK ‘The example shown was calculated on HTRI shell-and- tube computer program ST-4 which incorporates this — analysis. {t represents an actwal exchangerthat was built neglecting effect of leakage. The exclianger failed to give the desired output temperature of 75°F. and gave instead outlet temperature of 152°F. CONCLUSIONS 1. A mathematical analysis which enables the predic- tion of the effects of thermal and physical leakage in a wo-pass shell exchanger has been developed. 2. Thermal leakage affects only the theoretically ideal temperature profiles, whereas physical leakage affects... ‘also the flow and pressure drop along the shell. The physical leakage reduces the flow of the shell fluid across the tube bank and also decreases the overall pressure drop. 3. The actual flow, pressure drop, and temperatures in the shell can be expressed by two dimensionless numbers which combine all the known operation and design parameters of the system with the exception of the shell- to-bafMe clearance. NOTATION total heat transfer area, sq. fi. area of the longitudinal baffle. sa. f. clearance between shell and longitudinai baffle, fe — eS specific heat, B.t.u./(Ib.)CF.) diametér of the shell-ft. ~~~ thickness of the baffle, ft MTD correction factor A= Ape a« R/sec.? hy = heat transfer coefficient across longitudinal baffle, B.t.u./(hr.)(sq. ft.) °F.) ‘hy = film coefficient along the baifle, B.t.u./(hr.)(Sq. f.)CF.) = sit leakage resistance coeificient -onductivity of the baffle metal, Bitu./(hr )(ft CF) length of the baffle, ft. leakage parameter WL) Lk (Wo) _— —<— path length of leakage flow, in’ : jeakage per unit length, tb./(ht.)(ft.) leakage Reynolds number Vix Wig exer 2) factor denoting type of flow in the shell heat transfer area per unit length of exchanger, sq. ft./ft. Newton's law proportionality factor (g =-32.17), 20 HEAT TRANSFER—TULSA Py, = shell-side pressure along the exchanger, War/sa. ft. AP, = calculated overall pressure drop in the shell side for no leakage conditions, 1b.;/sq. Mt AP, = actual overall pressure drop in a leaking ex- APa R Ro e885 cont changer, Ib.;/sa. ft. pressure difference between top and bottom sides of the baffle, If./sa. ft, temperature ratio parameter. pa Dh Sah leakage variable Ro = OM Ca ONEP) Weo/2L thermal effectiveness bot a , temperature of shell fluid, *F. inlet temperature of shell fluid, “F. outlet temperature of shell fluid, °F. temperature of tube fluid, "F. inlet temperature of tube fluid, "F. outlet temperature of tube fluid, "F, overall heat transfer coefficient, Btu,/(hr.)(sq. ft.CF.) velocity of leaking stream, ft,/hr. total shell flow, Ib./hr. = flow in the shell side along the exchanger, Ib./hr, dimensionless flow AIChE SYMPOSIUM SERIES W, = flow 9: turnaround end of exchanger, Ib./he. X = distance along the exchanger, ft x = dimensionless distance x = X/L 2 = ratio of length leakage path to clearance = //a @ = thermal leakage parameter _ Ashe) AG) constants ii Equation (106) prefix denoting difference viscosity, centipoise density, Ib./cu. ft. dimensionless pressure For Py vane TOP, SUBSCRIPTS AND SUPERSCRIPTS leakage variables = no leakage variables = shell-side variables = tube-side variables = variables in the upper section of the shell «= variables in the lower section of the shell aware LITERATURE CITED 1, Whistler, A. M., Trans. ASME, 683-685 (Aug. 1947). 2. Palen, J, and J, Taborek, paper presented at ASME-AICE Heat Transfer Coat, Philadelphis (1968) 3, Taborek, J., Rept of the Phillips Petroleum Co, (1961)

You might also like