| How to Calculate Taki Heat Losses Well within normal engineering accuracy, this method resolves much empirical data into two working charts with simple equations Franco Bisi and Sergio Menicatti, SNAM Progetti, San Donato Milano, Italy ‘Tux rrontest oF EvaLUATING the heat losses in storage tanks and correctly sizing the elements for keeping the stored product at a convenient temperature, has often ‘been dealt with. Since the problem is rather complex, + attempts are generally made to arrive at approximate solutions, which are of great utility for the designer. ‘As a matter of fact, the degree of accuracy attainable is often sufficient for a'satisfactory definition of the engi- ncering problems. This article gives an exact metliod for the calculation. For clarity, the problem is divided as follows: (1)- gen- eral, (2) calculation of wall temperature, and (3) caleu- lation of heat losses during heating and under normal ‘operation. General, For this article, the study is limited to tanks without external thermal insulation. Moreover, we shall not consider any formation of deposits by the stored prod- uct. These, as it is known,’ somegimes form a layer of true insulation inside the tank. It is obvious that the influence of deposits can be taken into account by suitable values of a fouling factor which will add its own resistance to the heat flow from the inside.to the outside of the tank Calculation of Wall Temperature. ‘The basic relation- ships have been summarized.** Heat is generally trans- ferred from a heating source (such as a coil on the tank bottom) to the’stored product. This can be put into the tank at soaking\ temperature or at a higher temperature, or colder than ‘required by viscosity requirements for Pumping purposes. In such case it is clear that the prod- uuct will be heated to the desired temperature. Tn each case, the basic pproblem is the determination of the exact temperature of the metallic wall (supposed bare) of the tank; such temperature will vary with time. ‘The following values are indicated below Q=amount of heat transfer in unit time, Btu/be. = temperature of product, *F February 1967, Vol. 46, No. 2 Watt TeMpeRaTuRe AIR TEMPERATURE Tee tta(T-t) Fig. 1—Shows temperature drop across storage tank wall, mean temperature of air, °F T= temperature of (saturated) steam ‘used for coil heating, *F T,, = temperature of tank metal wall, °F coefficient of expansion of stored product, 1/91 viscosity, Ib./f. hr. thermal conductivity of stored product, Btu/ft. he. °F The following relationship applies at ae) e a * where 22 indicates the sum of the resistances opposing the heat flow from inside to outside garage ai a re ‘And assuming aS (Ty— he = (T= Tihs whence - ° and then the two relationships Ty=T—hy+ (T-/ty thy wo #4 A(T —O/0h, + h,); assuming that (2) Jthy +h) oO) 145HOW TO CALCULATE TANK HEAT LOSSES . . . “Grae pee we obtain Se (7 —0):T 7, = (1 —a)(T 0. Several surveys have been carried out by SNAM Pro- getti to evaluate the most convenient formulas to be used in the calculation of the coefficients Ay and he and to as- ‘certain the effect of wind on he. From the experimental results, the following relation- ships appear to hold: mK [ sre} In this article the first relationship will be used. For the: values of is, the effects of convection and radiation, ‘must be consideyed. Hedles gue = 0:8(T, — 1908 _ wf (04) where ¢ is the radiation factor of the outside surface of the tank. Several values of # have been published. For ‘a mean value we recommend « ~ 0.8. We obtain: + (eb 460) 100 70 ‘To take into account the wind effect in a suitable way, we could introduce a multiplying factor of the convection coefficient hey. Such factor X” would be a function of the wind velocity. In this article itis however considered Imre opportune to rer to @ conficient X which muli- aT, plies both Moore and fina SucPg factor has evidently no physical meaning, in contrast to Yhe factor 1’. Nonethe- less, it permits notable simplifications. Between the two coefficients there is the relationship N= 0% © hone heat)/ Cheon + hres) * Operating thus, we could obtain one family of curves of the type reported in Fig. 2 The difficulties now encountered in the evaluation of the wall temperature 7, assigned T and ¢, are due to the ‘very complex bond between a, T, t. This bond allows one to arrive at the value of i ‘means of successive appro ‘To obviate this inconvenience, We temperature drop across the’ tank metal wall