9.

FRICTION LOSS ALONG A PIPE

Introduction In hydraulic engineering practice, it is frequently necessary to estimate the head loss incurred by a fluid as it flows along a pipeline. For example, it may be desired to predict the rate of flow along a proposed pipe connecting two reservoirs at different levels. Or it may be necessary to calculate what additional head would be required to double the rate of flow along an existing pipeline. Loss of head is incurred by fluid mixing which occurs at fittings such as bends or valves, and by frictional resistance at the pipe wall. Where there are numerous fittings and the pipe is short, the major part of the head loss will be due to the local mixing near the fittings. For a long pipeline, on the other hand, skin friction at the pipe wall will predominate. In the experiment described below, we investigate the frictional resistance to flow along a long straight pipe with smooth walls.

Friction Loss in Laminar and Turbulent Pipe Flow

Fig 9.1 Illustration of fully developed flow along a pipe Fig 9.1 illustrates flow along a length of straight uniform pipe of diameter D. All fittings such as valves or bends are sufficiently remote as to ensure that any disturbances due to them have died away, so that the distribution of velocity across the
90

pipe cross section does not change along the length of pipe under consideration. Such a flow is said to be "fully developed". The shear stress τ at the wall, which is uniform around the perimeter and along the length, produces resistance to the flow. The piezometric head h therefore falls at a uniform rate along the length, as shown by the piezometers in Fig 9.1. Since the velocity head is constant along the length of the pipe, the total head H also falls at the same rate. The slope of the piezometric head line is frequently called the "hydraulic gradient", and is denoted by the symbol i:
i = − dh dl = − dH dl

(9.1) (The minus signs are due to the fact that head decreases in the direction of increasing l, which is measured positive in the same sense as the velocity V. The resulting value of i is then positive). Over the length L between sections 1 and 2, the fall in piezometric head is h1 − h2 = iL (9.2) Expressed in terms of piezometric pressures p1 and p2 at sections 1 and 2: p1 − p2 = wiL = ρgiL (9.3) in which w is the specific weight and ρ the density of the water. There is a simple relationship between wall shear stress τ and hydraulic gradient i. The pressures p1 and p2 acting on the two ends of the length L of pipe produce a net force. This force, in the direction of flow, is (p1 − p2)A in which A is the cross-sectional area of the pipe. This is opposed by an equal and opposite force, generated by the shear stress τ acting uniformly over the surface of the pipe wall. The area of pipe wall is PL, where P is the perimeter of the cross section, so the force due to shear stress is τ.PL
91

Equating these forces: (p1 − p2)A = τ.PL This reduces, by use of Equation (9.3), to
τ = A ρgi † P

(9.4) Now expressing A and P in terms of pipe diameter D, namely, A = πD2/4 and P = πD so that (A/P) = D/4, we obtain the result:
τ = D ρgi 4

(9.5) We may reasonably expect that τ would increase in some way with increasing rate of flow. The relationship is not a simple one, and to understand it we must learn something about the nature of the motion, first described by Osborne Reynolds in 1883. By observing the behaviour of a filament of dye introduced into the flow along a glass tube, he demonstrated the existence of two different types of motion. At low velocities, the filament appeared as a straight line passing down the whole length of the tube, indicating smooth or laminar flow. As the velocity was gradually increased in small steps, he observed that the filament, after passing a little way along the tube, mixed suddenly with the surrounding water, indicating a change to turbulent motion. Similarly, if the velocity were decreased in small steps, a transition from turbulent to laminar motion suddenly occurred. Experiments with pipes of different diameters and with water at various temperatures led Reynolds to conclude that the parameter which determines whether the flow shall be laminar or turbulent in any particular case is Re = ρVD µ = VD ν (9.6) in which Re = Reynolds number of the motion ρ = Density of the fluid

The term (A/P) which appears here is frequently called the "hydraulic radius" or "hydraulic mean depth", and may be applied to cross sections of any shape.

92

This lower value of Re is found experimentally to be about 2000. Fig 9. no matter how severe. Experiments made with increasing flow rates show that the critical value of Re for transition to turbulent flow depends on the degree of care taken to eliminate disturbances in the supply and along the pipe. obtained by dividing the discharge rate Q by the cross sectional area A µ = Coefficient of absolute viscosity of the fluid ν = µ/ρ denotes the coefficient of kinematic viscosity of the fluid Note that the Reynolds number is dimensionless. On the other hand. In each case the velocity rises from zero at 93 . as may readily be checked from the following: ρ [M L−3] µ[M L−1 T−1] ν [L2 T−1] V [L T−1] D [L] The motion will be laminar or turbulent according as to whether the value of Re is less than or greater than a certain critical value.V = Q/A denotes the mean velocity of flow.2 Velocity distributions in laminar and turbulent pipe flows Fig 9. Below this.2 illustrates the difference between velocity profiles across the pipe cross sections in laminar and in turbulent flow. experiments with decreasing flow rates show that transition to laminar flow takes place at a value of Re which is much less sensitive to initial disturbance. the pipe flow becomes laminar sufficiently downstream of any disturbance.

