Pump sizing: system curve, operating point and NPSH

What pump sizing is

Sizing a pump is not picking a model from its catalog flow rate: it is finding the real operating point, that is, the flow rate and head at which that specific pump will run once installed in that specific system. This point arises from the meeting of two curves on the head-versus-flow plane (H × Q): the system curve, which represents the energy the installation demands, and the pump curve, which represents the energy the machine delivers. Where they cross lies the hydraulic truth of the project.

The most common field mistake is to size by static lift and then discover, after start-up, that the pump delivers less flow than expected. The cause is almost always the same: head loss was underestimated, and the real system curve rose above the projected one.

The system curve

The system curve describes how much energy the installation demands for each flow rate. It has the form:

H_sys(Q) = H_geo + k · Q²

The H_geo term is fixed — it is the static lift plus any pressure difference between the reservoirs. The k · Q² term is variable and grows with the square of the flow rate, because both friction and minor losses scale with v² (and v is proportional to Q). At zero flow the pump only overcomes the lift; as the flow rate rises, friction takes over and the curve climbs steeply.

Friction losses come from Darcy-Weisbach:

h_f = f · (L/D) · v²/(2g)

and the minor losses (bends, valves, reducers, inlets and outlets) from the sum of the K coefficients:

h_loc = (ΣK) · v²/(2g)

The friction factor and the Reynolds number

At the heart of friction loss is the friction factor f. For turbulent flow (Re > 4000) it follows the implicit Colebrook-White equation, a function of the relative roughness ε/D and the Reynolds number Re = ρ·v·D/μ. Because Colebrook is implicit, it is solved by the Serghides estimator followed by 2 Newton steps on the transformed form — the explicit Serghides + Newton scheme, reaching machine precision (relative error < 7×10⁻¹⁶) with no variable convergence loop:

1/√f = −2·log₁₀(ε/(3.7·D) + 2.51/(Re·√f)) (Colebrook-White; solved by Serghides + Newton)

For laminar flow (Re < 2300), f = 64/Re is used; in the transition zone the value is interpolated. Choosing the correct absolute roughness ε of the material (≈ 1.5×10⁻⁶ m for PVC, ≈ 4.6×10⁻⁵ m for commercial steel) is decisive: it changes the friction and therefore the operating flow rate.

Total dynamic head (TDH)

The total energy the pump must deliver, in meters of fluid column, is the TDH:

H = (p₂ − p₁)/(ρ·g) + (z₂ − z₁) + (v₂² − v₁²)/(2g) + h_f + h_loc

For tanks open to the atmosphere, the pressure terms and the kinetic term cancel, and the TDH reduces to static lift + total losses. It is this value of H, at the design flow rate, that must be matched against the pump curve.

How the method finds the operating point

The calculation proceeds in stages:

1. Build the pump curve from three catalog points by Lagrange fitting, obtaining H_pump(Q).
2. Build the system curve by computing, for each test flow rate, the velocity v = 4Q/(πD²), the Reynolds number, the friction factor and the losses h_f + h_loc, adding them to H_geo.
3. Solve the intersection by bisection: evaluate the function g(Q) = H_pump(Q) − H_sys(Q) over a flow range and narrow the range until g(Q) ≈ 0. That flow rate is the operating point.
4. Repeat for three tank levels (full, mid and empty), because each level changes H_geo and shifts the point.

Bisection is preferred for its robustness: the curves are monotonic over the range of interest, guaranteeing convergence without numerical oscillation.

NPSH check and cavitation

Finding the operating point is not enough — you must ensure the pump does not cavitate. The NPSH available, the pressure margin at the suction above the vapor pressure, must exceed the pump’s NPSH required:

NPSHa = (p_atm − p_v)/(ρ·g) ± z_suc − h_f,suc ≥ NPSHr + margin

The critical condition occurs with the suction tank full and at maximum flow: that is when NPSHr is highest and NPSHa lowest. A margin of 0.5 to 1.0 m is recommended (the Hydraulic Institute suggests larger values for big pumps or hot liquids). Skipping this check is the most frequent root cause of impeller erosion and premature performance loss.

Practical design considerations

• Keep the velocity in range: 0.6–1.5 m/s in the suction (protects NPSH) and 1.5–3.0 m/s in the discharge (controls losses and water hammer).
• Check the position relative to BEP: the operating point should land between 70 % and 120 % of BEP to preserve efficiency and service life.
• Account for the level range: sizing only for the mid level hides the cavitation scenario (full) and the minimum-flow scenario (empty).
• Align standard and method: NBR 12214 governs supply systems; HI 9.6.1/9.6.3 rules NPSH and operating region; the friction criterion follows Colebrook-White (Serghides).

Following this chain — system curve, friction factor, TDH, operating point by bisection and NPSH verification — yields a sizing that is numerically rigorous and stands up to field reality.