Hydraulic

Pumps in parallel: why two identical pumps don't double the flow (operating point worked example)

Two identical pumps in parallel deliver the sum of their flows only at the same head — and the system will not stay at the same head. In the worked case below, the second pump adds 36.8% of flow, not 100%, and the third adds only 9.4% more.

The short answer

Two identical pumps in parallel add their flows at the same head. The system does not stay at the same head: friction loss grows with Q², so when the flow increases, the system curve demands more head, and each pump slides back along its curve to a lower flow. The result is always less than double. In the worked case below, one pump delivers 74.4 m³/h and two pumps deliver 101.9 m³/h — a gain of 36.8%. A third pump adds only another 9.4%.

How much the second pump adds is not a property of the pump. It is a property of the ratio between static head and friction loss in the system.

The two curves that set the operating point

Combined pump curve. In parallel, the pumps share the suction and discharge headers, so they work at the same head. At each head, the combined flow is the sum of the individual flows. For N identical pumps:

H_comb(Q) = H_single(Q / N)

If the single-pump curve is fitted as a parabola H = a·Q² + b·Q + c, the combined curve for N pumps is H = (a/N²)·Q² + (b/N)·Q + c. The shut-off head (c) does not change — only the flow axis stretches.

System curve. The head the system demands at a flow Q is the static head plus the losses:

H_sys(Q) = H_st + h_f(Q) + h_minor(Q)
h_f = f·(L/D)·(v²/2g)        h_minor = ΣK·(v²/2g)

With a nearly constant friction factor, this is the familiar H_sys = H_st + k·Q². The operating point is where H_comb(Q) = H_sys(Q).

Worked example: one, two and three pumps

All numbers below were computed with the same friction-factor module used by the LOGOS calculation engine (Colebrook-White, solved to machine precision), the internal diameters from the LOGOS pipe table (ASME B36.10M, Sch 40/STD), and a 3-point Lagrange fit of the pump curve, with the operating point found by bisection — the same method the LOGOS pumps-in-parallel calculator uses.

Inputs

ItemValue
FluidWater at 20 °C, ρ = 998 kg/m³, μ = 1.002 mPa·s
Pump curve (3 catalog points)0 m³/h @ 38 m (shut-off) · 60 m³/h @ 30 m (BEP) · 80 m³/h @ 24 m
Fitted curve (Q in m³/h)H = 38 − 0.008333·Q − 0.0020833·Q²
Static head18 m
Suction6 m of 8” Sch 40 (ID 202.71 mm), ΣK = 1.6 (entrance 0.5 + gate valve 0.2 + elbow 0.9)
Discharge1,000 m of 6” Sch 40 (ID 154.08 mm), ΣK = 7.3 (check valve 2.5 + gate valve 0.2 + 4 elbows 3.6 + exit 1.0)
Roughnessε = 0.045 mm (new carbon steel)

The model treats suction and discharge as common headers carrying the total flow; the short individual pump branches are neglected.

Results

Pumps runningTotal flowHeadFlow per pumpPer pump vs BEPGain vs 1 pump
174.4 m³/h25.8 m74.4 m³/h124%—
2101.9 m³/h32.2 m50.9 m³/h85%+36.8%
3111.5 m³/h34.8 m37.2 m³/h62%+49.7%

With two pumps, the discharge velocity is 1.52 m/s (Re ≈ 233,000, f = 0.0174), the discharge friction loss is 13.2 m, minor losses 0.9 m and the suction loss 0.08 m. Between the two operating points the head rises from 25.8 m to 32.2 m, and that 6.3 m difference is exactly why each pump falls from 74.4 to 50.9 m³/h.

The intuitive answer — “two pumps rated at 60 m³/h give 120 m³/h” — fails on a simple check: at 120 m³/h this system needs 37.4 m, but two pumps at 60 m³/h each deliver only 30 m. The combined curve and the system curve never meet there.

Near the design point, the system curve can be approximated as H_sys ≈ 18 + 1.37×10⁻³·Q² (Q in m³/h). That k is useful for a quick sketch; the calculation itself keeps f varying with Reynolds.

Steep vs flat system: a table of the real gain

To isolate the effect of the system curve, the table below keeps the same pump and forces every system through the same single-pump duty point (60 m³/h at 30 m, the BEP). Only the split between static head and friction changes; the discharge length of 6” pipe was solved for each case.

Static share of head at dutyStatic headEquivalent 6” length2 pumpsGain3 pumpsGain
90% (flat curve)27 m545 m89.8 m³/h+49.7%102.7 m³/h+71.1%
67%20 m1,972 m74.1 m³/h+23.5%78.1 m³/h+30.1%
33%10 m4,011 m68.1 m³/h+13.4%70.0 m³/h+16.7%
0% (pure friction)0 m6,050 m65.6 m³/h+9.4%66.9 m³/h+11.5%

Two conclusions follow. In a friction-dominated (steep) system, a second pump is a poor way to gain flow — a larger pipe usually buys more. And in the steep rows, each of the two pumps ends up at 55–62% of its BEP flow, below the preferred operating region.

