Predict storage-tank heat transfer precisely Use this procedure to determine the rate of heat transfer from a vertical storage tank when shortcut methods are inadequate. Jimmy D. Kumana and Samir P. Kothari, Hewingsom, Durham and Richardson, Tne D Heating or cooling storage tanks can be a major ‘energy expense at plants and tankfarms. Though many procedures for calculating such heat-transfer require- ments have been published [/,3,5,7,8,/0], the simplify {ng assumptions that they use can lead to significant ‘errors in computed heat-transfer rates. This is of con- ‘cern because efficient sizing of tanks, insulation, heaters and coolers depends on accurate estimates of heat trans- fer to and from the various tank surfaces. And the ulti- mate value of being accurate increases as energy costs ‘continue to rise. ‘The procedure presented here determines the heat transfer to or from a vertical-cylindrical storage tank seated on the ground—like the one in Fig. 1. It includes the effects of tank configuration, liquid level, ambient temperature and wind speed, as well as temperature variations within the tank and between air and ground, A partially worked example shows how to use the tech- nique, and how to do the calculations on a computer. The theory Storage tanks come in many different shapes and sizes. Horizontal-eylindrical and spherical tanks are used for storage of liquids under pressure; atmospheric tanks tend to be vertical-cylindrical, with flat bottoms and conical roofs as shown in Fig. 1. The example pre- sented here is forthe latter configuration, but the proce- dure applies to any tank for which reliable heat-tranafer correlations are available. For the sake of simplicity, we assume that the tank contents are warmer than the ambient air, and that we fare concerned with heat loss from the tank rather than heat gain, But the method may, of course, be applied to either case ‘Consider, then, the categories of surfaces from which ‘heat may be transferred across the tank boundaries: wet ‘or dry sidewalls, tank bottom, and roof. In the context ‘used here, “wet" refers to the portion of the wall sub- ‘merged under the liquid surface, whereas “dry” refers to the portion of the wall in the vapor space, above the liquid surface In general, the heating coils would be located near ‘the bottom of the tank, in the form of flat “pancakes.” ‘Therefore, the temperature of the air (or vapor) space above the liquid level may be expected to be lower than the liquid itself. Experience has shown that the average bulk temperatures of the liquid and vapor space may be significantly (i.e, more than 5°F) different, and they are treated accordingly in our procedure. Use of differ- cent liquid and vapor temperatures is an important de- parture from the traditional approach, which assumes the same value for both. (Our basic approach is to develop equations for calcu- lating the heat loss from each of the four categories of surfaces, and then add the individual heat losses to get the total heat Joss. Thus: For dry sidewall qy For wet sidewall 4y, JgAalTy — Ts) UAgT, ~ Ty) ay @ For tank botiom = U,A,(T, ~ Tq) (3). For tank roof 4 = UALTy — Ty) a) Toral Q=utet ary © ‘When using these equations in design or rating prob- lems, we either assume the various temperatutes for typ- ‘Typical vertical-cylindrical tankIndividual film heat-transter coetficients Table | ical conditions or determine them by measurement. ‘The area values are also easy to obtain: Ay = 2D(L—L,) © Ag = Dlg o A, = 0D*/4 (8) 4, = (9D/2D*/4 + a2) o The complications arise when we try to estimate the overall heat-transfer coefficients U,, U,, U, and U,, for the four surfaces of the tank. For the tank geometry