206 Chapter 7 Since all values on the right hand side of Eq. (7-30) are known, assuming a value of f allows a corresponding value of Npepi to be determined. ‘This value can then be used to check the assumed value of fusing the general expression for the power law friction factor [Eq. (6-44)] and iterating until agreement is attained 2. Bingham Plastic The basic dimensionless variables for the Bingham plastic are the Reynolds number, the Hedstrom number, and the friction factor. Eliminating Dg. from the Reynolds number and Eq. (7-25) (as above), the cost group is PB, 2 2 ets — () PPapnetia” 731 Ie *, 10(B + CY)uee> G3) Dj. can also be eliminated from the Hedstrom number by combining it with the Reynolds number: 4 [coprt? iteNRe = (3) ( (7-32) These equations can readily be solved by iteration, as follows. Assuming a value of f allows Npe to be determined from Eq. (7-31). This is then used with Eq. (7-32) to find Njj.. The friction factor is then calculated using these values of Nge and Nye and the Bingham plastic pipe friction factor equation [Eq. (6-62)]. The result is compared with the assumed value, and the process is repeated until agreement is attained. Graphs have been presented by Darby and Melson (1982) that can be used o solve these problems directly without iteration. However, interpola- tion on double-parametric logarithmic scales is required, so only approximate results can be expected from the precision of reading these plots. As mentioned before, the greatest difficulty in using these equations is that of ensuring con- sistent units, In many cases it is most convenient to use cgs units in problems such as these, because fluid properties (density and viscosity) are frequently found in these units, and the scientific system (e.g., cgs) does not require the conversion factor g.. In addition, the energy cost is frequently given in cents per kilowatt-hour, which is readily converted to cgs units (e.g., S/erg).. lll FRICTION LOSS IN VALVES AND FITTINGS Evaluation of the friction loss in valves and fittings involves the determina- tion of the appropriate loss coefficient (K;), which in turn defines the energy loss per unit mass of fluid: ep = Ki V7/2 (7-33)Internal Flow Applications 207 where V is (usually) the velocity in the pipe upstream of the fitting or valve (however, this is not always true, and care must be taken to ensure that the value of V that is used is the one that is used in the defining equation for K;). The actual evaluation of K; is done by determining the friction loss er from measurements of the pressure drop across the fitting (valve, etc.). This is not straightforward, however, because the pressure in the pipe is influenced by the presence of the fitting for a considerable distance both upstream and downstream of the fitting. It is not possible, therefore, to obtain accurate values from measurements taken at pressure laps immediately adjacent to the fitting. The most reliable method is to measure the total pressure drop through a long run of pipe both with and without the fitting, at the same flow rate, and determine the fitting loss by difference. There are several “correlation” expressions for Kr, which are described below in the order of increasing accuracy. The “3-K” method is recom- mended, because it accounts directly for the effect of both Reynolds number and fitting size on the loss coefficient and more accurately reflects the scale effect of fitting size than the 2-K method. For highly turbulent flow, the Crane method agrees well with the 3-K method but is less accurate at low Reynolds numbers and is not recommended for laminar flow. The loss coefficient and (L/D). methods are more approximate but give acceptable results at high Reynolds numbers and when losses in valves and fittings are “minor losses” compared to the pipe friction. They are also appropriate for first estimates in problems that require iterative solutions. A. Loss Coefficient Values of K; for various types of valves, fittings, ete. are tabulated in various textbooks and handbooks. The assumption that these values are constant for a given type of valve or fitting is not accurate, however, because in reality the value of Ky varies with both the size (scale) of the fitting and the level of turbulence (Reynolds number). One reason that Kr is not the same for all fittings of the same type (¢.g., all 90° elbows) is that all the dimensions of a fitting, such as the diameter and radius of curvature, do not scale by the same factor for large and small fittings. Most tabulated values for K; are close to the values of K, from the 3-K method. B. Equivalent L/D Method The basis for the equivalent L/D method is the assumption that there is some length of pipe (gq) that has the same friction loss as that which occurs in the fitting, at a given (pipe) Reynolds number. Thus, the fittings