As the Reynolds number increases. we might reasonably relate the wall ‡ The derivation of the parabolic velocity profile. When considering such experimental results. is given in many standard textbooks.5) from this equation leads to the result i = 32 νV gD 2 (9.12 as Re increases from 104 to 107. the velocity distribution is much flatter over most of the pipe cross section.9) Substituting for τ in Equation (9. In the case of laminar flow.24 to about 1.7) and the velocity gradient at the wall is given by du dr = R − 4U D = −8 V D (9. the velocity profile is parabolic‡. Typically. U/V falls from about 1. 94 . The ratio U/V of centre line velocity to mean velocity is U = 2 V (9. the profile becomes increasingly flat.8) so that the wall shear stress τ due to fluid viscosity is τ = 8µV D (9.the wall to a maximum value U at the centre of the pipe. it is not possible now to find a simple expression for the wall shear stress.10) which is one form of Poiseuille's equation. The mean velocity V is of course less than U in both cases. and of the Poiseuille equation. so the value has to be found experimentally. Because of the turbulent nature of the flow. the ratio of maximum to mean velocity reducing slightly. In the case of turbulent flow.

h2) between sections 1 and 2 of a pipe of diameter D. due to Blasius. which is due to Prandtl. and the following result is readily obtained: i = 4 f V2 D 2g (9. The results of many experiments on turbulent flow along pipes with smooth walls have shown f to be a slowly decreasing function of Re.5).11) The hydraulic gradient i may now be expressed in terms of f by use of Equation (9.15) 95 .13) where L is the length of pipe run between the sections. the head loss (h1 . along which the mean flow velocity is V. So a dimensionless friction factor f could be defined by τ = f. is seen from Equation (9.079Re−1/4 (9. Another correlation. Various correlations of the experimental data have been proposed. one of which is 1 f = 4 log( Re f ) − 0.4 (9.14) This expression. This is frequently referred to as Darcy's equation. is: f = 0.12) Therefore. ρV 2 (9.shear stress τ to the mean velocity pressure ρV 2 . fits experimental results well in the range of Re from 104 to 107.2) to be given by h1 − h 2 = 4f L V2 D 2g (9. although it does have the slight disadvantage that f is not given explicitly.

the value of f may be found theoretically from Poiseuille's equation. 96 . There is no corresponding theoretical for turbulent flow. Water from a supply tank is led through a flexible hose to the bell-mouthed entrance of a straight tube along which the friction loss is measured. in the case of laminar flow. the hydraulic gradient i may conveniently be expressed in terms of a dimensionless wall friction factor f. but good correlation of many experimental results on smooth walled pipes is given by equations such as (9. which reads head loss directly in mm of water gauge.16) In summary. Equating the expressions for i in Equations (9.10) and (9.14) to within about 2% over the limited range of Re from 104 to 105. We have seen that when the flow is turbulent it is necessary to resort to experiment to find f as a function of Re. or to a U-tube containing water and mercury to cover higher values of head loss.3.This gives explicit values which are in agreement with those from the more complicated Equation (9. These clear lengths upstream and downstream of the test section are required to ensure that the results are not affected by disturbances originating at the entrance or the exit of the pipe. Above 105. the Blasius equation diverges substantially from experiment.15). The piezometer tappings are connected to an inverted U-tube manometer. Piezometer tappings are made at an upstream section which lies approximately 45 tube diameters away from the pipe entrance. and approximately 40 diameters away from the pipe exit. Description of Apparatus The apparatus is illustrated in Fig 9.12): 32 νV g D2 After reduction this gives the result f = = 4f V 2 D 2g 16 Re (9.14) and (9. However. This factor has the theoretical value f = 16/Re for laminar flow along a smooth walled pipe. however.

The bench supply valve is then carefully opened and adjusted until there is a steady flow down the overflow pipe from the supply tank. Experimental Procedure The apparatus is set on the bench and levelled so that the manometers stand vertically. and may be measured by timing the collection of water in a beaker which is weighed on a laboratory scale or measured in a volumetric cylinder. so that it provides a constant head to the pipe under test. whereupon 97 . With the needle valve partly open to allow water to flow through the system. any trapped air is removed by manipulation of the flexible connecting pipes. The needle valve is then closed.3 Apparatus for measuring friction loss along a pipe The rate of flow along the pipe is controlled by a needle valve at the pipe exit. The water manometer is then connected to the piezometer by opening the tap at the downstream piezometer connection. (The discharge rate is so small as to make the use of the bench measuring tank quite impractical).Fig 9. Particular care should be taken to clear all air from the piezometer connections.

These readings should comfortably cover the whole of the laminar flow region and the transition from turbulent flow. It is recommended that the manometer is read several times during the collection period. check that the flow has been stopped absolutely. The first reading of head loss and flow may now be taken. and that it never becomes immersed in the discharged water. If they do not. and the rate of flow measured. The bleed valves and needle valve are then closed. further attempts should be made to clear the connections of air. The needle valve is then closed in stages. and the needle valve opened slightly. The water temperature should be observed as accurately as possible at frequent intervals. It is advisable to plot a graph of differential head loss against flow rate as the experiment proceeds to ensure that sufficient readings have been taken to establish the slope of the straight line in the region of laminar flow. the joints should be secured using hose clips. to provide readings at a series of reducing flow rates. The needle valve is opened fully to obtain a differential head of at least 400 mm. The bleed valves at the top of the U-tube are then opened to flush out any air in the connecting tubes. and a mean value of differential head taken. Readings of head loss and flow rate are now taken. is available. With the pump running. The supply hose from the overhead tank is disconnected and replaced by one from the bench supply pipe. starting at the maximum available 98 .the levels in the two limbs of the manometer should settle to the same value. During the period of collection. so that there is a moderate discharge from the pipe exit. If a suitable laboratory scale. the time required to collect a chosen volume is measured. If a volumetric cylinder is used. Otherwise. and that all air has been cleared from the piezometer connections. ensure that the outlet end of the flexible tube is below the level of the bench. the differential head and rate of flow may change. If it does not. the discharge is collected over a timed interval and then weighed. or may be depressed by pumping in air through the valve. The water manometer is isolated by closing the tap at the downstream piezometer connection. and a check is made that the U-tube shows no differential reading. To obtain a range of results with turbulent flow it is necessary to use water from the bench supply and to measure differential heads with the mercury-water U-tube. The height of the water level in the manometer may be raised to a suitable level by allowing air to escape through the air valve at the top. Manipulation of the tubes will help to fill the whole length of the connections from the piezometer tappings right up to the surfaces of the mercury columns in the U-tube. the bench supply valve is opened fully. Since the equipment will be subjected to the full pump pressure. weighing to an accuracy of 1 g. especially at the lower flow rates.

