Piping System Design

Pipe sizing starts from a required flow, not from a pipe you happen to have on the shelf. Once you know the volumetric flow Q the system must move, the diameter you choose sets the mean velocity: for a fixed Q, a narrower bore means a faster stream, a wider bore a slower one. The area–velocity relationship looks trivial, but the choice of design velocity is where the real judgement sits, because friction loss, erosion, noise and cost all swing on it.

Marine piping guides settle on velocity bands rather than a single number: suction lines are kept slow, typically 1 to 2 m/s, because a fast, turbulent stream arriving at the pump inlet strips away the margin of available suction head before the fluid has even reached the impeller eye. Discharge lines can run faster, commonly 2 to 3 m/s, because the pump is pushing rather than pulling and a little extra velocity head is cheap once the fluid is already pressurised. Seawater and other erosive duties tighten this further: at sustained high velocity the protective film on the pipe wall wears away faster than it can re-form, and metal loss accelerates at bends, branches and anywhere the flow separates from the wall — erosion-corrosion, a materials problem caused entirely by a hydraulic choice.

• Suction lines — kept slow so the pump's available NPSH is not eaten up before the fluid arrives.
• Discharge lines — allowed faster, since the fluid is already under pressure.
• Seawater or abrasive duties — tighter limits still, to hold back erosion-corrosion at bends and branches.

Once the bore is fixed, the pipe resists the flow with a pressure drop that Darcy and Weisbach's equation ties directly to velocity, length and diameter.

The friction factor f is not a fixed material property; it depends on the Reynolds number Re = VD/ν and on the pipe's relative roughness (wall roughness height divided by bore), and its value is read from the Moody chart, or an equivalent correlation, once both are known. Two regimes behave very differently. Below roughly Re = 2000 the flow is laminar — it moves in smooth, orderly layers, viscosity dominates, and the friction factor collapses to a simple closed form, f = 64/Re, falling as flow increases. Above roughly Re = 4000 the flow is turbulent — eddies mix the fluid across the pipe, wall roughness starts to matter, and f settles onto a Moody-chart curve that depends on both Re and relative roughness. Marine piping runs are almost always turbulent in normal service; the laminar formula mostly matters for viscous duties such as cold lubricating oil.

Straight pipe is rarely the whole story. Every strainer, bend, tee, reducer and valve disturbs the flow and dissipates energy as the stream separates and re-mixes downstream of the fitting, and this is accounted for as a minor loss.

The word "minor" is worth treating with suspicion. A long, unobstructed transit main really can be dominated by its straight-pipe friction, but a short engine-room run packed with a suction strainer, two or three bends around machinery, a non-return valve and an isolating valve can lose more head in its fittings than in all its straight pipe combined — exactly the situation a designer who only checks the straight-pipe Darcy–Weisbach term will miss. The practical fix is to convert every fitting's K factor into an equivalent length of straight pipe that would give the same loss, add that to the physical pipe length, and run the friction calculation once against the total, so nothing is left uncounted.

Valve choice matters here too: a full-bore gate or ball valve, wide open, contributes very little resistance, while a globe valve, whose flow path turns through the body even fully open, carries a meaningfully higher K in its best-case position. Where a valve is also used for throttling, its K factor at partial opening — not its wide-open value — is the one that belongs in the calculation for that duty.

A pump raises the fluid's pressure, and where the layout calls for it, its elevation, against everything the piping resists. The hydraulic power it delivers to the fluid, and the shaft power its driver must supply to produce that, are related through the pump's efficiency.

The system curve is the piping's own signature: the total head it demands as a function of flow, made up of a static term — whatever elevation and pressure difference exists whether or not fluid is moving — plus a friction term that grows with the square of flow, since velocity itself is proportional to flow for a fixed bore. Plotted against flow this is a rising parabola, offset upward by the static term, or passing straight through the origin where there is none, as in a closed cooling loop. The pump curve runs the opposite way: head falls as flow rises, because forcing more flow through a given impeller at a given speed costs pressure.

Net positive suction head is the margin, expressed as head, between the pressure at the pump's suction and the fluid's vapour pressure at that temperature — the cushion that keeps the liquid from flashing to vapour as it accelerates into the low-pressure eye of the impeller. If that margin runs out, vapour bubbles form and then collapse violently as they reach a higher-pressure region inside the pump; the resulting cavitation erodes the impeller, drops performance sharply, and is audible as a rattling or gravel-like noise.

NPSH available describes what the installation offers, built up from the source: atmospheric or tank pressure converted to head, plus or minus the static height between the source surface and the pump centreline, less the suction-line friction, less the vapour pressure converted to head. The sign on the static term is the detail that decides the whole calculation — with a flooded suction the source sits above the pump and that head helps, so it is added; with a suction lift the pump has to pull the fluid up to itself and that same head works against it, so it is subtracted. Treating a lift as a flooded suction overstates NPSH available by twice the lift distance, enough to turn a genuinely marginal installation into one that looks safe on paper.

NPSH required is a property of the pump itself at a given flow, taken from the manufacturer's curve, and the installation is only sound once the available figure exceeds it by a margin — commonly at least 0.5 to 1 m, more for hot or volatile duties — never merely equal to it.

When a centrifugal pump's running speed changes and its impeller geometry does not, three simple ratios link the old and new operating points.

The cube law on power has the practical payoff: slow a pump to 80 percent of its speed and it draws only about half its former power, which is the economic case for variable-speed drives on part-load duties like closed cooling-water circuits, where the system curve has no static term and the true duty point at reduced speed lies exactly on the path these ratios predict.

The same three ratios are sometimes reached for when an impeller is trimmed to a smaller diameter at constant speed instead of the speed being changed, and this is where marks are lost: the diameter-trim relationships are only an approximation, reasonable over small trims and increasingly unreliable as the cut gets larger, because trimming changes the impeller's internal flow geometry in a way a pure speed change does not. A trimmed-impeller duty point should be checked against the manufacturer's trimmed-diameter curve, not assumed from the speed-change ratios.

Specific speed groups pumps of very different physical size but the same hydraulic character, and its value at the duty point points toward whether a radial, mixed-flow or axial-flow impeller is the right family for the job, well before any single pump's curve is opened.

The three examples below carry a piping design through from a bare flow requirement to a checked, working system: sizing a main and totalling its losses, proving the suction side against cavitation, and re-rating a pump by speed rather than by guesswork.