The push a pipe needs
In Fluid statics and Bernoulli an ideal fluid traded pressure, speed and height along a streamline and lost nothing. Apply that to a level pipe of constant bore. Continuity says the mean speed is the same at both ends; the height is the same; so Bernoulli says the pressure is the same too, and water should glide along a kilometre of pipe for ever without being pushed.
Every water utility knows otherwise. The mains under a street in Nairobi need pumping stations; a district-heating loop in Copenhagen has circulating pumps working day and night; the oil in a car engine needs an oil pump to reach the bearings; your heart does real work to push blood through your arteries. Pumps and fans are among the largest users of electricity in industry and buildings, and much of that electricity goes to overcoming one thing: viscosity, the internal friction of a real fluid.
This lesson puts viscosity back: why fluid sticks to walls, why one dimensionless number decides between smooth and chaotic flow, how to compute the pressure a pipe and its fittings need, how a pump and a pipe system agree on a flow rate, and, in outline, the boundary layers and drag on everything that moves through air or water. Heat transfer and computational fluid dynamics build on all of it.
Layers that drag on each other
Stir a jar of honey and a glass of water with the same spoon. The honey resists: it takes force to make one layer slide over the next. That resistance to shearing is viscosity. It is not the same as density: olive oil floats on water, yet it is about eighty times more viscous.
The simplest case is a thin film between two large plates, the lower one fixed and the upper one dragged along at speed . The fluid touching each plate moves with that plate, the speed in between varies linearly from zero to , and the force needed to keep the top plate moving is proportional to and to the plate's area, and inversely proportional to the gap.
This no-slip condition is the heart of viscous flow. Fluid at a wall is at rest relative to it and fluid far away is moving, so there is shearing in between, and shearing a viscous fluid takes work. That work becomes internal energy, warming the fluid very slightly; engineers call it a loss, and the pump must make it good.
In 1883 Osborne Reynolds, in Manchester, ran water from a tank through a glass tube with a thin streak of dye at the entrance. At low speeds the streak stayed a clean thread; above a certain speed it wavered and burst into eddies that mixed the dye through the tube. He had made visible the two regimes of pipe flow: laminar, moving in smooth layers, and turbulent, with unsteady eddies of many sizes tumbling across the flow.
Analogy, not a formal construction
Laminar flow is like disciplined motorway traffic: every car keeps its lane, and the lanes slide past one another. Turbulent flow is like a busy market crowd, where people weave across the aisle carrying their momentum into slower regions, so speeds even out, but those next to the stalls are jostled far harder. Turbulence in a pipe likewise flattens the velocity profile and multiplies the friction at the wall.
Viscosity and the Reynolds number
Definition— dynamic viscosity (Newton's law of viscosity)
For a Newtonian fluid the shear stress on layers sliding over one another is proportional to the rate of shear, the velocity gradient across the layers:
The constant is the dynamic viscosity, measured in pascal seconds (Pa·s). Water, air, oils and most gases and simple liquids are Newtonian.
Definition— kinematic viscosity
The ratio of viscosity to density, in m²/s, is the kinematic viscosity. It measures how fast momentum diffuses through the fluid, and it is the combination that appears in most flow formulas.
Representative values near 20 °C and atmospheric pressure, rounded (take design values from property databases such as the NIST Chemistry WebBook):
| Fluid | (kg/m³) | (Pa·s) | (m²/s) |
|---|---|---|---|
| Air | 1.2 | ||
| Water | 998 | ||
| Olive oil | 910 | about 0.08 | about |
| Glycerol | 1260 | about 1.4 | about |
| Honey | about 1400 | 2 to 10 | or more |
Viscosity depends strongly on temperature: water's halves between 20 °C and 50 °C, and engine oil can be a hundred times thicker on a frozen morning than in a hot engine. Gases go the other way: their viscosity rises with temperature, because it comes from molecules carrying momentum between layers, and hotter molecules travel faster.