chosen, these can fortunately be calculated from the individual film heat-transfer coefficients in the conven- tional manner, using published correlations. ‘The overall coefficients “Table I shows the component coefficients for each ‘nieface, The overall heat-transfer coefficient forthe dry ‘idewall ofthe tank (U,) is calculated as the sum of the resistances of vapor film, fouling, metal wall, insulation {if any), and outside air (convection plus radiat ‘The outsideair heat-transfer coefficient (iga) i8 a function of wind velocity as well as temperature gradi- Effect of wind velocity and AT on heat transfer rate Fig.2 128 ent. Data on the effect of wind velocity and AT have been presented by Stuhlbarg [/0] and Boyen (2]. With a little bit of manipulation, their data were replotted, Yielding the “wind enhancement Factor” (W;) in Fig. 2. By definition: Wy = hale = har/bar (ao) ‘Therefore, once the outside-aie coefficient for still ait lige) is known, the overall dry-sidewall coefficient at vvarfous wind velocities can be computed as: 1/Ug = Whyy + tly + ty/hy + VW ge + bys) + Whpy (11) Similarly, the overall coefficients for the wet sidewall, bottom and roof surfaces are: Lliyge + ba Hho + tl hy + T/W hie + he) + Whine (12) ay + by/hn + Vig + Whig (13) ig g/g 1/(Wligy + hey) + Wty (14) Eq, 13 and 14 assume that the roofand bottom are not insulated, which is generally the case in temperate clic ‘mates, We shall now review correlations for the individ- tual heat-transfer coefficients needed to obtain the over- all coefficients. Individual film heat-transfer coefficients ‘The film heat-transfer coefficients may be divided into four categories: convection from vertical walls, convection from horizontal surfaces, pure conduetion, and radiative heat transfer. Within each category, cor: relations are presented for several flow regimes. ‘Vertical-wall film coefficients. These apply to the in- side wall (wet or dry) and the outside wall (still air). For vertical plates and cylinders, Kato et al. [6] recommend the following for liquids and vapors: Nye = 0.138 NOE(BT — 0.55) where 0.1 < Np <40 and No, > 10°. For isothermal vertical plates, Ede [4] reported the following for liquids Nyy = OABS(NgyNpy) 25 16) where Np, > 100 and 10 < (No, Np) <108, and for gases Nyy = 0.0295 NBM NRA (1 + 0.5 NBST)-240 (17) where Npy 5 and (Ng, Npy) > 10° For vertical plates taller than 3 ft, Stuhlbarg (/0] recommends: (13) 45 £19 (No, Np) where 108 < (Vey Nyy) < 108. Horizontal-surface heat-ransfer coefficients. These coefficients apply to the roof and inside-bottom surfaces of the tank. The bottom is assumed to be flat. For sur- faces facing up (3}: (18) Nyy = O14 (Noy Noy)? (19) For surfaces facing down: Nga = O21 (Ne, Ney)? (20)Nomenclature ‘Area of heat-iranser surface, fi; 4y for bottom, Ay for dey wall, for wer wall, A, for roof Specific heat at constant pressure, Btu /Ib-"F Diameter of tank, fe Height of conieal roof at center, fe ‘Acceleration due 10 gravity, 4.17 > 108 f/h# idual film coefcient of heat transfer, Biu/ftth-"Fs hy for air outside the walls, hy, for air above the rool, Hay and Hy for sil ai, for liquid between the walls, hyy for liquid near the bottom, fy for Vapor nea the walls hy for vapor near the oof ip Fouling coeficient, Btu/fhF; Ayy for Hquid at the walls hy for liquid at the bottom, hy for ‘vapor at the walls oF the root ‘hg Heattransfer coeficient for ground, Beu/f2h-"F Ay Hleattransfer coefficient for insulation, Btu/ft*h- +e fy leat transfer eneficient for metal, Buu/fith-"F ‘hy Heattranafer cooficient for radiation, Beu/f%h- *F: hyy for bottom, hyy for dry wall, hy, for wet wall, be, for root Thermal conductivity, Beu/ftch-"F; fy for ground, 4 