28 −1.069 mm2 = 7.698 28.0 223. a few readings should be taken below about 40 mm on the mercury-water U-tube.407 3.5 325.98 −1.962 0.691 0.3 1.128 0. The diameter D of the tube under test.195 3.48 −1.178 84.887 1.792 0.0 362.061 1947 10.3 0. Noting that 1 mm differential reading on the mercury-water U-tube represents 12.00 mm 2 Cross sectional area of pipe πD /4 A = 7. should be noted.916 0.4 h1 521.910 2.9 93.0 349.0 85.734 0.747 2.457 0.0 114.072 812 558 280 144 476. Students are encouraged to use a programmable pocket calculator or to write a suitable program for use on a computer for reducing the experimental data.61 −1.0 145.605 −0.071 −1.0 427.52 −1.083 2.1 and 9.385 3.248 0.1 Results with water manometer § § A great amount of repetitive computation is required to reduce the results shown in Tables 9.051 Table 9.296 −1.069 × 10−6 m2 Qty (ml) 400 400 400 400 400 300 300 300 200 150 85 50 t (s) 50.161 −0.444 3. Results and Calculations Length of pipe between piezometer tappings L = 524 mm Diameter of pipe D = 3.586 0.2.497 −0.0 θ V (mm) (mm) (°C) (m/s) i log i −0.893 −1.8 67.306 0.0 113.0 316.6 mm differential water gauge.315 −0.0 h2 56.20 −1.161 15.2 57.978 2954 10.8 340.8 61.590 0.0 15.319 100.5 174.0 375.980 1565 13.085 0.0 390.048 0.5 306. using the needle valve to set the desired flow rates .0 295. The water temperature should be recorded at frequent intervals.8 71.550 66.32 −1.discharge rate.0 500. It is desirable to provide some overlap of the ranges covered by the mercury U-tube and the water manometer.6 332.3 0.9 92.842 0.71 −1.0 282.8 54.875 1211 20.960 2552 10.5 129.724 −0.232 −0.70 −2.0 452.07 −1.765 103f log f Re log Re 3.998 2233 8.017 10.04 −1.471 3.974 2779 10.988 2428 10.101 −0.290 3.0 58.052 −0.0 263. and the length L between the piezometer tappings.106 0. and reducing in stages.0 15. 99 .41 −1.349 3.0 245.114 0.448 2.211 0.055 0.484 0.189 0.

420 0.206 1.044 −0.99 −2.02476 (θ − 20) + 0.658 Table 9.738 5 1.5 358.836 0.3 gives values of 106ν.0 254.3 Table of 106ν (m2/s) as a function of water temperature θ 0C Tables 9.055 0.898 3.5 2.058 6566 8.027 5414 9.675 2.0 62.607 351. and Table 9.683 9 1. expressed in units of m2/s.6 51.710 7 1.40 −2.754 0.633 1.7 68.5 i log i 0.5 283.809 2.214 0.271 0.110 0. they may be obtained from the empirical formula 106 ν = 10049 .854 0.232 2.0 362.0 15.0 377.265 5.929 2382 390.106 8704 8. Alternatively.307 1.625 7.046 5853 9.303 0.170 0.20 −1.Qty (ml) 900 900 900 900 900 900 900 600 600 600 300 t (s) 39.5 370.34 −2.859 2.082 0.00044(θ − 20) 2 (9.0 195. Values of ν.034 4813 9.893 0.2 Results with mercury manometer θ°C 10 20 30 0 1.734 3.0 214.627 0.4 48.5 54.0 42.955 3248 11.195 2.0 240.77 −1. in the range from 10°C to 39°C.512 3.25 −2.5 275.980 0.106 0.463 3.818 0.17) which fits experimentally measured values of ν very well over the range of θ from 15°C to 30°C.202 0.0 294.079 7285 8.753 4 1.817 3. 100 .1 and 9.767 3.9 1.696 8 1.784 2 1.031 2.5 3.236 0.862 3.557 0.37 −2.5 261.0 402.5 305.469 0.5 47. − 0.914 0. which are needed to compute Reynolds numbers.682 3.940 3.004 0. as a function of water temperature θ°C.008 1.700 3.029 0.028 5006 9.555 1.342 0. may be obtained by interpolation from this table.732 4.204 103f log f Re log Re 3.377 431.002 4187 11.75 −2.0 15.635 Table 9.957 0.884 0.769 3 1.140 0.03 −2.96 −2.83 −2.682 0.6 70.2 present typical results obtained using the water and mercury manometers.095 7913 8.935 0.7 58.946 2.5 15.873 0.787 2.0 340.9 46.0 331.0 h1 h2 θ V (mm) (mm) (°C) (m/s) 414.200 1.968 4.724 6 1.801 1 1.622 3.0 270.0 226.

