Professional Documents
Culture Documents
By considering the balance of forces between gravity or pressure pushing a flow and
shear stress due to friction resisting the flow a broadly useful equation is obtained. The
mean boundary shear stress for the whole pipe cross-section may be related to the head
loss due to friction by the formula,
0 gRS f 1
A different expression for shear stress is given by Newton's law of viscosity
du
2
dy
an expression that relates the variation of boundary shear stress to the velocity.
Integrating this around the pipe perimeter and equating the two expressions [Try it] yields
an expression for the parabolic distribution of axial velocity, u, across the section, and
ultimately,
gS f d 2
u 12 u max 3
32
So for a parabolic distribution the mean velocity is half of the maximum. It is well worth
noting that the mean velocity in laminar flow is proportional to the head loss.
The above results are fine for Re < 2000 or so but what about Re > 2000? For higher
Reynolds numbers we must resort to experiment. The most important pioneering
experiments were performed by one of Prandtl's pupils - Nikuradse. By painstakingly
LEC 0708 – updated 2017-18
1
ENG-7004B Pipes 01
gluing artificial sand roughness to the inside of a series of pipes he was rewarded with a
consistent set of graphs which suggested that for intermediate Reynolds numbers the
roughness elements were submerged beneath the laminar sub-layer and the friction
factor followed the Blasius' formula for smooth pipes [Figure 1]. For increasing Reynolds
numbers as the roughness elements began to protrude through the boundary layer the
curves followed distinctly different paths depending on the relative roughness - ks/d,
where ks was the sand grain size.
Commercial pipes did not show the same characteristic dip in the curves but always
seemed to exhibit a gradually declining friction factor with increasing Reynolds number.
However it was not at all clear how to relate the artificial sand grain roughness ks to the
non-uniform roughness elements of a commercial pipe. Later a rough equivalence for
non-uniform roughness was established and led to the now commonly used Moody
diagram [Figure 3].
Relying on a diagram suits some applications but in other cases, such as computer
models of distribution networks, it would be helpful to be able to represent the curves
mathematically so that the friction factor can be calculated.
Many accurate measurements of the boundary layer on flat plates suggest that the
velocity distribution is logarithmic.
Precise measurements of hf and Q for fixed d can be used to obtain estimates of 1/f.
Some early results are referred to as the Karman-Prandtl formulae:
1 Re f
Smooth 2 log 10
f 2.51
1 3. 7d
Rough 2 log 10
f ks
More recently Colebrook and White have demonstrated that these can be usefully
combined to produce the 'Colebrook-White Transition Formula' for commercial pipes.
1 k 2.51
2 log10 s
f 3.7d Re f
For high Reynolds numbers this shows that f depends only on the relative roughness
ks/d. For lower Reynolds numbers it provides the observed gradual transition to the
smooth pipe result.
On modern calculators it is relatively easy to calculate f since the C-W formula converges
using simple iteration. Nevertheless for engineering calculations useful charts have been
produced by Hydraulics Research at Wallingford [Figure 2].
Also the EPANET software makes use of a very good explicit approximation known as
the Swamee and Jain equation (although it quotes the wrong logarithm),
0.25
𝑓=
𝑘𝑠 5.74 2
[log10 ( + )]
3.7𝑑 𝑅𝑒 0.9
which can be rearranged for comparison with the C-W formula as,
1 𝑘𝑠 5.74
= ±2 log10 ( + 0.9 )
√𝑓 3.7𝑑 𝑅𝑒
1 k 2.51
2.0 log10 s
f 3.7 d Re f
or since u f 2 gS f d
u ks 1.775
2.0 log10 1.5
2 gS f d 3.7 d gS f d
The following table suggests ways to tackle the solution of the above equations. You
should attempt the various combinations until you are satisfied that you know how to
obtain a solution.
Note that because pipe diameters come in a limited range of sizes it is frequently better
to guess specific values of d if possible. Also it seems that the H.R. tables are the most
straightforward approach for the design of individual pipes. The most common mistake
in using the H.R. tables is to forget that the axis uses 100Sf.
Example:
A pipeline 10km long and 300mm diameter with an estimated equivalent roughness ks =
0.03mm conveys water from a reservoir (water level 850m A.O.D.) to a free outlet (700m
A.O.D.). Assume that all the head loss takes place in the pipe and hence find the
discharge Q.
Figure 2. Example Hydraulics Research chart based on the C-W equation for fixed
ks/d.
However in situations where pipe runs are short and many bends and valves are
encountered (such as in a water treatment works) the pipe friction loss may be totally
swamped by the losses at the fittings. In many texts these are referred to as ‘minor losses’.
Many an engineer has regretted assuming that this name meant that they could always be
neglected.
Entry loss
hf H
ū ū2/2g
The local loss, hL is usually expressed in terms of a k-factor for the fitting,
Example.
In a treatment works two tanks are connected by a 20metre length of 200mm diameter pipe.
The difference in water level is 5m. It is hoped that at least 0.2 m3/s will flow between the
tanks. The pipe has an abrupt entrance and exit as well as four 90 bends. Calculate the
discharge with and without allowing for local losses.
Figure 6. An abrupt expansion. Note that the HGL rises because the velocity had
reduces in the larger diameter pipe.
ū1
Figure 7. At an abrupt exit to a pipe the HGL lines up with the downstream water
surface because the whole of the velocity head is lost.
Figure 8. Two large tanks connected by small and large diameter pipes. Show HGL,
EL and losses from entrance, expansion, exit and pipe friction.
Figure 9. A large tank connected to a pipe which culminates in a nozzle. Show HGL,
EL and losses from entrance and pipe friction.
Figure 10. A large tank and a siphon pipe. which culminates in a nozzle. Show HGL,
EL and losses from entrance and pipe friction. Note that the streamline with which
these losses are associated is shown.