The idea
Wherever the local static pressure in a flowing liquid falls to its vapor pressure, the liquid
boils — even at ambient temperature. Vapor bubbles born at the impeller inlet ride into the
high-pressure interior and collapse there within milliseconds, focusing the surrounding liquid
into micro-jets with impact pressures on the order of \(10^{4}\) atmospheres. The symptom chain
is unmistakable: crackling "gravel" noise, rising vibration, pitting of the inlet vanes, and
finally a sagging head curve.
NPSH — net positive suction head — is the bookkeeping that keeps suction pressure safely
above vapor pressure. Available NPSH (\(NPSH_A\)) is a property of the installation;
required NPSH (\(NPSH_r\)) is a property of the pump at a given flow and speed. The
margin between them is the whole discipline.
The equation
All terms are heads of the pumped liquid. For a boiling liquid (\(p_{e,abs}=p_v\)) only the
static head above the pump remains — it must be positive.
The 3 % criterion and the margin ladder
Published \(NPSH_r\) values are almost always NPSH3% — the NPSH at which head has already
fallen by 3 %, meaning cavitation is thoroughly developed. First bubbles (inception,
\(NPSH_i\)) typically appear at 2–5× that value. So the ladder, bottom to top, reads:
\(NPSH_{3\%}\) → \(NPSH_{40}\) (the value giving ≈40,000 hours of erosion-free impeller life)
→ \(NPSH_i\approx5\times NPSH_{3\%}\). The margin ratio
\(R=NPSH_A/NPSH_{3\%}\) locates you on that ladder. For hot water and hydrocarbons the
manufacturer's cold-water \(NPSH_r\) may be reduced — weaker bubble collapses — but the
reduction is capped at the smaller of 50 % of the cold-water value and 3 m (10 ft).
How fast the damage grows
Erosion intensity grows with roughly the sixth power of the inlet tip speed — equivalently with
the cube of the NPSH scale at constant suction ratio. Raise speed 25 % at the same duty and the
erosion rate multiplies by about \(1.25^{6}\approx3.8\). Water is most aggressive around
38–49 °C; dissolved gas cushions the collapses; hydrocarbon vapors collapse so weakly that
erosion is rarely observed. Temperature also eats the margin from the \(NPSH_A\) side: water's
vapor head climbs from ≈0.24 m at 20 °C to ≈4.8 m at 80 °C. A single vapor cavity lives only
≈0.003 s from birth to collapse — but millions of them march through every minute.
Engineering criteria
| Situation | What it means | Verdict |
|---|
| \(NPSH_A \lt NPSH_{3\%}\) | head dropping, "gravel" noise, active erosion — unacceptable for continuous duty | 🔴 |
| \(1\le R \lt 5\) | performance looks fine, but erosion and pulsation proceed; pressure pulsation peaks near \(R\approx2\) | 🟡 |
| \(R\ge5\) (\(NPSH_A\ge NPSH_i\)) | at or beyond inception — essentially cavitation-free operation | 🟢 |
| \(NPSH_A \lt 0.6\) m | absolute installation floor for any liquid, whatever the arithmetic says | 🔴 |
| \(NPSH_{3\%}\le NPSH_A \lt S_A\cdot NPSH_{3\%}\) | margin below the recommended ratio \(S_A\) (from the selection curve) — risky especially at high eye tip speed or with corrosive media; seawater wants the upper margin | 🟡 |
| Quoted \(NPSH_r\) with no criterion named | inception / 0 % / 3 % / full-choking values for the same pump differ by a factor of several — the criterion is part of the number | 🟡 |
| Hot-water / hydrocarbon NPSH reduction | allowed reduction ≤ min(50 % of cold-water \(NPSH_r\), 10 ft / 3.0 m) — anything larger is cut back to that cap | 🟡 |
| NPSH reduction forbidden | entrained or dissolved non-condensable gas that can come out of solution; systems with strong transient pressure/temperature swings (add margin instead); liquids off the reduction chart (trial value only) | 🔴 |
| Entrained gas at the inlet, GVF ≤ 0.03 | commercial industrial pumps hold their head — the usual acceptance limit | 🟢 |
| GVF 0.04–0.07 | head rise still stable on a conventional single-stage pump, margin thinning | 🟡 |
| GVF > 0.07 | head/flow/power decay toward loss of prime — unacceptable without special design | 🔴 |
