The two-K method predicts This is a new technique that requires only two constants, plus the Reynolds number and fitting diameter, to predict the head loss in an elbow, valve or tee. It is accurate even for large-diameter and alloy fittings, and at low Reynolds numbers. William B. Hooper, Monsanto Co. C1 Forcing a uid through a pipe fitting consumes en- e=ay, which is provided by a drop in pressure across the fing. This pressure drop—or head loss—is caused by friction between the fluid and the fitting wall and by creation of turbulence in the body of the fluid. "The loss due to wall frietion is best handled by treat- ing the fitting as a picce of straight pipe, of the same physical length as the fitting. All common prediction methods, and the two-K method, do this. But each method predicts the remaining “excess” head loss a dif- ferent way. Equivalent length “The equivalent-length method adds some hypotheti- cal length of pipe 10 the actual length of the fing Yielding an “equivalent length’ of pipe (L,) that has the Same total los asthe fitting. The unfortunate drawback to this simple approach is that the equivalent length for a given iting isnot constant, but depends on Reynolds number and roughness, as well as size and geometry “Therefore, use of the equivalentlength method requires consideration of all these factors “The excess head loss in a fting is due mostly to turbu- lence caused by abrupt changes in the direction and speed of flow. Thus tf best to predict this loss by using 2 velocity-head approach. Velocity head ‘The amount of kinetic energy contained in a stream is the velocity head. An equivalent statement is that the velocity head is the amount of potential energy (head) necessary to accelerate a fluid to its flowing velocity. For example: Pressure gages on both sides of a grad- wal, frietion-free pipe entrance would show that the pressure in the flowing fluid is lower than the pressure in the feed tank by one velocity head. (This is why an ‘eductor works.) The potential (pressure) energy of the fluid in the tank is not lost; it has been converted to kinetic energy. The number of velocity heads (H,) in a flowing stream is calculated directly from the velocity of the stream (0) Hy = 0% 96 With this background, consider a square elbow. The centering fluid experiences a pipelike frictional head loss as it moves down the inlet leg. At the turn, the flow stops abruptly and starts in a new direction. Since the inlet velocity vector has no component in the outlet di rection, all of the inlet kinetic energy is lost, Thus, this part of the loss in a square elbow is clofe to one velocity head. The remaining losses are the frictional losses in the turn and the outlet leg. The total head loss in the elbow is the sum of the frictional and dircetional losses. The excess head loss (AH) is less than the total by the amount of frictional loss that would be experienced by straight pipe of the same physical length. (Of course, the actual frictional loss in the fitting will be different than the loss in a pipe.) The excess loss in a fitting is normally expressed by a dimensionless “K factor”: AH=KH, The two-K method = Kis a dimensionless factor defined as the excess head. loss in a pipe fitting, expressed in velocity heads. In gen- tral, it does not depend on the roughness of the fitting (or the attached pipe) or the size of the system, but it is a function of Reynolds number and of the exact geome- ty of the fitting. The two-K method takes these de- pendencies into account in the following equation: Nye + Kell + /ID) K for the fitting at Np, K for a large fitting at Internal dia. of attached pe, in. How Nie and fitting size affect K ‘Why two KS, when the literature usually reports a single K value? Most published X values apply to fully- developed turbulent low. This is convenient because K is independent of Ng, when Np, is sufficiently high However, K starts to rise as Nq, decreases toward 1,000, and becomes inversely proportional to Nq, when Np, is below 100,awa head losses 1n pipe fittings+ From (3.4) « Fam (11 2 Foe ntnos tes. ew