Pipe Friction

Friction depends on whether the flow is orderly or chaotic, and the Reynolds number tells you which.

In 1883, Osborne Reynolds ran water from a tank through a long glass tube and fed a thin thread of dye into its mouth. With the valve barely open, the thread ran down the tube as straight as a wire. He opened the valve further, and the thread began to waver, then burst into a cloud that filled the pipe.

Same tube, same water, only faster. The water went from sliding in orderly layers to tumbling in eddies, and the two rub against the pipe differently. That difference sets the friction loss hfh_f, which the energy equation has so far been handed.

Orderly or chaotic

Why should speed matter? Picture a small parcel of water knocked a little sideways by a bump in the wall. Its momentum carries it on into its neighbors, and they shove theirs, so a wobble tends to grow into an eddy. Viscosity fights back. Each layer drags on the layers beside it, and that drag smooths a wobble out before it can grow. The flow stays orderly as long as viscosity wins.

So compare the two. Inertia, the stubbornness of moving mass, scales as ρv2\rho v^2. Viscous stress is μ\mu times a velocity gradient, and across a pipe that gradient is about v/Dv/D, so it scales as μv/D\mu v/D. Their ratio is the Reynolds number:

Re=vDρμ=vDν\mathrm{Re} = \frac{vD\rho}{\mu} = \frac{vD}{\nu}

That's how the Handbook prints it under Reynolds Number, with the kinematic viscosity ν=μ/ρ\nu = \mu/\rho. Its Similitude section writes the same ratio as the inertia force over the viscous force, FI/FVF_I/F_V. A large Re means inertia wins and wobbles grow, and a small one means viscosity damps them. Re has no units, so it means the same thing in any fluid and any pipe. For flow in a pipe, the Handbook draws the lines here:

  • laminar for Re<2,100\mathrm{Re} < 2{,}100
  • transitional for 2,100<Re<10,0002{,}100 < \mathrm{Re} < 10{,}000
  • fully turbulent for Re>10,000\mathrm{Re} > 10{,}000
laminar: Re < 2,100 flow transitional: 2,100 < Re < 10,000 turbulent: Re > 10,000
Reynolds' dye thread at three speeds
Checkpoint

Water flows through a pipe in laminar flow. Which statements are true?

A.
Re weighs inertial forces against viscous forces, so a high Re means viscosity has lost control.
B.
Sending the same flow rate through a larger pipe makes turbulent flow more likely.
C.
Raising the velocity makes turbulent flow more likely.
D.
Switching to an oil with a higher viscosity, at the same velocity, makes turbulent flow more likely.
E.
Warming the water makes turbulent flow more likely.
F.
A larger pipe carrying water at the same velocity makes turbulent flow more likely.

Computing the Reynolds number

For water, read ν\nu from the Handbook's Properties of Water tables at the stated temperature, as in Fluid Properties. Here are a few rows of each.

Kinematic viscosity of water
Temp. (°F)ν\nu (× 10⁻⁵ ft²/sec)Temp. (°C)ν\nu (× 10⁻⁶ m²/s)
401.66451.518
501.410101.306
601.217151.139
701.059201.003

Source: FE Reference Handbook, Properties of Water (English Units) and Properties of Water (SI Metric Units), selected rows

The SI table writes out its zeros, so 1.003 here is its 0.000001003. Re comes out unitless only if its inputs agree: ft/sec, ft, and ft²/sec, or m/s, m, and m²/s. A diameter left in inches makes Re twelve times too big. If a problem gives μ\mu instead, use the vDρ/μvD\rho/\mu form, with ρ\rho in slug/ft³ or kg/m³.