used for graphical representation.052 To obtain the friction factor f.887 = − 0.0) 524 = 0.874 × 10-6 7.8 = 7.1. in the first line of the table. values of i are obtained simply from i = ( h1 − h2 ) L For example.874 × 10-6 m3 s Q = = so the velocity V along the pipe is Q A 7. the flow rate Q is Qty t 400 × 10-6 50.114 m s The velocity head is then V2 2g 1.12) may now be used to find f: i = 4f V 2 D 2g so that 101 .0 − 56. for example. we first compute the velocity head as follows: In the first line of the table.069 × 10-6 V = = = 1. is log i = log 0.887 and log i.0632 m Equation (9.In Table 9.114 2 (2 × 9.81) = = 0. i = (521.

are done in identical fashion.6 ( h1 − h 2 ) L In the first line of Table 9.17).00 × 10 −3 1 × 4 0.0632 = 10.52 × 10 −3 f Then log f = log 10.6.52 × 10-3 ( ) = − 1. to have the value ν = 1131 .978 Finally.00 × 10−3 1. for example.887 × 3.2.2.471 Calculations for Table 9.f = i D 1 4 V 2 2g and inserting numerical values for the first line of the table.114 × 3.3. containing results using the mercury-water manometer. × 10 −6 m2 s at the relevant temperature of 15. or from Equation (9. except that the differential heads recorded from the manometer need to be expressed as equivalent heads of water by multiplication by the factor 12. Hence Re = 1.131 × 10−6 = 2954 and log Re = 3. = 0. Reynolds number Re is obtained from the definition Re = VD ν in which ν is found by interpolation from Table 9. So the hydraulic gradient now becomes i = 12.3°C. 102 .

due to a water column of height (h1 − h2) and specific weight w. and recalling that pu = pv. and are connected hydrostatically round the bottom of the U-tube.0) 524 = 5.675 Fig 9. is pu − ps = w(h1 − h2) In the right hand limb. we see that hs − ht = (s − 1)(h1 − h2) The specific gravity s of mercury is 13. so the pressure difference between V and T is pv − pt = sw(h1 − h2) Subtracting these results. so the mercury is driven down to point U in the right hand limb and up to point T in the right. The difference of pressure between U and S in the left hand limb. Now the pressures pu and pv at points U and V shown on the diagram are equal. The pressure applied to the left hand limb of the mercury-water U-tube is greater that applied at the right.6.4 Diagram of mercury-water U-tube Fig 9. since these points are at the same level. the mercury column of height (h1 − h2) has specific weight sw.0 − 195.6 arises. where s is the specific gravity of mercury. The difference of levels of these points is (h1 − h2).4 shows how the factor 12. so that 103 . we obtain ps − pt = (s − 1) w(h1 − h2) Expressing this as a differential head of water of specific weight w.i = 12.6 (431.

412 s m Rewriting Equation (9. above which the proportionality does not apply.5 Variation of hydraulic gradient i with velocity V up to 1 m/s Fig 9.10) in the form ν = g i 2 D 32 V and inserting numerical values. the slope of the linear portion is found to be i V = 0.5 shows how the hydraulic gradient varies in proportion to flow velocity V over a range from zero to the critical value. From the graph.hs − ht = 12.6(h1 − h2) Fig 9. 104 . The critical value of Re for transition from turbulent to laminar flow (the experiment having been performed with decreasing flow rate) is 1950. which is a form of Poiseuille's equation. Equation (9.10). may be used to infer the coefficient of kinematic viscosity from measurements in the region of laminar flow.

The straight line corresponding to Equation (9. of course.15) for turbulent flow is also shown.6 shows logarithmic graphs of both hydraulic gradient i and friction factor f as functions of Reynolds number Re. viz Re = 1950.29 to 3. so confirming the validity of Poiseuille's equation in the laminar flow regime.00 × 10−3 32 ( ) 2 ν = 1.3 at the working temperature of 15. This follows.29.ν = 9. from the good correspondence which has been found between the value of ν obtained from Poiseuille's equation and the value obtained from standard data.81 × 0.43 (Re 105 .6 Variation of log i with log Re Fig 9. this agrees with the value ν = 1. Transition occurs at the value log Re = 3. and it is clear that excellent agreement with experiment is obtained.131 × 10−6 m2/s obtained from the standard data of Table 9.3°C. The straight line corresponding to the Blasius Equation (9.137 × 10−6 m2/s Within the limits of experimental error. Fig 9.412 × 3. In the range of log Re from 3.16) for laminar flow is shown on the figure.

f rises along a curve as Re increases. Using values of µ and ν taken from physical tables. Devise a simple method of producing a steady flow of air at a known rate by displacement from a closed vessel. 3. For higher values of Re. i.03 mm in measurement of D.from 1950 to 2710). What suggestions do you have for improving the apparatus? What percentage changes in the computed values of V. a fair agreement with the Blasius friction factor is found. Questions for Further Discussion 1. i) ii) An error of 1. Consider whether this could be measured using a water U-tube. f. calculate the likely critical velocity and the corresponding pressure drop. 106 .0 mm in measurement of L. A possible project is the adaptation of the apparatus to operation with air instead of water. ν and Re would you expect to result from. An error of 0. 2.

