Solutions Manual for CONVECTION HEAT TRANSFER Third Edition Adrian Bejan J. A. Jones Professor of Mechanical Engineering Duke University 2004Contents Chapter 1, Fundamental Prineiples 2. Laminar Boundary Layer Flow 3. Laminar Duct Flow 4. External Natural Convection 5. Internal Natural Convection 6. Transition to Turbulence 7. Turbulent Boundary Layer Flow 8. Turbulent Duct Flow 9. Free Turbulent Flows 10. Convection with Change of Phase 11, Mass Transfer 12, Convection in Porous Media This solutions manual was organized and typed by Linda Hayes Adrian Bojan January 2004 Durham, NC 10-1 Mel 12-Chapter 1 FUNDAMENTAL PRINCIPLES L1 We are told that the scales of the two major terms in the two groups of terms in eq. (1.5) or eq. (1.6) are measured experimentally: Dp pvv=0 ) . (~ p38) < scales ‘Therefore, if eq. (1.8) is to apply, then the first scale must be negligible, Be (93 a vB cpt in other words, the relationship between dp/0x and du/dx must be apex. p. Bulax Ay, Az — Az can be sketched as follows: Zz zZ rb 4 é x a eo In eq. (1.9) we have vy. Ye, 1 Vg , OVz BtLtt w+ 79 ax dy or, since v; + u, ve > v and v, > w, du, ov, aw _ rae yt =O (1.8)The momentum equations (1.21) have the same property; for example, the r equation (1.21a) can be written as ( Var __ 2B 2 avg , Hy, =- = +H ats in other words, 2u p(Se vu Mtv du ew Bu) = ~ 8 u(y By Be) ae, (1.19) The message of this exercise is that, through a simple transformation, the validity of equations in cylindrical coordinates may be tested based on the considerably more familiar Cartesian forms. Problem 1,6. In the r+ limit, the spherical coordinates sketched in Fig. 1.1 become in other words, Ar — Ax, r sin 6 A@ — Ay and rA —» Az. The mass continuity equation (1.10) can be expanded as: Oe 2¥r yd av% come, 1 ave 9 at tT rae . > 08 eo» One B Noting that vr u, vg > v and vg — w, the above equation reduces to gus ges gr no (1.8) Following the same procedure, the momentum equation (1.22a) reduces to eq. (1.192). 14Problem 1.7. Taking the specific kinetic energy into account, ¢ is replaced by e + V2/2 in €q, (1.25). This substitution generates two more terms: (ince des) (yyo- a oy (ou) Z [> ¥] Together, these additional terms combine into one new term on the left-hand-side ("LHS") of eq, (1.25), D(v72) LHS. = ax Ay p 2ST where we have assumed a constant-p flow. This contribution, ic. LHS., is cancelled by the last four terms in the work transfer group { }5: to see this, we combine the constitutive relations (2.15) and (1.16) with the last four work terms, and get 1 {cet ), = of au a +33) ay Bxdy \four terms. fax 2 352 * 5k (StH 0, cf. eq. (18) d fou , or ~H 0d (5 +e) oO Py, av 328 eal +B me Shay (e+ 3] 2 vd (y+ &) In the above expression we eliminate the pressure gradients using the momentum equations (1.19a,b) 2, ae 3) Bee(a 3) After some algebra, the last four work terms reduce to J_ {the ast } Bxdy \four terms! which amounts to a new term on the right-hand-side of eq. (25) 15piv?n) RHS, = Axdy p 5 ‘This additional term cancels LHSs. In conclusion, the energy equation (1.39) is valid even when the specific kinetic energy is accounted for in the derivation. Problem 18. Consider the general flow system shown in the sketch. The first and second laws ‘of thermodynamics state, respectively, OE, h— Yeah 2Eev W=Zaj+a9+ Doth Finke Sev _ yi Sgen= get Lal Po- Ene + Zs 20 ‘The reason for distinguishing qo from the remaining heat transfer interactions gj is thi Wis to vary with the degree of irreversibility (Sgen), then, according to the first law, at least one additional energy interaction must vary. Let qo be the energy interaction that varies when W_ varies. In the extreme case of reversible operation, Sgen = 0, the second law requires if ose" T(-25 Ser _y BAY ins E ss) In the same limit, the first law reads Wrov= Lait Goyer Z th Z sin 258 out The difference between the work output in the reversible limit, Wrey. and the actual work W is simply Wrev W = o,rev— do ‘However, from the second law for the general case (Sgen > 0), we recall that ) do=To(-Spn + 2534— EMBs + Be ‘Therefore, qo —do,ev = — To Sgen: in other words Wrey-W =To Sgen>0 0 Sgen: Wiost Note that To is the absolute temperature of the heat reservoir whose heat transfer rate qo varies (Eloats) as W and the degree of irreversibility (Sgen) Vary.Problem 19. Consider the two-dimensional control volume of size dxdy shown below. The entropy generation rate in this system is Spen dxdy: this quantity can be calculated by applying €q. (1.47) to the control volume dxdy, sup dy svpdx + 289 dxdy Neglecting the terms smaller than (dxdy), we obtain v1 (84, 29) a (oe Ty ge OT sant (B+) alc og > +o[B+udt+v¥) op zer0, cf. eq. (1. 17In conclusion, Ds_ 2 Du__P DP PDT Dt pt Dr and from the first law of thermodynamics, eq. (1.26), Viq"-PVv suo Du PD Combining the last three equations to eliminate Ds/Dt and Du/Dt, we find. Lowrie pe ge or, using the Fourier law of heat conduction, q" =—k VT, in which k is constant, (1.49) 4T heat transfer fluid friction inreversibility irreversibility Thus, S3¢q is dominated by fluid friction when kar Tp or when 18Problem In general, the energy equation represents the competition between three effects, Tongitudinal conduction, lateral convection and internal heat generation, Tr kA. &r nP(T—Tp) + q’”A=0 Tr fi dx’ SS Oe ad conduction convection —_—_generation. (plays the role of z wall, plaid “heat sink” term) 1 2) __ Inthe system of length x the three competing scales are kA AT, hP AT and q’”A, where ‘AT=Ta.—Tp, Inthe limit x >, the conduction scale becomes neBligibe relative to convection and generation. b) _Sutficiently far from the wall the balance is hP AT ~ q""A, and this means that qa bP -To~ ©) Sufficiently close to the wall we have kA a ~ q"A, therefore we conclude that d) The heat transfer rate into the wall is gp ~kA a - -A Qk aT)? Problem 1.12. The momentum equation (2.26) is du, 1 4Pey yu agtey oy “TP dk, * 3,2 ae ay inertia pressure friction 19Sly In the space selected for scale analysis, we have x~L, y~L, u~Us in other words, L i inertia pressure friction The ratio inertia/friction is of the order of 2 ia U2 12 _ Unk dnettia _ == _L friction L VU, YE If Rey = 103, then inertia > friction, meaning that the friction effect can be neglected in the L x L region. In that region, the flow is ruled by the balance between the remaining effects, namely, inertia ~ pressure. Observations: — the the" appl x L region is not the Boundary Layer region; \ertia/friction ~ Rey.” interpretation of the Reynolds number does not in the boundary layer region discussed in chapter 2. 1-10Chapter 2 LAMINAR BOUNDARY LAYER FLOW Problem 21. In order to derive eq. (2.52), construct the energy-equivalent of the bottom portion of Fig. 2.3: & 2,70 (because T#Ts, constant) Hh, Y rN Hew) me He a ye 0 icp Se" a 0 In the steady-state, the first law of thermodynamics for the control volume (c.v.) is, Y . Yow ” Het [ aidy—Hysac- Paha dy +45 ax + Hy =0 where Y ‘Y Hy =cpT.. dix, =f pudy; =f pepuTdy We assume T.. = constant. Dividing the first law by dx. we obtain ay at a ra eate -£ff pepuTdysopTa df pudy- ax Jp ax dy + q9=0 ‘The scales of the last two terms in the last equation are, respectively, 1 AT aly, «A Cleatly, if x > ¥ [ie. if the (x) x CY) region is slender] the last scale dominates, and the third term in the energy equation can be neglected. We are left with 24ef pepu(te-Thdy=K(3) ot, assuming p cp = constant, ap’ _Deay=0(2 Ef, wt--dey a(S) 0 (2.52) Problem 2.2. The Blasius profile problem reduces to solving 2" + ff" =0 (2.82) subject to £0) =f) =0 (2.83) (2.84) ‘Assuming the small-n expansion £() = ag + ayn + ayn? +agn3 +..... f'(q) = ay + 2agn + 3agn? + 4ayn3 +... from the 1 = 0 conditions (2.83) we leam that ag =ar=0 The Blasius equation (2.82) becomes 23 x2 Lag +4x3x2agn 45x 4x3 asn2+6x5x4a6 nF + + (aan? +az m4...) (2x lag+3x2a3n+4x3ayne+ or, in the form of a table to collect the same powers of 1, I2a3 + 48agn + 120asn? + 240agn3 2ajn? + Zagagnd + 6azayn> + v 4 L ag=0 120.a5 + 2ap2=0 L ul eo a3 240 ag + 8 aga3 =0 1520 22‘The coefficients of higher order, a7, ag,..., are determined in the same manner. The resulting series can be written as “ELS ant 2002) Gari where ot = (0) is the unknown curvature at the wall, and [13] Co=1 3=375 Cel C4 = 27,897 @= C5 = 3,817,137 ‘The unknown a. follows from the remaining condition, eq. (2.84). Unfortunately, the six-term Blasius series listed above is acceptable only when 1 is small, say, 1) <2 as illustrated in the sketch below. Nevertheless, the skeich of (cin) shows that [P = at large n] is achieved when ois a number between 0.3 and 0.4. The correct value (ot = 0.332) is found by matching the small-n, series with another expansion valid in the domain n > O(1): this procedure is outlined in Ref. 13]. 23Problem 2,3. We divide the n space into slices of thickness An. At any node i, the Blasius ‘equation (2.82) requires f ” 7 0 1 2 38 & Thus, the Blasius equation allows us to calculate fi+2 based on the values of f at the preceding three nodes, fea Biot 38 + fen SE i let +a -28) ‘The values at the first three nodes are fo =0, cf. condition (2.83) f1=0, so that fy nee = 05 condition (2.83) fy =(An)2a,,_ where a = fg is the unknown curvature at the wall, which must be determined from condition (2.84) f@)=1 As indicated in the second part of the problem statement, itis sufficient to shoot once. Taking = 1, we find that the slope f at large 7's (7| = 10) approaches the value. £(10) 2.088, Therefore, the comect guess for the initial curvature is = 2.088-3? = 0.3314 when the integration step is An = 0.0125. This value agrees very well with Blasius' well known number FO) 1.332‘step size An) Fd0) 0.2 . 2.12113563 OL 2.1019154 0.05 209333778 0.025 2.08946 109 0.0125, 2.08886862 Problem 2.4, ‘The special feature of the heat transfer configuration in this problem is that the wall temperature To must increase in the x-direction, in order for the constant flux q" to overcome the growing thermal boundary layer thickness, 57(x). ‘We are interested in the local Nusselt number Nast x TheO-T. k’ hence. in To — T.. or, from the graph, in &p(x). The energy integral for T.. = constant is & [noo as subject to *(5), = constant Assuming the simplest velocity and temperature profiles, 25vetek and MUSE we distinguish the following two limits: 8y<8(orPr> 1). The energy equation becomes where and, from Table 2.1, 46 x Rex)? Integrating the energy equation, and noting that its right-hand-side is a constant, we obtain oy 1 1 SE = 2.75 Prt Rexl? In conclusion, ey = 0.364 Pr! Re}? (see Table 2.1) &y>8(or Pr <1). In the extreme we have 8/57 — 0, so that u may be taken as equal to Use in the ayer of thickness 6p. The energy equation becomes 26& Pre-ct-ta(2 where the right-hand-side is again constant. The solution is 5r= (FE), In other words, Ni 0.707 Prl?? Rel? by Note that results for Pr << | are not listed in Table 2.1 Problem 2.5. Since Pr > 1, the local Nusselt number given by Table 2.1 is Nu = 0,332 Pr!3 Rel? ‘The Reynolds number follows from Crx = 0.0066 and Table 2.1, Cry = 0.664 Re;!? = 0.0066, namely, Rex = 1.01 x 104 (note that the boundary layer is entirely laminar, marginally; see Table 6.1). In conclusion, Nu = (0.332) (71/3) (1.01 x 10412 = 63.9 2.6 ‘The analysis that follows is a generalization of the analysis presented in the text, eqs. (2.93)-(2.107). The difference is that the boundary condition (2.95) is replaced by 0-198 an=0 ) where vee)? (Sa) @ ‘The J parameter is a constant if t(x) varies as t=Cxl2 3) The C constant is proportional to J, 27ey" ® ‘The analysis up to eq. (2.97) is still valid. The new boundary condition (1) replaces eq. (2.98) with c= @(n) - 6(0) = 60) f exp[ dy ‘) in which, cf. eq. (1), 90) =3 6°) 6) If we set 1) > 2» and 6(ce) = 1 in eq. (5) we obtain 1=9(0){5+ [expt Jan} ON ‘The integral that appears above has the value, cf. eqs. (2.99) and (2.102), = 1 ee =——_,. (Pr > 0.5) 8) fPewt Jan omni (Pr>0.3) ® ‘The overall heat transfer rate through the wall of length L is, ef. eq. (2.101) = 40-4. L=k (Te — Tp) 260) Ref? ) ‘The J = 0 limit of this result is obtained by integrating the local heat flux calculated with eq. (2.103), a’ (1=0) =k (T.,~Tg) 0.664 Pr! Ref? 10) To see the effect of the solid coating (J), we estimate the ratio g@ _ &© =O WW qU=0) © 0.332 Prt? “ eo 0.332 Pr! {) + [° el. jen} “” If we accept the high-Pr limit (8) this ratio becomes, a 1 - —- 13) YC=0) “0.332 Pr!3 J +1 © which is shown in the attached graph. ‘The coating reduces the heat transfer rate when J exceeds the order of 1.0.1 1 10 Problem 2.7. The problem statement for the thermal boundary layer over a constant-q" flat plate is ar, yat_ er x Ox TY ay = az qd) at y=0 Q) as yon @) ‘The flow solution is known: @) ©) © To construct a similarity solution to problem (1)-(3), we must choose the dimensionless temperature 0 in such a way that the transformed eqs. (1)-(3) depend solely on 8, 7, f(n) and Pr. The beginning of writing @ is easy, 29T-T,, we 7 ATycale o (n, Pr) The temperature difference scale AT scale must be chosen such that the wall boundary condition (2) reduces fo a statement involving only the similarity variables. We rewrite eq. (2) in an order of magnitude sense, aT __ATscaie eae ® and draw the key conclusion that Tyee = a)? O} ‘Combining eqs. (7) and (9), we find the needed transformation from T to @: THT+0-(RY-E (10) Beyond this point, the transformation of eqs. (1)-(3) is routine. ‘The partial derivatives of eq. (1) are’ ay (12) (13) Equations (1)-(3) become, in order, or +E (re-16)=0 (14) 6” ()=-1 (as) B(~)=0 (16) The problem represented by eqs. (14)-(16) was solved numerically for @(n, Pr). The similarity temperature profile for Pr = 0.72 is shown in the attached figure. ‘The key heat transfer result is the local Nusselt number an6 Pr=0.72 1 2 3 which is proportional to the inverse of the local temperature difference, cf. eq. (10), a1 pele Nu= gb Rel (1s) For Pr > 0.5 we expect a scaling of type Nu ~ Prl/3 Re,U/2, which would mean that (0) must vary as Pr-V3, Therefore we solve eqs. (14)-(16) for several large Pr values, and calculate the product 0(0) Pr!/3, to find the constant C in the proportionality 00) =C Pr! (9) The result is C = 1/0.463 if Pr > 10 [36]. For liquid metals, Pr << 0.5, we expect Nu ~ Prl!2 Rex!2, which means 6(0)=K Pr? 20) ‘We solve eqs. (14)-(16) for two small Prandtl numbers, and obtain K — 1/0.886 as Pr — 0 [38].In conclusion, the local Nusselt number asymptotes for laminar boundary layer flow over a constant-q" flat plate are Nu=0.463Pr'3Rel? — @r>0.5) Nu=0.886Pr!?Rel? = Pr-<< 0.5) Problem 28. a) We recognize, in order, the L-averaged shear stress, the total tangential force experienced by the wall, and the mechanical power spent on dragging the wall through the fluid: 664 p u(y? P=tL P=FU =0.664 p UL(f)!” = 0.664 p v2 S212 Tf ( )c and ( )h represent the “cold” and "hot" flow conditions, the dissipated power changes according to the ratio: Ph Pelee)" (Baie) In the cold case (no heating), Te = 10°C and Poel ey He =0.013 3 In the hot case, the heat transfer from the 90°C wall to the 10°C water takes place across a boundary layer with the film temperature Th = 50°C. The properties corresponding to Th are hy = 0.00548 aS ‘The power ratio Poof 9.00588)!" 9.65 P. “(1 “0.013 shows that the heating of the boundary layer decreases the dissipated power (as well as F* and 2) by 35 percent. b) By retaining the observation that the water density is practically constant, we find that the power that has been saved by heating the boundary layer is, 2121/2} Py Py = 0.664 p US2 LV [1-(4) | @ ‘The electric power spent on heating the wall is given sequentially by fe? Le 12 Rep RL = ef = 0.664 Pri Ref! yy!” ® ¥ 7L = 0.664 ky, AT Pri3( in which the properties (k, Pr, v) are evaluated at the film temperature Th = 50°C, ky=0.64 We, Pry =3.57 By dividing equations (1) and (2), we obtain (after some manipulation) a dimensionless measure of how effectively the heating of the wall has been converted into saved mechanical power: “oh ata a) -C)"] 4.18 kiékg-K, AT = 90°C - 10°C = 80°C and In this expression we substitute ch (vp/vo)!2 = 0.65, and obtain P,-P; stan) PL ‘This ratio is appealing (greater than 1) only when U > $16 m/s. ‘The length L, however, must be sufficiently small if the boundary layer is to remain laminar UL s Ve <5x10 L<5x 105 YB <5x 108 95 0.00554 ems <5%10° "51600 emis <0.54mm In conclusion, when L > 0.54 mm and the flow is laminar the electric power used to heat the wall is greater than the fluid-drag power saved. 213Problem 2.9. The film temperature in this configuration is (40°C + 20°C) = 30°C. The calculation of the total force F proceeds in this order: Ux => n —i = 5 Re, = 940.5 B 10m Ay 31x10 just barely) Op = 1.328 Rex}? = 0.00238 A=10m20m=200m? (roof area) FHA tyx=ACry 4p US k 2 = 200 m?0.00238 4 1.165 s (0.5 By = 0.069 N= .016 Idk In order to calculate the total heat transfer rate q, we must first evaluate the average heat flux: Ni, = = 0.664 (0.72) (3.1 x 105}? = 331.4 1.664 Pr Rel? fy =Na, R= 3314x0.026 We he ~ Ta} =0.862 (40-20 K 4=G,A=17.2 © 200m?=3.4kW m Problem 210. a) The local heat flux varies as Fx = EX" in which n= — 1/2, and Cis a constant. The L-averaged heat flux is therefore 214ae ey 12 w= Tem Sw = 2d = 2 CL ‘The position x where the local flux matches the L-averaged value is obtained by writing Cx 12 = 2 CL-V2 and this yields x = 1/4, The attached figure confirms this finding, because it is drawn to scale. o aL x b) The relationship between the local heat flux in the middle and the L-averaged flux is obtained by first writing daun=C(4)” ai.=200? Dividing side by side, we conclude that the mid-point heat flux is about 30 percent smaller than the average flux: Quin C2 gay, Sun _ C2 = 1 = 0. Gwe. | 2c 2M 215Problem 2.1. When the boundary layer is much thinner than the pipe diameter D, the wall friction force F and the total heat transfer rate q can be estimated based on the formulas for the flat wall: F=nDL ty =nDLT yy 4p UZ =RDL 1.328 Rept? 4p U2 a = 2.086 p U2, DL Ref’? q=mDL Gy. ="DLAT Ay, =mDL atk 0.664 Prl3 Ref’? @Q) = 2.086 k AT D Pr!3 Re} Dividing (1) and (2) side by side we obtain 4 pan &pAT pore, Problem 2.12. We evaluate the properties of water at the film temperature 20°C + 50°Cy2 = 35°C, and calculate in order: Ux 5 4 — Ss = 6.9.x 10° Regs 9> = 5 SP 100 em Tag oe Ganainat) Nu, = 0.332 Pr!3 Re? = 0.332 (4.87)"3(6.9 x 104)? = 148 h, Na, =" h nets exon ark T9186 ie B,=2h,=1833 YE Ww. WwW Oy = BAT = 183.3 30K = 5500 8 Since the total area of the wall of length x = Im is A=4x 20cm x Im =0.8 m2 the total heat transfer rate is G= yx A= 4400 W 2:16‘The velocity boundary layer thickness at the same location (x) 6 =4.92 xRexl2= = 4.92 x 1m (6.9 x 104)? = 1.9 em is much smaller than the 20 cm-side of the duct cross-section. In conclusion, the use of the flat wall boundary layer resulis is justified. Problem 2.13. The narrow-strip configuration isa case of wal wth non-uniform heat fox, in whicl 0, Ox) + Ax Substituting this expression in eq. (2.122) we obtain, in order: T,CO-T.