Most problems give a flow rate rather than a velocity. Put v=Q/(πD2/4)v = Q/(\pi D^2/4) into Re, and you get a form worth having:

Re=4QπDν\mathrm{Re} = \frac{4Q}{\pi D \nu}

A duct that isn't round needs a stand-in for DD. The Handbook's section on Flow in Noncircular Conduits uses the hydraulic radius, the flow area over the wetted perimeter, and the hydraulic diameter, four times as big: DH=4A/PD_H = 4A/P. A full rectangular duct aa by bb wets all four sides:

DH=4ab2(a+b)D_H = \frac{4ab}{2(a + b)}

Use DHD_H in place of DD in Re. The hydraulic radius, a quarter of it, gives an Re four times too small.

Worked Example The Reynolds number in a pipe
1 of 2

Water at 60°F flows at 0.50 cfs through a 6 in. pipe. Find the Reynolds number and classify the flow.

The largest laminar flow

Run Re backward, and it tells you how far a flow can go before it stops being laminar. Set Re=2,100\mathrm{Re} = 2{,}100, then solve for the velocity in a given pipe, or the diameter at a given velocity:

v=2,100 νDD=2,100 νvv = \frac{2{,}100\,\nu}{D} \qquad\qquad D = \frac{2{,}100\,\nu}{v}

The largest flow rate is that velocity times the area. For water, the answer is humbling. At 60°F in a 1 in. tube, v=2,100(1.217×10−5)/(1/12)=0.31v = 2{,}100(1.217 \times 10^{-5})/(1/12) = 0.31 ft/sec, about a foot every three seconds. Water in real pipes is nearly always turbulent, and laminar flow belongs to thick oils and thin tubes.

Checkpoint
1 of 2

An oil with SG=0.90SG = 0.90 and μ=0.12 Pa⋅s\mu = 0.12\ \text{Pa}\cdot\text{s} flows in a 100 mm pipe. What is the largest flow rate that keeps the flow laminar?

L/s

The Darcy-Weisbach equation

Now the loss itself. What ought to set it?

  • Every foot of pipe has wall that drags on the water, so the loss should grow with the length LL.
  • Faster water loses more. The eddies carry energy in proportion to the velocity head, as at a fitting, so the loss should follow v2/2gv^2/2g.
  • A bigger pipe has less wall per pound of water. The wall grows as DD and the water's area as D2D^2, so the loss should fall as 1/D1/D.

Put a pure number in front to cover everything else, and that's the Darcy-Weisbach equation, from the Handbook's Head Loss Due to Flow:

hf=f LD v22gh_f = f\,\frac{L}{D}\,\frac{v^2}{2g}

The friction factor ff has no units. Read L/DL/D as the pipe's length counted in diameters: each diameter of pipe costs ff velocity heads. With LL and DD in the same unit, hfh_f comes out in the velocity head's unit.

This hfh_f is the friction loss along the pipe, which is what the Handbook means by it. The Energy Equation let hfh_f stand for a line's whole loss, and fittings add their C v2/2gC\,v^2/2g to the friction found here. In a level pipe of one diameter, all of it comes out of the pressure, as the Handbook's Energy Equation section says: P1−P2=γhfP_1 - P_2 = \gamma h_f.

Problems often give QQ instead of vv. Substitute v=4Q/(πD2)v = 4Q/(\pi D^2), and you get a form the Handbook doesn't print but that's worth seeing:

hf=8fLQ2π2gD5h_f = \frac{8fLQ^2}{\pi^2 g D^5}

Look at the powers. The loss goes as the flow squared, through the velocity head. It goes as one over the diameter to the fifth: four powers because the velocity goes as 1/D21/D^2 and then gets squared, and one more from L/DL/D. A small change in diameter makes a big change in loss.

Worked Example The Energy Equation's missing 8 psi
1 of 2

The Energy Equation opened with a level mile of 8 in. main losing 8 psi. Say it carries 430 gpm with f=0.020f = 0.020. Find the head loss and the pressure drop.