Note
Many everyday fluids are non-Newtonian: their apparent viscosity depends on how fast they are sheared. Paint, ketchup and blood thin as they are sheared harder (shear-thinning); a thick suspension of cornflour in water stiffens suddenly when struck (shear-thickening); toothpaste does not flow at all until the stress passes a threshold (a yield-stress fluid). Their behaviour is the subject of rheology. Everything below assumes a Newtonian fluid.
Definition— Reynolds number
For a flow of speed past or through an object of size ,
It compares the momentum the fluid carries (inertia) with the viscous stresses that smooth it out. For flow in a round pipe, with the mean speed and the internal diameter, the flow is normally laminar below and turbulent above about 4000, with an unsteady, intermittent transitional range in between.
The Reynolds number spans an astonishing range: about for a swimming bacterium, which stops dead the instant it stops pushing; for water in a garden hose; for air past a car; for a large ship. Two flows with the same geometry and Reynolds number behave alike whatever their size, which is why models can be tested in towing tanks and wind tunnels.
The number 2300 is a convention, not a law. Laminar pipe flow becomes turbulent only when disturbances are large enough: with very smooth inlets and quiet tanks it has been kept laminar beyond , while below roughly 2000 every turbulent patch eventually decays. In a real plant, with pumps, bends and valves upstream, expect transition between about 2000 and 4000, and in that range design for whichever regime gives the larger pressure drop.
Laminar pipe flow, exactly
Laminar flow in a long straight pipe is one of the few viscous flows that can be solved exactly with a pencil, and the answer is worth knowing by heart.
Theorem— Hagen–Poiseuille flow
In steady, fully developed laminar flow of a Newtonian fluid through a round pipe of radius , with a pressure drop over a length , the velocity at radius is the paraboloid
the volume flow rate is
and the mean velocity is exactly half the centreline velocity.
Proof
Take as a free body a cylinder of fluid of radius and length , coaxial with the pipe. Because the flow is fully developed, the velocity profile is the same at both ends, so momentum flows in and out at the same rate and the net force on the cylinder is zero. The pressure difference pushes on its ends with force ; the slower fluid outside drags back on its curved surface with shear stress over area . Equilibrium gives : the shear stress rises linearly from zero on the axis to its maximum at the wall. The velocity falls as increases, so Newton's law of viscosity reads , and
where the constant of integration was fixed by no slip, . The flow rate is the integral of over the cross-section, . Dividing by the area gives , which is half of .
∎
The fourth power of the radius is the striking part. Halve the diameter of a laminar pipe, at the same pressure difference, and the flow falls by a factor of 16. That is why a small narrowing of a blood vessel matters so much to the heart, and why lubricant films and capillaries are so sensitive to their gap.
Rearranged in terms of the mean velocity, Hagen–Poiseuille reads
which is the Darcy–Weisbach equation below with friction factor . In laminar flow the pressure drop is proportional to the flow rate; that is about to change.
Head loss: Darcy–Weisbach, Colebrook and fittings
Between two points 1 and 2 of a pipe system, the energy equation for steady incompressible flow is Bernoulli's equation with two new terms, written as heights of fluid ("heads", in metres):
is the head a pump adds; is the head loss, the mechanical energy per unit weight of fluid turned into internal energy on the way. (It is a per-unit-weight form of the steady-flow energy equation of the first law.) A head of 1 m of water is a pressure of kPa.
Definition— Darcy–Weisbach equation
The head loss to friction in a straight pipe of length and internal diameter , with mean velocity , is
where the dimensionless Darcy friction factor depends on the Reynolds number and the relative roughness of the pipe wall.