= mite ("+ fee Low) xy tax 3/4)-28 “tere, by eas _ 0.623 fi (784 : kPrll3 Rel dy Problem 2.14. The flux q” covers only the front half of the plate length L. We use eq. (2.122) 28 0623 ue “(E473 Tw) Tee = FBS RET? Je-o |! @ ve ay ote 7 0.023 fro 1=m34)?8 am ‘The alternative calculation is based on eq. (2.121), in which we assume that the uniform flux q"/2 is spread over the entire length L: ¢ L (Ta) ~Telaoron = SaS5K PTB RET 2 ‘The relative goodness of this approximation can be seen by dividing eq. (2) by eq. (1): 27PFO. 30a Ty@)-T.. 0.623’ In conclusion, the approximation is not very good. Problem 2.15. The relevant properties of water at 20°C and atmospheric pressure are p =0.997 & Pr=7.07 - a = w v=o01 SB k=0.59-% ‘The L-averaged shear stress Tw, can be calculated in the following order, (@) = (40 U.7) 1.328 Repl? where the Reynolds number has a value that falls in the laminar range: Yel _o5m_ Lem _ Rey =F Torem ys 90° » Combining (a) and (b) we obtain twit 0997 5 a cosy = 1.328 (5000)? gm? N ° = 0.00234 £7 = 2340 By ‘The L-averaged heat flux can be calculated based on a similar sequence: @ 218Problem 2.16, The Prandt! number of air at 20°C is 0.72, therefore we can use eq. (2.121) as an approximate estimate of the local Nusselt number along the strip: Took £50453 Pr Rep? @ In this equation qj, is the heat flux to one side of the strip, qj, = (300/2)W/m2 = 150 W/m2, and x = Lis the position of the trailing edge where the temperature sensor is located, Tw(L) = 30°C. Noting further that k = 2.5 x 10-4W/cm K, we use eq. (a) to calculate Rej/?: ne Ww Ld Rel wil) Te, k 0.453 Prl8 = 180Wim? ____2cm —_. GB0~ 20°C 2.5x10-4 Wem K 0.453 (0.72) = 29.56 From this we deduce that Rey, = 873.5 = UaL/V, in which v = 0.15 em%s, The U.. velocity of the free stream is therefore 2 U,,= 873.5 21S emis - 9,66 Note that Rey, = 873.5 means also that the flow at x = L is laminar, ic. that the use of eq, (a) is justified Problem 2.17. a) Assume that To (unknown) is the uniform temperature of the interface between the two bodies in rolling contact. Consider first the temperature distribution in body no. 1, Itis a “thermal” boundary layer through which body no. I "flows" with the uniform speed U: -o, 2E uaa SF @ T=T at y=0 @ ToT, a you @ ‘The problem (1)-(3) is identical to that of the temperature distribution in a thermal boundary layer in the limit Pr— 0, When Pr is very small the velocity thickness 8 is much smaller than 8p. ‘This means that in the Pr ~> 0 limit the fluid flows with a uniform velocity through the thermal boundary layer region. The uniform longitudinal velocity (U) is a noteworthy feature of eq. (1), therefore, for the local heat flux we can simply use eq. (2.107): @A similar analysis of the thermal boundary layer that develops inside body no. 2 leads to an alternative expression for the same local heat flux qx: a x Ux) l2 Tp k) O54 a) © b) By dividing eqs. (4) and (5) we obtain an expression for the interface temperature. Tepe htyhh © in which r is the dimensionless group _ (ock)}? (pck)}? c) The local heat flux can now be calculated by eliminating To between eqs. (4) and (6), or between eqs. (5) and (6): a (gyre ® ‘The L-averaged heat flux that corresponds to this result is eo) 4) The preceding results apply as long as the two boundary layer regions are sufficiently slender. According to eq, (2.37), in the small-Pr limit the slenderness ratio (S7/L) is of order Pez!2.” There are two slenderness criteria to meet, because there are two boundary layers: (yr «, tanh x — 1, the gp expression of Problem 2.18 reduces to kg (Uso t)! ppyain]!? wate [nsznreis 4) 09] a) qp is proportional to b3/4, therefore if b increases by a factor of 2, qp increases by a factor of 234 = 1.68 b) Based on the above formula for qp, we conclude that (dee (2) (Ee) ga (oa) (23)ghy)"* = 13.62 2-22Problem 2.20. In the ease of non-zero longitudinal pressure gradient, the momentum equation reads a du _ 2 oe ay ge BUR vey or, introducing the streamfunetion yy = u, Wx = Wy Wyn Va yy = PUL + VW yyy Tis as Recalling that U.. = C x™, the similarity transformation consists of V= (Uv)! 1 =cl2 i231 gm) n =} (Us)Pay clay te Bt One by one, we evaluate the y derivatives, wy=Cxm ir a. Vyy =O veg iY Wyyy = C2 vl x2m—1 pv” aa - 12 mel citying ype B51 iPyingn-ty(G)"?r Substituting these into the momentum equation, the terms containing y (C/v)¥/2 drop out, and we are left with mey-meler = mse ee LHS RHS which can be rewritten as 26” + (m+ 1) ff” +2m(1-£%)=0 2.23‘We process the energy equation in the same way: Ev Eat LHS RHS The required @ derivatives are pol Oy= Cy x 7 e’Cy xml Substituting the @ and derivatives into the energy equation, the terms containing y(C/v)12 drop out from the left-hand-side of the equation, leg ea ~My te=fe" This result can be rewritten as 26” +Pr(m+1)f6'=0 Problem 2,21, The energy equation for a two-dimensional boundary layer with viscous dissipation is negligible, because the boundary layer region is slender, 8 << L 2-24‘The j1@ term is simplified based on the following scale analysis: vol By] + «at L + vu yt ve Ee these two terms are of the same size, U2/L?,. 2, 5 8 2 cf, mass conservation ie. neglect 2 (UJL~ vis) Ls Be Ox In conclusion, the 1. term has two scales, the first of which is much smaller than the second: Uz uh ornTy he sw We conclude that in the boundary layer 8 x L, ve in other words the energy equation reduces 10 HyyM- gFT , Huy “ox *¥ey ~ *az * 7a (9) SSF Ss SE convection conduction heat generation Un velocity heat % 1 generition ar Pomperature 2-25Continuing with the scale analysis of the 8-thick region, we seck to determine the wall temperature rise AT as a function of U.o and fluid properties. The energy equation involves three scales: u.at jt Ue -L Bop ge conv. heat gen. ‘Among these, the heat generation scale can never be neglected, because without the j1¢ term we do not have any heat transfer problem left. Two cases exist: i) convection ~ heat generation Writing AT~=2 or 6,=0(1) For what Pr values these results are valid, is shown by the conduction scale (neglected in this case): 1 prt 1 conv. cond. heat gen. In conclusion, case (i) corresponds to Pr > 1 fluids. ii) conduction ~ heat generation Claiming now a balance between the conduction and heat generation scales, 2 wAT | HUB BPR we find 2 ar HUs or 6,=0Pr) The three scales in the energy equation line up as follows: 2-26Pr 1 1 conv. cond. heat gen. Since in case (ii) the convection effect was neglected, the above results are valid for Pr < 1 fluids. ‘The remainder of this problem is routine. The energy equation must be transformed according to the definitions Teo y Un om=t =. nt oe ive usU.fm . iV neg where f(n) is the Blasius profile (see Fig. 2.6). We have ua) On the right-hand-side we obtain 2TH (aug Ua go, HB y2 gpm aS2 * Dep (3y) “%Aep St ae pe ae Equating the two sides we obtain -$8e = pes 2” conv. cond. _heatgen. or of + Be re+2Pre?=0 This equation is linear of the first order in @,(): its solution is, =f Lpvean eta Jim famead 2.27If all the integrals start from 7) =O, then the constant C must vanish so that the adiabatic wall condition 6, = 0 holds at) = 0, Aff gPctiman fE seee0e [Pore we a8 ee Integrating this expression from 1) = 7|to 7) =e, and using the condition @,(es) = 0 yields ayn=2ee[” {tee ‘The dimensionless wall temperature rise Q,(0) can be evaluated numerically for various values of Pr. These results should be compared with the scale analysis results for 8; obtained earlier in this solution, . The mass conservation and longitudinal momentum equations for a two- dimensional laminar jet, as a slender flow region, are +o=0 au, yu aP yy ou wey Ba pave J Pos = ox de” because P,, = const. jbl In terms of scales, the two equations dictate: 2.28In conclusion, we have one relationship between the unknown scales U and D, UD? ~ vx @ The second relationship necessary for determining U and D uniquely follows from the HINT given in the problem statement: the x-momentum equation can be written as K+ Fon-vy and integrated (at constant x) from y =~ 2° to y = +22, zeros, because in the far-field reservoir fluid u=0, uniform (note: Vie may not be zero.) We conclude that or f * \2dy = U3 Dp, constant ‘The scaling law implied by this result is UD ~ UZDp @ Combining eqs. (1) and (2) we find D~v25 x23 Ug? Dal? U~y8 18 Ud DzB 2.29Note: D increases as x3 from a “virtual origin" (v.o.) located upstream of the nozzle. Problem 2,23. This problem is a continuation of the preceding problem, ‘The energy equation for the same slender flow is, ar yy at PT ae ty Gy 5D Too In the Dy-thick region, the scaling law required by the energy equation is (AT = T - T=) AT AT . AT x" Dr De. Pr iT same size, from mass conservation in the x-Dp region In conclusion, 2.30where, according to the preceding problem, U ~ v-V3, x-13 Uy¥3 Dp23. Note that Dr varies as x22, just like D. The ratio Dy/D is Br _ ale _ 1 =(9)!? DMP DIE PaQBUAD AY in other words, Pr> 1 fluids Pr< 1 fluids (v.0. = virtual origin) To determine the x-variation of centerline temperature, we exploit the HINT and integrate the energy equation Zone Zon= over the entire x = const. cut across the flow, £f UPdyt vaTan vn T= 0(57) (3) T., Wa=V-ss), zeros, because Tis unknown to be uniform as y > +> determined from the mass conservation equation ‘The mass conservation statement means also that 2-31In conclusion, the energy integral reduces to a f Tay-T, £f = Kp ete a which means that d [* = £f u(T-T,.) dy =0, or £ u(T—T.) dy = Up (Tp - Te) Do, constant In terms of scales, we have (U) (AT) (the smaller of D or Dy) ~ Ug ATy Do “ To see why “the smaller of D or Dr" must be used in eq. (*), consider multiplying the u and AT profiles when D and Dr differ appreciably, say, when Pr > 1 r 1 ln of ender p+jo.. —2-. (werd, ), | ¥ Secause D_ is Phe smalle~ velocity, Temperature product of Dand D, “ dference, “(rr ) in this cave, TT, ‘Therefore, in order to determine the AT scale we consider two separate cases: Pr>L, or Dr Fey ten, * assuming that pyiguia >> Pair (for brevity Priguid = P)- 2-33As boundary conditions, we invoke 0 at y=0 (no slip at the wall) us & =0 at y=8 (zero shear free surface) The terminal velocity distribution is therefore uy) or, in dimensionless form, where U is the free-surface velocity Moving now to the heat transfer part of the problem, the energy equation for the film is or ai, eT US te By 8 Sr L zero, if terminal velocity is reached (for proof, Consult the mass conservation equation) 234reference Pemperatire, Great AT Very close to x = 0, ie. when &r is much smaller than 8, the velocity scale in the 5y-thick layer is iT vu). In the Sp-thick layer, the energy equation represents the following equivalence of scales Sry) AT _ 4 AT (0) at oa which means that xo8)!3 br (8) The wall-film heat flux near x = 0 scales as qd ~RAT at (205 ye A more accurate result is produced by integral analysis. First we write the energy equation as a a er Zon+ Zon Pa 72i0, because v =0 and integrate it from y= 0 to y = 6 “6- 2 ( wrayeal@) —o (2 & fy ote=0(3)),, (Fluo Nev hats zero, because the fluid above y = Sr is isothermal (T = To). ‘Assuming the temperature profile sketched earlier, 2.35tty ar[i-2¢ +(Z)| and using also the terminal velocity profile derived earlier (and skipping the algebra), yields an equation for B(x): a (& - 4°) 2a axte~ 30)" 4 DF Integrating from x = 0, where A = 0, we obtain aor s where A= —— > ax 0.04 Te Us? ‘The solution expires when A reaches 1, i.e. when the free surface first feels the heating effect. Let x=x1 where 8p = 8: which means thatDownstream from x = x) the integral solution must be modified to allow for the temperature variation along the free surface. As new temperature profile we can use: Osysé where Ts(x) is the free-surface temperature (Ts is the unknown in this integral analysis). Otherwise, the integral analysis is the same as for the 0.< x Oas y > 2. ‘The wall condition £(0) = 1 pinpoints ¢1, 0 1 cy -(f ew an} therefore 2 [> en gne f #f eT an=1-erf (n), and the near-wall velocity field is uo t-et(za) At the time t= ty, the wall temperature is changed to To. ‘The energy equation for the time- dependent boundary layer is 2-38a * a okt fo) ) at ‘The energy equation is analytically the same as the momentum equation solved earlier. A similarity solution can be derived exactly in the same way, by selecting T-T.. Trt 7 en) ‘The solution is “Geea") The velocity and temperature fields derived above are valid in "boundary layer” regions, because these regions are "slender" in the extreme (the wall length is infinite), and because the flow region slenderness was invoked early in the analysis, in order to simplify the momentum and energy equations. The temperature field is independent of the velocity field, unlike in Pohthausen’s problem. Problem 2,26. We write Ty, for the uniform temperature of the substrate, and To(x) for the nonuniform temperature of the wetted surface of the coating of thickness ((x). ‘The uniform heat flux q" can be related to the temperature difference across the coating, Ty — Tox) “mw w From this we deduce that q Ty Tao = To) Tea et) @ for which [To(x) - Te] is furnished by eq. (2.121). We find that 2.393) ‘The condition that C must be constant leads to the needed t(x) relation o_o)? 1) _ 0.453 @ L Er Rel? w ‘The L-averaged thickness of this coating is Finally, by eliminating C in favor of 7 using eq, (5), we obtain the coating thickness variation 2_pxyn woe, 3() « LOL k pl pel? 0.453 K Pr! Re} ky and the temperature of the isothermal substrate, Qt 1472q’L " Tw-Ta= ot + GB) ky rl Rel? ‘There are two terms on the right side of eq. (3°) because the transfer of q"L from Ty, to Te is impeded by two obstacles, the coating (the first term) and the laminar boundary layer (the second term). Problem 2.27. When Pr =0, 8/87 is zero, and this means that u = Uw in the Sy-thick region, In this limit the Pohlhausen problem statement is replaced by [33] Tey=T+t Ry" a) 12 c=y(¥e) @) veh (Cr-9=0 @) 2-40vO = t(@) = 0 4) Separation of variables is achieved by differentiating eq. (3) once, $ (6) fll v2 This equation is then integrated sequentially three times, and, after using eqs. (3) and (4), the analytical expression for the similarity temperature profile becomes 7) (6) va This shows that the value at the wall is (0) = 2/m' Nusselt number is = 1/0.886, which means that the local uy? x aX 2 = 9X 1 flow is Us, L ne Nuo 1 = 0.664 Pr!/3 (224) 6) or, in view of eqs. (2) and (4), Nup_p = 0.6647/? Bel’? © In scaling terms, the Nusselt number is the same as the Be, number raised to the power 1/3. 2-42Chapter 3 LAMINAR DUCT FLOW Problem 3.1. From eqs. (3.5) and (3.10) we have xD -2-2_ zz $(om-2-Z-16 nm) 12 CyRep= $ where m =Ud/U. The above system establishes a parametric relationship between (Cix Rep) and (x/D Rep), via parameter m. This relationship can be plotted (as in Fig. 3.2) by first assuming values of m, and then calculating both (Cj, Rep) and (x/D Rep). A few points along this curve are calculated below. x m Dep Ctx 1 0 00 duct entrance il 8.56 x 10-4 32.27 13 88x 103 15.02 15 2.6 x 10-2 12.00 beginning of fully- developed region Problem 3,2. The fluid occupies the annular space rj < r < ro; the appropriate momentum. equation for fully-developed flow is or‘The general solution and the two boundary conditions are =P 10 =F terns ey u=0, at u=0, at hence Omitting the algebra, the average velocity U follows from the definition un(2-2) f f° u@ rdrd9, and the result is The perimeter-averaged wall shear stress tayg is defined as Fave 2h (fo tH) = 2M ro To + 2K 5% g2 (fo +1) fo T+ 2k total ve where sal = on-4Q) = In the end, we obtain Tawe=— 5 FE lfo-9) ‘The above results can be condensed in the friction factor formula 32namely r=,i6 (=m? Tm Py | ¢m2y tom? In(m) where up, 4n(2—29) Rep, =v" and Da= eee ery 2 lm) The student will find that the algebra required by the [u(r), U, tavg, £] results is lengthy and, possibly, a source of errors. It is always a good idea to check the validity of “difficult” results by reducing them to well known limits. In the case of Hagen-Poiseuille flow through an annulus, two classical limits come to mind: i) round tube, rj = 0 or m= 0; in this limit the friction factor reduces to = 16 to! Rep” which is the correct formula for a round tube. ii) parallel plates, rj > ro orm — 1, or € > O where m= 1-e, 2 din, f= gl (-i+ey Pe Lede This is the correct formula for parallel plates. Problem 3.3. The scale analysis of eg. (3.30) indicates that the 22w/@y? term may he neglected relative to the 42/022 (note that this simplification does not mean that u is not a function of y) Lae Pu, ou H dx” ay? ta Lo ot vo?” oa 33flow direehon (U= average velocity ) Integrating Lap _ eu Hoax ~ a2" and claiming that at z=0 at z=ey yields the flow distribution P (ey) The cross-section-averaged velocity U can now be calculated: 2 b [rey eb L =-—LaP ,354 us -[ If veya ary = Tea’? cross sectional area ‘The perimeter-averaged wall shear stress Tayg is defined by » » Twe 2 =f + ays f 1 dy total 10 bottom wall 0 topwall perimeter where 34‘The result is ep wap te a dx Noting that ‘we find that the friction factor obeys the relation f Rep, = 12. Problem 3.4. The decomposition of problem (A) into (B) + (C) can be summarized graphically as shown below [note that the differential equations and boundary conditions of (B) and (C) add up to those of (A)]: + i ~ « 2 Ke A ? _ ; = uly.z) = u(y) + ua(y,z)The (B) problem is the same as the fully-developed flow between parallel plates; the end result is wd A-( The (C) problem accepts the general solution ug =K (os ay +0, sin cry) (cosh az +m, sinh az) in me (A2=oay=0) (M2020) hence up = K cos ay cosh ow. From up = 0 at y = +a/2, we obtain _Qn+in ee in other words p= 2 Kn C08 Oy ¥ Cosh Cy 2 To determine Ky we multiply both sides of the above equation by cos amy, and integrate them" from y= 0to y=al2, The right-hand-side yields a finite term only when m =n: ‘a2 ral2 f ~wy() e084 y dy =K cos AED ED cos? Gin y dy QT OA cos? (dummy) d dummy) — ~~ — 1 © demmy 3 (D[2n+1) F] (sce dimensionless dark area in the sketch) * while holding z constant; 3-6The left hand:-side of the last equation is integrated using the earlier result for uy(y) and the formula f x2 cos x dx = 2x cos x +(x2—2) sin x ‘The Ky result is therefore 18 4a2 gE Qn+ Dab] 2a Ky (2n +1)? 13 cosh Problem 35. After setting = PY ogg RZ sug cos Fp cos 42, eqs. (3.33), (3.35) and (3.36) become: abe ~4p.up(8+4) po 2r ab? Rep, (a+b)? ‘The friction factor has the limiting values: £ Rep, = 19.74, for flat cross-sections £Repy = 9.87, for square cross-sections Figure 3.6 indicates that the assumed cosine form of u(y,z) is not as "reasonable" as the parabolic form treated in the text. Problem 3,6, From tables of mathematical formulas we learn that the area of a regular hexagon is Aauct = 2.598 a2, hence and 3-7DRA Aguct = 0,907 flew ale ‘The table is completed as follows FRep, = (16) (0.907) = 14.51 3. 