Risk 1: the pump left alone runs to the right of its curve

Parallel stations are designed for N pumps running, but they also run with fewer — at start-up, during maintenance, after a trip. With one pump alone, the system head falls and the pump runs out along its curve. In the example, the single pump runs at 124% of BEP (74.4 m³/h), against 85% when paired.

Running to the right means:

  • Higher NPSH required. NPSHr rises steeply beyond BEP, while the available NPSH falls because suction losses rise. The NPSH check must be done with the fewest pumps running, not with all of them.
  • Higher power demand. On a radial pump the power curve keeps rising with flow; a motor sized at the paired duty point can be overloaded when the pump runs alone.
  • Outside the preferred operating region. ANSI/HI 9.6.3 places the preferred operating region (POR) typically between 70% and 120% of BEP flow for most rotodynamic pumps (80–120% at high specific speed). 124% is outside it.

Risk 2: unequal pumps and the weaker one at shut-off

With different pumps in parallel, the stronger pump sets the header head and the weaker one delivers whatever flow its own curve gives at that head. Same system as above, with the second pump replaced by a weaker one (0 m³/h @ 32 m, 50 m³/h @ 25 m, 70 m³/h @ 19 m):

  • Header head: 29.4 m.
  • Stronger pump: 62.3 m³/h (104% of its BEP).
  • Weaker pump: 28.5 m³/h (57% of its BEP).
  • Total: 90.8 m³/h, against 101.9 m³/h with two identical pumps.

Raise the static head to about 28 m (a fuller destination tank, a higher reservoir) and the header head reaches the weaker pump’s shut-off of 32 m: its check valve stays closed and it runs at zero flow, heating the liquid in the casing. The same happens with identical pumps whose curves are flat or droop near shut-off. Pumps for parallel duty should have matched shut-off heads and stable curves that rise continuously toward shut-off.

Parallel, series, a bigger pump or a variable speed drive

  • Parallel — when the demand varies in steps, when redundancy matters (N+1 with a standby), and when the system curve is flat (static head dominant). It is the strongest option in the top rows of the table above.
  • Series — when the system needs more head at about the same flow: high static lift, long transfer lines where one pump cannot reach the required head.
  • One larger pump (or a larger pipe) — when friction dominates. Several pumps fighting a steep system curve add little flow; reducing the losses changes the curve itself.
  • Variable speed drive — when demand varies continuously and friction is a large part of the head; reducing speed follows the system curve instead of burning head across a valve. With a high static head, the speed range is short before the pump reaches shut-off, so drives are often combined with parallel pumps.

What the LOGOS calculator does — and what it does not

The LOGOS pumps-in-parallel calculator takes three catalog points, builds the combined curve for N identical pumps, computes the system curve with Darcy-Weisbach and Colebrook-White plus the K of each fitting, and finds the operating point by bisection for three tank-level scenarios. It reports total flow and flow per pump, and it can check NPSH available against an NPSHr you enter. The physics runs on versioned server-side functions; the public page explains the method, the calculation is done in the logged-in area, and the result can be exported as a Word (.docx) calculation report. During the open beta, full access is free.

Limits to keep in mind: the calculator models identical pumps only — for unequal pumps, the curve-addition must be done by head, as shown above. It does not model the individual pump branches, the power curve, or transients on pump start and trip. The result supports, but does not replace, the engineer of record’s judgment and signature.

Standards & methods

  • ANSI/HI 9.6.3 — Rotodynamic Pumps, Guideline for Operating Regions (POR typically 70–120% of BEP flow)
  • Darcy-Weisbach with Colebrook-White friction factor
  • ASME B36.10M — welded and seamless wrought steel pipe (internal diameters)

Frequently asked questions

Do two pumps in parallel double the flow?

Only if the system curve is flat, which means the whole head is static and there is no friction loss. In any real pipeline, friction loss grows with the square of the flow, so the operating point moves to a higher head and each pump delivers less. Typical gains are 10% to 50% for the second pump, depending on how much of the head is friction.

How do you draw the combined curve of two pumps in parallel?

For each head value, add the flows of the pumps that can deliver at that head. For N identical pumps, the combined curve is the single-pump curve with every flow multiplied by N, so H_comb(Q) = H_single(Q/N). The operating point is the intersection of this combined curve with the system curve.

What is the difference between pumps in series and pumps in parallel?

In series, the same flow passes through both pumps and the heads add. In parallel, the pumps share the same suction and discharge headers, the head is the same and the flows add. Series helps high-static systems that need more head; parallel helps systems that need more flow and have a relatively flat system curve.

Can I put two different pumps in parallel?

You can, but check every operating scenario. The weaker pump sees the head set by the stronger one; if that head approaches the weaker pump's shut-off head, its check valve stays closed and it runs at zero or near-zero flow, which overheats the pump. Pumps in parallel should have similar shut-off heads and stable, continuously rising curves toward shut-off.

Is a variable speed drive better than adding a parallel pump?

For a system with varying demand and large friction loss, often yes, because reducing speed follows the system curve instead of throttling. For a high-static system, speed reduction quickly runs into the static head and the pump may reach shut-off, so a duty-assist parallel arrangement, often combined with a drive on one pump, is the usual solution.

Run this calculation

Reference tables

Read next