Cane mtb (6 spo or shar a owe 12 Car 0K etn = LT att bY i [Exces head ss (K), velocity heads oy b oF 0405 19 2 4 6810 2 a0 00% Internal ia of elbow 0, in ee 0 100 1,000 ~Yo.000 700.000 Reynolds ruber (Yq "The two-K method fits head-loss data for laminar, transitional and turbulent flow Fig. (Gas of show affects Fig. | isa plot of K'vs. Nyy for short-radius elbows 2} Constants for two:K method Note that the two-K expression, with 800 for Ky and (0.40 for K., fits the points accurately in all flow regimes. Ti this case, K, has no effect on the predicted K at Np, aia above 10,000; K. is negligible below an Ng, of 50. ‘Theoretically, K should be the same for all Sittings that are geometrically similar. In faet, smaller tings 2 fare more sensitive to surface roughness and have more abrupt changes in crossscction. Thus K is greater for Fieting woe |r x Standard (RIO ~ 1), serewedt Standard (R/O = 11, fangedisiged Longradivs (/D = 15), al types we MED ea smaller fittings of a given type. Mirred 2 Weld (45% anal) 035, 2 ee 3S ‘The 1/1D correction in the two-K expression ac- (ais ‘counts for the size differences: K is higher for small sizes, er SS Waid (18° angles! ozs bbut nearly constant for large sizes. Fig. 2 is a plot of K vs. pipe size data for long-radius (R/D = 1.3) elbows [J.3,4}. The solid line shows how the two-K correlation 8 fits these points; the other lines are correlations that will ae be discussed later. Stands (A/D 1) al types 45 Miered, 1 weld, 45" angie Mitered, 2 weld, 21" ales Standard (R/O = 1), srewed 180° Standard (R/D = 1), langed/welded Lona radius (A/D = 1.1, all tyees 11000 030 Standoed, seemed Long vais, screwed 3+ Standord, flanged oF welded S888 8888 888 Recommended values ‘The table lists values of K, and K. derived from plots of K vs, Np, and size (similar to Fig. 1,2). The reader is encouraged to keep this and use it, because it is the heart of the owo-K method. | ‘Three special cases are not listed in the table because 88 a Serre the ae coreeson of the twor equation doc not apply un d them. ‘The flowing equation applis pipe ex: through Flanged or weldee er Te eter een Oo OT ae me Geom rch 5 Gate, Full fine size, p= 1.0 Ka KY/Nge + Ke bale) Resend n=O : ug Reduced tim, p= 0.8 ‘The constants are (K. is the “classic” K): onal 1, Pipe entrances (Fig 3): Ky = 160; K. = 0.50 for ats Gib. ano Yam ‘normal” entrance, and 1.0 for “Borda” entrance. Buttery 2 Pipe exit: Ky = 0; Ke = 10. ce 3. Orifice: XK,‘ is variable; K, = 291 (1 — A) ae (0/8 = Va where ithe aco of ieee. wo pipe Tani inside di Two-K vs. equivalent length Why use the two-K method when the equivalent- length method is more familiar? and easier to use? This Note: Use A/D = 1. values for A/D = 8 98 TATRA TTNomenclature - Inside pipe dia, f D J Moody ffictian factor (= 64/Np, for taminar 4, Ne Reynolds number for flow (Np, Equivalent length of a fiting (L, = KD//). D+) flow) fn” Number of fitings of a given type fr “Standard” friction factor for head loss in fitting AP Pressure drop (AP = p3H/144), psi 2 — Acceleration due to gravity, 3217 ft/s? R/D_ Bend radius of an elbow divided by inside dia. of Hy Velocity head, fof Haid AH Head loss, feof fuid 1D Inside pipe dia. in, K Excess head loss fora fitting, velocity heads pipe Fluid velocity, f/s Ratio of erfice dia. to pipe inside dia K, for fitting at Ng, = 1, velocity heads Viscesty of fluid, ty 8 € Roughness of pipe wall, ft “ ° K. K for very large fitting at Ng, = 20, velocity, Density of uid 1b/° heads L Length of pipe, including physical length of fit- | i tings, fe t ‘Two-K method K-factor method [5] Porm Hf 2: R= KM Reh + 1/10) tan aio) oA, Find K for Bttings: Find K for fittings and exit: \ Thee 8K, 1 inn soe «| | 90° elbows: 6 800 1.20 90° elbows: 6 0.22 | Tatar 2 80 ‘a Bee ba ‘nae 4 he ha canes cas Te a8 ma : 3 | = 7,400/1,210,000 + 3.1001 3305 Find X for exit and straight pi ight pipe: i P an ea $1 or rpe dd Find head loss: heed We kd, pare , (5h +104 ossnssn Saison ] New Crane method [6] | Old equivalent-length method [1] | Form: AM = (/,/D) Hy Form: AH = ((/L/D) + K) Hy | ‘fy for this