3.7% relative error Nut-const. = (3.66) (0.907) = 3.319; 1.0% Nug'sconst, = (4.36) (0.907) = 3.954; 17% The usefulness of the new group 7Dy?/4/Aquct is illustrated further by Fig. 3.7, where ED2/4/Aduet varies from 0.5 to 1.67. Problem 3.7. We seek the proper value of Nu so that Nu=-2 (9,21 (3.73) where $(r,) is the solution to Li ual 2 (eo) =-2Nu(1-A)o (3.71) subject to conditions 6°=0, at ry=0 (3.72) 6=0, ata Making the initial guess 49(rs) = 1, eqs. (3.71) and (3.72) yield the first corrected guess 1(t4): 2 (94) =-2Nu(1-8)q) 38‘The second guess $2(r+) is obtained by repeating the above analysis Eleoy 2Nu(1—18)o4 (4) and the result is = 2 | lL 5 oy 4 Nu (-Sids sled Se ahs he B+ 0.02724) Tn order to use eg, (3.73), we calculate (@))-se1 =- 0.1146 Nu? hence, Nu =~ 2 (- 0.1146 Nu?) or Nuy = 4.364 This estimate is 19% higher than the correct value 3.66. ‘An improved estimate can be obtained by repeating this type of analysis: the equation E (0-65) =-2Nu(i A) on(04) yields 6 3, which is substituted in Nu=~2()r01 to calculate Nug, and so on. Problem 3.8. The analysis follows the steps outlined in eqs. (3.56)-(3.65) in the text. For fully developed flow and heat transfer in the parallel plate channel of Fig. 3.1, we must solve the energy equation uy) or _2r a Ox = 52 a in which Q@ 3)av _ ST ___20” x dx pepDU @ Tox) - Tey) _ TyG)= Tyo) “9 © To(x) — Tm(x) = AT, constant (6) Substituting eqs. (2)-(6) into eq, (1) we obtain a (8) and invoke the symmetry condition dg/dn = 0 at 1 = 0, to find that Cy = 0, We integrate eq. (8) once more, = Nu(Co-n?+ Ent) ® and invoke the wall condition T = To at y = D/2, which means @ = 0 at n= + 1. This condition yields Co = 5/6. ‘The final step is to determine Nu such that the definition of the mean temperature difference holds: Tp-T, 1 ppe [ppt to-Dey (10) ‘Combining this with eqs. (2), (5) and (9), and dividing by AT we obtain ab which yields Nu= 140 - 8,235 a2) 3-10Problem 3,9. The energy equation for fully developed flow between two parallel plates reduces rn Soe a where U is the uniform longitudinal velocity (slug flow), and y is measured transversally from y=—D/2 to y= D/2. The similarity temperature profile is defined as = To-T&y) = Tao @ where To = constant is the wall temperature, To repeat the analysis that led to eq. (3.70) in the text, we write the first law 2a @ ‘dx ~ pepDU where q" = h[To —Tin(x)], and obtain To=Tn(o) _ wD ToT exo(. Nu 2) @ with Nu= Be, Pep= WP 6) “The problem statement for the 6 profile is obtained by using eqs. (2) and (4) in eq, (1), —Nu g=9” N O=¢ © $0) =0, oe D=0 M where © = 6G), ()' = d( dy, and 8) ‘The solution to eqs. (6)-(7) is $=C-cos(F§), and Nu=n? 0) ‘The amplitude C is determined by writing the heat flux released into the fluid by the y = plate, D/2 31(10) which can be rearranged as, Nu ig 4 Cl) a Combined, eqs. (9) and (11) yield C = n/2. Problem 3.10. With reference to the sketch shown below, the total heat transfer rate per unit pipe Tength q{W/m] can be written as To-Ter ‘total thermal resistance’ per unit pipe length Or, keeping in mind that q” is the heat flux based on fo, anna’= Ty-T. "A Tn), 1 Qnky — hy ety we obtain Tota] thermal! fesistance 3-12Problem 3,11. We look at Fig. 3.11 and see that at each x along the pipe the heat flux q” is, impeded by two resistances in series, I/he and 1/hj, where wa nned Ty en - 22 -- general case he he ma 1 mm 1 and Tp = Tee and q” =h, (T..- Ty 0 a’ =he(To—To) ite. To = constant constant Consider the last limit, where he is negligible compared with the fixed quantity (k/ta). The. continuity of q” through the pipe wall requires q” = he (Teo ~ To) = hj (To- Tim) in which hj ~ k/ro. Therefore, if q” is finite, then (T..~ To) is infinitely larger than (To ~ Tm). Teo To > To- Tm The temperature variation AT measured along a fixed length of pipe L scales as ATo __k DT ihep (To-Tn) Returning now to q” = he (Two ~ To), we note that over the length L the temperature difference Tox —To changes by ATo. However, in the limit Bi > 0 considered here, ATo is negligible compared with T..— Tp because 3-13ATy__ T.-Tp fixed —0,asBi>0 In conclusion, along any finite length L the temperature difference Tw. ~ To is practically constant in the small-Bi limit, which means that constant Problem 3.12. In the case of slug flow, the energy eq. (3.96) assumes a simpler form 1 (yg) 28 ee, 1 BB 2 oe 9g 1h Ore Hagen-Poiseuille profile, must be replaced by 1/2 ‘The complete problem statement is (©) (Be 1) (BC 2) ac) Oe=1, at re=0 where nt te=thy 1p Pr Separation of variables is achieved by setting 04 (t,x) = R (14) +5 (a) and the general solution of the energy eq. (E) is, 3.148. =[A Jo (Ore) +.B, Yo (a re)]e4o?x+ J zero, because of (BC 2) From the first boundary condition, eq. (BC 1), we have Jp (@-1)=0, ie. on = 2.4048, 5.5201, 8.6537, ... n=l n=2 n=3 ete. In conclusion, Gx (rm, x4) Ag Jp (Oty te) eA OR 1 To determine the Ap coefficients, we use the initial condition, eq. (IC): I= Ze Jo (Ont) Multiplying both sides by re Jo (im r+) and integrating from rz = 0 to re = 1 yields [f= tolomree EAs [P Jol) Yolen) rede zero, unless or eH (On) = An 3 [5 (On) + Jon) hence Putting the solution together, we can write T=Ty > $ Joly") sore, Tin—To n= 14y J1 (Gq) which graphically looks as follows:(1X q (1reaces 4 =z X20 5 f* ° Le oO 4 7% Tb Note that this problem is analytically the same as that of transient conduction in a cylindrical rod with uniform initial temperature Tyy and constant wall temperature To; the energy equation for the conduction problem is which is the same as (E) (oh is in this case the thermal diffusivity). Problem 3,13. Let x, = 1; the Nux and Nug_x values depend on two series: S\= Ec Dare = (3,339) 10-7 + (9.739) 104 +... ae z S negligible, especially if x+ > 1 es = (4.566) 10° + (2.183) 104 +... = § Gn wd, S2= 2° ‘The numbers on the right-hand-side of S1 and S» are from Table 3.4. Next, we evaluate one by one the quantities defined by eqs. (3.102)-(3.104): 8x, = (8) (4.566) 10° = (3.653) 107 6.339) 107 (2)(4.566) 10° Nugx = 4 In —L—. = 3.706 Me 4 653) 107 Nu, 656 ‘The last two values are very close to 3.66, which is the fully-developed flow Nusselt number. 3-16Problem 3,14. Assuming that in any cross-section through the tube the twisted tape appears as a rectangle of thickness d much smaller than D, we have: = —0.884 2 Lo ond hi =0611 of! ones + ems 0 ($5) na nies) = 0611 D(1~0.884 4) Problem 3,15. We know the following quantities: th = 10 g/s q”=0.1 Whem? B + 1=D/2=1em ‘ ? 0.01 gia Ply a H=0.01 gf(ems) Z k = 0.006 W/(cm K)a) The Reynolds number is ‘We conclude that 2 ‘The product pU follows from pU ED. 3188 «> om. GMS. = 637 cm? s 001g Re= The flow through the pipe is laminar (see Table 6.1). b) The heat transfer coefficient follows from the definition of Nusselt number Nu =D = 4,364 (Table 3.2), therefore, h= (4.364)(0.006) -W —1_ = 0.0131 ¥ mK 2em om’K ©) From the definition of h, namely q" = h (To ~ Tn), we deduce Tq =0.1 Wom? K_ To Tn=0.1 3 0.0131 W 64K Problem 3.16. a) The effectiveness is defined as eo nck Ti-Tw hence Tz =T) - 0.85 (T) —Ty) = 6°C b) Treating the tube as a control volume, we find that the total heat transfer through the tube surface is a= fr a" () RD dx = tie (Ty —T) 3-18wee CS raas | 1 + | Z heat fle distribution Numerically, Q= 100 § 4.182 x (40 -6)K = 14.22 kW; this calculation is based on neglecting the enthalpy change associated with pressure drop through the tube (see Table 1.1 for incompressible fluids). c) In order to find L1/L when e1 = €, we must determine the relation between € and geometry (D, L). The heat transfer per unit tube length is “L Zz dQ =h (aD dx\(T-T,,) =- the dT. which integrated from T(x = 0) = Ty yields ~beDx TO)-Tw=(Ty-Tye and, in particular, ~bgDL T=Ty+(T-Twe Based on the above, it is easy to show that 319e=l-e ™ Next, we reason that if € = €1, then ADL = hyDyLy in which hD = hyDy because BD 23662 Pt 3 66=E In conclusion, the new length must be the same as the old one, L=L; Problem 3.17. As the fluid is extruded through the tube it tends to heat up because of friction. ‘The cooling effect provided along the tube maintains the tube wall temperature at Tg. To find the fluid temperature distribution T(r) and Q, we must solve the energy equation we Qe) Gy constant Leeyade WH Qfiatts] for tube length L subject to the boundary conditions ote = = GP=0 ar=0, and T=To at r while keeping in mind that, from eqs. (3.22), =2uf (27) =O vaau[t GP] md Ua CR Since the algebra associated with solving for T(r) is straightforward, I show only the final result: 3.20‘The total cooling rate for a pipe of length L is Q=2n ry Lq’= 8x uU?, which can be rewritten as Q=8mLpUxU note: v= 82) - (_ ap =u Leg tivp AP (pressure drop across length L) In conclusion, we arrive at Q= gk “ To see the thermodynamic significance of this result, let ns assnme that the fluid is incompressible (see Table 1.1), and that the pipe of length L is part of the flow circuit (cycle) sketched below, The cycle consists of two processes: (1 > 2) = constant temperature cooling of the stream th (29 1) = reversible adiabatic (ic. isentropic) compression through a pump 3-21(Tend % are bule stream tempentures) ‘The first and second laws of thermodynamics for the control volume that contains the entire cycle state that: Q=Woump 1} Q Sgen= 7? @ It is not difficult to apply the first law to the pump alone, to show that Wpump = AP/p; thus, eq. (1) becomes a faster way of arriving at the conclusion labeled (*) on the preceding page. “The thermodynamic significance of Q = th AP/p is that it represents the pump work (per unit time) used to extrude the fluid through the pipe. This work is irreversibly lost, hence Q= Whos, and from eq. (2) above, ‘Wiost = To Sgen- 3-22Problem 3.18. The relevant properties of water at SO°C are Ww g£ k= 0.64 p =0.988 =, mK om? 2 _ om: eae y=0.00554 cpa We begin with the calculation of the Reynolds number, to be sure that the flow is laminar: Dn=2D=2em UDy om 8 Rep, = =3.2 2 2em—+_, “ny s 0.00554 cm? = 1155 (laminar) ADy Nus——= E7436 fully developed) dT _P dw . 2W dw ‘& “A pepU WD pepU 800 2,2. —___™ __ Olm 9.998 8, 4.183.290 om: ek s aie =1215C 3-23For the thermal entrance length, we use ¢q. (3.90): Xp 0.1 Rep, PrDy =0.1 x 1155x3.57x2.em=8.2m In conclusion, the length of the parallel-plate channel must be considerably larger than 8m if the above calculations are to be valid. Problem 3.19. a) The hydrostatic pressure distributions Pc(y) and Pp(y) must cross at y = H/2 so that the height-averaged pressure is the same on both sides of the door, rol) =Pel ‘The pressure difference that drives the air leak through the bottom gap is AP =P. (0)-Py ()=pee # - Pye H H a) = Ape H, where Ap =pe-Pp ‘The gap is a parallel-plate channel (D thin, W wide, L long), therefore Q) @ ‘The warm chamber (Th) loses energy by convection, because of the th counterflow, ‘warm over the top of the door, and cold under the bottom: = Hija tipo (A= specific enthalpy) @) =thep (Ty To) b) For the numerical part of the problem we have 3.24T.= 10°C T,=30°C = kg = kg Pea l247 Spy 16575 H=22m, D=0.Smm, W=1.5m, L=Sem We evaluate the other air flow properties (V, cp, p) at the representative temperature of (10°C + 30°C)/2 = 20°C, so that we perform the flow calculations only once (instead of doing them for cach air gap separately). We obtain in order Ap = pe~ Py = 0.082 * m (0,0005m)? 1.5m Im oi? £7 19 x015 OO o.05m = 1.843 x 10 8 = = 5 kg ki q=thep(Ty-T) = 1.843 x 10° S 1.006 Eg 20°C =0.37W Itcan be verified that the air flow in the gap is laminar (Rep, = 1.4), and that itis also fully developed over most of the gap length L. The important conclusion made visible by eqs. (3) and (4) is that the air leak th and the heat leak q are proportional to the gap spacing cubed (D3), For example, if D is 2mm instead of 0.5mm, the heat leak q jumps to 24 watts. 3-25Problem 3.20. If on either side of the hot blade the group hAwhi cp is much greater than 1 (Say, greater than 3), then the outlet temperature is practically the same as Ty. For the two sides of the blade we write 41 = ¢p(Ty—To) 42 = thy cp (Tw — To) In the case of fully developed laminar flow, we have in which W is the width perpendicular to the plane D x L. We write next Dy=yD and. Dp =(1-y)D and calculate the total heat transfer rate q=q1+q2 The result is 12uL Ty-To pcpWD? AP =yh+(l-y? The following table shows that the thermal conductance q/(Tw ~ To) is the smallest (i.e. the worst) when the hot blade is positioned in the middle of the channel. The thermal conductance is four times greater than this minimum value when the blade is attached to one of the walls of the channel: 3-26Problem 3.21. ‘The relation between the total heat transfer rate (q) and the largest temperature difference (Ty ~ To) is provided by a combination of eqs. (6.103")-(6.105), hAy q = thep = T9))-o[- } a) ihep where pWD? AP ° nL W = width perpendicular to the plane D x L. @) Nu = 4.86, (Table 3.2 iu (Table 3.2) @ (note: Dy = 2D) n-Ne kK Aci xw (6) 2D bA Lt)’ w po =.= 6Nu HE | E 6 thep "AP? (3) ° Next, we write the symbol Be, for the dimensionless pressure drop 2 Be, = SPE o ho. and substitute eqs. (2)-(7) in eq. (1): a= z k (Ty, — Ty) Bel! * 8° [1 - exp(— 29.168-4)) (8) where 6 is the dimensionless spacing, D pel’ 3-2 Be 9 r Pee ) ‘The maximization of q with respect to 5, namely Gq/@8 = 0, yields the equation 4 4 exp(a) = 1+ 3° with a = 29.166 (10) The solution of this equation is asp, = 0.5502, which means that 3.9: = 2.70, and 3.27D, ' Ht = 2.70 Be! L ay By substituting eq. (11) in Eq, (8), we obtain the corresponding maximum thermal conductance 4 a =_ = 0.693 Wk Be 12) [= 4 =| et (12) Or, if we write {" = q/WL for the average heat flux, we obtain finally [ets © = 0.093 Bel/4 (3) Tm) k In conclusion, when the, pressure difference AP is specified, the optimal spacing has a certain value that varies as AP", and the maximum conductance increases as AP", Problem 3.22, (a) The highest temperature of the constant-q" board occurs at the trailing edge, Th=Tw@=L) where the relationship between the wall temperature and the fluid outlet temperature is a’ Py _ — Nu = 5.385 (Table 3.2) Th ~ Tou * We note that D, = 2D, and conclude that q2D i— 1 kNu 0 Ty ~ Tout = ‘The outlet temperature can be calculated based on eq. (6.39), at, mo P _4 2) dx — A popU where p = W, A= DW, W = width perpendicular to the plane D x L, and APD? @ 12pL By integrating eq. (2) from T ~ Tp (at x = 0) to T = Toy (at x = L) we obtain aL Tout — To = (4) out ~ To = SoD 3-28‘Next we climinate T,« between eqs. (1) and (4), and then use eq. (3). The result can be arranged as follows 3 Th=T _ 12 {L) , 2D 6) q'L/k ~ Bey, Nuc where 2 Be, = PE oy no The excess temperature [eq. (5)] can be minimized with respect to the spacing D, but since L is fixed, it is easier to perform the minimization with respect to the dimensionless parameter 6 = D/L. The optimal value of this parameter is v4 ore -(s Nu 4 = 3.14 Be;!/4 0” and, after substituting it in eg, (5), the corresponding minimum Tis given by Th -To k a4 Th=T] K Lissap 8 (28) sag » Turmed upside down, this result spells out the maximum heat transfer rate that can be removed by the stream when the trailing temperature of the circuit board must not exceed the (specified) ceiling temperature Ty: a L Th — To k (b) The corresponding result for the board with uniform temperature Ty is (see the preceding problem) 0.664 Bel! + 1) 14 0.693 Bet! (10) ‘When the board ceiling temperature is fixed in both designs, Ty = Ts, the maximum heat transfer of the isothermal board exceeds by 8 percent the maximum heat transfer made possible by the board with uniform heat flux. The reason for the 8 percent difference is the leading section of the isothermal board, which is considerably warmer (ie. higher above Ty) than the leading, section of the constant-flux board. 3-29Problem 3,23. _ Examine the cross-section through the flow passages and fins, which is shown in the attached figure. Assume that the heat transfer from the wall (Ty) to the fluid flow (bulk temperature T)) is due mainly to the fins. In other words, assume that the heat transfer through the unfinned wall patches of width D is negligible. ath +~De For cach fin, the heat transfer rate (per unit length normal to the figure) can be calculated with the formula fin =2X2XL(Ty-T)1 a) where the factor 2 means “two side figure. The fin efficiency is, e. the wetted perimeter per unit length normal to the ~ tanh (mL) mL ® where hp 2 melee) * ® In eq. (3) we substitute p = 2W and Ag = 1W, where W is the dimension normal to the figure. We also assume that the flow is laminar and fully developed, and that L >> D. According to Table 3.2, the Nusselt number hD, Nup, = * 4) 3-30varies between 7.54 and 8.235 depending on how we model (approximate) the thermal boundary condition over the lateral surfaces of each fin. The key here is that Nup, is a constant (called Nu for short), whose order of magnitude is known, Remembering that Dp = 2D, we conclude that ke h=Nugy (5) and ie L L=(Nust} Lb 6 " ( ky) yl? ° ‘The total number of fins spread over the wall of breadth B is B/(t + D). Therefore the total heat transfer rate released from the wall to all the channel streams is B t+D =_B Ler By Nakek (ty-TIN fin a” It is most convenient if we nondimensionalize the total heat transfer rate by dividing eq. (7) by kw(Tw - TOBIL, and call it Q, 5-18 * ® Equation (7) can be written sequentially as ® (10) ap and x is the dimensionless fin thickness, 331D (2) The fin thickness t (or x) is the dimension that must be chosen optimally, while holding the other parameters fixed. That Q has an optimum with respect to x can be seen by considering the following extremes. When x << 1, Q increases as x12, and when x >> 1, Q decreases I/x. ‘The function Q of eg. (10) can be maximized numerically for a given b, and the results are b Xopt Qrnax q Ol 0.057 0.0089 0.945 02 0.113 0.0322 0.896 0.5 0.270 0.152 0.775 1 0.498 0.419 0.627 2 0.809 0.971 0.439 4 0.989 1.999 0.248 10 0.999 5 Ol 20 0.999 10 0.05 In conclusion, when b2 2 te opimal fn thickness is epproximately the same asthe channel spacing D. When b is smaller, say b $ 1, the optimal fin thickness is approximately proportional to b, = 0057 a3) ope _ Le (14) Said another way, when b £ 1 the optimal fin thickness is such that the slendemess of the fin profile equals the quantity listed on the right side of eq. (14). Problem 324, When the board is positioned off center, say, Dy > D2, the ny stream will be larger (and-a better coolant) than the tiny stream. ‘The board surface cooled by the my stream (surface no. 1) will have a temperature T(x) that is lower than the corresponding (aligned) temperature of the surface cooled by the th stream, T2(x). The local temperature difference Tz =Ty drives a conduction heat flux through the board, from surface no. 2 to surface no. 1, wk, a= *#(T)-T) @ The heat flux generated by the electronics mounted on each surface is q" = constant, The heat flux removed by the tiny stream is larger than this, because of the qz contribution, qed’ +a @ 3-32For the same reason, only a portion of the