In laminar flow, and roughness does not matter. In turbulent flow there is no exact theory, but there is an excellent fit to experiment, published by Cyril Colebrook in 1939 from his own and Johann Nikuradse's measurements:
In a smooth pipe () only matters, and falls slowly as it grows: the viscous layer at the wall buries any small roughness. At very high the second term vanishes and the pipe is fully rough: the bumps poke through the thin viscous layer and depends only on . The equation is implicit, with on both sides and no algebraic solution. Engineers of the 1940s read from Lewis Moody's chart; today we solve it numerically in microseconds, or use explicit approximations such as Haaland's (1983), within a couple of per cent:
Typical roughness heights, for orientation (pipe makers publish values for their products, and roughness grows as pipes age and corrode):
| Pipe | (mm) |
|---|---|
| Drawn copper or plastic (PVC, PE) | about 0.0015 to 0.007 |
| Commercial steel | about 0.045 |
| Cast iron | about 0.25 |
| Concrete | about 0.3 to 3 |
Minor losses
Bends, valves, tees, entrances, exits and changes of section disturb the flow and cause extra losses. Each is described by a loss coefficient , so that , and the system's total loss is
Indicative values, which vary with size, design and manufacturer: a sharp-edged entrance from a tank, about 0.5; a well-rounded one, below 0.05; any exit into a large tank, 1.0 (all the kinetic energy is dissipated); a 90° elbow, roughly 0.2 to 1; a fully open gate or ball valve, below 0.2; a fully open globe valve, around 5 to 10. In a plant room full of fittings, "minor" losses can be most of the total. One can be derived exactly: a momentum balance on a sudden enlargement from area to gives the Borda–Carnot loss on the upstream velocity, which becomes the exit loss as .
Computing it: Colebrook by iteration
Write . Then Colebrook's equation is , which already has the form , ready for fixed-point iteration: guess , evaluate the right-hand side, repeat. It converges because is small (the logarithm flattens everything), well under 0.1 for turbulent flow, so each iteration gains at least one correct digit.
The program below solves the worked example (water at 20 °C, 3.0 L/s through 100 m of 50 mm commercial-steel pipe), compares Haaland's approximation, and prints a small Moody chart as a table.
import numpy as np
def colebrook(Re, rr, tol=1e-12):
"""Darcy friction factor from the Colebrook equation (turbulent flow).
Re: Reynolds number; rr: relative roughness eps/D."""
x = 8.0 # x = 1/sqrt(f); f = 0.016 is a fair first guess
for it in range(100):
x_new = -2.0 * np.log10(rr / 3.7 + 2.51 * x / Re)
if abs(x_new - x) < tol:
break
x = x_new
return 1.0 / x_new**2, it + 1
def haaland(Re, rr):
return (-1.8 * np.log10((rr / 3.7) ** 1.11 + 6.9 / Re)) ** -2
# Worked example: water at 20 C, 3.0 L/s through 100 m of 50 mm commercial steel pipe
rho, nu = 998.0, 1.00e-6 # kg/m^3, m^2/s
Q, D, L, eps = 3.0e-3, 0.050, 100.0, 0.045e-3
V = Q / (np.pi * D**2 / 4)
Re = V * D / nu
f, n_it = colebrook(Re, eps / D)
hf = f * (L / D) * V**2 / (2 * 9.81)
print(f"V = {V:.3f} m/s, Re = {Re:.0f}, eps/D = {eps / D:.4f}")
print(f"Colebrook f = {f:.5f} after {n_it} iterations; Haaland f = {haaland(Re, eps / D):.5f}")
print(f"head loss {hf:.2f} m of water = pressure drop {rho * 9.81 * hf / 1e3:.1f} kPa")
# A text Moody chart: f against Re for several relative roughnesses
rrs = [0.0, 1e-4, 1e-3, 1e-2]
print("\n Re laminar" + "".join(f" eps/D={r:<6g}" for r in rrs))
for Re in [1e3, 2e3, 4e3, 1e4, 1e5, 1e6, 1e7, 1e8]:
lam = f"{64 / Re:8.4f}" if Re < 2300 else " -"
turb = "".join(f" {colebrook(Re, r)[0]:12.4f}" if Re >= 4000 else " -" for r in rrs)
print(f" {Re:8.0e} {lam}{turb}")The mean velocity is 1.53 m/s and the Reynolds number about 76 000, well into turbulence. The iteration settles to in a dozen steps, and Haaland's formula gives 0.02223, about 1% lower. The 100 m of pipe costs 5.35 m of head, a pressure drop of 52 kPa.
The table is the Moody chart in numbers. In a smooth pipe keeps falling, from 0.040 at to 0.006 at ; at it levels off at 0.038 from about on, where the pipe is fully rough and viscosity has stopped mattering. Notice also the jump at transition: at a laminar flow would have , while Colebrook's smooth-pipe value there is 0.044, so a flow that trips into turbulence can double its pressure drop with no change in flow rate.