Laminar friction

In laminar flow, the friction factor has a formula. The Handbook's Moody diagram, coming up next, labels its laminar line with it:

f=64Ref = \frac{64}{\mathrm{Re}}

Put that into Darcy-Weisbach, and something changes. Re carries one power of vv, which cancels one from the velocity head, so the loss goes as vv, not v2v^2. As a pressure drop:

P1−P2=32μLvD2P_1 - P_2 = \frac{32\mu L v}{D^2}

That's Newton's law of viscosity scaled up to a pipe. In orderly layers, drag follows the velocity gradient, so twice the speed means twice the drag. Roughness doesn't appear at all: whatever a bump on the wall stirs up, viscosity damps out. The Handbook prints the same result solved for the flow, as the Hagen-Poiseuille equation under Pressure Drop for Laminar Flow: Q=πD4ΔPf/(128μL)Q = \pi D^4 \Delta P_f/(128\mu L). Check that Re is below 2,100 before using either one.

Checkpoint

Oil with SG=0.90SG = 0.90 and μ=0.0020 lbf-sec/ft2\mu = 0.0020\ \text{lbf-sec/ft}^2 flows at 0.050 cfs through 500 ft of level 2 in. pipe. What is the pressure drop?

psi

Turbulent friction and the Moody diagram

Turbulent flow has no tidy formula for ff. The Handbook writes f=f(Re,ε/D)f = f(\mathrm{Re}, \varepsilon/D): it depends on Re and on the relative roughness ε/D\varepsilon/D, the height ε\varepsilon of the wall's bumps compared with the pipe's diameter. A bump that matters in a 2 in. pipe is lost in a 10 ft tunnel. The Handbook lists ε\varepsilon by material under its Moody diagram.

Absolute roughness of pipe materials
Materialε\varepsilon (ft)ε\varepsilon (mm)
Glass, drawn brass, coppersmoothsmooth
Commercial steel, wrought iron0.0001–0.00030.03–0.09
Galvanized iron0.0002–0.00080.06–0.24
Cast iron0.0006–0.0030.18–0.91
Concrete0.001–0.010.30–3.0
Riveted steel0.003–0.030.91–9.1

Adapted from the FE Reference Handbook, Moody, Darcy, or Stanton Friction Factor Diagram (roughness table), selected rows

Real pipes vary, so each material gets a range. Use a stated ε\varepsilon if there is one, or else the middle of the range, and put ε\varepsilon and DD in the same units before you divide.

Why should roughness matter only some of the time? Even in turbulent flow, the water touching the wall is held still, and a thin layer beside it keeps sliding in orderly layers. This viscous sublayer thins as Re rises. While it buries the bumps, the turbulence never feels them, and the pipe acts smooth. As Re climbs, the bumps poke through and shed eddies of their own. Once they all do, the bumps do the churning, viscosity stops mattering, and ff depends on ε/D\varepsilon/D alone.

ε turbulent core viscous sublayer lower Re: bumps buried,pipe acts smooth ε viscous sublayer higher Re: bumps pokethrough, fully rough
The viscous sublayer at a pipe wall, at lower and higher Re

The Handbook's Moody, Darcy, or Stanton Friction Factor Diagram puts all of this on one log-log chart: ff up the left side, Re along the bottom, and one curve for each ε/D\varepsilon/D, labeled at the right.

6 8 103 2 4 6 8 104 2 4 6 8 105 2 4 6 8 106 2 4 6 8 107 2 4 6 8 108 0.008 0.009 0.01 0.012 0.014 0.016 0.018 0.02 0.025 0.03 0.035 0.04 0.05 0.06 0.07 0.08 0.09 0.1 Reynolds number, Re = vD/ν friction factor, f relative roughness, ε/D 0.05 0.02 0.01 0.004 0.002 0.001 0.0004 0.0002 0.0001 0.00001 laminar flow,f = 64/Re smooth pipes laminar flow transition zone complete turbulence, rough pipes critical zone
The Moody diagram
About half of the Handbook's roughness curves, recreated. The dashed path is the reading in the Worked Example below.

