Pump Suction Piping — NPSH and the Eccentric Reducer

The pump that passed its calculation and still hammered

A condensate transfer pump, 180 m³/h, water at 95 °C from an atmospheric tank whose level sits 2.5 m above the pump centreline. Twenty-five metres of DN150 suction, three fittings. The process engineer's datasheet says NPSHa = 2.6 m, the vendor's curve says NPSHr = 2.2 m at duty. Margin positive. Signed.

Six weeks after start-up the pump rattles like gravel in a drum, the discharge pressure wanders, and a borescope shows the vane inlets already pitted. Nothing on the datasheet has changed. The level is where it should be, the temperature is where it should be, the flow is where it should be.

Guess before reading on: what is wrong?

Two things, and the calculation could not see either of them. The margin was 0.4 m when it needed to be 1 m and 1.1–1.3× — so the pump was living on the NPSHr point, which is already the 3 %-head-drop point. And the fabricator installed the eccentric reducer flat-side-down, building a 54 mm-deep roof at the top of the pipe (D − D_n = 168.3 − 114.3 for this DN150 line into its DN100 pump nozzle) where vapour collects and then sheds into the impeller in slugs.

The misconception, stated plainly

"Reducer orientation is a detailing preference. It's a piece of pipe; the fluid doesn't care which way up the taper is."

It is a seductive belief because the arithmetic agrees with it. A concentric reducer and an eccentric reducer have essentially the same K-factor. Flip an eccentric reducer over and not one number in the NPSH calculation changes — not the static head, not the vapour pressure, not the friction. The orientation is invisible to the maths.

It is visible to the liquid. An NPSH calculation is a statement about the average pressure at the suction flange. The reducer decides what the impeller eye sees instant by instant — and an impeller does not fail on averages.

Symbol key — every symbol on this sheet

• NPSH — Net Positive Suction Head: how much pressure the liquid has above its own boiling point, expressed as a height of that liquid · m
• NPSHa — NPSH available: what the piping and the source deliver to the suction flange · m
• NPSHr — NPSH required: what the pump needs there, from the vendor's test · m
• margin — NPSHa − NPSHr; also quoted as the ratio NPSHa/NPSHr · m, −
• P_s — absolute pressure on the liquid surface in the source vessel · bar a
• P_v — vapour pressure of the liquid at the pumping temperature · bar a
• h_s — static head: liquid level minus pump centreline (negative = suction lift) · m
• h_f — friction head lost between the source and the suction flange · m
• ρ (rho) — liquid density at temperature · kg/m³
• g — 9.81 m/s²
• v — velocity in the suction line · m/s
• f — Darcy friction factor · −
• L_eq — equivalent length: straight pipe + fittings converted to pipe · m
• D — inside diameter of the suction line · m, mm
• D_n — nominal bore of the pump suction nozzle (usually smaller than D) · mm
• FSU / FSD — eccentric reducer installed flat side up / flat side down
• σ (Nss) — suction specific speed; the number that says how greedy a given impeller is

NPSHa: four terms, and the three that steal from you

NPSHa = (P_s − P_v)/(ρ·g)  +  h_s  −  h_f

Read it as an accounting of how far the liquid is from boiling, measured in metres of itself:

• (P_s − P_v)/(ρg) — the pressure account. What the surface pressure gives you, minus what the liquid's own vapour pressure takes back. This is where temperature bites. Cold water in an open tank: 10.3 m. The same water at 95 °C: 1.78 m. At 100 °C in a vessel at its own boiling point: exactly zero, no matter how high the vessel pressure is. A saturated liquid brings nothing to the table — every deaerator, every reflux drum at bubble point, every LPG sphere is in this category.
• h_s — the static account. The only term you can buy with civil work, and the only one that is reliably positive. It is also the reason a deaerator sits on a 12-metre stool.
• h_f — the friction thief. Scales with v², so it is a flow-squared thief: at 140 % flow it takes twice as much. This is why suction lines are one or two sizes larger than the pump nozzle, kept short, and stripped of fittings.
• And the silent fourth: NPSHr itself is not a constant. It climbs roughly as Q^1.8. The available curve falls as the required curve rises, and they meet at a flow beyond which the pump cannot go — which is exactly what happens on start-up against an empty line, or when an operator opens the discharge valve wide "to get it going".

Watch the two curves run at each other: ▶ open the interactive: piping layout npsh calc — set the condensate preset, then slide the flow up. The crossing point is the real capacity limit of that suction line, and it is usually well below the pump's rated runout.

The worked example, and the two ways out of it

The condensate pump from the opening, run through the model:

One line size on the suction recovered 1.2 m of NPSH — more than lifting the tank by a metre would have done, and far cheaper. That is the general lesson: friction is the cheapest metre of NPSH to buy back, because it falls as roughly 1/D⁵ at fixed flow while the pipe only costs about D^1.15.

The other lever is temperature, and it is brutal near the boiling point. Hold everything and slide the temperature from 95 °C to 105 °C in the calculator: P_v passes atmospheric, the pressure account goes negative, and an open tank can no longer feed the pump at all.

The last three diameters: straight run and the reducer

Everything above is about the number at the suction flange. The last two or three metres of pipe decide whether that number means anything.