electrically generated q” is removed from surface no, 2 by the ty stream, =a @ ‘The heat fluxes (qj,.q3) and surface temperatures (T1, T2) are functions of longitudinal position, x. The temperatures increase with x, reaching their highest values at the trailing end (=D). TQ)=Tha TyL)=Th2 @ The larger of these two values is the most critical item in the design: our objective is to minimize it. We obtain the temperature distributions T(x) and T2(x) by making the simplifying assumption that the temperature increase along each surface [for example, T1(L) - T1(0)] is. considerably greater than the local temperature difference between the surface and the corresponding stream. This assumption becomes better as the D x L channel becomes more slender. It means that we approximate the local stream temperature as being almost equal to the temperature of the neighboring spot on the surface. Consequently, the first law of thermodynamics for a dx slice of the Dj subchannel is written as tiny cp dTy = qj W dx oo) in which W is the width perpendicular to the D x L plane, Note that only one surface of subchannel no. 1 transfers heat to the stream, ‘The corresponding first-law statement for the second subchannel is thy cp dT, = 93 W dx © ‘The mass flowrates thy and thy are driven by the same pressure drop AP, which is ‘maintained across the entire assembly. Assuming that the flow is laminar and fully developed along most of the length L in both subchannels, we write a, PW AP) 3 =e Pl a” fay = PW AP ps fy FB @®) At this stage, we have all the ingredients that are necessary for integrating eqs. (5) and (6) away from the entrance, where T,(0)=T,(0)=Tp 9) This operation is considerably simpler (and safer) if we restate the problem (1)-(8) in terms of the following dimensionless variables: 5=§ longitudinal coordinate 3.33= temperature of surface and stream no. 1 AT scale 0= Tr-To temperature of surface and stream no. 2 AT scale scale of longitudinal temperature tise [picked like this in order to "clean up” eqs. (5,6) to appear as in eqs. (10,11)] spacing of subchannel no. 1 l-y= Po spacing of subchannel no, 2 The problem reduces to integrating for ;(E) and 62(€) the two equations 8; PEABO -e) (10) ~B (2-6) ay by starting from the inlet, where @1(0) = 62(0) = 0. The dimensionless group B accounts for the transversal conductance of the board (ie. its substrate), 2 sais =12 a ae (2) First, we eliminate @; by adding eqs. (10) and (11), 1-38-52) a3) We substitute this into eq. (11) and obtain a single equation for 8, a +8 «ay & la-y ‘The solution to this first-order linear ordinary differential equation is (recall that 82 = 0 at § = 0) 3-34ox =(B-B) lex v1 + FS as) with the shorthand notation 1 wre “ 1 a ay (7) =—2B__ owe “8 ‘The highest temperature of surface no. 2 is at the trailing edge, ®y,2 = @2(1), namely sr2=(p-B\L oc ol+p a9 ‘This temperature is a function the position of the board (y) and the board conductance number (B), as shown in the attached figure. 3-35The highest temperature for surface no. 1, 1, is obtained by switching y and (1 —y) in the 4,2 solution (19). Graphically, this is the same'as superimposing on the attached figure another set of curves (for p,1) that are the mirror image of the ®},2 curves (the mirror is the y = V/2 vertical line). On the composite graph that results, we seek the design (board position yy that results in the lowest 0}, and ),2, when B is specified, ‘The answer depends on whether the board substrate is a good conductor: a)_ When B is of the order of 1 o larger, the @),1 and Qp,2 curves are bell shaped and fall on top of each other. The lowest temperatures are registered at'y = 0 and y = 1, i.e. when the board is positioned close to one of the insulated walls of the channel. The worst position is in the middle of the channel, y = 1/2, where the highest temperature rise (8,1, or O4,2) is about four times greater than when the board is mounted close to one of the insulated walls. b) When the board is a poor thermal conductor, such that B is smaller than the order of 1, then 6,1 and 8,2 curves intersect forming a cusp (a V-shaped valley) at y = 1/2. That intersection corresponds to the lowest (8,1 = ®},2) values, indicating that the best position for the board is along the midplane of the D XL chanel. ‘The student may wish to examine eq, (19) more closely, to determine the exact B that marks the transition from the optimal design for a) highly conducting boards, yopt = 0,1 ) poorly conducting boards, Yopt = 0.5 That “critical” B_value is obtained by setting 0p,2(1/2) = @p,2(1), and solving for B. Itis absolutely fascinating that the optimal design for boards of type (b) [namely y = 1/2] is the same as the worst possible design for boards of type (a)! This observation stresses the importance of the dimensionless number B: this must be calculated early, to determine the problem character, (a) versus (b) Problem 3.25. The analysis begins with the calculation of the heat transfer rate in the two asymptotic regimes noted above. Later, the optimal cylinder-to-cylinder spacing is determined by intersecting the two asymptotic estimates obtained for the tootal heat transfer between the bundle and the free stream. For the sake of concreteness we assume that the centers of adjacent cylinders form equilateral triangles, although other arrays can be treated in the same way. ‘The approximate optimal spacing results developed here for the equilateral triangle array are applicable in and order-of-magnitude sense to any other array type. Consider first the limit where the spacing S is sufficiently large that each cylinder acts as if it is alone in its own cross-flow of free-stream velocity Us. The total heat transfer rate experienced by the bundle is q=nqi a where qi is the heat transfer associated with a single cylinder, a= $ Nur DL(T,,-T..) @Q 3-36and n is the total number of cylinders, HW nm (S+D)* cos 30° % ‘We are assuming that W is considerably greater than (S + D). In the range 0.7 < Pr < 500 and 40 . Sonus = 79K ES On) Fe 12) This estimate shows that if the spacing is small, the total heat transfer rate increases as S2 when the spacing increases. o Le ! Sopt 3-39Intersection of the asymptotic regimes. To summarize the results determined until now, we found that q varies as S~” when S is large, and as S’ when S is small. These asymptotic trends are sketched in the attached figure. The actual (unknown) curve q(S), which is indicated by the solid line in the figure, has a maximum where the spacing $ is approximately the same as the S value obtained by intersecting the two asymptotes. The S,q value obtained by eliminating q between Eqs. (7) and (12) is given implicitly by Av &, 25H a3) et pape This result shows that the optimal spacing increases with the length of the bundle, and decreases with the applied pressure difference and the Prandtl number. The strongest effect is due to the bundle flow length H . It is also interesting that in Eq. (13) the exponents of P and Pr are almost the same (note that 3/14 = 0.21). This means that instead of the product B*/4 pr®-!8 we may use approximately Bey 4, au? (4) Bey * where the Bef 4 group is defined as “Ds Bep = ae = Bpr (1s) In this way Eg. (14) becomes similar to the dimensionless results reported for optimal plate-to- plate spacing in forced convection with Pr > 1 fluids, where the relevant dimensionless pressure ‘group is Bep not P. 3-40Jem 3.26 The analysis follows the same steps as in eqs. (3.122)-(3.134) in the text. The only difference is that instead of eq. (2.110) we use the small-Pr equivalent, eq. (2.111), such that €q, (3.129) is replaced by v2 =1 vaspr!?2(YeE) v ‘The rest of the analysis is replaced by apl3 3 Qb = 2.053kH (Ty - To) ————— W276 ge pa Dept 3.33 Bel’ prl/® L Ginax. $ 0.92 (DAP)? Pr Hop (Ty ~ To) Problem 3.27 An analytical expression for the peak excess temperature (Teak ~ To) can be developed when ko is small, and the aspect ratio Ho/Lo is sufficiently smaller than 1 such that the conduction through the heat-generating material (ko) is oriented perpendicularly to the fluid channel. If we also assume that Do << Ho, the temperature drop between the hot-spot comer (Tyeax) and the wall spot near the channel outlet in Fig. P3.27 is Tpeak — Tw = 9'""HG /(8kp), in accordance with the steady conduction analysis reported in [8]. The increase experienced by the bulk temperature of the stream from the inlet to the outlet is Toy ~ To = q"HoLo/ (thy cp). There is also a temperature difference between the bulk temperature Toy, and the duct wall temperature (Ty) in the plane of the outlet: temperature differences of this kind are neglected in this study based on the assumption that the flow is fully developed and the channel spacing is sufficiently small In conclusion, the peak excess temperature is given by a two-term expression that can be nondimensionalized in the form of the overall resistance of the elemental volume, AT = = 2) 0 FF, (2) where At = = @) a" Ag/ Ko M = th'ep AY? / ko, constant (4) 3-41In this notatign the size constraint HoLo = Ay becomes Figo ‘The right-hand side of eq. (2) is equal to Hj /8 + 1/(MHg), and shows that the overall resistance ‘AT can be minimized with respect to the external shape parameter Hy. The results are . 4y3 . vate Boe =(2)” tom = (8) 6 (i). Noteworthy is the optimal extemal shape (Ho/Lo)op., which is independent of the channel size Dp. The elemental volume is more elongated when the M parameter is large, i.e., when th” and Ap are large and ky is small. The starting assumption that Ho/Lo << 1 means that the above solution is valid when M >> 4. The assumption that Ty, ~ Tow is negligible (relative 10 Tyeax — T,) means that Do/Hy << Nu k/ko, where k is the thermal conductivity of the fluid and Nu is a dimensionless constant of order I (the Nusselt number for fully developed laminar flow in a parallel-plate channel). This last inequality comes from Tw — Tax << Tpak — Tw and the (6) definitions Nu = 2Dph/k, h= q"(Tw ~ Tou) and q” = q""Ho/2. Problem 3.28 Substituting AP ~ ; pU? we find, in order, 2 212 Be, - PE LLU 1 =~ Ref Pr wa =o 2 va 2 D, - = 3.25Pr M4 Rey? imax $ 0.26 pepH UPr!!? (Ty, — Ty.) 3-42Problem 3.29 For Pr ~ | fluids, the hydrodynamic and thermal entrance lengths have the same seale, X~Xr~DRep a) where Rep = UD/v. In the X region the wall shear stress behaves as in laminar boundary layer flow, v pu (él ° Let AP be the pressure drop along the entrance region. The global force balance in the longitudinal direction requires +X ~ DAP (G) By eliminating U and t between eqs. (1)-(3) we obtain 22 x _| Dap Dd [ ° D | ww which for Pr~ 1 is the same as x 1/2 = ~ Be} 5 D D 6) where according to the Be definition (3.133), D> AP Bep 6 3-43Chapter 4 EXTERNAL NATURAL CONVECTION Problem 4,1. The momentum and energy equations for any point inside the slender boundary layer region are . + oy sy oy oP z Qe Working on the momentum equation first, we write it as, £ w+ $= veg Pe + p(T T.) and then we integrate it from x = 0 10 x= X, x x d [* axev(&) —y( - ux x! %9 +3 vWdxev v(3), +28. (T-T.} dx ols In conclusion, we obtain aX ofp ve ‘The energy equation is handled the same way, 41RoprZon=08F ar ux Tx wo Tors f vTdx= 0 Too where ux is the so-called "entrainment" velocity; we find it by integrating the mass continuity equation, se + ay namely x ux 9 +2f vdx=0 0 Back in the energy equation, we now have tok [ vane vrds=-a() in other words, X Hf --Mo=0(F) ‘Note that immediately outside the boundary layer the reservoir fluid is not purely motionless, as the entrainment velocity is finite and negative. Problem 4,2. The analysis consists of substituting T-T=AT(1-3) and v=V&(I -3 into the integral equations derived in Problem 4.1. Note that in the above profiles it is assumed that 8p is the same as 8. We work first on the momentum equation: x oy x $f ve =), +28 [ (t-Ta)ax 2 v5 { m2(1—m)'dm x ans { (1—m)2dm oo oo 1 1 Tos 3 ‘The result is therefore 42(x8) dy \105. +4aT deb ™ Tuming now our attention to the energy equation, we have x a _ a) af MTT) dx 4 oe aed 1 -vars f, m{1-mjex -24T 2 i 30 in other words A (yg) = 0 ay WO=5 ©) At this stage we have two equations, (M) and (E), for two unknowns, V(y) and &(y). The y- dependence of V and 8 is already known from scale analysis (see Table 4.1): V=Cyyl2, B=Cg yl Constants Cy and C5 are obtained by substituting the above into (M) and (E), and the results are 12 12 =5.17v (Pr+ 20 (23) ce 118 93 Prt (Pr ny (23) ¥ In particular, 8(y) can be written as $-s93( 0/21. 5 1)4 paid 21 4.1)!" Ras} For the local Nusselt number, we write =o ¥ ok (TAN ¥ 9 ¥ No=apeo aT k78 = 0.508 (2021 4 1)! glia = 0.508 (2021 +1)" Ra} The above is Squire's result, which is shown also in Fig. 4.4. In spite of the fact that the 5p = 8 assumption makes Squire's analysis valid for Pr = QC1) only, the heat transfer results of this analysis scale correctly throughout the Pr domain (Table 4-1), 43Nu 0.508 Ra}/4, ifPrae Nu0.514 (PrRay)!4, ifPr+0 This apparent contradiction is resolved if we recognize that the analysis must be judged based on its heat transfer and fluid flow results, and that itis the flow part of Squire's analysis that fails in the two Pr extremes. For example, in the case of Pr > 1 fluids the overall thickness of the velocity boundary layer should be greater than Squire's. The reason Squire's solution yields the correct Nu scales is that the temperature profile is correct (regardless of Pr), in spite of the fact, that the velocity profile is not. Cthe sold curves are Squire's profiles ) Problem 4.3, ‘The similarity transformation is qeCxyt4, T-T.=ATO(n), y=aC y34 Fm) where ‘Term-by-term, the energy equation can be written as(Zacy*-4 a Chay MP) (are cy M4) -(ac?y!?F) ¢ i ATO'Cxy4) = LATO” C2y12 which reduces to factary re =ac? sty ie" hence Broce’ re’ =6" Similarly, the momentum equation 3 FR oe is written term-by-term as “(Racy rb acta Me (acy) +(c?y!2F) (a c2y WF 4 a Cxy A" =-vaC'F”+gh ATO Two terms on the left-hand-side drop out, and we are left with obct{-3 P43 F*)=—vacte + gBAT 0 in other words, ™) Probiem 4.4. Since air and water both have Prandtl numbers of order 1 or greater, the Nusselt ‘number for room heat leak depends only on the Rayleigh number based on wall height (Table 4.1). So, ifthe air natural convection heat leak is to be simulated in a water experiment, the room and the experiment must have the same Rayleigh number. For the room, we have Ragom = Sig ATH? = LOL. (15K (300m)? = (4.3) 101° Mom = BF Aa US KX = (4.3) Note that this Rayleigh number is high enough to be in the transition regime, Table 6.1. Similarly, for the water experiment we have 45_ (14.45) 10° experiment ~~ gm3 K (water) Ra. (10 K) Hater = (4.3) 10° which means that Hater = 66 cm. In conclusion, the air convection heat transfer may be simulated in a water apparatus roughly one fifth the size of the air system. Problem 4.5. Consider the fluid pool heated from the left through a vertical wall of finite thickness, W. If the wall is to be regarded as "isothermal", then the heat flux applied to the lower end of the wall must find it easier to reach (by conduction) the upper end of the wall than the fluid reservoir, The two thermal conductances available to q” are shown below. a Ww @ 2 i aw 4 Plaid Ay g esi ; tls, @ 2 7 O38 ‘Therefore the wall may be modeled as isothermal when ky WS kH in which H/Sy ~Raf{*, therefore Peles The isothermal wall model deteriorates as Ray increases. Problem 4.6. This problem is similar to Problem 4.2, the only difference being that this time the wall temperature is a function of altitude. Assuming the profiles T-T.=[Tg9)-Tal(1-4) vo) ¥(1-X2 v=VO 5 (1-3) ‘we have the additional relation 46‘The integral equations are (Problem 4.2) ki 5 oe v¥ + S88 ry.) on PlW0(T-Tal=P (Fo-T-) ® ‘There are three unknowns (V, 8, To) and three equations, namely (M), (E) and the q relation. The solution form is suggested by scale analysis [see eqs. (4.69)-(4.71)], constant B=Cgy",— Ty-Ta=Cry"8 V=Cyy5 and the three constants (C3, Cr, Cy) are determined from the three equations. In particular, the &(y) and To(y) results are [15] hence the local Nusselt number u Nu =0.616(45 +1) 5 Ray (4.75) In the two Pr extremes, this analysis predicts Nu=0.616 Ray), asPr3eo Nu=0.644(PrRayy)"®,asPr30 47Problem 4,7. Based on the scale analysis presented in the text, for Pr > 1 fluids near a vertical wall with uniform heat flux (Fig. 4.7.b), we select the similarity variables shown below: = bar)" us n-(ie *y —T = yus T-T.= Vey 9m, Pr) weaCySF(n, Pr), where ‘The energy equation Ha-RE reduces to dre'-tFo= et ©) Ls convection conduction Equation (E) shows that the boundary layer is ruled by the balance between longitudinal convection and transversal thermal diffusion, Both terms in (E) are of order O(1) because the similarity transformation is based on the correct scaling laws of the flow (for more on this, see the end of this solution). ‘The momentum equation ay v5 teb(T-T.) reduces to