for insulation, hy for metal wall 1 Total length for heat-transfer surface, Ung Total length for wetted surface, f Grashof number, 3p%—AAT/u® Nusselt number, AD/& or hL/& Prandtl number, t/k "Rate of heat transfer, Blu/h q Individual rate of heat transfer, Beu/h; g for bot- tom, gg for dry wall y for wet wall, 4, for root T Temperature, "F; T, for ambient air, 7, for bulk liquid, Ty for vapor, Tq for ground, 7, for inside wall, Tye for outside wall ‘Temperature difference, “F | 1 Surface thickness, ft; ¢ for insulation, ty for metal U Overall heattransfer coefficient, Btu/h-°F; Uy for bottom, Ui for dry wall, Ui, for wet wall, U, for roof Wind enhancement factor ‘Volumetric eoeficient for thermal expansion, Viseority of fuid, 1b/tesh Density of uid, Ib Emissivity & Sea st eRS Both equations apply in the 2x wre (NarNpy) <3 10 Equivalent coefficients for conductive heat transfer. ‘The wall and insulation coefficients are derived from the thermal conductivities: fg = Kael hy = hy/ty range a 22) ‘The coefficient for heat transfer to and from the ground is the coefficient for heat conduction from a semi- infinite solid [9]: hig = Bkg/aD es | Fouling coefficients. The coefficients pp, fing ad hyp apply to the vapor and liquid at the wall, and the liquid at the bottom of the tank, respectively. ‘These are em- pirical, and depend on the type of fluid and other fac- tors such as tank cleaning. Generally, hyy is the greatest of the three, and fy, the least, indicating that the great- cst fouling resistance is at the bottom of the tank. Equivalent coefficient for radiative heat transfer. The coefficient for sidewalls and roof depends on the emissivity of these surfaces, and is given by (8) be = OPI (Tay +460) (Ty + 460)¢ ele) -Caar)] eo With these relationships, we now have the tools to cal- culate heat transfer (o or from the tank. Example ‘ABC Chemical Corp. has a single manufacturing plant in the U:S., and exports a high-viscosity specialty oil product to Eurape. The oi is offioaded in Port City, and stored in a flat-bottom, conical-roof tank rented from XYZ Terminal Co, Ltd, The tank is located out- doors and rests on the ground. It is equipped with pan cake-type steam-heating coils because the oil must be ‘maintained above 50°F in order to preserve its fluidity Other pertinent data are: tank diameter is 20 ft: tank height is 48 ft (to the edge of the roof); roof incline is 4, in. per foot; tank sidewalls are grin. carbon stel; nse lation is Hein, fiberglass, on the sidewall only XYZ Terminal Co. does not have metering stations on the steam supply to individual tanks, and proposes to charge ABC Chemical for tank heating on the basis of caleulated heat loses, using the conventional tables [1], and assuming a tank wall temperature of 30°, The project engineer from ABC Chemical decided to inves tigate how XYZ's estimate would compare with the ‘more elaborate one described in this article First, the engineer collected basic data on storage and climate. Oil shipments from the U.S. arrive at Port City approximately once a month, in. 100,000-yal batches. Deliveries to local customers are made in. 8,000-gal tanktrucks, three times a week on average. The typical variation in tank level over a 30-day period is known from experience. The ambient temperature goes through a more com- plex cycle, of course. Within the primary cycle of 365 days, there are daily temperature variations. But in the seasonal cycle, heat supply is required only during the ‘winter months, when temperatsres Fall well below 50°F, Wind conditions at the storage site are not as well defined, and therefore much harder to predict, How- ever, we can assume that the wind speed will hold con- stant