system is 0.013 (p. A26) Find equivalent nt 1] cet ad tok | xs | 90" eos ‘oe | al | | 90: ets vas & i300 | Teer Gate utes) 2 . | “Tees (rice outies) 0.790 2 | Sar ai 2 8 | Geasniva oir 2 ome | | matt ' » Exit 10001 Sig ie Pe = | Tal arn | | Find tr si i Find head loss: 4 ight pipe | | Kn Q/D =093 evn Mi = /Ds, (0.0122 (687/1.302)(1.554) Find head loss: . | 928 f (JL/D) + K) Hy 0.957 + 4.328\1558) lak | ETAT RETR TAT 99Inward-orojecting "Norma, squareedged (Gorda) entrance Pipe-entrance K values do ‘not depend on diameter Fig. 3 classic method, in which each type of fitting has one “equivalent length,” is reliable for 1-6 in. earbon-stee! piping in normal runs (see the dashed line in Fig, 2). In large, complex alloy systems, the method could predict head losses 1.5-3 times too high. That means oversized pumps and a large waste of energy and capital. In lami- nar flow, on the other hand, it could predict head losses ‘a whole order of magnitude too low. ‘The equivalent-length concept alo contains a booby trap for the unwary. Every equivalent length has a spe- cific friction factor (/) associated with it, because the ‘equivalent lengths were originally developed from K factors by the formula L, = K D/j. This is why the lat- ct version of the equivalent-length method (the 1976, edition of Crane Technical Paper 410 [6)) properly re- quires the use of two fiction factors. The first is the actual friction factor for flow in the straight pipe (/), and the second is a “standard” friction factor for the particular fitting (/,). Thus the two-K method is as easy to use as the updated equivalentlength method. And the results are as accurate. ‘What about the widely-used K-factor graphs pub- lished by the Hydraulic Institute? (See [5] for a good presentation of these graphs.) The graphs are good for 1-8 in. pipe in fully turbulent flow (see dotted line in Fig. 2), but extrapolation to larger sizes can cause er- rors. For example, the K-factor line in Fig. 2 shows a K’ of 0.075 for a 36-in. elbow, but the actual K is about 0.200. OF course, these charts greatly underestimate laminar head losses, and should not be used for Ng, below 10,000, Example Consider a 164n. Sch 108 stainleststel system as shown in Fig. 4 The system contains 100 actual fof Pipe; 6 long-radius (normal for most systems) elbows, Sde-outlr tees 2 gate valves and an exit into a tank The uid has a visosty of | cP, a specie gravity of 1 and i owing at 10 fs What isthe head fos through this sytem? ‘Lets frst calculate and convert the given data to get the needed information 1 x 6243 = 62.43 Ib/ft? 1 672 x 10 = 6.72 x 10 Ib/fies 100 Pine and itings: 100 fof 16in, Sch10S staniessatel pipe © Slongradius (270 = 1.5) elbows side outlet tees 2 gate valves 15.624 in, for Sch 108 pipe D = 15.624/12 = 1.302 ft Nae = (10)(1.302)(62.43)/(6.72 x 10-4) = 1,210,000 Hy = 07/2g = 10°/64.34 = 1.594 ft of uid Given ¢ = 0.00005 ft for stainless pipe, we can find f from the Colebrook equation: / = 0.0122. Thus, L/D = (0.0122)(100)/(1.302) = 0.937 (this is the K value for the pipe itself) ‘The four boxes (on p. 99) show how to calculate the total head loss by the two-K method and three other methods. The results: 1. Two-K method: AM = 8.15 ft 2. Old equivalent-length method: AH =9.28 ft (14% high). 3. K-factor method: AH = 6.52 ft (20% low), 4. Revised Crane method: AH = 8.18 fk Note that flow was fully turbulent in this example, For laminar flow, the equivalent-length and K-factor meth- ods would have been off considerably more. ‘Merk Lips, Edioe References 1 Freman, JR “Experiments Upon the Flow of Water Pipe ‘iim "Ameican Sx nf Mechanic! Engin: Ne ort 2 Rare Pe Row Rena one for aaa aS Frou Loma Tyg Pipe in Figs Tom ASME Pag RJ 8, Loses Pipe aed in, Te ASME. 79 2. 2. Sippon LL, Sine ie poe plan Che Eg J Pi shat Vs” Crane Tec Pape 10, ih ising The author {lam 8 Hooper en Moana Depe of Momane Cos 0 Eialverg Bla'Sc Louse MO 63408 ttm ini pce ‘ifr vale in designing antoe fettded stone, Syte eaes ed vin plata Hope beds BS gre ie chanical eget ee the Unterty of Olan Imemter f ACME: and ep Protioalcagines Sut