| Cavitation surge (1–6 Hz low-frequency oscillation) | a recirculation–vapor-lock cycle; avoided by running above \(Q_{min}\) — rare on low-\(N_{ss}\) impellers | 🟡 |
| High-energy pump judged on \(NPSH_{3\%}\) alone | at high inlet tip speed the 3 % criterion is nearly meaningless for damage — size the margin to \(NPSH_{40}\) (≈40,000 h impeller life) or a "no visible cavity" test | 🟡 |
| Cavitation life endpoint | erosion through 75 % of vane/wall thickness ends the life count: \(Life[\text{yr}]=0.75\,t[\text{mm}]/MDPR[\text{mm/yr}]\) | — |
| High suction energy \(SE = D_e \times N \times N_{ss} \times \text{sp.gr.}\) | high-SE pumps need a larger NPSH margin, and a still larger one below BEP flow | 🟡 |
| Two 90° elbows in different planes directly upstream | distorted inlet flow — a documented cause of through-wall cavitation damage in months | 🔴 |
| Minimum-flow temperature rise | ΔT ≤ 100 °F (56 °C) general cold-liquid pumps; ≤ 50 °F (28 °C) modern boiler feed pumps — beyond that a minimum-flow bypass is mandatory | 🟡 |
| Hot-liquid pump start | warm the pump to working temperature before starting (unless designed for fast start); never run a hot pump at shutoff — a metered-orifice bypass is required | 🔴 |
| Inducer-equipped impeller above rated flow | the combination's \(NPSH_r\) rises steeply past rating — keep operation at or below rated flow (variable-pitch inducers excepted) | 🟡 |
Rules of thumb — quick estimates
| Name | Rule / formula | Applies to |
|---|
| Inception multiplier | \(NPSH_i\approx(2\text{–}5)\times NPSH_{3\%}\); use 5× as the typical value | conventional centrifugal pumps |
| Erosion-rate scaling | rate \(\propto V^{6}\propto(p_1-p_v)^{3}\propto NPSH^{3}\) at constant cavitation number — and \(\propto N^{6}\) on speed | same liquid, same geometry |
| Material dependence | erosion intensity \(\propto 1/(\text{tensile strength or hardness})^{2}\) | material comparison |
| Worst water temperature | cavitation damage in water peaks at 100–120 °F (38–49 °C) | controlled-test result |
| Bubble life and collapse | cavity life cycle ≈ 0.003 s; collapse pressures on the order of \(10^{4}\) atm — every material succumbs under long exposure | all liquids |
| NPSH3% correlation | \(NPSH_{3\%}=k_1 V_e^{2}/2g+k_2 W_{1,sh}^{2}/2g\); \(k_1=1.69,\ k_2=0.102\) (typical mid-size), \(k_1=1.4,\ k_2=0.085\) (large pumps) | zero pre-swirl inlets |
| Cavitation coefficient form | \(\tau_{3\%}=NPSH_{3\%}/(U_e^{2}/2g)=(k_1+k_2)f_e^{2}+k_2\), with eye flow coefficient \(f_e=V_e/U_e\) | inlet design checks |
| Typical installation NPSHA | ≈ 60 % of the barometric head — roughly 20 ft (6 m) of water | first-pass condensate estimates |
| Vapor head of water | ≈ 0.24 m at 20 °C, ≈ 4.8 m at 80 °C — temperature eats margin fast | all \(NPSH_A\) arithmetic |
| Gas and chemistry effects | erosion falls as dissolved-gas content rises (cushioned collapse) and as temperature rises past the peak; it grows with corrosivity and is rarely seen on hydrocarbons | liquid comparison |
| Compression heating | reversible \(\Delta T_c\): hydrocarbon fuels ≈ 3 °F/1000 psi (0.24 °C/MPa); 350 °F feedwater ≈ 1.6 °F/1000 psi — subtract it when inferring efficiency from temperature rise | thermodynamic efficiency tests |
| Anti-swirl ribs | 3–4 radial ribs reaching ≈ ¼ of the inlet diameter into the suction run suppress pre-rotation at less NPSH cost than full-length vanes | part-load swirl control |
| Typical NSS by inlet style | see the reference table below — pick the lower end when eye tip speed (and erosion risk) is high | metric NSS, BEP, 3 % basis |
Reference data tables
Typical suction specific speed by inlet configuration
Metric NSS with \(N\) in rpm, \(Q\) in m³/s per eye, NPSH in m, evaluated at BEP on the 3 %
head-drop criterion. The three columns are the usual lower / typical / upper values; convert
with \(N_{ss}(\text{US})=51.6\times NSS(\text{metric})\):
| Inlet configuration | Low | Typical | High |
|---|
| End-suction, axial inlet | 190 | 230 | 270 |
| Shaft through the eye, moderate blockage | 170 | 200 | 240 |
| Multistage pumps with stage heads > 500 m | 150 | 180 | 220 |