has been neglected, and the expression 146 hasbeen replaced by a simpler one without serious ero proceeding as follows: Epa RUAGAG (3), by weplecamens ot Keedl.ba PReRep 7 ESP em where F-= ((T, +480)/100}* — [(4 +460) /100}+; with a fair approximation in the interval o 009) > os oar asas0 i FumcTin fon cALcuLATION oF waLL s we Temreetnose abeyageat ete "ig a " S003 cos ons oe eea70 oo i 8 5 6 6 #00 tee 2 bi nin oki al empresa ii cy castes tepemapeenpeare cl'gatel'prokcce ae 200 . al Nee ro rm oe 0 r E i hm a aes - > o O oO 0 00 ‘00 een ol heeds el Sided Sc vow ttpir “7 February 1967, Vol. 46, No. 2HOW TO CALCULATE TANK HEAT LOSSES . It is normal to use the’ relationship [2B rn] a ne where d is the outside diameter of pipe coil expressed in inches It is opportune to introduce also a certain fouling fac- tor for deposits which might accumulate externally on the tubes: the resulting coefficient of exchange will be, indicating by ff the fouling factor, oo hae "TIE ‘As regards amount of the exchange surfaces of the tank, on the basis of experimental surveys it has been found sufficiently logical and concordant that the side surface can be considered as completely active to the thermal exchange, whereas only a portion of the roof sur- face should be considered since there is usually a gas Chamber at rest between the liquid and the roof, at least in the cases of fixed roof tanks. For these it can be as- sumed that only half of such a surface exchanges heat ‘The tank bottom can even be considered adiabatic since below the plates easily covered with deposits one finds a soil layef which has slight thermal conductivity If His the height (feet) and D the diameter (feet) of tank, the’ heat loss will be given by the expression o(us2) qr fasfer—o Js lta] C389) The amount of heat used to increase by dT” the fluid temperature will be, in the time d?, ¢ ple = DMP vce pat One arrives thus at the final relationship of the ther- bend aul! oa Be s Gs) 1 10 © ar oy Equation (6) is a first order differential equation with separable variables, the integration of which presents re- markable difficulties, not being known an analytic rela- tionship which explicitly gives « as function of 7. We 48, Eg About the authors s FRaNco Bist is in charge of SNAM Progetti’s computation conter in Milan, Maly. His duticn include that of © ‘mathematician and programer in de- ‘ign and engineering problema, Mr. Bisi received his degree in pyre physics ‘at the University of Pavia and has ‘worked as an assistant professor at the Faculty of Experimental Physics at te Polytechnic of Milan. Smncio Mentcartt és én the mechanicat ‘equipment ofice of SNAM Progetti, Milan, Italy. His duties include thermo. diwanicant mechanical design of he exchangers, furnaces ‘and pressure ceived hin che Polytechnic In: member of the Engineers’ Order of Milan. have then proceeded numerically, replacing the differen- tial Equation (6) with an equation by finite differences, determining for each value of T the corresponding value of a through the graph at Fig. 2 and assuming as an initial condition the following: T=? where T is the pre-established temperature of fluid at the time o=0 In Equation (6), after having established T, t, the physical characteristics of the product and the action of the wind, there remains only the parameter A, and it is then possible to build up a set of curves which for values of A give the progress of Tas a function of the time. was done with the help of an electronic computer. Figure $ shows the curves 7 =\T (4, 9) forthe fol lowing case: Ha 42 feet 85 bt D=50 fect 0.52.Btu/b. °F @=238 4104 1/°F T, = 356° =13"F 5.75 tb./t. hr 0.075 Beu/tt, he. 33 T inti From Fig. 3 it appears evident that, if it is not of im- portance the time required to bring the fluid, for instance to T = 95°F, the necestary surface required is 400 ft. Should it, however, be required to perform the heating to the same temperature in a time of 44 hours, then the necessary surface would increase from 400 to 600 ft. ‘As it can easily be seen from the graph, there are tem- peratures assumed as initial, so that with a given value ‘of the coil surface there occurs an initial bil cooling and ‘a subsequent stabilization at a lower operation tempera- ture cae avid, Parte Refer, Vol. 38, No 4, Ape 1958, p. Me McGraw Hl Bock Go, New Beh ar an 8 1G arte Tin "Kine chart Ry Perelcame Rife, Va 38, No. 5, May 199, p. 27.236 ig Reem, St Corns, Gemomaions-10, Dven-4, Hyprocannon’ Processino