the water flows along a pipe of some other diameter Dd. For long pipes with few fittings. the overall loss is dominated by wall friction. Over this region.1 Schematic representation of loss at a pipe fitting Fig 10. The figure indicates the variation of piezometric head along the pipe run. the distribution of velocity across the pipe remains unchanged from one cross section to another. In the experiment described below. and at fittings such as bends or valves. far upstream of the fitting. then the principal losses are those which are produced by disturbances caused by the fittings. the pipe is short and there are numerous fittings. loss of head along a pipeline is incurred both by frictional resistance at the wall along the run of the pipe. If.10. the piezometric head falls with a 107 . however. this is the condition of fully developed pipe flow which is considered in Chapter 9. LOSSES AT PIPE FITTINGS Introduction As described in Chapter 9. as would be shown by numerous pressure tappings at the pipe wall. along which the velocity of flow is Vd. Measurement of Loss of Total Head at a Fitting Fig 10.1 shows water flowing at speed Vu along a pipe of diameter Du towards some pipe fitting such as a bend or a valve. In the region of undisturbed flow. typical of those which are used frequently in pipe systems. Downstream of the fitting. but shown for simplicity as a simple restriction in the cross section of the flow. we investigate losses at various fittings.

a loss of piezometric head ∆h due to the fitting is found.1) It is convenient to express this in terms of a dimensionless loss coefficient K. as the velocity distribution gradually returns to the condition of fully developed pipe flow. caused by rapid changes in direction and speed as the water passes through the fitting. In the downstream region. Because the velocity varies across the pipe cross section. The result is K = ∆H Vu 2 2g or ∆H Vd 2 2g (10. so we may simplify the definition to K = ∆H V 2 2g or ∆h V 2 2g (10. by dividing through by the velocity head in either the upstream or the downstream pipe (the choice depending on the context. the velocity head also varies over the cross section. these disturbances die away. there are sharp and substantial local disturbances to the piezometric head. however. as we shall see later).uniform.05 to 1. from zero at the wall to a maximum at the centre. † 108 . being typically 1.2a) For the case where Du = Dd. To obtain results of high accuracy. If the upstream and downstream lines of linear friction gradient are now extrapolated to the plane of the fitting.1 it is clear that V2 V2 ∆H = ∆h + u − d 2g 2g (10. are needed to establish with certainty the relative positions of the linear The velocity head V2/2g used here is based simply on the mean flow velocity V. mild gradient. (of 60 pipe diameters or more). as a result of constant friction at the pipe wall in the fully developed flow. The mean value of velocity head in this non-uniform flow is somewhat higher. To establish the corresponding loss of total head ∆H it is necessary to introduce the velocity heads in the upstream and downstream runs of pipe. and the line of piezometric head returns asymptotically to a slight linear gradient. Close to the fitting. the flow velocities in the upstream and downstream pipes are identical. long sections of straight pipe.07V2/2g when the flow is turbulent. From Fig 10.2b) where V denotes the flow velocity in either the upstream or the downstream pipe run†.

An estimate is then made of the friction head loss ∆hf which would be incurred in fully developed flow along the run of pipe between the piezometer tappings. The radius of the bend is R. sudden enlargement and sudden contraction Fig 10. As the flow leaves the bend. Such long upstream and downstream lengths are impracticable in a compact apparatus such as the one described below. The curvature of the flow as it passes round the bend is caused by a radial gradient of piezometric head.2(a) illustrates flow round a 90° bend which has a constant circular cross section of diameter D. These show the piezometric head loss ∆h' between the tappings. these heads start to equalise as 109 . measured to the centre line.3) Characteristics of Flow through Bends and at Changes in Diameter Fig 10. so that the piezometric head is lower at the inner surface of the pipe than at its outer surface. one placed upstream and the other downstream of the fitting. just two piezometers are used. Instead.sections of the piezometric lines. at sufficient distances as to avoid severe disturbances. The piezometric head difference ∆h across the fitting is then found by subtraction: ∆h = ∆h ′ − ∆h f (10.2 Flow in a bend.

For this case. The vigorous turbulent mixing.the flow loses its curvature. and which generates further losses that are attributable to the bend. causes a loss of total head. starts at the same value as in the pipe immediately upstream. are illustrated in the figure. so generating mixing losses in the subsequent turbulent reattachment process. and inward components near the pipe walls. Assuming that the piezometric pressure on the face of the enlargement to be equal to that in the emerging jet. we would expect the value of K to fall. the loss of total head may be shown to be ∆H = ( Vu − Vd ) 2g 2 (10. making the bend less sharp. the radial gradient of piezometric head sets up a secondary cross-flow in the form of a pair of vortices. forming a jet which diffuses into the larger bore. the value of K falls. As R/D increases. Fig 10. as this ratio increases.2 as R/D increases up to 2 or 3. so that the piezometric head begins to rise along the inner surface.2(b) shows the flow in a sudden enlargement.4. and that the momentum flux is conserved. The flow separates at the exit from the smaller pipe.5. These changes in total and piezometric head.5) The loss coefficient K is in this case best related to the upstream velocity Vu so that 110 . which persists for a considerable distance in the downstream flow. so rising across the enlargement. the value of the loss coefficient K will be a function of the geometric ratio R/D. reducing to values which may be as low as 0. The piezometric head in the emerging jet.4) The corresponding rise in piezometric head is ∆h = 2 Vd ( Vu − Vd ) 2g (10. When superimposed on the general streaming flow. Additionally. having outward directed velocity components near the pipe centre. the value of K is usually about 1. and increases through the mixing region. neglecting the effect of friction gradient. however. Clearly. The smallest possible value of R/D is 0. the result is a double spiral motion. This rise causes the flow to separate. There is also a slight dependence on Reynolds number Re. and which reattaches to the wall some distance downstream. resulting from the separation and reattachment of the flow. for which the bend has a sharp inner corner.