| Industrial inducers | 350 | 500 | 700 |
Cavitation-erosion resistance of materials
Weight loss in a standardized 2-hour magnetostriction cavitation test — lower is better:
| Material | Weight loss, mg / 2 h | Standing |
|---|
| Rolled stellite | 0.6 | best (costly, hard to machine) |
| Welded aluminium bronze | 3.2 | |
| Cast aluminium bronze | 5.8 | |
| Welded stainless (17Cr-7Ni, 2 layers) | 6.0 | |
| Hot-rolled stainless (26Cr-13Ni) | 8.0 | |
| Quenched & tempered rolled stainless (12Cr) | 9.0 | |
| Cast stainless (18Cr-8Ni) | 13.0 | |
| Cast stainless (12Cr) | 20.0 | |
| Cast manganese bronze | 80.0 | |
| Welded mild steel | 97.0 | |
| Steel plate | 98.0 | |
| Cast steel | 105.0 | |
| Aluminium | 124.0 | |
| Brass | 156.0 | |
| Cast iron | 224.0 | worst |
NPSH reduction for hot water and hydrocarbons — worked examples
| Case | Vapor pressure | Chart reduction | Cap check | Applied NPSHr |
|---|
| Cold-water \(NPSH_r=16\) ft, pumping propane at 55 °F | ≈ 105 psia | ≈ 9.5 ft | 9.5 > 8 (= 50 % of 16) → capped at 8 ft | 16 − 8 = 8 ft |
| Same pump, propane at 14 °F | ≈ 50 psia | ≈ 5.7 ft | 5.7 < 8 → full reduction applies | 16 − 5.7 ≈ 10 ft |
| Liquid off the chart, vapor pressure 30 psia at 100 °F | 30 psia | ≈ 2.3 ft | trial value only | reduce by ≈ 2.3 ft |
Common mistakes
- Quoting or comparing \(NPSH_r\) without naming the criterion (inception / 0 % / 3 % /
breakdown) — the same pump can carry values differing by a factor of several.
- Computing \(NPSH_A\) with cold-water vapor pressure for a hot or volatile liquid: \(p_v\) is
set by the pumping temperature, always.
- Forgetting suction-side friction and fittings, or referencing levels to the wrong datum on
vertical pumps — both quietly consume the margin.
- Applying the full hot-water/hydrocarbon NPSH reduction without checking the 50 %-of-value /
3 m cap — and without asking whether dissolved gas or transients forbid any reduction.
⭐ Deep-data appendix
The finer print of cavitation engineering: how the required-NPSH figure is supposed to be
defined, what the pulsation spectrum tells you, the improvement-measure ladder, inducer
capability numbers, and a measured inception-to-breakdown anchor series.
Additional engineering criteria
| Situation | What it means | Verdict |
|---|
| \(NPSH_r\) quoted for a selection | the correct figure is the larger of the performance-protection value (no significant head loss) and the damage-protection value (erosion life) — on high-energy stages the damage value governs | 🟡 |
| Pressure pulsation vs margin ratio | cavitation pressure-pulsation amplitude peaks near \(R=NPSH_A/NPSH_{3\%}\approx2\) and its frequency rises with \(R\) — a field diagnostic of where you sit on the margin ladder | 🟡 |
| Using a suction-specific-speed limit as a selection criterion | not recommended — an \(N_{ss}\) ceiling alone predicts neither suction recirculation nor field behaviour; the NPSH margin and the operating flow range decide | 🟡 |
| Inception-coefficient design target | conventional vanes: \(\tau_i\approx1\) at BEP (higher below BEP); aerodynamically optimized inlet vanes reach \(\tau_i\approx0.5\); holding \(NPSH_A\ge NPSH_i\) (\(\tau_A\le\tau_i\)) means zero cavity length — genuinely cavitation-free | 🟢 |
| Inducer + impeller combination | runs at ≈ 2 × the suction specific speed of a conventional impeller; the combination's \(NPSH_r\) is ≈ 50 % of the bare impeller's; the inducer contributes ≤ 5 % of total head — valid only at and below rated flow (variable-pitch inducers excepted) | 🟢 special |
Additional rules of thumb
| Name | Rule / formula | Applies to |
|---|
| Improvement-measure ladder | curing an NPSH shortfall, in order of preference: special suction impeller > double-suction first stage > inducer > booster pump; inducers and suction impellers must be pulsation-free at part load | retrofit decisions |
| NPSH-curve shape vs flow | \(NPSH_{inc}\) is at its minimum near the shockless-entry flow and rises to a part-load maximum from inlet recirculation — the worst cavitation activity is usually below BEP, not at runout | reading NPSH–Q curves |