Pumps, system curves and operating points
A centrifugal pump spins an impeller that flings fluid outwards; its casing turns much of the fluid's speed into pressure. At a fixed speed the head it delivers falls as the flow rises, and the maker publishes this pump curve , measured to a standard such as ISO 9906, with the efficiency and power at each flow.
The pipework has its own curve. To push through it, the pump must supply the static lift (plus any difference in surface pressures) and the losses, which grow roughly as :
Pump and pipes must agree: the flow settles at the operating point, where the curves cross. The shaft power the pump then draws is
where is the pump efficiency at that point.
Example— Lifting water to a header tank
A pump lifts water at 20 °C from a sump to a tank 12 m higher, through 150 m of 80 mm commercial-steel pipe with fittings whose loss coefficients add up to . The pump, at its rated speed, has (as a model fitted to its published curve) , with in metres and in L/s, and its best efficiency, 72%, at 14 L/s. Find the operating point and the power. Then the plant needs only 10 L/s: compare closing a valve (throttling) with slowing the pump.
Because depends on , the equation has no closed-form solution, so we solve it by bisection: the difference is positive at and negative at large , and halving the bracket repeatedly pins down the root. For the slower pump we use the affinity laws: at speed ratio , flow scales with , head with and power with at corresponding points, so the curve becomes .
import numpy as np
rho, nu, g = 998.0, 1.00e-6, 9.81
D, L, eps = 0.080, 150.0, 0.045e-3 # commercial steel pipe
dz, K_sum = 12.0, 8.0 # lift (m) and sum of minor-loss coefficients
A = np.pi * D**2 / 4
def friction(Re, rr):
if Re < 2300:
return 64.0 / Re
x = 8.0
for _ in range(50):
x = -2.0 * np.log10(rr / 3.7 + 2.51 * x / Re)
return 1.0 / x**2
def system_head(q): # q in L/s -> head the pipework needs (m)
if q <= 0:
return dz
V = q / 1000 / A
f = friction(V * D / nu, eps / D)
return dz + (f * L / D + K_sum) * V**2 / (2 * g)
def pump_head(q, s=1.0): # a model pump curve; s = speed / rated speed
return 32.0 * s**2 - 0.060 * q**2
def pump_eff(q, s=1.0): # best efficiency 72 % at 14 L/s (rated speed)
return 0.72 * (1 - ((q / s - 14.0) / 14.0) ** 2)
def bisect(fun, a, b, tol=1e-10):
fa = fun(a)
while b - a > tol:
m = 0.5 * (a + b)
if (fun(m) > 0) == (fa > 0):
a, fa = m, fun(m)
else:
b = m
return 0.5 * (a + b)
print(" Q (L/s) pump H (m) system H (m)")
for q in range(0, 24, 3):
print(f" {q:6.1f} {pump_head(q):9.2f} {system_head(q):11.2f}")
q = bisect(lambda q: pump_head(q) - system_head(q), 0.0, 23.0)
H, eta = pump_head(q), pump_eff(q)
P = rho * g * q / 1000 * H / eta
print(f"\noperating point: Q = {q:.2f} L/s, H = {H:.2f} m, efficiency {eta:.1%}, shaft power {P / 1e3:.2f} kW")
# The plant now needs only 10 L/s. Option 1: close a valve (throttle).
qt = 10.0
Ht = pump_head(qt)
Pt = rho * g * qt / 1000 * Ht / pump_eff(qt)
print(f"\nthrottled to 10 L/s: pump head {Ht:.2f} m, of which the valve burns {Ht - system_head(qt):.2f} m;"
f" power {Pt / 1e3:.2f} kW")