1. The boundary conditions for eqs. (E) and (M) are F=F’=0 and 6=-1 at n=0 F>0 and 0-30 as N= 48‘The scale-based similarity formulation presented above can be compared with Sparrow and Grege's original formulation [15]: where oe eng] 5k v2, The resulting equations are 06 + Pr (4 669 ~ @gF6) =O, sa) Fp ~3 (Fe) +4 FoF 80=0 Msc) Unlike eqs. (E) and (M) seen earlier, eqs. (Esq) and (Msg) do not reveal to the reader the actual “balances” that rule the flow. Problem 4,8. The procedure for devermining the integral momentum and energy equations is the same as in Problem 4,2 of this chapter, with the only difference that this time To -T.» of Problem 4.2is replaced by To-Ta= Ty-Tao -YY ae AT, constant Dividing by AT, we see that (To~T».AT (which was equal to 1 in (M) and (E) of Problem 4.2) is this time replaced by Bt (Bgl “The momentum and energy equations are therefore a (v6 Vul B[MB]=-v¥+ Lovee -b99 a lvl -byal= 2 by.) (2) 49or, introducing 8+ = é Ral{4pand Ve=V H Rag, 1d fy Ve oe eR a td) = Retz ll-by) ™ yy Ve Be -by,) =PU-dy.) ® In the Pr —> © limit the inertia term drops out from (M), hence, V. can be eliminated between (M) and (F) to yield 8 (sd _240+8 bat dys 1 —bys This equation can be integrated analytically from 5. = 0 at ys = 0, 2041 —by.) 8-1 From the formula given in the text, the overall Nusselt number can be calculated numerically Nuo-u _ (1 2(1—bys) ay. Ralf! & fot which 8 is provided by the preceding equation. The numerical integration is hampered somewhat by the fact that the integrand blows up at ys =0 (because 5. = 0); to get around this difficulty, we can break up the integral: 1 2(1=bys) ay _ (© _2dys 1 2(1 =bys), 5 Y= Jo Gyo le Noting that im 8+ = 2401/4 yl, we have 1 2(1—bys) f (1 by4) ay, 1 ergs [ 2(L—bys) ay, e Evaluating the second integral by dividing the [e, 1] interval into equal steps of size Ays =€, Tcan construct the following table for the case b 4-10Nup-sy/Rajf* ‘Ayaze Problem 4.9. The conjugate boundary layers and the wall may be viewed as three thermal resistances in series, hence =_TucTe WH RRS where Ray t giant lee” al 4 wath J a 4 1 ee iM q OW oo gH ® RR oty is the thermal resistance of the wall itself. The other resistances follow from the result for the local Nusselt number; for example, for the left face of the solid wall we can write ay" oxy)" ave - = g 8)" (eb US us Ty T_y)= 1.623 L(1 +98) et) oy Averaging this temperature difference from y = 0to y = H we obtain 4-1TT af (14 98)" (BE) ae =(H 1352(1 +o 08)"" Rag} Ry orRy Putting everything together, THT BUI 15 cllS SB qH= and noting the definitions gH = 1K palif Nuoa= Gy -TgR’ ont allt we find that U5 5 Ret . Nugy =[2.704 (1 +98)!" Ral +o Ray ' ) In view of the relationship between Rayy and Rap, Rasy = eBHtg” _ aw eq. (*) provides an engineering estimate for the function Nug_ (Ray, @). In the limit @ > 0, this function is Ralf Nubhe Nugay=0.37(1+ in other words, 08)" pats Nup = 0.288(1+ 98) Ralf This estimate compares favorably with the alternative calculations displayed in Fig. 4.9. 4-12Problem 4.10 a) Since Prai = O(1), and assuming laminar flow (Table 6.1), from Table 4.1 we ave, Nu~Rajf* where =H _ 2 Nu= otk kT Ti) "| pt f Q Lwin forced in. b) Rewriting the Nu ~ Ray! relation as electrically ology" we conclude that AT4 EP 4 = constant, or AT = S28 H ‘Thus, if a design change calls for Hnew/Hold = 2, then AT, 3/5 Minew «9-5 2. ate =? 0.66 Problem 4.11. Let qy represent the local heat flux from the natural convection boundary layer to the solid-liquid interface Tin, a ky y= y (ToT) Noy ) where y is measured downward, from the upper edge of the wall, and Nuy is the local Nusselt hhumber, The same heat flux must penetrate by conduction the solidified layer of local thickness Ly), 413Zs ” Ta-Ty gy =k, (b) The solid-layer thickness is therefore equal to Ks Tm=Tw _¥_ LL ToT Nu, © for which Table 4.2 recommends (recall that Pr = 55.9): Nuy = 0.487 Ral/# @ Equations Oe and (d) demonstrate that L is proportional to the local boundary layer thickness scale y Ray! Before substituting numerical values in eq. (c) itis convenient to first nondimensionalize this equation by dividing both sides by H, 4 @ The Rayleigh number based on the overall height is gb (Teo-Tan) HP _ _9.81m 8.5% 104K"! (35 - 27.5) K @.1m)> OW Rag= 2 9x104em%s 0.05 cms = 1.39 x 108, (laminar flow) which means that eq. () becomes 6, 27.5-20 ay H 0.15 35-275 0.487 (1. 39x 108] 4H = 0.045 ()"* ‘The shape of the L-thin solidified paraffin layer has been drawn to scale in the attached figure. 414Problem 4.12. The total heat transfer rate to the stream rising through the duct is Q=pepVAAT=pH apy hence Hao _ ya _VDp KAT ~ PO “do, ‘The problem reduces to finding V, which is the cross-section averaged velocity through the vertical duct. ADP e natural convection _forced convection we learn that the two flows are equivalent if we set eB AT Vv in other words if s =pef AT, constant All we have to do then is exploit the solutions presented for forced flow in Chapter 3. From the definition of friction factor, 415we obtain y2_SBATDy _ gBATDy VDy 2 2(FRep,) or sh ATDR 2(FRep,)V In conclusion, the Nusselt number formula becomes Hap 4 = “EAT = Hap) AD = ORD, For example, if the cross-section is circular we find that the numerical coefficient (?) is =. -1L O= 506 ~ 8 Problem 4,13. Using scale analysis only, we start with (To-1)< (To- Te} and with Combining the two equations, and recalling that we obtain the criterion 416H Rap oy Next, we make the observation that Ray= and continue with eq. (1), which yields .132 Ray 43 Ray = 0.219 Ray 4 = 392.3 Ty, = 25°C + 8.4°C = 33.4°C The correct film temperature (25°C + 33.4°C)/2 = 29.2°C is nearly the same as the value assumed at the start. In order to verify that the boundary layer is indeed turbulent, we evaluate the AT-based Rayleigh number Ray: — Ra 3x10! _ 10 Ray= Fue “gg 726% 10 and the corresponding Grashof number, — Ran _ 26x10! = 9 Pr 54g —=48%10? (wrbulent, barely) Problem 4,16. The film temperature is (10°C + 20°C)/2 = 15°C. The air properties that will be needed are = =0.025 ¥. 8B 116 Pr=0.72 k= 0,025 4 ov = He ‘The 5°C/m stratification of the room air means that the top and bottom edges of the window sec the following air temperatures: 418T., (top) = 20°C + 0.5m 5 EF =22.5°C T... (bottom) = 20°C — 0.sm5> Nair, the overall h is 4-20h~hairside~ p(k Ralf)yie= (Constant) HM Tn conclusion, = (Ha) _ gua_ -(z) =5!4= 1.50 i.e. the cooldown in the vertical position requires a time that is 50% longer than the cooldown in the horizontal position. So, if you would like to drink this beer cold and as soon as possible, choose position 2. Problem 4.18. Focusing on the H x H region, and claiming the following scales (Ss) the momentum equation shows that av ay uXey v2¥ 4 eB(T-T.) ox ® 278 Ya inertia friction buoyancy 2 -F “He 7 eB aT Invoking the balance friction ~ buoyancy, w. WW ~gBAT w we obtain eBH?AT — (™) and inertia v2 _ SBATH _o, friction ~ yyy? ve 4-21However, the energy equation HW. get vEvE = oh A convection diffuision VAT _oAT H Hw suggests a different scale for v: vee ©) ‘The fact that two equations (momentum and energy) provide conflicting answers for the unknown v, should be expected: the decision to adopt scales (S) leaves only one unknown (v) for a system of two equations (momentum and energy). This case of mathematical incompatibility should tell the problem solver that one additional scale, which in (S) was assumed known, is in fact an unknown of the problem. That additional scale can only be the scale of x, ie. the boundary layer thickness. The error in the scale analysis that culminated with (M0 and (E) was the selection of the domain (x,y) in which the scale analysis was performed. (See the first rule of scale analysis, Chapter 1.) I composed this problem in order to demonstrate the artificial origin of the Grashof number Gry used so often in the current literature, To claim that the Grashof number is interpreted as the parameter describing the ratio of buoyancy to viscous forces is as erroneous as claiming that the flow fills the H x H region entirely, or that the H x H region is the boundary layer. Problem 4.19. In the Pr—+ = limit, eq. (4.65) yields Nay = 0.671 Raj/4 a This result holds for an isothermal wall (Ty) exposed to an isothermal reservoir (T..). It should hold approximately also in the case of the linearly stratified reservoir of Fig. 4.8, provided the Nuy and Ray of eq, (1) are based on the average wall-reservoir temperature difference: 4 2B y3 AT ay, ) 671 (8224To @ dwoy ¥ ATayg In order to form the groups Nuy; and Ray: defined by eqs. (4.83) and (4,81) in the text, in the above equation we set y = H, and multiply and divide by AT max: o- 3 4 4 Sina HH 29,671 (88!* AT ma) (“Toe ® Tag Tra w AT In short, 4-221/4 ATmax Tg Fae Nuy =! o.on ait(St] @) which means that AT yg) at) mal © Tmax Using the definition of b, eq. (4.81), we reach the conclusion that the ratio NujyRaj{4 must indeed decrease as the stratification parameter b increases: Noy =0.671 ( - ay Ralf* © ‘The curve NujyRa}{* recommended by eq. (6) falls below the Pr —» ee curve of Fig. 4.8, the largest discrepancy (12.8 percent) occuring at the right extremity of the graph (b = 1). Problem 4,20. The air properties evaluated at the film temperature (20°C + 40°C)/2 = 30°C are 90.7 = = we eB Pr=0.72 k=0.026 at a) Vertical plate of height H= 4 em: 8B 3 av H (Tw-T) = -20:7_ (4 em)3 (40-20) K = 1.16 x 10° om K Ray The Nusselt number correlation is eq. (4.107): Nig = 0.68 +0.515 Ralf? 68 +0.515(1.16 x 109) *= 10.19 Now & Way WwW _1__ xe = 10.19 x 0.026 WE gay = 6.62 This i value applies to both sides of the plate. ‘The total heat transfer rate is =f sides) H (Ty Te.) =6.62 -W_ - = 10.64 = 6.62 2x 0.04in (40-20) K = 10.6 Ty 423b) Plate inclined at 45°C: the calculation of q! consists of repeating part (a) in which g is replaced by g cos (45°) = 0.707 g. The results are, in order, Ray =8.2x 104 .68 +0.515 Raff =9.4 A Pd (this applies to both sides of the plate) m i Qsides) H(T,,—T..)=9.77 ©) Horizontal plate: in this case we distinguish between the upper surface (hot, facing upward), and the lower surface of the plate (cold, facing downward). For the upper surface, we begin with the characteristic length of the rectangular surface: Hxlength _ b= Fxlength 2em Ra, = 2819 (r,-T,)= 1.45 x 104 Equation (4.119) delivers the Nusselt number: Nuy = 0.54 Raf!# = 5.93 Bapper= Ma, = Ww _1__77_W_ = 5,93 x 0.026 mK 002m V7 mK ‘The lower surface has the same L and Rey. As Nusselt number correlation we use eq. (4.121): this choice is an approximate one because eq. (4.121) is more accurate when Rar, exceeds 105. We obtain: = Gupper* Giower = Rapper H (Tw-T..) + Dower H (Tw — Tex) = (Bupper* Fiowes) H (Tw - Tex) = (7.74 3.85) ¥— 0.04 m (40-20) K mK 29.24 =9.24 In conclusion, the best position for heat transfer is the vertical one, and the worst the horizontal. The difference between these extremes is relatively small, only 13 percent. 4-24Problem 4,21. The properties of air at the film temperature (10°C + 30°CY/2 = 20°C are = = wy ‘8B __107 Pr=0.72 k=0.025 -W oe a) Sphere with the diameter D = 3m " 1G1_ (300 em)? (30-10) K = 5.78 x 10" cm? K Next, we substitute this Rap and Pr = 0.72 in eq. (4.124): Nup =2 40.455 Rajf# = 225.2 fi=Mipk= i =Mip k= = wai =225.2x0.025 WE at q=hxb*(T.-T,) = 1.88 SL r (3m? GO- 10) K = 1.06 KW mK 1.88 m2K b) Horizontal cylinder with the diameter d= D-sphere, the cylinder must have the length: 5m. To have the same volume as the 4,(D) $7(3) ae *(g Rag= £8 (Tas Ty) = 7.22 x 10? D> & =8m whe The overall Nusselt number is supplied by eq. (4.122), in which Pr = 0: (0.6 +0322 Ra) =217.3 i Nag k = WY 1-3. 2173 0.025 Te Tyg = 3:62 FE ‘The total area of the horizontal cylinder is 4-25AzndL +2n(a)° =nd(L+d) = 1.5m (8m + 0.75m) = 41.23 m? Since the area of the two end discs is only 8.6 percent of the total area, itis safe (ie. a good approximation) to assume that the same h holds over the entire area: a=hA(T..-Ty) =3.62 4 2 (30 = 3.62 mk 1123 (30-10) kw ¢) The cylinder will be heated three times faster than the sphere, therefore the preferred shape is the sphere. Problem 4,22. The film temperature and the water properties that will be needed are Tram = (BOC + 20°C)/2 = 50°C = =0.64 2B _ 51410 Pr=3.57 k=0.64 a3 a) Focusing on ¢q. (4.127), we calculate in order l=lem+2em+lem=4em eB Ray= By P (Tw-Te) = 51410 (4.cm)3 (80-20) K = 1.974 x 108 daminar) om3 K Nu, =0.52 Raf! = 61.64 b) If we rely on eq. (4.129) and Table 4.3, we proceed as follows: 4-26L=A!2=[6(2. omy]? = Ra, = 28 L3 (Ty—T..) =3.63 x 10° (laminar) 0.67 x 0.951 Raf!* [r-(osepn” 9 cm Nay = 3.388 + = 3.388 + 0.562 Raj'4 = 80.95 ¢) On the other hand, if we use the simpler formula (4.130), we obtain Nay, = 3.47 +0.51 (3.63% 108)" = 73.87 W. 64 73.97 mK - w h=73.87 7049! 964. mK If we regard estimate (b) as reference, we see that estimate (a) is off (smaller) by only 6.7 percent, while estimate (c) is off by 8.8 percent, These deviations are insignificant in heat transfer calculations (think of the uncertainty in the values used for water properties), therefore all three methods are satisfactory. Problem 4.23, The six surfaces of the parallelepiped are of four types (A1,A2,A3,Ay), therefore =] Ay +43 2Ag +43 2A3 +94 Ag = (pA +2RAy + 2G Ag + RgAy) AT All the air properties are iV evaluated at the film temperature AY (0°C + 20°Cy/2 = 10°C; 3 D namely: Ay k=25x104_W, omK 4.27‘The Aj surface is cold and faces upward, The heat transfer coefficient follows from eg. (4.121) in which the length scale L is = 30emim___30 im Tm + 30cm) ~ 2(130) = 11.S4em Therefore, the Rayleigh number based on Lis Ra, = 2BL AT __125_ (44. 54)3 om3 20K LG aK = 3.84 x 106, 1.95, The hy coefficient is 25) 104Ww 4 195 omK 11.54cem a ‘The average Nusselt number is given by eq. (4.106), 67 Raf{+ [1+ (0.49201) which in the case of air (Pr = 0.72) reduces to Nig =0.68 + Nay =0.68 +0.515 Raff ‘The Rayleigh number based on H is 125— 30% em3 20 K = 6.75 x 107 therefore Nuy = 47.36, and 4-284 47.36 2A mK 30cm =3.95x 104 W_ = 3,95 W_ om? K mK Finally, the Aq surface is cold and faces downward. It has the same L = 11.54 em and Ray, = (3.84) 106 as the Ay surface, therefore eq. (4.119) yields 54 Ra! =23.9 N which translates into the average heat transfer coefficient ig= Na, k=239 25104 Ww 1 L cemK —11.54cm =5.18x 104 W_ = 5.1g_W_ om? K mK In summary, the q expression listed at the start of this solution becomes q=(2.59%1x0.3+2%3.95x (03+ 42x 3.95% 1x 0.34518 1x 0.3] Bm? 20°C mK = 108.24 W The instantaneous melting rate is proportional to q, q=th he where the latent heat of melting is hef = 333.4 ki/kg, therefore 9 _ 108.24W _ Ww ingg 7 333.4 kkg = 0775 Tig 0,325 8 4.29Problem 4,24. _In view of the elongated shape of the ice block, the length 1 used in Lienhard's formula Nuy =0.52 Raj/4 (4.127) can be approximated by measuring half of the perimeter of the smaller (square) cross-section: = 23m + 0.3m +230 = 0.6m Evaluating all the air properties at 10°C, we calculate in order Ray = 125 20°C (60)? em? = 5.4 x 108 om? K 52 [(5.4) 108]! = 79.27 i=Na k=79.27 goo Sur 'w wm The total area of the ice block is A= 2(0.3m)? + 4(0.3)(1)m? = 1.38m2 and the total heat transfer rate can now be evaluated by simply writing Ww 138m? 20°C mK q=hADT= =91.1W This quick estimate is only 16 percent lower than the more rigorous (and tedious) result of the preceding problem. Problem 4.25. Let AT be the unknown of the problem, ic. the temperature difference between wire and ambient air. The conservation of energy in each cross-section requires q'=haDAT=Nup xk AT therefore AT can be calculated with the formula q aT= D a) ‘The average Nusselt number is given by eq. (4.122), 4-30— yo? Mip-lo6+ 0207 Ral [1+ @ssomn”™ which in the case of air (Pr = 0.72) reduces to Nip = (0.6 + 0.332 Raff)? 2 ‘The Rayleigh number depends on the unknown AT, B D3 ar 8 If we do not have access to a programmable calculator, the temperature difference AT can be calculated by trial and error, executing the following steps: i) assume a AT value (labeled AT), ii) calculate sequentially Rap and Nup, using eqs. (3) and (2), iii) calculate the AT value (labeled ATc), and compare it with the ATa guess. iv) if AT¢ differs greatly from ATa, go back and repeat the (i)-(iv) sequence. In the first iteration of this calculation, we assume ATa=10°C and calculate the air properties at the reservoir temperature of 20°C, = 4 Ww. 2B _ 107 k225x104 We, wR Normally, we would be evaluating these properties at the film temperature (20°C + 30°C)/2 = 25°C, however, this first iteration is "rough", and the above properties are accurate enough. From eqs. (3) and (2) we obiain: Rap = 1.07, and Nip = 0.857 Finally, eq. (1) yields aT, = 20LW. emK 25x 104 W0.857 = 14.86 K = 14.86°C In conclusion, AT; > AT; the second iteration begins with assuming a larger AT, value: ATa= 20°C This time we evaluate the properties at the film temperature, (20° + 40"/2 = 30" 431AT, = 13.46°C IS This time, the calculated AT ai is smaller than the assumed value, ATs. In summary, the two iterations carried out so far allow us to draw a line in the plane (AT, AT2). Intetsecting this line with the theoretical line AT = ATe (see the attached sketch), we obtain the answer to the problem: sto) 4 13 AT = 143°C 10 If we have access to a programmable calculator, we can solve eq. (2) directly for AT. We still have to assume the film temperature, in order to evaluate the physical properties. To begin with, we evaluate all the properties at 20°C, and eq. (2) becomes AT ae a5) with the solution AT = 13.96°C. This answer suggests that the film temperature is higher than the assumed 20°C. Therefore, as a second try we evaluate the air properties at 30°C; eq. (2) changes slightly, 12.24°C ar_yv6P ‘aT pps+asse(5; 36°C. and the new solution is AT = 13.77°C. In summary, the table below shows that the true film temperature is approximately 20°C + AT/2 = 26,9°C, Trim 20°C 30°C AT 13.96°C 13.77°C, Interpolating between the values listed in the table, we obtain the final answer, AT = 13.83°C, which is not far off the trial-and-error value determined in the first part of this solution, ATS 14.3" 432Problem 4,26. For the warm side of the single-pane window, Table 4,2 recommends the following formula for the height-averaged heat flux (note: Pr = 0.72 for air, and Nuy = 4/3 Nup) Nuy=0.517 Raff, — (Pr=0,72) Recall that both Nuyy and Ray are based on the reservoir-wall temperature difference, which in this case is Th - Tw: x [x6 (Ty=Tw HP S17 aia) 1" T-Ty mio ‘Symmetry requires that the glass temperature Ty, be situated half-way between Th and Te, therefore Ta-Tw=3(Tn-Td) and the average heat flux formula based on Ty ~ Te becomes Problem 4.27. For the warm side of the glass layer, eq. (4.75) recommends the local Nusselt number: ow yaylS ss0(s84°") » (Pr=0.72) ‘The corresponding overall Nusselt number is obtained by replacing y with H, Tw(y) with T,,, and the coefficient 0.53 with 0.53, Tay 7283 where (~ 1/5) is the exponent n in the y-dependence of the local heat transfer coefficient, or, after some algebra, 433gf _HL [:6(= Tor a0 a ve)" ‘The average temperature difference Ty ~ T,, can only be equal to half of the side-to-side temperature difference, Ty-Tw=4(T-Td therefore, the wanted relationship between q" and (Th -T2) is cna v 4 fH Th-Te k Note that the 0.252 coefficient is only 16 percent greater than the 0.217 coefficient derived based on the constant-Ty model (see the preceding probiem).. Problem 4,28. The thickness of the window glass varies linearly with altitude, = 1 5=5+b(L-2) a where y is measured downward, x = 8 Zé b= @ ‘The H-averaged thickness 5 is fixed, in other words, the total volume (or weight) of the window is fixed. The taper parameter b is one variable in the design of the window, The local heat flux must overcome two resistances in series, the air boundary layer on the room side (I/h), and the glass pane itself (S/ky): @) The total heat leak through the window (per unit length in the lateral direction) is »_ fH ATA: of Ee ® h*ky Assume now that the heat transfer coefficient decreases in the downward direction, as in laminar boundary layer natural convection, 4.34h=bain(jf) © In this equation, n = 1/4, and min is the value of h at the bottom of the window, y = H (ie. the smallest value). By combining eqs. (1)-(5), we can nondimensionalize the total heat leak as 1 a = (6) a f espifi+s(t-2] © in which q) Bi=Pand ® ® ‘The objective is to minimize q with respect to the "shape" (dimensionless-taper) parameter S, while Bi is fixed by glass volume and air natural convection constraints. The integral (6) can be evaluated numerically (recall that n = 1/4); Ss O01 Bist 0 1.31005 0.5605 0s 1.30860 0.55605 0.535 *0.55603 06 0.55610 10 1.30716 1.14540 1S 1.30576 1.13876 2.0 ¥1,30437 *1,13307 ‘The asterisks indicate the suallest value reached by G as the taper parameter $ changes. At low Biot numbers, @' decreases monotonically as the taper increases, so that Gnin = ¥ 6 =2) ‘The smallest heat leak occurs when the taper is the most accentuated (S = 2), thickness at the top of the window is 25, and zero at the bottom, when the glass Athigher Biot numbers (e.g. Bi ‘occurs at an intermediate (optimal) taper, ) the @ behavior changes. The minimum heat leak in this case at $ = 0.535. 