Adapted from the FE Reference Handbook, Moody, Darcy, or Stanton Friction Factor Diagram

Read the chart from left to right.

  • Laminar flow: one line, f=64/Ref = 64/\mathrm{Re}, for every pipe.
  • Critical zone: about 2,100 to 4,000, where the chart gives no ff.
  • Transition zone: the curves slope down, so ff depends on both Re and ε/D\varepsilon/D. The name means the shift from smooth-pipe to fully rough behavior. To classify the flow itself, still use 2,100 and 10,000.
  • Complete turbulence: each curve levels off, and ff depends only on ε/D\varepsilon/D.

To read ff, go up from Re on the bottom axis to the curve for your ε/D\varepsilon/D, estimating between curves if you need to, then go straight left.

Worked Example A friction factor from the chart
1 of 2

Water at 60°F flows at 0.60 cfs through 1,000 ft of 6 in. galvanized iron pipe. Using the middle of the Handbook's roughness range, find the head loss.

The chart also says when the powers from the Darcy-Weisbach section hold exactly: in complete turbulence, where ff doesn't change with Re. Elsewhere it drifts a little, so scaling questions say to hold ff constant.

Checkpoint
1 of 2

A long pipeline loses 10 ft of head. With ff held constant, what is the loss if the flow doubles? And if instead the original flow goes through a pipe of half the diameter, with the same length?

A.
20 ft if the flow doubles; 40 ft in the smaller pipe
B.
40 ft if the flow doubles; 160 ft in the smaller pipe
C.
40 ft if the flow doubles; 320 ft in the smaller pipe
D.
40 ft if the flow doubles; 80 ft in the smaller pipe

Running Darcy-Weisbach backward

The exam also runs the equation backward. It ties together five quantities, so given four you can find the fifth.

  • For the velocity at an allowable loss, v=2g hfD/(fL)v = \sqrt{2g\,h_f D/(fL)}, and then Q=vAQ = vA.
  • For the diameter that keeps a flow within an allowable loss, solve the flow-rate form, D5=8fLQ2/(π2g hf)D^5 = 8fLQ^2/(\pi^2 g\,h_f), and take the fifth root.
  • For the friction factor from a measured pressure drop in a level pipe, turn the drop into a head, hf=ΔP/γh_f = \Delta P/\gamma, then f=hf/[(L/D)(v2/2g)]f = h_f \big/ \left[(L/D)(v^2/2g)\right].

An allowable pressure drop becomes a head the same way. In real design, ff depends on the very vv and DD you're solving for, so you'd guess ff, solve, and check it on the chart. Here, ff is given.

Worked Example Sizing a pipe for an allowable loss
1 of 3

A 1,500 ft pipeline must carry 2.0 cfs of water with no more than 15 ft of head loss. Taking f=0.020f = 0.020, what is the smallest diameter that will do?

Recap

  • Re=vD/ν=vDρ/μ\mathrm{Re} = vD/\nu = vD\rho/\mu weighs inertia against viscosity. Pipe flow is laminar below 2,100 and fully turbulent above 10,000. A full duct uses DH=4A/PD_H = 4A/P.
  • The largest laminar velocity, flow, or diameter comes from setting Re = 2,100. Water in real pipes is nearly always turbulent.
  • Darcy-Weisbach: hf=f(L/D)(v2/2g)h_f = f(L/D)(v^2/2g), and in a level pipe P1−P2=γhfP_1 - P_2 = \gamma h_f. With ff fixed, the loss goes as Q2/D5Q^2/D^5.
  • Laminar flow has f=64/Ref = 64/\mathrm{Re}, with no roughness. Turbulent flow takes ff from the Moody diagram with Re and ε/D\varepsilon/D, and in complete turbulence ε/D\varepsilon/D alone sets it.
  • Run the equation backward for the velocity, the diameter, or a measured friction factor, turning any pressure into a head first.