however. Equation (10. viz.4). then there is in effect no contraction and there will be no separation of the flow.6 approximately.0. Following the result of Equation (10.6 = 1667 . so the loss coefficient K becomes K = Vd − -1 Vc 2 = Ad −1 Ac 2 (10.8) then gives a zero value of K.7) The obvious choice of reference velocity in this case is Vd.0. The losses are therefore expected to be those due to an enlargement from the contracted area Ac to the downstream pipe area Ad. so Ad/Ac = 1.K = ( Vu − Vd ) 2g 2 Vu 2 2g = V 1− d Vu 2 = A 1− u Ad 2 (10.8) Consider now the probable range of values of Ad/Ac. Beyond this contracted section there is a region of turbulent mixing. the contraction coefficient has the value 0.0 (the case when there is no enlargement). We might then reasonably expect the flow at the entry to the downstream pipe to resemble that from a large reservoir through an orifice of area Ad. that is if Ad/Au = 1. the expected loss of total head in contraction is ∆H = ( Vc − Vd ) 2g 2 (10. to 1. If the value of the pipe contraction ratio is 1. For such an orifice. then the upstream pipe tends to an infinite reservoir in comparison with the downstream one. The flow separates from the edge where the face of the contraction leads into the smaller pipe. forming a jet which converges to a contracted section of cross sectional area Ac. The losses occur almost entirely in the process of turbulent diffusion and reattachment. If. Ad/Au → 0.0 when Au/Ad falls to zero. 111 . in which the jet diffuses and reattaches to the wall of the downstream pipe. the contraction is very severe.6) This indicates that K increases from zero when Au/Ad = 1. Consider lastly the sudden contraction shown in Fig 10. so that Ad Ac = 1 0.0.2(c).

Description of Apparatus Fig 10.3 has the 112 .Substituting this value in Equation (10.44 It might therefore be expected that K would rise from zero when the pipe area ratio Ad/Au = 1 to a value of about 0.44 as the ratio Ad/Au falls towards zero. incorporating selections of fittings in various configurations. The particular equipment illustrated in Fig 10.3 Arrangement of apparatus for measuring losses in pipe fittings Several arrangements of apparatus are available.7) gives K = 0.

and the control valve at the exit is opened to allow water to circulate through the pipework. the air valve at the top of manometer is slackened or removed completely. at clear lengths of 4 pipe diameters. The system may be purged of air by venting to atmosphere through the manometer. as shown on the mimic diagram. The pump is then started. and up the the the the the 113 . The control valve at the exit is then partially closed. It may be operated from the H1 Hydraulic Bench. It provides a run of pipework. along the connecting tubes. made up of components manufactured in rigid plastic material. and through a vent valve at the highest point of the pipe run. so that pressure inside the pipework drives water out through the vent at the top of pipework and through the piezometers. and is discharged at the exit to the measuring tank. The flow rate through the equipment may be varied by adjusting the valve near the pipe exit. so that the discharge from the equipment flows into the measuring tank of the bench. Experimental Procedure Details of procedure will vary according to the facilities provided by the particular equipment in use. The supply hose of the Hydraulic Bench is connected to the pipework inlet. To ensure that all air is expelled from the system. The tappings are connected to a glass multitube manometer which may be pressurised using a bicycle pump. The following description applies to the equipment illustrated in Fig 10. supported in the vertical plane from a baseboard with a vertical panel at the rear. A further hose is fixed to the exit pipe. Water is supplied to the pipe inlet from the hydraulic bench.advantage of portability. and the vent valve at the top of pipework is opened. In the run of the pipe there are the following fittings: • • • • • 90° mitre bend 90° elbow bend 90° large radius bend Sudden enlargement in pipe diameter Sudden contraction in pipe diameter Piezometer tappings are provided in the pipe wall. upstream and downstream of each of the fittings. are noted. The diameters of the pipes and dimensions of the fittings.3.

When this is complete. the levelling screws are then used to set the scale of the manometer board perfectly horizontal. The cycle pump is then used to drive the water levels in the manometer tubes down to a convenient set of heights. this may at any time be checked by opening the air vent for a short time. If it is thought that air might have collected at the top of the pipework. so ensuring that the system is thoroughly purged of air. Results and Calculations Dimensions of Pipes and Fittings Diameter of smaller bore pipe Diameter of larger bore pipe Radius to centre line of elbow Radius to centre line of bend Length of straight pipe between piezometer tapping and fitting D1 = D2 = Re = Rb = 22. alternatively the bench valve may be used to effect part of the flow reduction.1 mm 4D1 or 4D2 If the measured flow rate is Q l/s. When the maximum feasible flow rate is reached. then the velocities V1 and V2 along the pipes of cross sectional A1 and A2 m2 areas are: V1 = 10−3 Q/A1 m/s and V2 = 10−3 Q/A2 m/s 114 .98 × 10−4 m2 29.0 mm 69. The apparatus is now ready for use. to show a uniform reading across the board. the air vent valve should be closed. With the exit valve closed.6 mm A2 = 6.88 × 10−4 m2 35. This flow of water will carry air bubbles along with it.5 mm A1 = 3. while the water levels are observed in the manometer tubes. while the collection of a known quantity of water in the measuring tank of the bench is timed.manometer tubes. and the manometer air valve replaced and tightened. Air is admitted or released as necessary to keep all the readings within the range of the scale. It may be necessary to pump in more air to the manometer to keep the readings within bounds as the exit valve is closed. These measurements are repeated at a number of rates of flow.e. The control valve should be closed sufficiently as to produce vigorous flow out of the air vent valve and the manometer. differential piezometer readings across each of the fittings are recorded. The exit valve is opened carefully. i.