4.35‘The top line in the table (S = 0) shows the reference case, ic. the heat leak through the glass pane of constant thickness. ‘The relative merit of the tapered giass design is indicated by the ratio: Bi=0.01 Bi=0.1 Bi 0.9957 0.9752 0.9920 Grer % reduction 0.43, 2.48 08 The bottom line in this table shows that the q’ reduction associated with the tapering of the glass is minimal. This is especially true at small Biot numbers, which is the normal range of operation of regular-size windows. In order to see this, consider the following order of magnitude calculation: glass pane data: air side data: ‘equation (4.65): Biot number: In conclusion, if we consider the added difficulty of manufacturing tapered glass, and the unrealistic assumption that the bottom edge of the window can have a knife edge (zero thickness, or $= 2), the marginal heat leak reduction calculated above does not recommend the tapered- glass window as a viable energy conservation feature for building design. Problem 4.29. We first identify the large S and small $ asymptotes, and intersect the asymptotes to locate Sopr Large S. When the spacing S is sufficiently large each horizontal cylinder is coated by a distinct boundary layer. We are assuming that (H,W) >> (D + S), and that Rap >> 1, where Rap = g8D3(Ty—T.-(av). ‘The heat transfer from one cylinder is 4-36a1=f NupaDL(T,-T..) o Nup =cRaff* @ and c is a constant of order 0.5. The total number of cylinders in the bank of cross-sectional area Hx Wis a aw Q) (S+D)* cos 30° therefore the total heat transfer from the bank is q = nqy, or Garges = 2S; LW. (1, -T..) Ralf Oo) 0s 30° (4D This result shows that when the spacing is large, the thermal conductance q/(Tw ~T..) decreases as $ increases. Small S. Consider now the opposite extreme when the cylinders almost touch, and the flow is almost cut off. In this limit the temperature of the coolant that exits slowly through the upper plane of the bundle (L x W) is essentially the same as the cylinder temperature Ty. The heat ansfer from the bundle to the coolant is equal to the enthalpy gained by the coolant, q=thep(T,,~T..), where th is the mass flowrate through the L x W plane. To obtain an order-of-magnitude estimate for the flowrate, we note that th is composed of several streams [total number = W/(S + D)], each with a cross-sectional area $ x L in the plane of one horizontal row of cylinder axes. The thickness of the channel traveled upward by cach stream varies between a minimum value (S) at the row level, and a maximum value at a certain level between two rows. The volume-averaged thickness of one channel is pb? $=S+D-0907 0 (5) however, we may adjust this estimate by using 1 in place of the factor 0.907 to account for the fact that the channel closes and the flow stops when the cylinders touch ($ = 0): S+2D S+D © When § is sufficiently small, the flowrate through each channel of cross sectional area SL and flow length H is proportional to the pressure difference that drives the flow. The pressure difference is AP = pgHB(Tw — T..), or the difference between the hydrostatic pressures under two H-tall columns of coolant, one filled with T.}, fluid, and the other with Ty fluid. The mean velocity of the channel flow, U, can be estimated using the Hagen-Poiseuille solution for flow between two parallel plates (spacing 5, flow length H), namely 437=i CO) The total flowrate through the bundle, tn(pUSL)-W/(S + D), leads to the total heat transfer rate th cp (Ty —T.-), which can be summarized as Siw 4 = k(T,-T..) Ra (8) small = 75 1 (G4) )Rap ‘The key feature of this estimate is that when $ — 0 the thermal conductance q/(Tw ~ T=) decreases as $3, Intersection of the asymptotes. The attached figure summarizes the trends uncovered so far. The actual thermal conductance would be represented by the solid curve sketched in the figure. The peak of this curve corresponds to a spacing (Sopt) that can be approximated by intersecting the two asymptotes, qiarge S = dsmall §- The result of eliminating q between eqs. (4) and (8) is 24S op/D D (1+SopeD"? (32555) (Ro ® This formula shows that the optimal spacing has a certain value (of the order of D) when the cylinder diameter and the bundle height are specified. The ratio Sopy/D increases almost linearly with the group (H/D)"3Rap~!4, which means that Sop is approximately proportional to H3 D-1/12, i, almost insensitive to changes in the cylinder diameter. The optimal spacing inereases with the height H, and is mainly a function of H. smal] Si limit- L large S limit oF 4-38Problem 4.30. The first part of the problem statement has two parts. First, we are told that the stream flows isothermally: Tin=Tow =T a) Second, the flow system operates reversibly, i.., with zero entropy generation Seen = Gout ~ Sin) @) Next, we use what is always true: written for the same system, the first law requires that Q- W + talbin ~ hou) =0 % For ideal gases, h is a function of T only, and Eq. (1) means that bin = fg @ Consequently, the first law (3) yields W=Q (5) ‘The second law (2) yields Q= mT (our - Sin) (6) Because the stream carries a single-phase fluid, we have dh=Tds + vaP a for which dh=0 (ideal gas, T = constant) (8) v=RT/P (ideal gas) (9) 439Equations (7)-(9) yield Sout (10) and, in combination with Eqs. (5) and (6), ayy In the second part of the problem, the cycle executed by the atmospheric air stream consists of four processes: 1-2 isothermal heating and expansion, at Tj 2-3 isobaric cooling, at P, 3-4 isothermal cooling and compression, at T, 4-1 isobaric heating, at Px In the ideal limit, the thermal contact between the two isobaric streams is perfect, T=T T3=Ts (12) and the regenerative heating released by the P, stream is absorbed completely by the Py stream, th (hy — h3) = rh(hy — hg) (3) According to eq. (11) we can write Ou = Wy = mi RTy in PH «aay PL Q. = W, = mkt, In z (1s) ‘The net power produced by this flow system is Waet = Wy - Wy = tik (ty - Tin PL (16) DF 440The heat engine efficiency is the Camot efficiency: Woes = - Te a7) Qu Ta 7 This is not a surprise, because the model of this flow system is based on the assumption that irreversible flows are absent. The pressure drop from 4 to 1 is zero, and so is the pressure drop from 2 to 3. Furthermore the regenerative (internal) heat transfer between the P,, stream (2-3) and Py, stream (4-1) occurs with zero temperature difference from stream to stream. 441Chapter 5 INTERNAL NATURAL CONVECTION Problem 5.1. Focusing first on the left-hand-side of eq. (5.7) we eliminate one by one the terms identified as relatively negligible: 2nd w LHS = 2 (ree +y x)- | oy} ay Lot oy, fy, uv v2) - 1 fu uw? wu! Sle oH Hle op Using the mass continuity scaling wy ~ v/H, the six scales shown above become fy ye ye vor v8p vbr (& te) . (ez ap! FF] very negligible large relative to relative these to this The surviving terms (scales) are (aoa ay ir)! or, multiplied by “2, 1 H): +H. OR): much smaller than I, as t(and v) > 0 ‘Note that (vt) is the vertical travel along the wall, in a fluid initially at rest, In conclusion, the left-hand-side of eq. (5.7) is dominated by the first term, whose scale ia v/(t 81). ‘The right-hand-side of eq. (5.7) can be analyzed the same way: 5-1mee 830-808 ead) Aa) 8 bis term is always Se prevent because it "drives" the problem ve yo i HBF very large lative fo this 2 relative to Ho ‘Therefore, on the right-hand-side the surviving scales are vw/S73 and gBAT/S7. Putting the pieces together, we find that the momentum equation is the competition between three scales Yo, wo | gp AT me Ba inertia friction buoyancy Problem 5.2. The boundary layer is distinct in the steady state if its development as “convective” layer ends before the fluid layer is penetrated entirely by pure conduction, 'e < ‘onduction vH_ yz 2 (ere) oe Multiplying the last inequality by a/H? yields «(hy which is the same as the criterion for distinct layers given in the text. Problem 5.3. If the heat transfer rate is dominated by pure conduction, then from the energy equation we extract the following scaling law: © ven =o 2-24) LBs convection << conduction @AT vo HD a) 52Note that the above analysis refers to the entire square region (x ~ H, y ~ H) because the state of pure conduction fills the entire space between differentially heated side walls. The unknown v scale appearing in eq. (1) follows from the momentum equation, in which we invoke the balance friction ~ buoyancy [see eq. (5.8)] vx Sb AT 2) Eliminating v between eqs. (1) and (2) yields Ray<< 1 as criterion for conduction dominated heat transfer in a square enclosure heated from the side. Problem 5.4. Integrating the mass conservation equation across the boundary layer (ue, Ove _ af f (Sut+$e ods > wnat f} va de and recalling that we obtain Using Gill's substitutions dya=Fl-@[ltivTs2q] and Tee where p(y.) = even function, and q (ye) = odd function, yields even function dyn 53Since didy. (even fen.) is an odd function of ys, the core velocity tes is an odd function. This means that the horizontal core flow changes its direction as we move from positive y.'s to negative y.'s. Problem 5.5. The three scales of the momentum equation are 2 a a AT mY BT inertia friction buoyancy Assuming that a balance exists between inertia and buoyancy, we obtain u~(gBHAT)I2 ‘The assumption “inertia ~ buoyancy" implies that the friction scale is negligible relative to either inertia or buoyancy; let's see whether this is true: friction, _ vi? | pap ini Buoyancy ~ BRAT ~ BH H [substituting u ~ @BH AT)!?] We learn that in the shallow enclosure limit H/L —> 0 the ratio friction/buoyancy would blow up. This means that the friction scale is not negligible, hence, the assumption inertia ~ buoyancy is inadmissible. The correct assumption is the one made in the text, friction ~ buoyancy. ‘To the mind that, is stil influenced by the scaling results developed in Chapter 4, the balance friction ~ buoyancy may "seem" to be valid only for Pr > 1. In reality, no Prandtl number restriction exists on the core flow solution sketched in Fig. 5.10; recounting the analytical steps that led to the shallow Core solution, we see that it applies strictly in the H/L —> O limit, Rayy and Pr being arbitrary but finite (fixed). Like any other flow, the core solution breaks down during the transition to turbulent flow (the stability of the S-shaped core velocity profile is discussed on page 111 of Ref, [40]). ‘Transition criteria can be obtained quite easily by applying the scaling conclusions of Chapter 6, namely the local Reynolds number of order 102, Problem 5.6. The equations for the conservation of mass, momentum and energy are 4X0 © re ar ™) S4® alti 4—— Assuming that the flow is slender, Lx >> H, the balance convection ~ conduction in (E) yields WAT AT . HL, 2 «) ‘The balance friction ~ buoyancy in (M) yields AT vw Be ay Hy Eliminating y between the last two relations we obtain 8 AT 4x -RalP, where Ray = "80 AT H ‘The overall heat transfer rate between cavity and reservoir can be evaluated in two ways. One way is to focus on the counterflow through the mouth of the cavity, and to say that the heat transfer rate matches the enthalpy flowrate into the cavity, Q="tncpAT" = 4,67 wee AT wo ~p y Hepat ~kATRaj{? ‘Therefore, the Nusselt number scale is a) 55Similarity solutions for this problem, as well as experiments, are presented in Ref. [60]. For example, in the limit Pr —> ee the similarity solution yields ts = 0.029 Ra}(? Nu = 0.053 Ra}{? These results validate the trends discovered via scale analysis. ‘The other way to evaluate Q is to integrate the heat flux through the horizontal wall of the cavity, Qk ATL, ~KATRalP, (2) which is the same as eq. (1). Problem 5.7. With reference to the governing equations (C), (M), (E) listed in the preceding solution, the balance friction ~ buoyancy in (M) yields in other words y ~ gBAT L3/v. It has been assumed that the flow region is slender, Ly >> L. ‘The balance between upward convection and lateral conduction in (E) yields TearEliminating wy between the last two expressions, we obtain Ly~L Rat For the overall heat transfer rate we write Q~(kAT)L, ~KATRay, hence Nu~Ray heat flux into Ly-tall wall Lighthill [61] reports an integral solution for the vertical penetration of natural convection into a vertical tube with one end open; that solution shows that Ly ~ ro Rag and Nu ~ Rar, where fo is. the tube radius. Problem 5.8, The water properties that wil be used are evaluated at the film temperature of = em? 8B _ 28.982 a=0.00085= Gy = ASE Assume that in the beginning the water is motionless. The water that comes in contact with the hot surface develops a “conduction” layer the thickness of which increases as 8~(an! a) This 5-tall layer is heated from below and, in accordance with a criterion similar to eq. (5.72), it becomes unstable when its Rayleigh number based on height exceeds the order of magnitude Rag~ 10° £8 §3(r,,-7.)- 103 ~103 Kt cm? 8-109 0583 GRIN THR ~ 0018 or, approximately, 8 ~ 1 mm, Tn view of eq. (1), this result means cnt ~ 0.0146 cm? In conclusion, thermals will rise from the same region of the heated surface at time intervals of the order of ten seconds. 57Problem 5.9. Since each cylinder is surrounded by its own boundary layer, the heat transfer is impeded first by the boundary layer resistance between T; and the water reservoir (Tuo = (Ty + Tz)/2 = 25°C], and later by the corresponding boundary layer resistance between Tos and T, The two resistances are equal and in series. They are equal because of symmetry and the assumption thatthe water properties are the same (well, nearly the same) inthe two Boundary layers, Tq Gt Z ‘t £0 2 g ww oo ws 4 aL hrD haiD The heat transfer rate from the cylinder T; to the cylinder Tp, expressed per unit length in the direction perpendicular to the cross-section, is T-T 2 wT 2 hed hed KRU (T}-7) where Nup can be estimated based on eq. (4,122). The Rayleigh number in that correlation, Rap, is based on the temperature difference across one boundary layer, e.g. AT = T Tw. ='5°C, and on properties evaluated at the film temperature [(30°C + 25°CV/2 = 27.5°C, or (25°C + 20°Cy/2 = 22.5°C]. We assume that the two sets of film properties are approximated well by the properties evaluated at the in-between temperature of (27.5°C + 22.5°CV2 = 25°C, =0.6 ¥ = 1981x108 k=06% Pr=6.21 rd and calculate, in order, Rap = eB D3 AT 3 = bee (dom)? 5K = 6,34 x 10° om: = (0.6 + 2.382413.6)? 39.45 (4.122) =206 - = 237 =F 0.6 WE 30.45 (30-20) K = 287 58Problem 5.10. An approximate way to determine the relationship between the heat flux from Ty to Te (namely q", uniform) and the difference Th ~ Te is to invoke the conclusion reached in Problem 4.26 (single-pane window, uniform q"): 4 He 28 (T= Tore) H® Hs 02x mw a qa Th~Toore Here Tyore is the average temperature of the air core contained between the two boundary layers inside the cavity. In real life, Teore increases from Te at the bottom of the cavity, to Th at the very top. Its value at midheight is clearly H Teor (y= H)=4 (14+ T9 @ ‘The average core temperature Tore is also equal to (Th + T¢)/2, therefore Th-Teore= 3 (Th- Te) 8 and, in terms of Ty ~ Te, €q. (1) becomes us Hz o,196[88 (T= TI¥"] Th-Te k~ @ This formula applies when the cavity is wide enough so that the innermost boundary layers are distinct. When estimating the thickness of one of the inner boundary layers, in order to see whether they are distinct, oy- (sO ead] 6 keep in mind that the glass-core temperature difference is approximately one fourth of the overall temperature difference, Too ®t (Ty-T) © . The original fluid layer has the height H, and the bottom-to-top temperature difference AT. The horizontal partition divides this layer into two sublayers, with the following thicknesses and bottom-to-top temperature differences (see the figure) Hi=xH AT] =x AT ay H)=(1-x)H AT2=(1-x) AT Q ‘The number x is between 0 and I, and describes the relative position of the partition. For example, x = 1/2 means that the partition is inserted at midheight. It is important to note also that the AT; and AT? expressions written above are based on the assumption that there is no convection in either sublayer. 59Cold well Packitor 4% z= Het wall Our objective is to determine the optimal partition location x so that the regime of pure conduction (no convection) is extended to the largest possible Rayleigh number B av Ra AT @) This is the "external" Rayleigh number, which is based on the overall temperature difference AT. ‘The objective then is to maximize the overall AT while preserving the state of pure conduction, i.e, least heat flux in the vertical direction. Recall that the formation of convection cells augments the heat flux relative to the heat flux present in the no-convection state. If we assume (for simplicity) that the partition is isothermal at some intermediate temperature between the top wall and bottom wall temperatures, the top sublayer is without convection currents as long as its own Rayleigh number is smaller than the critical value, 28 war, <1708 @ ‘The same can be said about the avoidance of convection in the lower sublayer, 28 iar < 1708 6) Convection is suppressed in the entire layer (H) when conditions (4) and (5) are satisfied simultaneously. Now itis a simple matter to use the x-based formulas (1,2) and the external-Ra definition (3) to rewrite conditions (4) and (5) as fa < 1708 © Rac we Q =x Conditions (6) and (7) are satisfied simultaneously by all the (Ra, x) points situated under the roof-shaped line described by Ra = 1708/x4 and Ra = 1708/(1 — x)4 in the attached figure. It is clear that the design with the maximum external Rayleigh number for "no convection” corresponds to 5-101 1 ® i.e. to a partition installed at midheight. That maximum critical external Rayleigh number is (approximately, of course) . 9) (uy ° 100 Ra 1708 10 ne convection 0.5 1 ‘The modelling of the thermal boundary condition on the partition does not affect the optimal location determined in eq. (8). For example, if the partition is modelled as a wall with uniform flux (instead of uniform temperature, the only item that changes inthe analysis isthe numerical value (of the order of 103) that must appear on the right-hand side of the inequalities. (4) and (5). This value, however, drops out in the steps that follow en route to eq. (8). ‘The design can be refined by accounting forthe temperature dependence of the property group g/crv. If this dependence is significant, the optimal location of the partition will diffe from xope = 1/2.