| Backflow recirculator | a passive de-swirl guide-vane device that returns a controlled backflow can suppress inducer cavitation instability over the whole shutoff-to-runout range | inducer-equipped machines |
| Thermodynamic effect | near-critical fluids (e.g. liquid hydrogen) tolerate a suction specific speed several times the cold-water value at identical geometry — bubble collapse is weakened by the thermal properties of the fluid | cryogenic service |
| Cavitation coefficient \(\tau\) | \(\tau=2g\cdot NPSH/U_e^{2}\) with \(U_e\) the eye tip speed; conventional vanes \(\tau_i\approx1\) at BEP, optimized ≈ 0.5, and \(\tau_i\) rises below BEP flow | inlet design checks |
| Boundary-layer blockage | turbulent displacement thickness \(\delta^{*}/L=0.0462/Re_L^{0.2}\); count \(2\delta^{*}\) on blade surfaces (adverse gradient) and \(\delta^{*}\) on hub and shroud (secondary-flow scrubbing) | passage blockage estimates |
Measured inception-to-breakdown anchor (boiler feed pump)
Cavitation coefficient \(\tau=2g\cdot NPSH/U_e^{2}\) read off a variable-NPSH test at fixed
speed and flow; the 3 % head-drop point is where the cavity reaches the throat of the
adjacent blade passage:
| \(\tau\) | Event |
|---|
| 0.61 | near visual inception (\(\tau_i\) of this impeller) |
| 0.41 | the plant's available value \(\tau_A\) for this duty — inside the cavitating band |
| 0.39 | developed cavitation, head still unaffected |
| 0.29 | further developed, head still unaffected |
| 0.20 | 3 % head drop (H = 97 % of the cavitation-free value) — cavity reaches the adjacent blade throat |
Engineering criteria — finer print
| Situation | What it means | Verdict |
|---|
| Acceptance on a sub-3 % criterion (0 %, 1 % head drop) | head-drop points below 3 % carry large measurement scatter — not recommended as acceptance criteria; quote \(NPSH_{3\%}\) | 🟡 |
| Running far below BEP | suction/discharge recirculation drives pressure pulsation whose peak-to-peak can reach the order of the stage pressure rise itself; inlet-edge pitting plus shroud bulging is the confirmed red-flag pair | 🔴 |
| NPSH datum plane on large machines | place the datum at the elevation most likely to develop cavitation (e.g. the top of a propeller blade) — referencing a lower plane quietly overstates \(NPSH_A\) | 🟡 |
How published \(NPSH_{3\%}\) chart families are parameterized
Two separate families exist, one per bearing arrangement; \(NPSH_{3\%}\) climbs along each
speed curve as flow rises. These families are the base data of any \(NPSH_r\) estimate:
| Family | Computed at | Horizontal axis | Family parameter |
|---|
| Overhung (end-suction) impellers | \(N_{ss}=200\) | flow \(Q\) | speed \(N\) |
| Impellers between bearings | \(N_{ss}=200\) | flow \(Q\) | speed \(N\) |
Rules of thumb — finer print
| Name | Rule / formula | Applies to |
|---|
| Coated-impeller life | \(Life=MDPR_{coat}\times t_{coat}+MDPR_{base}\times t_{rem}\), with \(t_{rem}\) the remaining allowed erosion depth — judge coatings against the same 75 %-penetration end point | protected impellers |
| How \(NPSH_r\) curves are born | fixed-speed, variable-\(NPSH_A\) tests (e.g. a 1.5-in single-stage pump at 3,470 rpm on 70 °F water): each \(NPSH_A\) level yields one H–Q curve; adjacent curves separate at the 3 % head-drop point | reading test data |
| Inlet W-profile convention | design the average relative velocity \(W\) to drop ≥ 10 % entering the cascade — it minimizes the sum of incidence and friction losses at BEP and keeps NPSH lower above BEP flow | radial impellers |
| Incidence convention | positive incidence 2–3° at the mean/rms streamline; shroud vane angle ≈ 1° below the local flow angle (slightly negative — efficiency and suction win together) | inlet vane setting |
| Documented erosion case | 3/8 in (9.5 mm) of stainless wall penetrated in 3 months of service — cause traced to two 90° elbows in mutually perpendicular planes directly upstream | inlet-distortion diagnostics |
Try it interactively — compute available and required NPSH and watch the margin verdict:
NPSH Calculator
Related