# Option 2: slow the pump down until its curve passes through the system curve at 10 L/s.
s = np.sqrt((system_head(qt) + 0.060 * qt**2) / 32.0)
Ps = rho * g * qt / 1000 * system_head(qt) / pump_eff(qt, s)
print(f"variable speed: {s:.3f} x rated speed, head {system_head(qt):.2f} m, power {Ps / 1e3:.2f} kW"
f" ({1 - Ps / Pt:.0%} less)")The curves cross at 11.6 L/s and 24.0 m of head, near the pump's best efficiency (70%), and the pump draws 3.9 kW. To get 10 L/s with a valve, the pump must climb back up its curve to 26.0 m, and the valve throws away 5.0 m of it: the power hardly changes (3.85 kW). Slowing the pump to 92% of its rated speed instead delivers the same 10 L/s at the 21.0 m the system actually needs, for 3.0 kW, 22% less. That saving is why so many pumps and fans now run from variable-speed drives, the power electronics of power electronics feeding the induction and permanent-magnet motors of transformers and machines.
A pump also needs enough pressure at its inlet. If the pressure in the impeller falls to the liquid's vapour pressure (2.3 kPa for water at 20 °C), vapour bubbles form and collapse violently further on: cavitation, which sounds like gravel, wrecks performance and pits the metal. Makers publish the net positive suction head required (NPSHr), and the designer makes sure the system provides more, with a margin. Nor should a pump run for long against a closed valve, where all its power heats a small volume of trapped liquid; installations have minimum-flow lines and controls for this reason.
Boundary layers and drag, in outline
Outside pipes, the same no-slip condition shapes the flow around every object. In 1904 Ludwig Prandtl explained why fluids with tiny viscosity still feel friction: at high Reynolds number the viscous effects are confined to a thin boundary layer next to the surface, outside which the flow is nearly ideal and Bernoulli works. On a flat plate the laminar boundary layer grows as the square root of the distance from the leading edge, with thickness about , where ; for air at 10 m/s, that is about 4 mm at 0.5 m along the plate. Near it usually becomes turbulent, thicker and with more friction.
If the outer flow slows down (as it does around the back of a cylinder, where the pressure rises again), the slow fluid deep in the boundary layer can be brought to a halt and pushed backwards. The boundary layer separates from the surface, leaving a broad, eddying wake at low pressure. The force a body feels in a stream is the drag, the sum of skin friction (shear stress on the surface) and pressure or form drag (the low pressure of the wake pulling backwards). It is written with a drag coefficient:
with normally the frontal area. Streamlined bodies avoid separation and have below 0.1; a sphere in the usual range has –0.5; most cars of 2026 are between about 0.2 and 0.35; an upright cyclist is around 0.9 to 1.1 on a frontal area of about 0.5 m². At about a smooth sphere's boundary layer turns turbulent before it separates; the more energetic turbulent layer clings on further round the back, the wake narrows, and falls by a factor of three or more: the drag crisis. The dimples on a golf ball trip the boundary layer early, so the ball enjoys this low drag at the speeds it actually flies.
At the other extreme, for , George Stokes found in 1851 that a sphere of diameter feels , that is, : the law of fine dust, mist and silt. Between the regimes engineers use fitted correlations. The program below finds how fast quartz sand settles in still water, balancing weight minus buoyancy against drag with the Schiller–Naumann correlation up to and beyond: the calculation behind settling tanks in water treatment.
import numpy as np
g = 9.81
rho, mu = 998.0, 1.00e-3 # water at 20 C
rho_p = 2650.0 # quartz sand
def drag_coefficient(Re):
"""Sphere drag: Schiller-Naumann up to Re = 1000, then roughly constant."""