Straight run. Vendors and HI 9.6.6 ask for a minimum of about 5 pipe diameters of straight pipe between the last fitting and the suction flange; 8–10 D is the safer target, and for a double-suction pump an elbow whose plane is parallel to the shaft is a specific prohibition — it splits the flow unevenly between the two halves of the impeller and produces axial thrust and vibration that no NPSH margin fixes. Where geometry cannot give you the run, a straightening vane or a suction diffuser (with its strainer maintained) is the compromise.

The reducer. The pump nozzle is almost always smaller than the suction line, so a reducer sits in that last straight run. On a horizontal suction, use an eccentric reducer, flat side up:

• Flat side up (FSU) — correct. The crown of the big pipe and the crown of the nozzle are one continuous line. There is no local high point, so vapour that comes out of solution in the low-pressure suction line is swept forward into the eye and away with the flow. All the taper is on the bottom, and the invert still falls back toward the source, so the line drains.
• Flat side down (FSD) — the defect. The bottom of pipe is continuous, so the crown steps down by (D − D_n) at the reducer. That step is a roof. Vapour and non-condensables collect under it, the pocket grows, and when it grows past the crown of the small pipe it spills — the impeller swallows gas in slugs. Symptoms: crackling noise, wandering discharge pressure, a head curve that will not hold, and pitting on the vane inlets. And the NPSH calculation still passes, because a pocket is a local phenomenon and NPSHa is an average.

Flip it and watch the pocket form: ▶ open the interactive: piping layout npsh 3d — the flip button moves the upstream run up or down by the eccentricity, rebuilds the reducer, and shows the trapped vapour lying along the crown in the wrong orientation, with bubbles shedding into the impeller.

The flat-side-down exception. The rule is about where the trap is, not about reducers:

• On a vertical run — a vertical can pump's suction, a riser, or the expander just above a horizontal pump's discharge nozzle — there is no top and no bottom. An eccentric reducer buys nothing there and a concentric one is correct.
• On a discharge line the fluid is well above its vapour pressure and the line is full, so a gas pocket is not the failure mode. What bites there is a liquid or solids trap at a step in the invert. If an eccentric is used, it goes flat side down — continuous bottom of pipe, so the line self-drains and nothing settles at the step.
• The same logic makes FSD correct on a suction line that runs continuously downhill into the pump from a source above: the falling crown cannot trap vapour, but a step in the invert would leave a puddle that never drains.

Rule of thumb that survives all three cases: put the flat side on the surface that must stay continuous. On a horizontal suction that surface is the top, because the enemy is vapour. On a draining or discharge line it is the bottom, because the enemy is standing liquid.

Common pitfalls

• Treating a positive margin as a pass. NPSHr is the 3 %-head-drop point. Aim for ≥ 1 m and ≥ 1.1–1.3×, more for high-energy pumps.
• Using vapour pressure at the normal temperature, not the worst one. Start-up, summer, upset and hot-standby cases all matter; for a saturated source the pressure term is zero by definition.
• Checking NPSH only at rated flow. NPSHa falls as Q² and NPSHr rises as Q^1.8. Check runout, and check the start-up transient against an empty discharge line.
• Forgetting the strainer. A temporary start-up strainer that nobody removes, or a permanent one nobody cleans, is a friction term that grows with time and has sunk many commissioning campaigns.
• An elbow hard against a double-suction nozzle, especially with its plane parallel to the shaft. Uneven split, thrust, vibration.
• A concentric reducer on a horizontal suction. Same maths, same K, and it builds a high point on both sides of the pipe.
• A high point anywhere in the suction line — a pipe that rises over an obstruction and comes back down is the same defect as a wrong-way reducer, built at a larger scale. If one is unavoidable, it needs a vent.
• Buying a low-NPSHr pump instead of fixing the piping. Low NPSHr means a large eye, high Nss and a narrow stable window. Fix the suction line first.

Outcome

• NPSHa = (P_s − P_v)/ρg + h_s − h_f: the pressure account (which a saturated liquid zeroes), the static account, and the friction thief that scales with v².
• NPSHr is the 3 % head-drop point, not the onset of cavitation. Design to ≥ 1 m of margin and ≥ 1.1–1.3× ratio; NPSHa falls as Q² while NPSHr climbs as Q^1.8, so the two curves cross at a real capacity limit — usually below the pump's rated runout.
• Worked case: 95 °C condensate, DN150 → margin 0.47 m (unsafe); one size up to DN200 → 1.67 m. Friction is the cheapest metre of NPSH to buy back, because h_f falls roughly as 1/D⁵.
• The last 5–10 D of straight pipe before the flange is part of the design, not detailing; for double-suction pumps an elbow in the plane of the shaft is a defect on its own.
• Eccentric reducer flat side up on a horizontal suction — continuous crown, no vapour trap. Flat side down where drainage governs (discharge, downhill suction); concentric on a vertical run where there is no top or bottom.
• A wrong-way reducer never changes the NPSH arithmetic. That is exactly why it gets built and exactly why it is dangerous.
• Interactive: ▶ open the interactive: piping layout npsh calc; 3D: ▶ open the interactive: piping layout npsh 3d.

Open items

• Add a real vendor NPSHr curve (rather than the Q^1.8 idealisation) and an Nss readout
• Extend the calculator to hydrocarbons and to the NPSH reduction credit for high-vapour-pressure liquids (HI 9.6.1 Appendix)
• Transient case: start-up NPSHa against an empty discharge line, and the level-swing case
• Photograph a real FSU/FSD field error and add it to the 3D as a before/after

Educational teaching material — simplified and illustrative. Not for engineering design use.
© EPiC Circle · provided by PipeAxis