for a short period of time, and calculate the heat loss for this unit period under a fixed set of conditions. ‘The wind speed to be used must be based on the known probability distribution of wind speeds at the site. “The procedure for determining the annual heat los consists of adding up the heat losses ealeulated for each unit period (which could be an hour, 12 hours, 24 hhouts, or 30 days, as appropriate) This example dem- onstrates the calculation of heat less for only one unit period, of 12 hours, using an ambient temperature of 129Data for ABC Chemical Co. example Physical properties Density, bite? 0.08 Specific heat, Btu 0.25 oor. 0.007 Thermal conductivity, ape Buulten ona 01s 0.018 Coeticens of volumetric ‘expansion per" F 1x10 09002 0.002 ‘Assumed fouling eoetficionts ry wall 1000 Brusti2h Wer vat ‘200 Root 1000 Bottom 500 “Thermal condutivitos ‘Metal walls roauten Insulation | 0.028 Ground, 0.80 Surfsce omisivity Will and root 09 Temperatures Vapor in tank o'r “Liguie in tank 55, Outside a 38 Ground 40 ‘since the liquid hat tow volalty, the vapor space Is msumed 10 bbe mostly at Hoat-transfer coefficients after first iteration Table HI Lente Diywell Wats Root —_Racon REL esasEN = 3 2 rs Pree eiecray = ss Ay = = eat pate is 5 z eee m . es a 02 tae i eae tae e See eet r= aw OS oi . = hgot 18818 = = rm HO Cr) oo u oz ozs = nr) 001000 seo tm 07868 oreo Oost — a o1sie 0828 0.1457 0.0033 Second iteration yields closer v “Termporature let wall Root Bottom Tylimisel,"F 2480 3595 53.7 1 425 425478 Tye (ouskiel, °F 2 38.9 3575 1 3878 Pee Table tt 95°F, a wind velocity of 10 mph, and a liquid level of 50%. The other data required are given in Table I Note that the liquid temperature is controlled at 55°F to provide a 5°F margin of safety. Since the Prandtl and Grashof numbers occur repeat edly in the film heat-transfer coefficient equations, and remain relatively unchanged for all the conditions of interest, let us first caleulate their values. Thus, for the liquid phase: Noy = L3pteB AT/ ut No = Cyi/k = 484 Similarly, for the vapor phase, No, = 1.90 x 107L8 AT, and Np, = 0.28. We can now calculate the individual film heat-transfer coefficients, using the ap- propriate L and AT values in the Grashofnumber equations. This is an iterative process that requires ini- tial estimates for wall and ground temperatures, plus wall temperatures Coefficient for vapor at wall (hy). As an initial ap- proximation, assume that the wall temperature is the average of the vapor and outside-air temperatures: T, = (50 + 35)/2 = 425°R. Then find the Grashof number: 75 LAT Noy = 1.90 X 10% ~ Ly) Ty — To) 90 x 107(24)8(7.5) 97 x 10 Employing Eq, 15, find the Nusselt number and then the coefficient (£ = 0.0151, L = 48 ft, Ly = 24 fh Ny = 0:138(N,)958 (MBI — 0.55) = 921.1 yg = Q2LAYA)/(L — Ly) = 0.581 Btu /ft@hetF Coefficient for liquid at the wall (z,). Here, neither Npp nor (Ng, Np) falls within the range of the applica- ble correlations (Eq, 16,18). Let us try both, again using an average for Ty Ty = (Ty, + Ty)/2 = 45°F Noe = 974714(T,, — Ty) = 1.35 X 107 Using Eq. 16 and 18, we get wo estimates forthe heat- transfer coefficient (i = 0.12, Np, = 484) (0.4954/1,.