return 24.0 / Re * (1 + 0.15 * Re**0.687) if Re < 1000 else 0.44
def settling_speed(d):
V = g * d**2 * (rho_p - rho) / (18 * mu) # start from Stokes' law
for _ in range(200):
Re = rho * V * d / mu
V_new = np.sqrt(4 * g * d * (rho_p - rho) / (3 * rho * drag_coefficient(Re)))
if abs(V_new - V) < 1e-12:
break
V = 0.5 * (V + V_new) # damped fixed-point iteration
return V_new, rho * V_new * d / mu
sig3 = lambda v: f"{v:#.3g}".rstrip(".") # three significant figures
print(" d (mm) Re V (mm/s) Stokes V (mm/s) time to sink 1 m")
for d_mm in [0.01, 0.05, 0.1, 0.2, 0.5, 1.0, 2.0, 5.0]:
d = d_mm / 1000
V, Re = settling_speed(d)
Vs = g * d**2 * (rho_p - rho) / (18 * mu)
t = 1.0 / V
when = f"{t / 3600:6.1f} h" if t > 3600 else (f"{t / 60:6.1f} min" if t > 60 else f"{t:6.1f} s")
print(f" {d_mm:5.2f} {Re:8.3g} {sig3(V * 1000):>9} {sig3(Vs * 1000):>14} {when}")Fine silt of 0.01 mm sinks at less than 0.1 mm/s and takes about three hours to settle through a metre of water, exactly as Stokes' law predicts. By 0.1 mm the Reynolds number is near 1 and Stokes is already 13% too fast; for a 1 mm grain it is six times too fast, and for 5 mm gravel, at , it would be wrong by a factor of 45. One formula, applied outside its range, would have ruined a settling-tank design.
Pipes and pumps in professional practice
Piping is designed by codes as much as by equations. Pressure piping in many countries follows the ASME B31 codes (B31.1 for power piping, B31.3 for process piping) or, in Europe, EN 13480 under the Pressure Equipment Directive, which set wall thicknesses, materials, supports, inspection and pressure testing. First sizing uses rules of thumb (pumped water lines are commonly sized for about 1–3 m/s, balancing pipe cost against energy, noise and erosion), checked with the calculations of this lesson. Real systems are networks: a city's water supply has thousands of pipes, loops, tanks, pumps and valves, analysed by solvers that apply mass conservation at every junction and the energy equation round every loop, solved by Newton's method. The best-known open-source tool is the US Environmental Protection Agency's EPANET; commercial packages build on similar methods.
Transients matter as much as steady flow. Close a valve suddenly and the moving column of liquid must stop: a pressure wave runs back along the pipe at the speed of sound in the pipe-and-liquid system (around 1000–1400 m/s in a water main), raising the pressure by roughly , Joukowsky's estimate. Stopping water flowing at 2 m/s can add 20 bar or more. This water hammer bursts pipes, which is why valves on large lines close slowly, pumps are stopped gradually, and surge tanks and relief valves protect long pipelines. Any system containing fluid under pressure is designed, tested and inspected under pressure-equipment rules, with relief devices sized for the worst case, and only trained people work on it.
Pump selection follows the method above: plot the system curve, choose a pump whose best-efficiency point sits near the duty, check NPSH, and consider how the duty varies over a year; the Hydraulic Institute and pump makers publish guidance and selection software. Pumps that move water between reservoirs also store energy for electricity grids (see renewables and storage in Lumenary EE and wind and water turbines later in this course). And when the geometry is too complicated for loss coefficients (a new valve, a pump inlet, an artery with a stent) engineers turn to computational fluid dynamics, and then test the result in a flow rig.
A short history
The German engineer Gotthilf Hagen measured water flow in brass tubes in 1839, and the French physician Jean Léonard Marie Poiseuille, studying how blood moves in capillaries, found the same fourth-power law independently in experiments published in the early 1840s. Julius Weisbach in Saxony proposed the form of the friction law in 1845, and Henry Darcy, who had built Dijon's clean water supply, confirmed it with careful measurements on pipes of many materials in 1857. Osborne Reynolds's dye experiment of 1883 identified the two regimes and the number that separates them. Prandtl's boundary layer of 1904 reconciled ideal-flow theory with real drag, and his student Johann Nikuradse glued graded sand inside pipes to measure roughness in the early 1930s. Colebrook combined the smooth and rough laws in 1939, and Moody's chart of 1944 put them on every engineer's desk until computers made iteration free.
Common misconceptions
Misconception
"The pressure falls along a pipe because the fluid speeds up." In a pipe of constant bore, continuity keeps the mean speed constant. The pressure falls because friction turns mechanical energy into internal energy; the pressure gradient is exactly what balances the wall shear.
Misconception
"Double the flow, double the pressure drop." Only in laminar flow. In turbulent flow the pressure drop rises roughly as in smooth pipes and as in fully rough ones, so doubling the flow multiplies it by about 3.4 to 4, and the pumping power by about 7 to 8.