‘TRANSITION TO TURBULENCE, According to Table 6.1, the most conservative condition for the survival of Taminar boundary layer flow is Rey <2 104, where Re, = Zak In terms of local Reynolds numbers based on local transversal length scales, the above condition becomes: Re based on displacement thickness, 5* 8 = L73L Rel? U.. 8 Re based on momentum, thickness, 0 9* = 0.664L Re; _Usd* 9. Bt _ U8 _ Rege= 9 = Re, & Reg==g~ =Rey 2 =1.73 Re}? = 0.664 Ref? Reg < 1.73 (2x 104)? Reg < 0.664 (2x 104)!” < 245 <94 In conclusion, both results show that the boundary layer remains laminar if the local Re based on transversal length scale is less than O(102). Problem 6.2. From Table 4.1, the relevant scales of the wall jet of Pr > 1 fluids are 2, My ears F Ray, b~Prly Ray, 61therefore, the local Reynolds number scales as Re~¥D ~pr Ral Table 6.1 indicates that when Pr = 6.7 the boundary layer remains laminar if Gry < 1.3 10%, which means Ray < (1.3 x 10°) (6.7) = 8.71 x 10°, in other words, when Re < (6.712 (8.71 x 109) <118 In conclusion, the transition of the buoyant wall jet from laminar to buckling (turbulent) flow occurs when the Re based on transversal length scale is of the order Re~ 102 Problem 6.3, Ina buoyant air plume (Pr ~ 1) the scales are D~y Raj!4, v~ $ Ray? hence, the local Reynolds number is a ar\ls vD _ palit (s8y AT. Re~¥P ~Raj! aby" at) To eliminate AT (which depends on y), we invoke the conservation of energy, q~pepD2 vat, which yields ‘Therefore, the local Re scales as 6-2and, since Fig, 6.2 shows that the transition is marked by Rag ~ 1010, the transition Re is of order Re ~(1019)1/4 = 316 in other words, Re=0 (102), ‘An important observation concerns the dependence of the local Reynolds number Re on the longitudinal position y. Since Re ~ Rag4, Re increases monotonically as y¥/?. This means that alLlaminar plumes become eventually turbulent, as they develop downstream. Note that the ‘same observation applies to the laminar boundary layer of Problem 6.1, and to the laminar buoyant wall jet of Problem 6.2. Problem 6.4. If the stream with constant V and Po is bent in a bending apparatus (duct), then the velocity distribution becomes such that the flow is the fastest near the inner wall and the pressure is the greatest near the outer wall. These features follow from the Bernoulli equation along one streamline v@) ous P@) 63ZPV?+Po= 4 p ve) +P(2) coupled with the force balance in the radial direction _ pve) “Re a" where the flow has been modeled as inviscid. In the limit 2/Res —> O (i.e. D/Rez ~> 0), these two equations yield V@=V(lt e+ . 2 P@)=Po-pR- In any cross-section, we identify nonuniform distributions of normal stresses due to P(z) and the reaction caused by momentum flow through the cross-section, pv2(z). Note that these nonuniform distributions tend to rotate the cross-section in opposing senses. The net bending 7 PY@) “4 oth oe 64m= ff, (pv? +P)zdA= “hb b-® 242 2 =f, Pv aa=PY where t= ff 2aAa A Ro A pv?z| R Po- zdA= Note that the bending moment has the same sense as in the cross-section of an elastic beam of radius of curvature R.. (which is why the inviscid stream possesses a buckling property analogous to that of elastic beams). ‘The same result, M = p V2V/Re., is obtained for an inlet cut through the stream (try it). Problem 6.5. The solution to this problem is given as part of the solution to Problem 2.25: vt To find the location of the knee in this velocity profile, we recall that, 2-2, [* em? ert) = ia lo em dm hence d lero ¢ 2ht free 0 > ° Lo (On the above figure, we have 2 ar) g in other words In conclusion, the knee is located in the vicinity of y y —, 089, of =—,=00). avg! (vy? 6-5Problem 6.6. If Np = O(1), then ty ~ tp, or Dp Tov V2 Rearranging, we obtain VD 3 XP ~32 3 and, if the stream is two-dimensional [A/D = 7/32, eq.(6.11)],

Problem 6,7. a) The stability analysis of an inviscid broken-line shear flow may be found in Ref. (35]. The main steps are: region(1) U=Up region(2) U=Up% GQ) tsAek 66The continuity of # across the y =h and y = —h planes requires Akh =Bekhy Cobh @ Dekh =B ekh + Cokh ) Finally, the pressure continuity conditions at y = +h, eq. (6.26), yield 2(6+k Ug) CP LO(B eM + Cott) =0 © 2(o-KUg Bet + W0(B el + Ce) =0 @ Note that the last two equations are the result of eqs. (6.26) at y = +h, in which A and D have been eliminated using egs. (a) and (b). Eliminating the ratio B/C between eqs. (c) and (d) yields 2 = U0. [(2 kh 1)?— eth] o qlee 1)? - ehh] The shear flow is unstable if o is imaginary, i.e. if the quantity in the square brackets is negative: o Os 0.6392 ‘The instability condition is Kh < 0.6392 or, since k = 2n/h, aL Gy 7492. b) The stability analysis of an inviscid broken-line wall jet is outlined in Ref. [45]. Here are the highlights: 6-7region(1) U=0 region (3) U=Up Bs ‘The ¢ solutions in each of the three regions are () % =e +Feby with F=0, because lim, = QF =CekT+Doekt Q) = Acky + Beby , with A +B =0, because 0(0) = 0. ‘The continuity of 7 across y = D/2 and y = D requires A ckD2 4B ekD2 = ¢ KD 4p ekD2 @ CekD +Dy kD = Be kD () ‘The continuity of pressure across the same interfaces requires Dol Fp +1) etD2+cekP2=0 © [I-gtp le rkuale +[1+ gpg (e+ Ua Dot gh (6 +k Ua)(A—B)=0 @ Egs. (a)-(d) and the condition A + B = 0 are homogeneous. A non-trivial (A,B,C,Do,E) solution is possible if the determinant of this system is zero m2 +m Qy2+kD-3-Y)-4 (1 +kD) +27 =0 where Whether or not o is imaginary depends on the character of m; since the m equation has the form am? + bm +c = 0, imaginary roots are possible if A = b? — 4 ac < 0, hence the coefficient Dp should not be confused with the jet thickness D. 6-8A= (QP +kD-3- yh? + dy (1 + kD) - 872 <0 Itis found numerically that the above condition is met by 1.337 4 > 1.833. Note that the minimum unstable wavelength, 1.833D, is nearly identical to that of a free jet of triangular profile (Fig. 6.6). Problem 6.8. From the solution to Problem 2.22 we know the following scales D~v23 x23 UG? Dgl Uv 18 Uf DBF The local Reynolds number is hence Since Re increases monotonically with x, the jet is always composed of a laminar section (length) followed by a turbulent section, “All laminar jets buckle and become turbulent sufficiently far downstream, ie. where Re > 102, 69R 4 2 Slams 3 buexiin g te é Cherbulont) jer Bi Problem 6.9. According to eq, (2.85), the scale of the laminar boundary layer thickness on the moving lid is 8 ~ SDRepl? a) where Rep = UD/v, According to the local Reynolds number criterion, eq. (6.15), the flow ‘ceases to be laminar when WS ot @) ¥ Eliminating 8 between eqs. (1) and (2) we anticipate that the transition occurs when Rep ~ 400 B) 6-10Chapter 7 ‘TURBULENT BOUNDARY LAYER FLOW Problem 7.1. The rules of time-averaging algebra are all based on the definitions Consequently, we can also write: (+uW+¥) d= ee W+ivevutuv a0 1 if? asad f° va svi [* dra [* uv'dt ty Jo o i) 0 0 =iv+uV; When u = v, this last rule becomes P+ (ue Next, we have: a1 [ mg 2[t [ ya] = a0 @-sh wo-ael va) 98 =0, because @ is not a function of time ot THB=2 [P Baw -, ty Jo to because 9 —> e whereas u(tg) can only be of the same order as u(0). The time-averaging of the energy equation is analogous to that of the x- momentum equation, which is outlined in chapter 7: rr ee w& - o( 2p 252) wy ——~] dt o+ 2 (a1) +2) +2. (07) =aVT Each term of type d{UT)/ax yields three terms: S)-£e-200-26) Problem 7.3. The van Driest model consists of using as mixing length Qexy(1-e¥7') which yields the eddy diffusivity eye ey2(1-eva"P| 5 The constant-tapp postulate becomes 72or, in wall coordinates, 42(, _e-yt/A*)? dat] dut [pss -eorm ate Solving for du*/dy* we obtain gut ay AI afi ean? [1-ev'P) which is also listed in Table 7.1. Integrating this numerically from ut(0) =0 yields the dash-line curves shown below. Clearly, the value of A* must be between 20 and 30, for the van Driest model (o smooth the transition from eq. (7.37) to eq, (7.41). a a c ® 4 40 + 100 13Problem. If we solve for du*/dy*, eq. (7.47) yields 1/2, ut _-1+(1+4x2y"7] ay* Dey Using the transformation do. 2x cos? a. tan + mo, or dy the dut/dy* expression becomes xdut = 49 4 —da_ sin? — cosarsin? a and utp” = [S08 + (- +intan (E42) 0 *|'sina sin 4 2g hence Note that im, a0 ‘The above u*(y*) formula has been plotted on the next page for x = 0.4 and 0.2 (dash lines). Itis clear that the dash lines will never mimic the profile constituted by eqs. (7.37) and (7.41), regardless of the x value (because the dash-line curves do not have a "knee"). 14too Problem 7.5. Using wall coordinates, we can write: ion pady= put usdy* ut dy* constant Conclusion: the flowrate through the viscous sublayer is independent of the longitudinal position. 15Problem 7.6. Starting out with the momentum equation (7.25), ayy He Pe ox Yay “Pax and recognizing that immediately outside the boundary layer yields (M) Invoking mass conservation, 2W/0x + 9¥/@y = 0, the left-hand-side (LHS) can also be written as Y Performing the integral J (LHS)dy, where ¥ > (boundary layer thickness), yields i (LHS)ay = fue h ( (ey afr. vy- m0 u.gey where, from the mass conservation integral Yi og 2, B) ay 0 Lge vy is given by Tn conclusion, the LHS-integral is Ya dU, 2 2 —y a lye [¥ aayl- leo i (LHS) gy =-4.] 02, ft (eyo undue [Y Gav]|-v- Gey b Wal” oven Next, from the definitions of the displacement and momentum thicknesses, s=[* (je. o=[" é(-ee 76we eliminate the integrals denoted by —~—~ in the LHS-integral: Y + +) dU. ff cus)ay-2 [ub(v-s-9]-u..2 [v..(v-3')]-v. Ge ¥ =(_s§* dU. 2 de =(-8*-29) u.. G2-uL 2 Integrating the right-hand-side of eq. (M), Y 1 f (RHS) dy =- 9 and setting ¥ Hs) 4 = (RHS) ¢ f y=], ly, leads to the wanted result ‘This result is also valid for laminar boundary layer flow (note the identical forms of eq. (M) and the 8°, @ definitions in laminar flow). Problem 77. Taking the curve fit fy = 8.7(y*)7, we obtain ut =87y*,at anyy UL=8.784, at y=8 Dividing side by side, leads to a (y\!7 Un 3) which is the velocity profile to be used in the momentum integral 5 4 f° aU, -aay = 2 ff, sw--ow=% 1 d|y2 W(Q— nl = a vrai al? (1-1) an} = 32 2 m2 1In conclusion, we obtain Ro ™ which is one relation between &(x) and to(x). ‘The second relation between 8 and to follows from writing u* = 8.7 y* aty =6, namely fd Ua 2 5.7(5v0)"*) (cop)? y which yields, to uly Penge ® Eliminating to/p between eq. (1) and eq. (M) we obtain 14 dB = alls oB - yi __ 72 tw.) 78.7)" which, integrated from 8 = 0 at x =0, yields 3. v_yus $=0373 (we For t(p Un?),€g, (1) yields the following estimate —- u,5y) "4 pure gt)" 0.0227 Finally, for §* and 6, we refer to their definitions o= f° (aja 26 [Un an=8 =f al In conclusion, we obtain a =f n!7(1—n')dn = 8 =88°=22 5=85 = 7 8 78Problem 7.8. Taking u+ = C(y+)!/m, and following exactly the same steps as in the preceding problem, we obtain (™) Eliminating To/(pU..2) between the last two equations, we have 8 int Din + 2S oe med ‘These results show that the numerical coefficients in the to/(pU..2) and 8/x expressions depend on the particular curvefit, ie. on the constants C and m. Problem 7.9. We start with the relations Gipp=40 and Tapp = To and divide side by side Sin SePoUAT Ty, constant ‘app 5 Cee Ue, leo Problem 7.10. The definition of L-averaged Nusselt number tells us what we have to do: o-LL Nip = 2% where . Seong flete ® aot if and + [ diane} onA A MA L- ‘Ye lamioa~ turbulent For the laminar section we write iam 0) = 0.332 KAT ppl Rel’, Pol hence f" am 4X = 0.664 k AT Pr! Rey? We do the same for the turbulent section, usp =P ep Un. AT Sty Poi —— pr? 4 Cex (Colbum) —_ 0.0296 Rex! (Prandtl) hence Lite» sw = 0.037 p ep Ue ATP UAT (Lng) In conclusion, we obtain L EE = Pr!8 [0.664 Rey? + 0.037 (Rey -Rey,}*] 7-10Problem 7. a) Starting with the local Stanton number definition (7.76), we write in order a) b) According to eq. (2.92), the right side of the Colburn analogy (7.78) is 4. Cy = 0.332 Rexl? 2) where Rex = Uxx/V. The left side of the Colburn analogy can be estimated using eq. (1) and the Nux formula (2.103), which holds for Pr2 0.5 fluids, Nuy St, PreB = Re, Pr Pes pols Re, Pr = 0.332 Rez! @) Equations (2) and (3) show that the Colburn analogy Sty Pr23 = (1/2)Cfx applies to the laminar section of the boundary layer, provided the fluid is such that Pr 2 0.5. Problem 7,12. a) The properties of water at 20°C are ik =4ig-kL y=001 92 pais co=4i8 he v=001 k=059-% — Pr=7.07 mk Inorder to calculate y, ‘we must first evaluate Tw,x/p. For this we use eq, (7.60): 0.0296 U2, Rexl5 in which TAL=1.2x106 (turbulent) The result is =0.0018 U2 12 "0.0424 x20 2 = 0,849 2 Tyx\ HP voto) = ems, =2.7x0.01 5S oMoem =°3™™ b) The boundary layer thickness can be evaluated based on eq. (7.58): 8=0.37 x RezS =0.37 x 6m (1.2 109)" = 13.5. em Inthe laminar regime, the corresponding thickness would be the 8 given by eq. (2.85): 8=4,92 x Rey? 612 = 4.92 x 6m(1.2x 109)" =2.7 em We sce that the laminar boundary layer would have been much thinner (at x) than the real (turbulent) boundary layer. c) Finally, for the x-averaged heat transfer coefficient we rely on eq. (7.78'): Nu, = 0.0296 Pr! Ref!S Nu, = 0.037 Pr’3 Ref’> = 0.037 (7.07) (1.2% 108)" = 5184 =m, & = W_ 1 Ww = 5184 0.59 Se bm OO aK ‘A more accurate estimate can be obtained by accounting for the laminar tip of the boundary layer, as in Problem 7.10. TlProblem 7.13. a) We begin by recognizing the wall averaged shear stress, the total tangential force experienced by the wall, and the mechanical power spent on dragging the flat surface through the fluid: 0.037 p U? Rez =tL =FU= 37 (US P=FU=0.037 p U’ L(gz) If ( )¢ and ( )y represent the cold-wall and hot-wall conditions, the dissipated power changes according to the ratio Hp)” b) When the wall is heated, the water film temperature is (90°C + 10°C)/2 = 50°C, with the corresponding properties My = 0.00548 =o ‘The ratio 1 o.00s48)"/5 _ 0.013 ) = 0.84 shows that by heating the wall to 90°C we can reduce the drag power by 16 percent. c) The power that has been saved by heating the water boundary layer is P,P, = 0.037 p UMS L45 35 ay The electric power spent on heating the wall is given sequentially by EL 0.037 Px} (ay” k 4/5, @ FL = 0.037 ky AT Pr}? (&) in which (k, Pr, v) are also evaluated at the 50°C film temperature, ky = 0.64 Pr, =3.57 By dividing eqs. (1) and (2) we obtain a dimensionless measure of how effectively the heating of the wall has been converted into P savings: TAB%e) ust, vy)" Ga)" -) In this expression we substitute c = 4.18 kI/fkg-K, AT = 90°C — 10°C = 80°C and (vi/Ve)"/5 = 0.84, and obtain y ‘This shows that the drag power savings are much smaller than the heating power investment, when the ship speed is of order 10 m/s. he Pa = (seh Problem 7,14. The relevant properties of air atthe film temperature (20°C) are p=1.205x 103-8, , v=0.15 om cm “When the iceberg drifts steadily, the sea water drag over the bottom surface matches the air drag over the top surface, Therefore, we can write in order Fpair=Fp,water (tw), = (uA), 2(_Va \U5_ 2 Vw \WS 0.037 p,U’ (ox) = 0.037 owe (ger) and the last expression yields wae -(20)fip)” Uwater \Pa {tke amd )” (aust \° © [(10cm)> 1.205 10g] 0.15 cm/s =324 and since Uwater = 10 cm/s, we conclude that the wind velocity is = om 3.94 m_ 4,7 km Upie= 32.4 x 10 $2 = 3.24 = 11,7 KE TdProblem 7.15. The air properties are evaluated at the film temperature, 2 v=015, — k=25x104 WE, Pr=0.72 ‘The air boundary layer is turbulent, Re, = 3.24 2 100m (2.16) 107 0.15 sont therefore, the Nuy, formula is eq. (7.78"): Nay = 0.037 Pr! Ref = 0.037 (0.72)#2(2.16 x 107)" 44 x 108 ‘The corresponding L-averaged heat flux into the top surface of the iceberg is a. =i, AT= Ruy FT = ia 4 Wo $2.44 x 102.5 104 SB Tag AOC 4 W_ __1___ =o, em K 0.184 cm/s om?) pep==24x In order to arrive at the heat transfer coefficient for the external surface of the window, we calculate in order: 715Ux Re, == = 15 2 60m—*—” ET is Oe 0.132 amis = 68x 107 4c, = 0.0296 Rez! =8.03 x 104 Stq=4 Cp, Pr? = 8.03 x 10 (0.72778 20.001 = Jism St, P epU = 0.001 x0.0013 15 3 = 19.6 Y mK ‘This hy value is a lower bound for the actual heat transfer coefficient because i) equation (7.60) underpredicts the skin friction coefficient at large Reynolds numbers, ii) The external surface of the building is not smooth, and iii) the free steam U.. is not completely smooth, i.e. without eddies, Problem 7.17, a) The properties of 10°C water are £ =0.013 pees n=0.013 =0.58 = k=058 He Pr=9.45 We calculate the drag force by using Fig. 7.9: UD 120.15m ——S—— . 0.013 x 104 m? 215 10° 3 (Fig. 7.9) A=LD =0.75m 0.15m Fp=CpAdru 3 = 1, 107 ks 7) mee. =13x0.75m0.15m 5 1 ema ll 8) =73N 7-16b) How large is a force of 73 N? The weight (in air) of the same portion of the leg (Lx, the portion that would be immersed in the river) is W= Peat D?L = 1060 ¥8 & (0,15m)?0.75m9.81 m4 Pd =138N The drag force is about half the weight of the immersed portion of the leg. c) The heat transfer coefficient is furnished by eq. (7.100), in which we substitute Rep = 1.15 x 105 and Pr = 9.45: Nup = 0.3 +380 (1 +0.571)"5 = 545.7 i= Mp 5 =545.7x058 3 tim =2110 WE ®) q=bAjger AT =f aDLAT= 2110 £0.15m 0.75m 1K, mK =746W Problem 7.18. a) The properties of air at 20°C are 2 p=1.205 XE v=0.15 m ‘The drag force can be estimated based on Fig, 7.9: U.D Rep = —G— = 35.76 2 0.074m ——S » 0.15 10 m? =1.76x 10° Cp=042 Fig. 7.9) Fp=CpA4p Uz (wnere = 042 F (0.074m)* 51 205 $B a3 05 76 =139N TTb) The weight of the baseball is comparable to the drag force: mg =0.145 kg 9.81 an 142N ©) Between the pitcher's mound and the catcher's mitt, the kinetic energy of the ball [(1/2) mU?] drops by an amount equal to the work done against the atmosphere (Fp:x). The first Jaw of thermodynamics for the process 1 (pitcher) —> 2 (catcher) is -Fpx=4muj-4mu} uj-u}=2 Fpx = 2 _ (m)2 Das eg 139 N 185m =354.7(B) U} = (35.76 BP 354.7 (@) =924.1 (BP U)=3042 ‘The final horizontal velocity is 85 percent of the initial value. Problem 7.19. a) Ineq. (7.104) we evaluate all the air properties at Tas = 20°C, except Hw = (Tw = 30°C): k= 0.025 W a. = 1.81 x 10° 5. mK sm 2 5 ke Veo = 0.15 SE Hy = 1.86% 10 5 Pras =0.72 ‘The Reynolds number corresponds to the upper end of the Rep range in which eq. (7.104) is valid, Rep= 22? =2235 B00n s = = 22.35 2 007m —_8__— “De Ves . 0.15 x10 m? = 1,043 x 10° therefore we obtain, in order, 718us Nap =2+(0.4 Reff? + 0.06 Re{}) POA (=) = 0.4 (1.81) 4 = 24262.1 (0.72)°4 (L8L) =2303 ~ ke w R= Nup 7 =2803x0.025 Wt = 82.25 W_ mK A=nD?=2 (007m)? =0,0154 m2 =i = —w 2 (30— G=RA (Ty ~Ta)=82.25 9 0.0154 m? (30-20) K =127W ‘The time of travel and the total heat transfer (ball —» air) are t= X= 18.5m__ 9) 935 22.35 mis, Qz=qt= 12.7 W0.83s = 10.53 b) The depth to which the cooling effect penetrates into the leather cover is 2 12 3~(Cheaner!® ~(0.001 28" 0834} ~ 0.29 mm ‘The temperature drop that would be experienced by the 0.29 mm skin of the ball is Q Q ATy=—@—-—_2 4 O)skin— ODteatner AB eC ke 5 2 860-75 15 pg 000154 m? 0,00029m This temperature drop is small compared with the ball-air temperature difference Tw — Teo = 10°C, therefore the assumption that Ty is constant between the pitcher and the catcher is reasonable. TAOProblem 7.20. The air properties that are required by a calculation based on eq. (7.104) are =2x 10348. Hy =2x 105 58 = y-5 ke Hoo = 1.76 x 107 em We begin with the Reynolds number, Rep=—> =22 0.06m—__s___ s 0.141 x 104 m2 =8511 which falls in the range where eq. (7.104) is valid: V4 54.58 Mp =...