\(Noy Np) = 0.704 Bru/fh-"F (OABL/LQ7)(No, Np,)9 = 1415 Beu/ftthe*F ve Nie To be conservative, we use the higher value: gg = VALS Buu /fth-°F, Coefficient for vapor at roof (ty,). We consider this a flat plate, with a diameter of 20'ft, and use Eq. 20, again with an average T, of 42.5°F (k = 0.0151) No, = 1.9 X 10°D8(Ty ~ Ty) = 1.14 x 10 hy. = (0.27H/D Noy Np) = 0.154 Btu/ft2he*F Coefficient for liquid at tank bottom (hy). Assume that the ground temperature (Ty) is 5°F above ambi- cent, and use an average of liquid and ground tempera- tures as a first approximation for the tank-bottom tem perature: T, T, + Te? (1, + Ty + 5/2 = 475°F 130 SRRTAT NRE ARGH HT‘Thea, igure the Grashof number, and use Eq. 19 t0 get the coefficient Noy = 9747 for Nop = 2.83 X 108 dngy = 1.105 Buu /f2h-*F = Ty) = 5.85 x 10% Coefficient for outside air at roof (i,,). Assume TT, sinoe the roof is uninsulated, and get the coef- fidient for still air from Eq, 19: Nop = 1.9 X 10'D%(Tyy — Ty) iy, = 0.663 Beu/ft2he"F Coefficient for outside air at wall (fi,)- Assume that the temperature drop across the film is one-fourth of the drop from the inside fluid to the outside air (averaged for the wet and dry walls), and use Eq. 15 to find the coefficient 1.14 x 10 AT = 175/4 Nor =4375°F 9 x 10718 4’ 51 Beu/fi?h-"F 9.19 x 10! Conduction coefficients for ground, metal wall, and insulation (i, hy and f,). These are straightforward, from Eq. 21-33: by /ty = 640 Beu/fh-*F /ty = 0.224 Bru/fteh-*F g/aD = 0.102 Bru/fe2h-"F Radiation coefficients for dry and wet sidewall, and oof (less hie: hye) AS for the outside-air film coef cients, “assume that Ty = Ty + 0.25. (Tyuy — T)> Where Thuy is the temperature of the liquid’ or vapor inside the tank, ifthe surface is insulated. For the unin- sulated roof, assume that Toy = Ty + 0.5(Ty — Ty). Then Ty = 98.75°F for the (insulated) dry sidewall, Te, = 40°F for the wet sidewall, and T., = 425°F for tthe roof. Using Eq, 24, find the coefficient for each of the three cases 797 Buu/fith-*F 758 Buu/ft?h-*F 765 Bu/ft?h- Pee Closing in on results ‘Table HI summarizes the heat-transfer cocffici just calculated, ineluding the corrections for wind—hyy and ky, are multiplied by 33 and 3.1, respectively, ‘based on data for 10-mph wind in Fig. 2. Substituting these individual coefficients in Eq. 11-14, we obtain the U values listed in Table IIL What remains to be done? When we began the calcu- lations, we assumed that the outside-wall temperatures were related to the bulk-fluid temperatures by: T,, = Ty + 05(Tyay Ta) for uninsulated surfaces T, + 0.25(Tyax ~ T,) for insulated surfaces In order to calculate accurate coefficients for heat trans- fer, we must now obtain better estimates of these wall ‘temperatures. This requires an iterative procedure that can be programmed and run on a computer. For dry wall, the rate of heat loss is given by all three Of the following: 94 = Ug Ag(Ty — Ta) (25) Ihe Aa(Ty ~ Tr) (26) na + haw Aa Fry — Ta) 7) Solving Eq, 25 and 27 for Ty yields: Tae = UaHhna + haw Ty — Tr) + Ty (28) Similarly, solving Eq. 25 and 26 for T, yields: Ty = Ty — Wal tye Ty — Ts) (29) Using the same approach, now calculate T, and Ty for the wet wall, and T,, for the roof and bottom of the tank. To find the correct wall temperatures, use the initial estimates of U and A values in Eq. 28 and 29 (and in the parallel equations forthe other surfaces) to get new T,, and Ty, Values. Table IV shows these temperatures after a second iteration. Using these new temperatures, re- compute Grashof numbers, individual heat-transfer coefficients and overall coefficients, and then iterate again to get a new set of Ty and Ty, values, When the current and previous iteration’s temperature estimates are the same (within a specified tolerance), the iteration is completed, Table V lists the individual and overall coefiients after the second iteration. Although itis clear that adi- tional iterations are needed, let us accept these Valles as sufficiently accurate for the present purpose. Then we ‘ean obtain the total heat-transfer rate (Q) by using the U values in Eq. 1-5 and summing. Table VI shows the calculated heat-transfer rates through each boundary, and the total rate. Note that