0 28.986 0.394 1.6 135 131 109 91 78 65 46 88 79 71 59 49 45 31 25 21 20 18 19 12 9 30 29 21 16 16 12 8 31 30 28 24 21 12 11 Table 10. Loss of Total Head ∆H (mm) Q (kg/s) V1 (m/s) V2 (m/s) V12/2g V22/2g Mitre Elbow Enlrg’t Cont’n Bend (mm) (mm) 1-2 3-4 5-6 7-8 9-10 0.453 Q m/s Differential Piezometer Readings and Loss of Total Head Differential Piezometer Head ∆h' (mm) Qty (1) Time (s) Q (1/s) Mitre 1-2 Elbow 3-4 Enlrg’t 5-6 Cont’n 7-8 Bend 9-10 24 24 24 12 12 12 12 43.0 19.1 30.161 1. tappings 1 and 2 are upstream and downstream of the mitre bend.671 0.554 0.3 45.6 36.392 0.827 0.2 68.427 0.1 gives a typical set of results as recorded in the laboratory.4 23.621 0.762 0.524 0. viz. Note that the reading for the enlargement is negative.7 26.747 0.570 0.293 1.8 46.6 16.5 34.5 85.9 33.5 11.514 0.462 0.or V1 = 2.329 154 148 126 104 90 75 53 113 102 93 77 64 58 40 -28 -26 -25 -19 -12 -14 -10 109 100 89 71 63 52 36 62 58 55 45 39 28 22 Table 10.514 0.8 49.318 1.392 0. and so on.5 0.329 1.806 0.427 0.0 29. showing an increase of piezometric head at the enlargement.478 99. Differential piezometric heads ∆h' between piezometer tappings are tabulated in sequence in the direction of flow.074 0. 3 and 4 upstream and downstream of the elbow.5 28.554 0.462 0.515 Q m/s and V2 = 1.8 58.1 Piezometric head losses at various rates of flow Table 10.2 Total head losses at various rates of flow 115 .0 88.524 0.

515 × 0. then for the smaller bore pipe: D1 = 22.45 = 1.132 m/s so Re1 = V1 D1 ν = . due to friction between piezometer tappings. However. to find the piezometric head loss ∆h from Equation (10. there is a change in velocity from upstream to downstream.5 6. Several options are available.0 5.55 × 104 116 .3) shows that the total head loss ∆H is the same as the piezometric head loss ∆h.3). If.4 ( ) (9. and the friction loss is generally fairly small in relation to the measured value of ∆h'. and the choice used here is the Prandtl equation quoted in Chapter 9: 1 f = 4log Re f − 0.5 6. This is the case for the mitre. The computations first use an estimate of the head loss ∆hf.67 Table 10. Equation (10. elbow and bend.1. however. and assuming the value ν = 1.45 l/s. since f varies only slowly with Re. as computed from the measurements of ∆h' in Table 10.14 3.3) is used to compute total head loss ∆H from the piezometric head loss ∆h.Table 10.3 Friction factor f for smooth walled pipe It would be possible to evaluate friction factors for each individual flow rate. which is close to the mid range of Table 10.5 × 10 −3 .5 5. it suffices to establish the value of f at just one typical flow rate. 1132 × 22.96 2.3.00 × 10−6 m2/s for the coefficient of kinematic viscosity.48 2.88 3.2 shows the head losses ∆H across each of the fittings. 10-4 Re 103 f 1. Choosing the typical flow rate Q = 0.0 6.14) Typical values derived from this equation. The friction head loss is estimated by choosing a suitable value of friction factor f for fully developed flow along a smooth pipe.0 7. If the velocity downstream of the fitting is the same as that upstream. then Equation (10.1.5 mm and V1 = 2. × 10−6 100 = 2. are presented in Table 10. at about the middle of the range of measurement.73 1.

00611 and f2 = 0.3). 2g = 0196 .654 × 29. consider the mitre bend.4 . The pipe diameter is D1. and the distance between the piezometers. measured along the pipe centreline.13) in Chapter 9. ∆h = ∆h' − ∆hf = 154 − 19 = 135 mm 117 .6 mm and V2 = 1. according to Equation (10.00654 These are the values to be used to correct the observed differential heads ∆h' in Table 10. For example. gives the frictional head loss ∆hf as L1 V12 ∆h f = 4 f1 D1 2g Inserting numerical values: ∆h f = 4 × 0. V12 2g In the first line of Table 10. × 104 = 194 The values of friction factor at these two Reynolds numbers may be found from Table 10.1. × 99 = 19. presented as Equation (9.6 × 10−3 .654 m/s so Re 2 = V2 D2 ν = 0.0061 × 8 V12 2g = 0196 . × 10 −6 100 .3 by interpolation to be f1 = 0. is given by L1 D1 = 8 Now Darcy's equation.453 × 0.45 = 0. say 19 mm The piezometric head loss ∆h across the mitre is then. therefore: ∆h f V12 = 0196 .2.Similarly for the larger bore pipe: D2 = 29.