=2+61.91 0.72)°4(126 i-mp wit A=Mip Fy =54.58x 0025 I A=nD? =n(0.06m)?= 0.0113 m? q=hA(Ty- Tes) = 22,74 W_ 0.0113 m2 (60-10) K = 12.85 W mK Problem 7,21, When the glass bead reaches its terminal speed, the weight of the bead (mg) is balanced by the drag force (Fp). This balance provides a relationship between the drag coefficient Cp and the unknown speed Us Fp=CpA 4 p,U2=mg 2in which we substitute p= 2800 ke/m3 (density of glass) Pa = 1.205 kg/m? (density of air at 20°C) g =9.81 mist D =0.0005m ‘The end result is Cp=1s2( ey a) Fig. 79° Sond felation involving the unknown Un is the definition of Rep on the abscissa of ig. 7.9, Rep= ue in which va = 0.15 cm/s is the kinematic viscosity of air at 20°C. Numerically, the Rep definition reduces to Rep =333 22 2) ep = 333 = @ Now we must find the proper Uss value so that the Cp and Rep values calculated with ‘eqs. (1) and (2) represent a point on the “sphere” curve in Fig. 7.9. We do this by trial and error, proceeding from left to right in the following table: pick calculate Cp with Tead Rep off calculate U~ with Us eq. (1) Fig. 79 eq. (2) 3.5 mis 1.24 85 2.44 mis 3.8 m/s 1.05 123, 3.7 mis 3.9 mis, 1 138 4.14 mis Ams 0.95 150 4.5 mis ‘The correct Use value is located between the second and third guesses, more exactly at forwhich Rep = 127 @) In vacuum, the bead velocity would be increasing as gt. The approximate time when the terminal speed is reached (ie. when the air drag effect becomes important) is when gt is ‘comparable with the just calculated Use, 721The bead falls for a total of at least 10m/(3.8 m/s) = 2.6s, therefore it is reasonable to assume that during most of this travel its speed has the terminal value Uce. ‘We can now evaluate the average heat transfer coefficient using eq. (7.104), in which we substitute Rep = 127 and Hw = 3.58 x 10°5 ke/sm (air viscosity at 500°C) Hoe = 1.81 x 10-5 kg/s:m. air viscosity at 20°C) Pr=0.72 (air Prandtl number at 20°C) In the end we obtain Nup = 6.47, which means that = Ka yy = 0.025W _647_ _ 394 W. he Nua K 0000Sm ~ 4 2K ‘Treating the glass bead as a lumped capacitance of temperature T, we write the first law for the bead as a closed thermodynamic system, a(r-T, p¥o— 5 —t =-hA, (T-T.) and recognize that V = (4n/3)(D/2)3, and As = xD2. Integrating this equation from t= 0, where T= Tj = 500°C, we obtain pe D, von in which t = 2.6s and, for glass, c = 0.8 ki/kg-K. Numerically, this yields in order T-T. Tae =ex9(451)=0011 T= 20°C + 0.011 (500-20)'C =25.3°C. In conclusion, by the time it falls to the ground, the glass bead has almost the same temperature as the surrounding air. 7:22Probl . a) The air properties are evaluated at the bulk temperature of 300°C, = Ww = k=0.045 Pr=0.68, 2 n=0.481 Pry = 0.72 (air at 30°C) Figure 7.11 shows that Cy = 1 for an array with 20 rows. For the aligned array we calculate, Rep= Umax =4.67x 100 —4em— — 3881 0.481 cm/s: Nip = 0.27 C, Re} Pr see) $0.27 x1 x (3881) (0.68)°36 (0.68)! = 42.26 = AW. 42.26 a = 0.045 Te XGim 74754 b) When the cylinders are in a staggered array, we must decide which space (flow area) is narrower, between two cylinders in the same row X,-D=3em or between two cylinders aligned withthe direction "a" in Fig. 7.10 (right), 2 2a(dxy] 7 ls +(pxy]-D ‘The space between two cylinders in the same row is the narrower; this means that Umax and Rep are the same as in the first part of the problem. By using eq. (7.112) we find that the heat transfer coefficient in the staggered array is approximately 30 percent larger than in the aligned array, 83. om 7-23The tube array is characterized also by + Xt _ 248mm X= = T0:7mm + X1_ 20.3 mm X= = 107 mm ~ XU/X = 1.22 (Fig. 7.13 insert) x ‘The number of tubes in the transverse direction is =_0.5m_ = = Fos48m = 79 In order to calculate Umax, We must first determine the minimum free-flow area Ac. The latter depends on the spacings between two adjacent tubes: transverse (vertical) spacing St = (24.8 - 10.7) mm = 14.10 mm diagonal spacing So= post ( 248) | “mm 10.7 mm = 13.09 mm ‘The flow blade that passes through S_ must pass through 2Sg. Since Sy <2 Sg, we conclude that the transverse spacing pinches the flow the most, Ac= Sy (width) ny ).0141m x 0.5m x 20 = 0.141 m 7-24S00kg m3 1 mex" pAc 36008 0.746kg 0.141 m’ =3.96 0 UnaxD Rep = Sax? _ 3.96 B 0.0107m ——1____ s 0.346 x 104 ms =1225 42 (Fig. 7.13) Now we have all we need in order to use eq. (7.113): AP=n1£% 4? Uinax =25x042x 1x4 x 0.746 *8 (3.96 BP m3 N =614—5 m Problem 7.24. a) The properties of air at 100°C are p=0.946 ¥8 v=0.23 m2 m ‘The pressure drop along the square array can be calculated by using eq, (7.113) and Fig. 7.12: =3m _9cm__675m =32 @-5)om 6.752 Rep = Umax 2 = 6.75 & —0.05m___ = 1.47x 104 v * 0.23 x 1074 m/s 9em _ Som 8 £20.25 ig.7.12) Xt, xXy-1 X=1 Fig. 7.12insert 7:25AP=nj£%4p Uhox = 1 kg 2 eee Ry =134, m 'b) The staggered array with the same X; and Xj will have the same Umax, Rep and X;* as in part (@): Umax=6.75B, Rep=147x104, Xp =18 ‘The f and x factors provided by Fig. 7.13 will have different values: £20.33 = 102, because Xt =1 x Equation (7.113) yields aP=21x0.33% 102% 4 x0.946 XE i675 BP =152N, m2 In conclusion, the pressure drop along the staggered tubes is 35 percent greater than the pressure drop along the aligned tubes. Problem 7.25. We have four unknowns (to, 49, V1, 8) and four equations: Bohemrs w 0.0366 (Tp - tT) a (1 3-20 P % 2) t= 0.0225 p V3 (may @) P28 20, ® PVT An expedient way to solve the problem is to set 7-26Vj=Ay™, S=By", t=CyP, qg=Dy4 and to use eqs. (1)-(4) to see what values (m, n, p, q) work. This does not take much effort, and the conclusion is Vi=Ay!®, B=By70, qy=cy70 gh ap ylls Using the above substitutions, the original system (1)-(4) becomes one for the new unknowns (A,B, C, D), which are all constants: g = psBAT or D_ P= 0.0439 aT ep sor ca74g14 0.0225 p v4 or DA-pr are or c P where x1, x2, x3, and xq are used as shorthand. The solution emerges rapidly: C=xB A= (Ky x4/xq)!? 3 (x1 Xa) ere De xP xg xf nf ‘The last result is what we need to evaluate the Nusselt number: += 0.0391 PrUS Rays ‘The wall shear stress to can be evaluated in the same way: =mi0 t= Cy 70 = y 70 (9710570 4/5710) wea ee constant Since 7/10 < 1, the to(y) distribution is as shown below. 7-27Problem 7.26. This problem can be solved by following the steps outlined in Problem 7.25. To begin with, the left-hand-side of eq. (7.118) is ad fl y2 8 - 2 ef v38(1-6) #7 4p =0.275 4 (v}8) The four equations needed for V1, 8, to and qg are 2.28 (via)=bepar ro) (0.0366). are 4 9-8 @Q 9 = 0.0225 p V3 (V, &vy 4 @) woe 10! pl? ay"? 4 Seeking power-law expressions in y, we find Vp=Ayl2, B=BylI8 ty=Cy%, qg=Dy"3 and, from eqs. (1)-(4), A= 0.498 (gp AT)? , or A=known De De Fg = 0.0448 p cpAT sor CAT Blt. pel? -9.1 Pe plZepaT, or DCM =xy or CAM BM, Out of this system we extract D: D= AS xf x49 x from which we determine the Nusselt number 9% ¥__¥_pyt 10-2) pel Rad Bie (- 102) Pr!” Rad which differs somewhat from Eckert & Jackson's Pr —> 0 limit, Nuy 0.0295 Prl!™5 Razls Note that both results depart from the Churchill-Chu correlation, eq. (7.116). 7-28Problem 7.27. We must compare the empirical correlation 13 Nup_y = 0.15Rajy>, (Pr > 1) with the H-averaged version of eq. (7.126): Nug_y = 3 (0.039) Pro!’ Ray® To find the range of + 5% agreement between these two estimates, we write in order 5 (0.039) Pr’ Ray> 0.95 <*— aa <8 0.15 Raby 4.38 < Pel Ralf!S < 4.85 (4.19) 10° Pr’ < Ray < (1.9110! Pr? For air, Pr = 0.72, the range is (1.56) 10° < Ray <(7.13) 10° and for water, Pr = 6, (9) 10" < Ray < (4.13) 10? Problem 7.28. If the horizontal round jet is driven solely by momentum, its centerline velocity decays as ay where c; = 7.46K", and K is given by eq, (9.33). Ifthe round plume is vertical and driven solely by buoyancy, its centerline velocity decays @)where and & = 0.12, The energy strength of the plume is t = D6 Uo ep (To ~ Tx) s As an approximation of the ‘bend’ where the horizontal jet turns into a vertical plume, we write y(x) for the centerline trajectory in the bend region, and G4) where t is the time of travel experienced by the fluid packets on the centerline. Combining these with eqs. (1) and (2), and integrating from t = 0 where x = 0 and y= 0, we obtain , 44 = (2cyt)!/? ye @ as] (5, 6) Eliminating t between eqs. (5) and (6) we find the trajectory 3/4 ey (5) ns ax3/2 Let (xi, ys) be the location where the trajectory reaches a 45° angle, dy. dx o) Equations (9) and (7) yield 2/7 27 4 (2) yea (2) (0,11) which are related by the proportionality ys/x) = 2/3. 7:30Chapter 8 TURBULENT DUCT FLOW Problem 8,1. We begin by writing ut = 8.7 y+!” at y = 19, namely Ue. g7(tou4)!7 where us = (to/p)"!2. To obtain to in terms of U (not Ue), we recall the definition of average velocity an ngu=f Re ardrdo uf? 1 = 2m (8.7) ee eh [nl nya 2m aa f al (1-n)dn which yields = 120 y= 1,224 Ue= 20 U = 1.2240 Recognizing further the f-definition, 0 Toe 2° the first equation on this page yields =0.077 Rey where Rep = UD/V. 81Problem 8.2. Assume a fully developed pipe flow with fixed Rep in the range 103 - 104, We can imagine this flow to be either laminar or coarsely turbulent (just turbulent), depending on the transition triggering mechanism. For laminar flow, the friction factor would be f= 16 Rep Atiransition, the flow "buckles" as a centerline jet of constant thickness, flowing through a fluid sleeve slowed down by the pipe wall. The wall friction mechanism is replaced by the intermittent wall-fluid sliding contact sketched in Fig. 7.5. Each contact spot is characterized by UL ~ 194 lamina ineipient Zz Suctlieg —— echanieon £79.75) and, since the Rep range is 103 ~ 104, this means that L ~D. ‘The largest eddies have diameters of order D: these eddies govern the maximum distance between direct contact spots. Therefore, during transition the density of contact spots is O(1): spotlengih(-L) Spot-spot distance (=D) Note that this prediction is consistent with 1, values in transitional boundary layer flow (imagine Unex/v ~ 104 to the left of Fig. 7.6). Since and since n ~ I, we have in other words fi 2 fiusturbutent~ 10 As shown in the next figure, there will be a jump in the f value as the flow shifts from laminar to turbulent, if the laminar flow is relatively free of disturbances (ic. if the natural transition is delayed). 82TRANSITION Note that the above discussion is a theoretical argument for predicting the transition in laminar duct flow (Table 6.1). Note further that the predicted f ~ 10-2 during transition, is a numerical order of magnitude supported strongly by the empirical evidence presented in the Moody chart (Fig. 8.2; note "4f" on the vertical scale). Problem 8.3, The M function may be written as 1@) M=1® To) where the integral I(r) is =2f'a 1W=3 J, ede In Hagen-Poiseuile flow i is replaced by 2U[1-(]. ul BaP 4G] andIn turbulent flow with & = Ue (y/rp) 7 we have l= uf (3) "bap where y = ro ~b. The end result for M is 11S 74 8 7 (=m? with the dimensionless notation The two M(x) curves have been plotted below: unlike in laminar flow, the M() function for Prandi!'s 1/7% power law velocity profile is not a strong function of r, thus justifying the M = I approximation used in eq. (8.24). 84,Problem 84. Consider the parallel-plate channel and the coordinate system shown below. The energy equation for fully-developed time averaged flow is, pepa d= 2 (ai) a—__ 2% We integrate this equation from y = 0 to any y, and from y = 0 to y = a, while keeping in mind that symmetry requires Gayp = Oat y = 0: (see top sketch for Ginn sign convention) 8-5Problem 8.5. If we set ii = U, constant, the energy equation becomes with Gipp= 0 at =O yields _ pep U tm Ea ai where. igo =-Pep(arey) (note: y=np- 9) By writing the energy equation as ueTm to-¥ __ at 2 dx @+8q™ oy and integrating from y =0 to any y, yields y Aaylty ay dy*=T* Da ® where Tt=(To-T) 2B ‘Note also that the friction velocity is 8-6Equation (1) can be integrated as described in the problem statement, esi. ye Leyit Tre PPO 1 ays y YE ay f pet Jew. ma with ey/e = Pry and ew/v = xy*, and the result is T=Pr +P (fn “ts yési+ Pr Ys KG At the tube centerline y* = 9%, we have Tr=Pryts.+P ( wf dois), F) €=Prygsp.+ Pilg ln a ry ©) where ro* is given by =U(g ef)” ‘The Te* equation, labeled (E), is the source of the expression for Stanton number because ih T v2 a(t) Beets ToT. 2)" Te=(To-T) Toh a The ratio (Ty -Tq)/(To- Ty) follows from the definition 13(To—T,) U= anf (t-Tarar which can be combined with 1/7" power profiles for both& and (Tg-7) to yield: ‘The result is, (ore that Problem 8.1 yielded ve ‘The Stanton number formula is therefore (2)? 0.833 RAS)" where [RHS] is the right-hand-side of equation (E). St= 87Problem 8.6. The pumping power is proportional to the product rh AP, namely P=} maP 0) where =f4b 1p y2 ap=t4h Lpu 2 Since ti, L, D and the fluid do not change as the flow regime switches from laminar to turbulent, the Reynolds number and the mean velocity also do not change, =-UD pep Rep="y> 4 Equations (1) and (2) show that the pumping power changes in the same direction (and to the same degree) as the friction factor: At transition Rep ~ 2000, therefore the pumping power experiences a jump of about 50 percent, Pour Plam 48 Problem8,7._In the following chain, we see that the temperature difference (Ty ~ Tm) varies as the inverse of the Nusselt number, because qi, and the fluid are fixed: (Tw Tad _ début iam (w= Taian Ge/iam — Prarb — Neda 4.364 Nupur 0.023 Pr4 Reh =2164Ropy* if Pr=0.72) In the vicinity of Rep ~ 2500, the temperature difference drops significantly if the laminar flow breaks down into turbulent flow: (fPr=0.72, Rep = 2500) 8-8Problem 8.8. The relevant properties of water at 20°C are 2 = =0.998 2 k=0,59 We p= 0.998 Sa _ = kJ Pr=7.07 op 4.182 5 v=0.01004 ae a) Pressure drop: kg us 05 Ep? o90g2_E 2 PED o998-E & (02m = m = 15952 595. —_0.02m_ 0.01004 x 10-4 =31773 (turbulent) .046 Rep =0.00579 [eq (8.14): note Rep range] =f 4k Lpy2 apart Lp’ = 146905 =0.145 am im’ b) Heat transfer coefficient based on the Colburn analogy: dp lb sew BE 200058 PPB 7.9723 =7.86 x 10-4 aah sesty 89h=Stp cpU=7.86 x 1040.998 8. soa 4182 ge 1.595 B c) Heat transfer coefficient based on the Dittus-Boelter correlation: Nup =0.023 Pr Reff5 = 0,023 (7.07) (31 773)8 =201 0,59 Win K 201 Tam =5930 mK The h estimate obtained in part (b) is 12 percent lower than Dittus-Boelter prediction, which is recommended. 4) Temperature difference across the stream 5x 104 Wim? 5930 Wim? K 43 K ©) Temperature increase in the longitudinal direction: th opd T= qy MD dx GT _ aw PD _ 5x10'W 002m __keK dx “thep ~m? gg Ke 4.182 x 1087 3 =15°C =1S6 aT yp Trou Tin EL = 15S 10m =15C 8-10Problem 8,9. We begin with the properties of air, which can be evaluated at the end-to-end averaged temperature (30°C + 90°C)/2 = 60°C: 12 v=0.188 p= 1.06 ¥8 m = - i Pr=07 cp= 1.008 FB k=0.028 W_ mK The required length of the pipe is given by eq. (8.46), for which we first calculate the heat transfer coefficient: Rep = UD = 5 ™ 9,04m ——_S8—__. ee 0188 x 10 me = 1.064 x 104 (turbulent) Nup = 0.023 Pr? Ref 0.023 (0.7)0-4 (1,064 x 104)0.8 = 33.21 he Nok = 33,2) 0.028 W - w 33.21 OOEW T= 2325 Ineq, (8.46) we substitute AsED? and n=aD and obtain 1.06 *8 0,04m 5 8 1,008 x 103 pDUcp ,, Ty=Tin m Kg |, 100-30 th Ty Tout 4x 23.25 W_ 100-90 mK =4.47m Since the flow is turbulent, the entrance length X (or Xq) ~ 10D 4m is less than one tenth of the pipe length L. In conclusion, the assumption of full development over the entire length L is justified. 8Problem 8,10. The relevant properties of water at 20°C are. = Ww 2 K=059 2 v=0.01004 Pr=7.07 p=0.998 5. ‘We calculate, in order, the heat transfer coefficient, the Nusselt number, the Reynolds number, and finally the mass flowrate: Nup =0.023 PrO4 Ref Regis - Nu 5 = = 2106.4 0.023 (7.07)"* Rep = (2106.4) = 14 270 m=p%D*U =0.998 = 3 F 0.025m)0.573 B an = kg = 0.2819 8-12Problem §.11. The information we know amounts to the following: U=1 ms v= 0.01 cms (water at 20°C) D=1cm yt= 116 ‘The wall layer thickness that corresponds to y* = 11.6 is syt—L_ a? or, after using eq. (8.10), yey’ wM —n__ u(e2)!? In order to determine f, we must first calculate the Reynolds number, Rep = YP =1m tom Toms 10 (turbulent) This Rep value recommends the use of eq, (8.13), £ = 0.079 (104) -4 = 0.0079 and, in the end, eq, (1) yields =116 20lem? s (_2_\W2 y= 16 00} at 5 (oars) =0.18 mm Problem 8.12. a) The properties of water at 80°C and moderate (nearly atmospheric) pressures 2 v= 0.00366 $ The frictional pressure drop per unit length is given by AP

> 1, We follow the method shown in section 3.6. 8-18a) When the board-to-board spacing is sufficiently small, the outlet temperature of the coolant is the same as the board temperature, and the total rate of heat transfer removed from the package is a) The mass flowrate is ri’ = npUD, where U is the mean velocity through cach D channel with fully developed turbulent flow, U ee” a ‘The friction factor f depends on the channel Reynolds number, as we will see in eq. (11). In conclusion, if we combine eqs. (1) and (2) with ri’ = npUD we obtain the D > 0 asymptote of the overall thermal conductance: (ratte), Tee Oe ® y" iL 1+ b)_In the opposite extreme each board is lined by boundary layers, while the core of the channel of spacing D is filled by coolant of temperature T.. and free-stream velocity U.. The latter is dictated by the force balance on the entire stack with fixed AP, HAP=2ntL @) where 7 is the L-averaged shear stress on the board surface. In writing eq. (4) we have assumed that the board thickness is small enough so that the force experienced by each board is dominated by skin friction over the L-long faces. This assumption is equivalent to writing that in an order of magnitude sense, Spuni<< 2tL ©) which, in view of the definition of average skin friction coefficient Cr =t/(p U2/2), means that wwe are assuming 4 t 0.5) ay (1 + t/Dopr) The corresponding maximum value of the overall thermal conductance is obtained by substituting D = Dap: in eq. (10) oF eq. (3): va a4 gt fer) psly, ©) pt gee [ima - ol. ( f } Pr (: * =] Bel’? (Pr 2 0.5) (12) where Be, = AP - L'(ct). The inequality sign is a reminder that if q! is plotted on the ordinate and D on the abscissa, the peak of the actual q' vs. D curve is located under the intersection of the asymptotes (3) and (10). In spite of this inequality, the right side of eq. (12) represents the correct order of magnitude of the maximum overall thermal conductance: Smooth surfaces. Beyond this point in the analysis we must make certain assumptions. regarding the values of the fiction factor and skin-Liiction coeflicient. If all the board su: are smooth, we can use the standard correlations aus 4 6 F = 0.046 Rep! (o < Rep, <10 ) (13) (w* < Rey, < 8) (14) 2D, Rep, = 2DU/v and Rey, = UsL/v. These allow us to relate U and Uz, to AP, by ‘combining eqs. (2) afd (13) for U, and (7) and (14) for Us: 5/9 U = 598 D3 (2) (as) pL 8.20