are208 Chapter 7 conceptually replaced by the equivalent additional length of pipe that has the same friction loss as the fitting: 4f L @- 6), (7-34) where f is the friction factor in the pipe at the given pipe Reynolds number and relative roughness. This is a convenient concept, because it allows the solution of pipe flow problems with fittings to be carried out in a manner identical to that without fittings if 2,4 is known, Values of (L/D)gq are tabulated in various textbooks and handbooks for a variety of fittings and valves and are also listed in Table 7-3. The method assumes that (1) sizes of all fittings of a given type can be scaled by the corresponding pipe diameter (D) and (2) the influence of turbulence level (ie., Reynolds number) on the friction loss in the fitting is identical to that in the pipe (because the pipe f values is used to determine the fitting loss). Neither of these assumptions is accurate (as pointed out above), although the approximation provided by this method gives reasonable results at high turbulence levels (high Reynolds numbers), especially if fitting losses are minor. C. Crane Method The method given in Crane Technical Paper 410 (Crane Co., 1991) is a modification of the preceding methods. It is equivalent to the (L/D)eg method except that it recognizes that there is generally a higher degree of turbulence in the fitting than in the pipe at a given (pipe) Reynolds number. This is accounted for by always using the “fully turbulent” value for f(e.g., fx) in the expression for the friction loss in the fitting, regardless of the actual Reynolds number in the pipe, ie., KV? ea. where Kp = 4fr(L/D)ag (7-35) The value of fy can be calculated from the Colebrook equation, 0.0625 fy = 2005 _ 7-36) J TeaG@aD/OF in which eis the pipe roughness (0.0018 in. for commercial steel). This is a two-constant model [fr and (L/D)qq), and values of these constants are tabulated in the Crane paper for a wide variety of fittings, valves, etc. This method gives satisfactory results for high turbulence levels (high Reynolds numbers) but is less accurate at low Reynolds numbers.Internal Flow Applications 209 D. 2-K (Hooper) Method ‘The 2-K method was published by Hooper (1981, 1988) and is based on experimental data in a variety of valves and fittings, over a wide range of Reynolds numbers. The effect of both the Reynolds number and scale (fit- ting size) is reflected in the expression for the loss coefficient + x.(1 + a) (7-37) Here, ID; is the internal diameter (in inches) of the pipe that contains the fitting. This method is valid over a much wider range of Reynolds numbers than the other methods. However. the effect of pipe size (c.g. 1/ID;x) in Eq. (7-37) does not accurately refect observations, as discussed below where Ky = E. 3-K (Darby) Method Although the 2-K method applies over a wide range of Reynolds numbers, the scaling term (1/ID) does not accurately reflect data over a wide range of sizes for valves and fittings, as reported in a variety of sources (Crane, 1988, Darby, 2001, Perry and Green, 1998, CCPS, 1998 and references therein). Specifically, all preceding methods tend to underpredict the friction loss for pipes of larger diameters. Darby (2001) evaluated data from the literature for various valves, tees, and elbows and found that they can be represented more accurately by the following “3-K” equation: Kiog Ky “Nat «(' to (7-38) Values of the 3 K’s (Kj, Kj, and Kg) are given in Table 7-3 (along with represesentative values of [L/D].,) for various valves and fittings. These values were determined from combinations of literature values from the references listed above, and were all found to accurately follows the scaling law given in Eq. (7-38). The values of X; are mostly those of the Hooper 2-K method, and the values of K; were mostly determined from the Crane data, However, since there is no single comprehensive data set set for many fit- tings over a wide range of sizes and Reynolds numbers, some estimation was necessary for some values. Note that the values of Ky are all very close to 4.0, and this can be used to scale known values of K; for a given pipe size to apply to other sizes. This method is the most accurate of the methods described for all Reynolds numbers and fitting sizes. Tables 7-4 and 7-5 list values for K; for Expansions and Contractions, and Entrance and Exit conditions, respectively, from Hooper (1988)So oo 0001 (st=a/s4) Vv & oF z1o0 0001 (L=a/4) pebuely 2 oo 80 001 os G=a/) pusg winjai esojo ‘papesiy| 08k §& ov zs00 00s 9 ($2) Spiam Z ov 980°0 00S SL (sy) lem | paeuyy ov 250°0 00s (st =a/4) snipes 6u07 ze 1400 00s ot (=a/s) prepueis pepeau,, sy ze seo 008 8 (oe) spiom € by 990°0 008 st (sp) Spjam Z ov 220 0001 09 (08) PIO | palo ay sZ0°0 008 Zt (9=a/4) ee 990°0 008 vb (y= a/4) ee 9S0°0 008 z (@=a/4) ov 1600 008 oz G=a/) spueg ‘papjem ‘pebuel4 zy 1200 008 ot (st=a/4) snipes Buoy ‘papeesy ov blo 008 og (=a/s) Prepueys ‘pepeeiy| 06 SMOqIy » be(a/7) Bums 210 ‘SUSU U} J8}aWWEIP [eUTWOU ayy SI “q e1ayM) (Ca/P + Dt SIN / bY = 4 sBunu.4 pue senjen 104 sjustoy909 $607 10} SIUBISUOD rE EL TTAant Internal Flow Applications oz 9y0 690 z10°0 L800 veo va0'0 bo ont 690 so L100 b60"0 ve0 azo vho yezo 007 ost 001 oe oe oe oof 00s. Oost ool os6 ook ost oz ool oo8 008 00s 009 rd oz 09 si l(gh/“arlor*“n 49849 UT an lGi/Fapelse = “A yooyo Bums edh-weg wiberydeig b= ‘piepueig —onea jie, L=¢ ‘piepueig anien a129 (uBnowy Moy) Kem-2ay snowy wyBreng ‘Moy Youeig eajen Brig 1 = ‘prepueig enpen eqoip | =¢/‘9z18 aun ny .06— | =@ ‘9z1S aul |} .sy — anfen ajBue sonjen uoueag ugg (sass) pebuel, (sal) papas, yBnoays uny youeig urams. (L=a/s) pabuely (st=a/s) (=a/s) Papeary, (mogye se) youeig-yBnouyy soo212 Chapter 7 TaBLe 7-4 Loss Coefficients for Expansions and Contractions ik to be used with upstream velocity head, VE/2. A= 4/D Contraction 0c 45° Nes < 2500: ven talie eg ]L sand Non > 2500: = ssio0+ 1a] ang Nae < 2500: 160) [1 ay? = here] be Nea > 2500: =], 9)" k= [0.6 + 0.484] = [3] Expansion - a « = i" > NM + 04s" Ne. 4000: K=5.2(1~p") sing Nos > 4000: K = 2.6(1 + 9.24)(1 — By? sin $ o> 45 Non < 4000: K = 21-6) Non > 4000: Ky = (143.24)(1 — 9?)* Nno is the upstream Reynolds number, and f; is the pipe friction factor at this Reynolds number. Source: Hooper (1988).Internal Flow Applications 213 TABLE 7-5 Loss Coefficients for Pipe Entrance and Exit KiNre + Ke Entrance Inward projecting (Borda) ; K, = 160, K.=1.0 i Flush (rounded) K, = 160 rd r ae \e 0.0 (sharp) 0.5 5 0.02 0.28 a g 0.04 0.24 0.06 0.15 0.10 0.09 0.15 &up 0.04 K, 0.0 (1 = 841 ~ By 2.91 » Kee (1 - 8) 25 Onifice: pr (BG — BY C38 Source: William B. Hooper, Chemical Engineering, p. 97, 1961. The definition of K; (ic., Kr = 2¢¢/V) involves the kinetic energy of the fluid, V?/2. For sections that undergo area changes (e.g., pipe entrance, exit, expansion, or contraction), the entering and leaving velocities will be different. Because the value of the velocity used with the definition of K; is arbitrary, it is very important to know which velocity is the reference value for a given loss coefficient. Values of Kr are usually based on the larger velocity entering or leaving the fitting (through the smaller area), but this should be verified if any doubt exists. A note is in order relative to the exit loss coefficient, which is listed in Table 7-5 as equal to 1.0. Actually, if the fluid exits the pipe into unconfined space, the loss coefficient is zero, because the velocity of a fluid exiting the pipe (in a free jet) is the same as that of the fluid inside the pipe (and the214 Chapter 7 kinetic energy change is also zero). However, when the fluid exits into a confined space the kinetic energy is dissipated as friction in the mixing pro- cess as the velocity goes to zero, so the loss coefficient is 1.0. In this case the change in the kinetic energy and the friction loss at the exit cancel out, IV. NON-NEWTONIAN FLUIDS There are insufficient data in the literature to enable reliable correlation or prediction of friction loss in valves and fittings for non-Newtonian fluids. As a first approximation, however, it can be assumed that a correlation similar to the 3-K method should apply to non-Newtonian fluids if the (Newtonian) Reynolds number in Eq. (7-38) could be replaced by a single corresponding dimensionless group that adequately characterizes the influence of the non- ‘Newtonian properties. For the power law and Bingham plastic fluid models, two rheological parameters are required to describe the viscous properti. which generally results in two corresponding dimensionless groups (Ne p and n for the power law, and Nee and Nye for the Bingham plastic). However, it is possible to define an “effective viscosity” for a non- Newtonian fluid model that has the same significance for the Reynolds number as the viscosity for a Newtonian fluid and incorporates all the appropriate parameters for that model and that can be used to define an equivalent non-Newtonian Reynolds number (sce Darby and Forsyth, 1992). For a Newtonian fluid, the Reynolds number can be rearranged as follows: _PVe pv? # EVID %/8 Introducing ty = m{(8V/D)(3n + 1)/4n}" for the power law model, the result is (7-39) Re zo (gn) (7-40) Nee = tepER (Fn4 7 which is identical to the expression derived in Chapter 6. For the Bingham plastic, the corresponding expression for the Reynolds number is: ———__4@e ___ Ne _ TDi + 2D %y/3204,) 1+ Nue/BNee (7-41) Re.BP This is determined by replacing 1, for the Newtonian fluid in Eq. (7-39) with 1 + Hoots and using the approximation 7, = 8V/D. The ratio Nye/Npe Dto/ Vi is also called the Bingham number (Ng). Darby and Forsyth (1992) showed experimentally that mass transfer in Newtonian and non-