Misconception
"A bigger pump gives more flow." The flow is set by where the pump curve meets the system curve. A pump much bigger than the duty ends up throttled by a valve, wasting energy, or running far from its best efficiency, where it wears faster. Matching the pump to the system is the job.
Misconception
"A perfectly smooth pipe has no friction." Even a smooth pipe has the viscous shear of the no-slip condition: its friction factor is if laminar and given by the smooth-pipe limit of Colebrook if turbulent. Roughness adds to friction only once the bumps poke out of the viscous layer at the wall.
Exercises
1. Viscous flow and pipes
Warm-upHumanWater at 20 °C ( m²/s) flows at a mean speed of 0.8 m/s through a pipe of internal diameter 25 mm. What is the Reynolds number? (Then decide for yourself whether the flow is laminar or turbulent.)
2. Viscous flow and pipes
CoreHumanA hydraulic test rig pushes oil ( kg/m³, Pa·s) through a straight tube of internal diameter 10 mm and length 2.0 m at 2.0 mL/s. Assuming fully developed laminar flow, what pressure difference is needed, in pascals, to the nearest 10 Pa? Check that the flow really is laminar.
3. Viscous flow and pipes
CoreHumanThe flow rate through a long, straight pipe is doubled, with the fluid unchanged. Which statements are true? Select all that apply.
4. Viscous flow and pipes
CoreHumanWater ( kg/m³) flows at 2.0 m/s through 50 m of 100 mm pipe with friction factor . The fittings on the line (entrance, bends, two valves and the exit) have loss coefficients adding up to . The pipe is level and has the same diameter throughout. What is the total pressure drop, in kPa?
5. Viscous flow and pipes
CoreHumanA centrifugal pump running at 1450 rpm delivers 20 L/s at a head of 30 m and draws 8.0 kW. A variable-speed drive slows it to 1160 rpm. Using the affinity laws (and assuming the efficiency at the corresponding point is unchanged), what shaft power does it now draw, in watts?
6. Viscous flow and pipes
CoreHumanA dimpled golf ball flies further than an otherwise identical smooth ball. What is the main reason?
7. Viscous flow and pipes
CoreComputationalWrite colebrook(Re, rel_rough) that returns the Darcy friction factor for turbulent pipe flow by solving the Colebrook equation
numerically, to at least eight significant figures. The tests check that your satisfies the equation for smooth and rough pipes from to , and reproduce the lesson's worked example.
8. Viscous flow and pipes
StretchComputationalWrite operating_point(pump, system, q_max, tol=1e-9) that finds the flow rate where a pump curve meets a system curve. pump and system are functions of the flow rate that return heads; assume that if a crossing exists in , then is positive before it and negative after it. Use bisection and return the crossing as a float, to within tol. If the pump cannot even overcome the system at zero flow (, a pump too weak for the static lift), return None.
Further reading and resources
- Turbulent flow in pipes, with particular reference to the transition region between the smooth and rough pipe laws · C. F. Colebrook · 1939Journal of the Institution of Civil Engineers 11, 133–156. The implicit friction-factor equation used throughout the industry.
- Friction factors for pipe flow · L. F. Moody · 1944Transactions of the ASME 66, 671–684. The paper that turned Colebrook's equation into the chart on every hydraulic engineer's wall.
- EPANET · United States Environmental Protection AgencyFree, open-source software for modelling water distribution networks — pipes, pumps, valves and tanks — used by utilities worldwide.
- Hydraulic InstituteThe North American pump industry association; publishes pump standards and guidance on pump selection and system efficiency.
- ASMEPublishes the B31 piping codes (power piping, process piping and others) used for pipework design in many countries.
- ISO, the International Organization for StandardizationPublishes ISO 9906, the international acceptance test for the hydraulic performance of rotodynamic pumps.
- MIT OpenCourseWare 2.25, Advanced Fluid Mechanics · Massachusetts Institute of TechnologyLecture notes and problem sets on viscous flow, boundary layers and pipe flow at graduate level.