the roof and bottom of the tank account for only slight heat loss, despite being uninsulated This, of course, is for the unit period of time, when wind speed is 10 mph, the tank i hal full, and the air is 35°F. Table VIL shows how the results of unit-period SRNR RGEERING KORG HT 131 Rovised coet Table Vv Coettciont Drywall Wetwall Root Bottom Ve 0.463, = a a oe - 098 s a nye - - 0. S Lb s - - ono ne. z = = 0.102 Nar - - oar 4 ae = - 0968 = Paw os7 asi? = » haw 10671087 z but 640 640 640 640 oy o7e ome | = ne 1,000 800 1000 500 oe 07500 0.7514 0.7500 a 01392 0.1655 0.1636 (0.0875 *For 10.moh windRate of heat transfer during unit period Table VI Surt wlth Dry welt 0.1392 348.7 Werwall 0.1655, 49915 Roof 0.1636 7730 Bottom 0.0875 3418 4124 Toral 385 9253 ‘Summing losses for unit periods yields heat loss for 30 days Table VII Period Liquid level, % T,°F Wind speed, mph Heat fre, Btu 1 50 35 10 111,908 2 50 2 5 392407 3 “a 2 ° 42591 a 93 5s 30 ° 59 58 a 20 12368, 60 «0 60 5 o otal fr 90-dey period 8,388,080 heat losses can be tabulated and added to get the cumu- lative heat loss for a month or year. Of course, this re quires climatic data and tank-level estimates for the overall time-period. Comparison with other methods Aerstin and Street [1] offer a very simple method for caleulating heat loss from tanks. For a tank with 1.5 in. of sidewall insulation, and a wind speed of 10 mph, the recommended overall U (based on & = 0.019 for the insulation) is 0.14 for AT=60°F and 0.14 for 10°F, Adjusting these values for # = 0.028 and 17°F, as in our example, yields an overall U of 0.206 Beu/fi2h-*F. The total exposed surface is 3,331 2 (tank bottom not included), and thus the overall rate of heat transfer by their method is Q = 0.206 x 3,881 X 17 = 11,666 Beu/h This compares with a heat loss of 8275 Beu/h (for the exposed surface) caleulated by the procedure of this article—see Table VI. Thus their method yields a result 31% too high in this case Stuhlbarg [70] takes an approach similar to that pro posed here, but his method differs in how the outside tankwvall film coefficient is computed, Stublbarg recom ends the use of a manufacturer's data table, and docs not explicitly distinguish between the bulk liquid tem- perature and the outside-wall surface temperature in calculating the proper heat-transfer coefficient The algebraic method of Hughes and Deumaga [5] resembles the one presented in this article in many ways, But it does not recognize differences between liq uuid and vapor temperatures inside the tank, nor does it account for the interaction between MT and wind speed in calculating a wind-enhancement factor, Finally, even though their procedure requires iteration, the focus of the iterative efforts is to get better estimates of fluid properties, not tankwall temperatures, Conclusions ‘Our engineer at ABC Chemical was able to negotiate significant reduction in the heating charges proposed by the XYZ Terminal Co., which had used a shorteut method for its estimate, because the procedure pre sented here is rational and defensible. A rigorous solu- tion of the iterations can easily be reached on a digital computer or even a programmable calculator, and the effort pays off in better design or operation criteria. Mark Lipoes, Ei References nergy Rezvey "John Wiley ons Now Yor, 4 Blea 6 Kae, Nihal, ad ia, of ton Mus Ti, V1 we Hea Tran” McGraw Hl, New York 1950, p 8 GH, "Chemica Engen’ Handi” th 0, Robe, WM and Hartnet, JP Handbook of Heat Tatler ‘The authors Veigetr st Hennings Bess sod Ree ear the Pon lite of Tbe, od Srataeecses ey Sami Roa ws reine hardy Ie then i ail wa toon ‘Rots bes Ba Sop stan Mabe Ge Ua at Me eerie Boge a i and 132 SIRT RSRTERTNS ART ET