Fig 10.4 Illustration of positions of piezometer tappings Since there is no change in velocity from upstream to downstream of the mitre. and the figure 135 is therefore entered in the first line of Table 10.2. using the relationship shown in Fig 10.4: L1 D1 = 8 + πR1 2 D1 118 . this is also the loss of total head ∆H. Similar calculations are made for the elbow and bend.

2 = − 41.3). Equation (10.0 = 31.257 × 99.31.257 V12 for the elbow 2g In the first line of results.0 = 31.00654 × 4 2 2g 2g In the first line of Table 10.0 = 25. 2g = 0. for the enlargement: ∆h f = 0.0.098 × 99.This leads to ∆h f and ∆h f = 0. 2g 2g = 4 × 0.4 = 87. say 88 mm for the elbow ∆hf = 0.6. from Equation (10.0 mm so ∆H = 62 .2 mm To derive the change ∆H in total head.0 + 0105 . ∆hf = 0. for example.25. × 33.313 V12 for the bend. say 31 mm for the bend In the case of the enlargement.313 × 99.4 mm so ∆H = 113 .2 mm The change in piezometric head is then.1) is used: 119 .00611 × 4 V12 V2 + 4 × 0.098 2 V12 V2 + 0105 . then.2.0 = 13. the sum of friction losses in the pipes of diameter D1 upstream and of diameter D2 downstream is ∆h f = 4 f1 2 L1 V12 L V2 + 4 f2 2 D 2 2g D1 2g Noting dimensions from Fig 10. ∆h = ∆h ′ − ∆h f = − 28 − 13.4 and inserting numerical values: ∆h f or ∆h f = 0.

Similarly for the contraction.2 = − 95.2.8 . where ∆hf has the same value as for the enlargement. say 25 mm = − 412 This is the value entered in the first line of Table 10.0 = 24.8 mm ∆H = ∆h + 2 V2 V2 − 1 2g 2g . Derivation of Loss Coefficients K 120 .8 .0 − 33.0 − 99. + 99.∆H = ∆h + V12 V2 + 2 2g 2g .0 = 29.2. say 30 mm = 958 which is also shown on the first line of Table 10. + 33. The computation is: ∆h = ∆h ′ − ∆h f = 109 − 13.

5 Total head loss ∆H in 90° bends of various radii 121 .Figure 10.

6 Total head loss ∆H at a sudden enlargement and at a sudden contraction Figs 10.5 and 10.6 show the total head losses ∆H plotted against velocity head for each of the fittings. the tube diameter is 122 . elbow and bend.Figure 10. In the cases of the mitre.

32 1.34 0.25 0. However. In this case.5 mm. For the contraction. In each case. elbow and bend show that the loss coefficient K falls substantially as the radius of the bend is increased. the relevant value is the velocity head in the pipe of smaller diameter.6) provides a theoretical value of K.76 Du/Dd = 1.07 Du/Dd = 0. The slope of the line gives the value of K for the fitting. from a value of about 1. The results are collected in Table 10.4 for the mitre bend.8) may be used to calculate Vc/Vd from the measurements: 123 . there will be a noticeable effect on the resulting value of K.86 0. which again is V12/2g. if there is significant error in the computed effect of pipe friction. Fitting K 90° mitre 90° elbow 90° bend Enlargement Contraction R/D = 0.56 R/D = 3.32 0. there is no theoretical value of K. For the enlargement and for the contraction. to a value around 0.22. The measured value is significantly higher.3 when R/D = 3.25. For the enlargement.98 6. the value of ∆hf is in this case about one half that of ∆H. Many previous experiments have indicated that. Therefore. Moreover. the results lie reasonably well on a straight line through the origin. K would be expected to reduce.4 Experimental values of loss coefficient K Discussion of Results The results for the mitre.50 R/D = 1. this theoretical value is A 1− u Ad 2 K = = 1− 3. at 0. Perhaps the piezometer tapping downstream of the enlargement is placed too close to allow the full recovery of piezometric pressure to take place . with a value of Re around 2 × 104.4. Equation (10. so the appropriate velocity head is obviously V12/2g.28 Table 10. The values obtained in this experiment are in good agreement with these expectations.88 2 = 018 . Equation (10.

0 discussed earlier. 4. and it lies between the extreme values of about 0. What suggestions do you have for improving the apparatus? No correction has been made to the differential piezometer readings for the difference in heights between the piezometer tappings. what percentage error in the value of K for any of the bends would result from an error of 0. Would a better result be obtained by estimating the friction effect over the range of flow rates? 3. from standard pipe friction data. Can you explain why it would be wrong to make any such correction? What are the sources of error? In particular. 124 . 2.65 This is a plausible value for the contraction coefficient of the jet at entry the contracted pipe. The effect of wall friction over the length of pipe between the piezometers has been estimated.1 mm in the measured diameter D1? (1. Questions for Further Discussion 1.6 and 1. at a single flow rate.Vc viz −1 Vd 2 = 0.28 from which Ac Ad = Vd Vc = 0.8% approximately).